Datasets:

haritzpuerto
/

the_pile_00_PubMed_Central

Dataset card Viewer Files Files and versions Community

Split (3)

train · 190k rows

text stringlengths 4 2.52M	meta dict
INTRODUCTION ============ Bee venom (BV) from the honeybee (Apis mellifera L.) possesses a variety of different peptides including melittin, apamin, adolapin and mast cell degranulating peptide [@B015]. In addition, it contains biologically active amines (histamine, epinephrine) and a few non-peptide components including lipids, carbohydrates and free amino acids [@B010]. BV has been used as a complementary medicine to treat such conditions as rheumatoid arthritis [@B011]; [@B015] and cancerous tumors [@B009]; [@B008]; [@B016]; [@B014]; [@B012]. Recently BV also has been used as a cosmetic ingredient for antiaging, anti-inflammatory and antibacterial functions. Pure BV is generally obtained by collecting a large amount of BV by electric stunning using a BV collector without harming the honey bees, removing impurities from the collected BV, and lyophilizing the resultant. We previously reported skin photoprotective action of BV through reduction of protein levels of matrix metalloproteinases which are main contributors to photoaging processes was found in our another study [@B004]. For the purpose of accessing BV further as a cosmetic ingredient and a potential external application for topical uses, we performed studies for the skin sensitization. Assessment of skin sensitization potential is an important part of any toxicology program for new consumer products to safe guard human beings against the possible adverse effects [@B017]. MATERIALS AND METHODS ===================== *Bee venom.* Colonies of natural honeybees (Apis mellifera L.) used in this study were maintained at the National Academy of Agricultural Science (NAAS), Suwon, Korea. BV was collected by a bee venom collecting device (Chunglin, Korea) in a sterile manner under strict laboratory conditions. In brief, the bee venom collector was placed on the hive, and the bees were given enough electric shock to cause them to sting a glass plate from which dried bee venom was later scraped off. The collected venom was purified by method of [@B005]. Purified BV was stored in a refrigerator for later use. *Animals.* Experiments were performed on fifty healthy, young 5 weeks old male guinea pig (Hartley Guinea pig, weight 318\~369 g, Samtako Bio, Osan, Korea). Animals were visually examined at the time of receipt and housed for 11 days in the animal room. They were maintained under controlled environmental conditions (temperature 23 ± 3℃; relative humidity 55 ± 10%; 12 : 12 hours light : dark cycle; ventilation rate 10\~20/hours; illumination 150\~300 lux), providing ad libitum access to a commercial guinea pigs diet (Harlan Teklad, USA) and water. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) in Korea Institute of Toxicology, KRICT and conducted in the facility approved by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International. *Skin sensitization test (Buehler test).* Dose levels for the study were decided the result of the preliminary study (not shown). HCA (Hexylcinnamaldehyde, CAS no. 101- 86-0, Sigma) solution of 10% was chosen for positive control. [Table 1](#T001){ref-type="table"} is showed group assignment of animals. The induction and challenge BV was dissolved in vehicle at 10 mg/ml for the administration. The positive control was dissolved in 80% ethyl alcohol for the induction and in acetone for the challenge, respectively. Induction was performed on the left flank and challenge on the right flank. To use topical (patch) application, application site was cleared of hair (closely clipped). A cotton pad (2 × 2 cm^2^) was fully loaded with test or control item solution and held on the application site by an occlusive patch for 6 hours exposure, the patch was removed and the application site was washed with tap water. The induction period was once a week total three times. The challenge period was 2 weeks after the last induction. Individual weight of guinea pigs was determined and they were observed for any clinical signs. Mortality was recorded during the observation period. The challenge sites were cleared of hair 21 hours after removing the patch. Approximately 3 hours later and 24 hours later, skin responses were recorded according to the grading scale as [Table 2](#T002){ref-type="table"} and [3](#T003){ref-type="table"}. After the observation of skin response, all animals were euthanized with CO~2~ overdose. ###### Experiment group assignment in guinea pigs ----------- -------------- ---------------- ----- Group No of animal Administration Induction Challenge G1^a)^ 20 BV BV G2^b)^ 20 HCA^d)^ HCA G3^c)^ 10 Ethyl alcohol HCA ----------- -------------- ---------------- ----- ^a)^Group G1: sensitized with BV and challenged with BV. ^b)^Group G2: sensitized with the HCA and challenged with HCA. ^c)^Group G3: sensitized with ethyl alcohol and challenged with HCA. ^d)^HCA(Hexylcinnamaldehyde) is positive control. ###### Evaluation of primary skin response ------- ------------------------------- Score Skin response 0 No visible change 1 Dispersed or blotchy erythema 2 Moderate diffused erythema 3 Marked erythema and oedema ------- ------------------------------- ###### Classification of skin sensitization level ------------------------ ------- ---------------- Sensitization rate (%) Grade Classification 1\~80 I Weak 9\~28 II Mild 29\~64 III Moderate 65\~80 IV Strong 81\~100 V Extreme ------------------------ ------- ---------------- *Statistical analysis.* The body weights collected during the study were analysed with F-test to examine variance homogeneity. The t-test was conducted to determine whether the BV-sensitization group was significantly different from the vehicle control-sensitization group. The level of significance was taken as P \< 0.05 or 0.01. Statistical analysis was performed by using Path/Tox System (ver 4.4.4, Xybion medical Systems Corporation, USA) according to SOPs of KIT, KRICT. RESULTS ======= *Mortality, clinical signs and body weight.* The BV application to the guinea pig skin revealed no appreciable clinical signs throughout the observation period of 31 days and there was no mortality seen ([Table 4](#T004){ref-type="table"}). Also there was no significant change in body weight of the guinea pigs from BV application during the observation period ([Fig. 1](#F001){ref-type="fig"}). In the skin sensitization test, no fur, erythema, scab or any other reactions were observed in BV ([Table 5](#T005){ref-type="table"}). *Observation of application sites.* Approximately 30 and 54 hours from the start of the challenge application (24 and 48 hours from the removing patch), the skin reaction was observed. The average skin reaction scores in the groups BV, positive control, and ethyl alcohol at 30 hours were 0.0, 0.55, and 0.0, respectively ([Table 6](#T006){ref-type="table"}). The sensitization rates at the same time point were 0, 40, and 0%, respectively. The average skin reaction scores in the group BV, positive control, and negative control at 54 hours were 0.0, 0.25, and 0.0, respectively. The sensitization rates at the same time point were 0, 20, and 0%, respectively. ###### Mortality of skin sensitization in guinea pigs -------- -------------- ----------------- ----------- ----------- --- ------ Group Dosing Phase Final mortality 1 day ≤ 1 weeks ≤ 2 weeks ≤ 3 weeks ≤ 4 weeks G1^a)^ 0 0 0 0 0 0/20 G2^b)^ 0 0 0 0 0 0/20 G3^c)^ 0 0 0 0 0 0/10 -------- -------------- ----------------- ----------- ----------- --- ------ ^a)^Group G1: sensitized with BV and challenged with BV. ^b)^Group G2: sensitized with the HCA and challenged with HCA. ^c)^Group G3: sensitized with ethyl alcohol and challenged with HCA. ![Body weights of skin sensitization in guinea pigs. Group G1; sensitized with BV and challenged with BV, Group G2; sensitized with the HCA and challenged with HCA, Group G3: sensitized with ethyl alcohol and challenged with HCA.](toxicr-28-1-g001){#F001} DISCUSSION ========== BV therapy is a treatment modality that may be thousands of years old [@B013] and involves the application of live bee stings to patient skin or, in more recent years, the injection of BV into the skin with a hypodermic needle [@B002]; [@B001]. BV also has been reported to be effective in treating allergies, scarring, burns, and skin diseases [@B004]. The EU Research Project CAESAR was responsible for developing robust QSARs for five toxicological endpoints of regulatory importance, one of which was skin sensitization. A skin sensitizer is a substance that will induce an allergic response following skin contact. Substances are classed as skin sensitizers, if there is evidence in humans that the substance can induce sensitization by skin contact in a substantial number of persons, or where there are positive results from an appropriate animal test [@B003]. This study was performed to determine the skin sensitively of BV. In Buehler test of a skin sensitization test using guinea pigs, the induction with patch application was repeated three times at intervals of one week and the challenge application was performed 2 weeks after the last induction application. The average skin reaction score and sensitization rate in the BV group at 30 hours were 0 and 0%, respectively. The skin response on the same site at 54 hours showed the average skin reaction of 0 and sensitization rate of 0%. ###### Clinical signs of skin sensitization in guinea pigs ----------------------------- ---------- -------- -------- ----- ------ --- Group G1^a)^ G2^b)^ G3^c)^ \#^d)^ \% \# \% \# \% Fur, Hair, Coat loss of fur 0/20^e)^ 0 10/20 50 0/10 0 Skin Scab 0/20 0 20/20 100 0/10 0 Erythema 0/20 0 20/20 100 0/10 0 ----------------------------- ---------- -------- -------- ----- ------ --- ^a)^Group G1: sensitized with BV and challenged with BV. ^b)^Group G2: sensitized with the HCA and challenged with HCA. ^c)^Group G3: sensitized with ethyl alcohol and challenged with HCA. ^d)^Number of animals. ^e)^Number of animals with sign/Total of animals observed. ###### Results of skin sensitization in guinea pigs -------- ---------------- --------------------------------- ------------------------ -------------------- ---------------- ----- ---------- Group No. of animals Average sensitization score^a)^ Sensitization rate^b)^ Maximization grade 30 h 54 h 30 h 54 h Grade Classification G1^c)^ 20 0.0 0.0 0% 0% I Weak G2^d)^ 20 0.55 0.25 40% 20% III Moderate G3^e)^ 10 0.0 0.0 0% 0% I Weak -------- ---------------- --------------------------------- ------------------------ -------------------- ---------------- ----- ---------- ^a)^Average sensitization score = (ΣSensitization score of each animal)/Number of animals observed. ^b)^Sensitization rate(%) = (Number of animals with positive response/Number of animals observed) × 100. ^c)^Group G1: sensitized with BV and challenged with BV. ^d)^Group G2: sensitized with the HCA and challenged with HCA. ^e)^Group G3: sensitized with ethyl alcohol and challenged with HCA. In conclusion, skin sensitization level of BV was classified as Grade I. BV is not considered to induce skin sensitization since the skin sensitization ratio was 0%. Since BV has recently been reported to possess antibacterial effect against acne-inducing bacteria and effect of wound healing [@B006]; [@B007], it is timely and appropriate to endeavour toxicological approach to BV for the possible adverse effects with the intent of using BV in cosmetic and medical applications. This work was supported by a grant from BioGreen21 Program, Rural Development Administration (Code \#: PJ008130012011). Republic of Korea.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== 1.1. Alzheimer\'s Disease: A Well-Known yet Untreatable Age-Related Neurodegenerative Disorder {#sec1.1} ---------------------------------------------------------------------------------------------- Alzheimer\'s disease (AD), the most common neurodegenerative disorder, as well as dementia type, is characterized by extracellular senile beta-amyloid protein (Aβ) plaques and intracellular neurofibrillary tau tangles \[[@B1]\]. There are two types of AD: (i) the sporadic form of AD (SAD) whose onset occurs above the age of 65 and (ii) the familial AD forms (FAD) that are more rare with less than 1% occurrence among the AD cases and whose onset starts at a younger age (\<65 years). The biological system of aging is the major risk factor of SAD \[[@B2]\]. The familial forms (FAD) bear inheritable mutations in the amyloid precursor protein (APP) and presenilin 1 and presenilin 2 genes \[[@B3], [@B4]\]. The symptoms of AD are the same in SAD and FAD \[[@B5]\]. There are different types of age-related cognitive diseases which differ in severity. SMI (subjective memory impairment) is the condition when nondemented aged people express subjective complaints related to their memory but have no organic or identifiable condition \[[@B6]\]. SMI is discussed as an early predictor of dementia \[[@B7]--[@B10]\]. The concept of mild cognitive impairment (MCI) defines an intermediate stage between normal aging and dementia. MCI patients show mild but measurable changes in cognitive tests and thinking abilities that are noticeable to the patients and to family members, but they are able to carry out everyday activities. Approximately 15-20% of people aged 65 or older have MCI. This group of people represents a population at increased risk for developing dementia, especially MCI involving memory problems \[[@B11]\]. The occurrence of MCI in the population is 3.2%, of which 11.1% of the cases convert to dementia within 3 years \[[@B12]\]. It has been indicated, yet not conclusively, that SMI is a precursor of MCI which can then lead to dementia or AD \[[@B4], [@B13]\]. Dementia is a more severe condition compared to SMI and MCI which affects aged people and interferes negatively in the performance of everyday activities. It is described as a cluster of symptoms related to mental, cognitive, and memory decline \[[@B12], [@B14]\]. There are different forms of dementia, such as AD, the most common type, and vascular dementia. Vascular dementia (VaD) is the second most common form of dementia and occurs as a cognitive decline due to a reduced blood flow in the brain (e.g., due to brain injury or stroke). However, sometimes different kinds of dementias coexist and their discrimination is difficult due to overlapping clinical symptoms. Moreover, many of these patients also suffer from psychiatric or behavioural problems that are sometimes referred to as BPSD (behavioural and psychological symptoms of dementia) or NPS (neuropsychiatric symptoms), including irritability, anxiety, psychosis, and aggression \[[@B15]\]. 1.2. Mitochondria and Neuroplasticity {#sec1.2} ------------------------------------- Mitochondria are essential yet independent organelles contained in eukaryotic cells, and they are responsible for numerous functional activities within the cells. However, they are not always an intrinsic structure of eukaryotic cells. They occur through the endosymbiosis of an alpha-proteobacterium into a prokaryotic progenitor, and this is why they contain their own DNA, namely, the mitochondrial DNA (mtDNA) \[[@B16]\]. Regarding the structural characteristics of these organelles, they contain two structurally and functionally distinct membranes, the outer and the inner membranes. The inner membrane encapsulates the matrix and also carries the electron transport chain (ETC) where oxidative phosphorylation (OXPHOS) is taking place. mtDNA is located in the matrix encoding 13 proteins which are used as structural components of the ETC complexes \[[@B17]\]. Mitochondria have obtained the title of "powerhouse of the cell" due to their ability of producing the energy, mainly through OXPHOS, required for the survival and functioning of the cell. Actually, they are more than just a "powerhouse" as they are the ultimate multitaskers which define the cell fate. Apart from the production of energy in the form of ATP, mitochondria are the key modulators of brain cell survival and death by controlling calcium (Ca^2+^) and redox equilibrium (which in turn affects neurotransmitter release and neuronal plasticity), by producing reactive oxygen species (ROS), and by controlling cell apoptosis \[[@B17]--[@B19]\]. The brain is an organ which requires a considerable amount of energy in order to operate, maintain, and enhance neuronal functions and plasticity. Neurons are postmitotic polarized cells with significant energy demands. OXPHOS, taking place in mitochondria, is the main energy provider in the form of ATP, and neurons depend almost solely on this procedure in order to satisfy their energy needs \[[@B20]\]. In particular, neurons direct this energy into the formation of interconnections, the synapses. The number and strength of these neuron interconnections define synaptic plasticity, which is responsible for cognitive function \[[@B21]\]. Synaptic plasticity is a crucial mechanism by which the neural activity generated by an experience alters synaptic transmission and therefore modifies brain function \[[@B22]\]. Neurite outgrowth is a process wherein developing neurons generate new projections as they grow in response to guidance cues. Nerve growth factors (NGF), or neurotrophins, are one family of such stimuli that regulate neurite growth \[[@B23]\]. Brain-derived neurotrophic factor (BDNF) exerts several actions on neurons ranging from the acute enhancement of transmission to long-term promotion of neurite outgrowth and synaptogenesis \[[@B24], [@B25]\]. Synaptic plasticity includes the dynamic regulation of long-term potentiation (LTP), spine density, and the number and length of dendrites and axons (neuritogenesis), as well as neurogenesis. Adult neurogenesis generates functional neurons from adult neural precursors in restricted brain regions throughout life \[[@B26]\]. These plasticity processes need a high energy requirement, and this is why mitochondria play such a pivotal role in the well-being of neurons especially when neurons need to adapt to periods of pathologically reduced functions. 1.3. Mitochondria, Oxidative Stress, Aging, and AD {#sec1.3} -------------------------------------------------- However, while mitochondria regulate the functions of healthy neurons, they are also largely affected during aging and pathological states such as age-related neurodegenerative diseases. Mitochondria are not only the regulators of energy metabolism but are also the main ROS generators. ROS are immensely reactive species which are produced in mitochondria mainly by complexes I and III of the ETC when there is a leak of electrons. They are chemical species including hydroxyl radical (^∙^OH), superoxide anion (O~2~-), and hydrogen peroxide (H~2~O~2~) which can interact with and damage DNA and proteins and lipids which can compromise cell survival leading to aging and to vulnerability to several diseases \[[@B27], [@B28]\]. When they exist at normal levels, they constitute signalling agents in many physiological processes, such as redox homeostasis, cellular death, cellular senescence, and cell proliferation, and they can also trigger immune responses, synaptic plasticity, and cognitive enhancement \[[@B20], [@B27]\]. ROS are neutralized by antioxidant enzymes such as superoxide dismutase, which transforms the radicals (O~2~-) into H~2~O~2~, and by catalase, glutathione peroxidase, and thioredoxin peroxidase, which diffuse H~2~O~2~ \[[@B27]\]. In a healthy state, there is a balance between ROS production and neutralization. Nevertheless, when ROS are produced in excess, e.g., during aging, they directly affect mitochondria since mitochondrial membranes consist of long polyunsaturated fatty acids which are easily oxidized. Also, mtDNA is found in close proximity to the ROS source and is susceptible to mutations resulting in the production of faulty ETC proteins, leading back to the production of more ROS \[[@B18], [@B28]\]. It could be said that mitochondria are the main organelles in aging and neurodegeneration by being both generators and targets of ROS. It has been shown that aging is characterized by a rise in oxidative stress, a decline in antioxidant defense systems, and an impairment of the OXPHOS. So, aging is characterized by energy deprivation and a shift of the redox state towards oxidation. Mitochondria are at the center of these hallmarks \[[@B20]\]. Neurons, which highly depend on OXPHOS to satisfy their energy demands, are particularly susceptible to energy hypometabolism \[[@B20]\]. In addition, taking into account that they are nondividing cells, neurons are almost as old as the entire organism and are not replaced during life with the exception of the hippocampus that continuously generates new neurons during adulthood \[[@B20], [@B29]\]. This means that neurons accumulate oxidative stress and therefore defective mitochondria during aging \[[@B20], [@B30]\]. This is particularly important since mitochondrial dysfunction represents an early event in AD pathogenesis \[[@B20], [@B28], [@B31]\]. Intense oxidative stress and decreased brain energy metabolism are common characteristics of both normal aging and AD, although to different extents \[[@B20]\]. Of note, mitochondrial abnormalities are observed in FAD and SAD brains \[[@B32], [@B33]\]. On one hand, recent data obtained from AD models, in which mitochondrial failure is a prominent feature, implicate tau hyperphosphorylation as well as Aβ overproduction and deposition. On the other hand, Aβ and tau target mitochondria synergistically, thereby possibly amplifying each other\'s toxic effects. This interrelationship of Aβ, tau, and mitochondrial function constitutes a vicious cycle \[[@B34]\]. The mitochondrial cascade hypothesis postulates that mitochondrial dysfunction represents the most upstream pathology in AD \[[@B28]\]. According to this hypothesis, arresting brain aging will prevent the development of AD \[[@B32]\]. 1.4. Mitochondria-Directed Natural Compounds {#sec1.4} -------------------------------------------- The current mitochondrial cascade hypothesis postulates mitochondrial dysfunction as a central pathomechanism in age-related degenerative disorders \[[@B28], [@B35], [@B36]\]. Taking into account their primary role in aging and in the early stages of AD, mitochondria constitute promising targets for therapeutic strategies. For this reason, pharmacological studies are directed in enhancing mitochondrial functions such as ATP production and respiration or in reducing mitochondrial harmful by-products such as ROS \[[@B36]\]. To date no drugs are able to cure or stop the progression of age-related neurodegenerative disorders. Most of them may be beneficial in delaying the progression of AD and only partially treat some of its symptoms (e.g., confusion and memory loss). Many drugs including whole plant extracts and single compounds originate from natural and botanical sources. Two single compound AD drugs are derived from plants: (i) the acetylcholine-esterase inhibitor, galanthamine, derived from the Galanthus species (Galanthus caucasicus and Galanthus woronowii) and (ii) rivastigmine, a physostigmine analogue (physistigmine was isolated from the Calabar bean, Physostigma venenosum) \[[@B37], [@B38]\]. In addition, the phytopharmacon GBE that is used as antidementia medicine was shown to improve mitochondrial function emphasizing the concept of targeting mitochondria as an emerging and promising therapeutic approach \[[@B35], [@B39]\]. Therefore, we focused our search on natural compounds that possess mitochondria-enhancing properties based on our own past and ongoing research as well as on research of other groups. Standardized Gingko biloba extract (GBE), resveratrol, phytoestrogens, and the natural neurosteroid allopregnanolone fulfilled our criteria. Common targets of these agents ([Figure 1](#fig1){ref-type="fig"}) have been reported, such as ROS, mitochondrial membrane potential (MMP), Aβ, tau protein, antiapoptotic protein (Bcl-2), and OXPHOS (Figures [2](#fig2){ref-type="fig"} and [3](#fig3){ref-type="fig"}). Accordingly, in this review we aimed to summarize the molecular modes of action of these natural agents with special focus on mitochondria, their mitochondrial function-enhancing properties, and their antioxidant properties. We discuss evidence on their mechanism of action from preclinical as well as clinical studies. Especially regarding clinical trials, there is a different level of existing evidence for each phytochemical. GBE, resveratrol, phytoestrogens, and allopregnanolone appear in a descending order according to their clinical evidence level. The databases PubMed and Google Scholar, as well as the database ClinicalTrials.gov were used for our search with a focus on the years 2000--2018. For the clinical evidence, we considered randomized, double-blind, placebo-controlled trials as well as ongoing trials, systematic reviews, meta-analyses, and Cochrane analyses. 2. Pharmacologic Features of Natural Substances in Alzheimer\'s Disease {#sec2} ======================================================================= 2.1. Gingko biloba {#sec2.1} ------------------ Gingko biloba has existed for over 250 million years and is a native from Japan, Korea, and China; however, it can be found worldwide. Traditional Chinese clinicians originally utilized GBE for a variety of applications \[[@B40]\]. There are several Ginkgo biloba extracts sold on the market, including standardized and nonstandardized extracts, which are also used in studies. The standardized extracts have to meet specific criteria regarding their manufacturing process, the quality of the plant material, and their composition, which is not the case with the nonstandardized extracts. Many products have been reported on the market which are not standardized and are even adulterated. These products not only reduce the efficacy of GBE, but they can be potentially harmful \[[@B41]\]. GBE contains two main groups of active constituents ensuring its medicinal effects: terpenes (including bilobalide and ginkgolides A, B, and C) and flavonoids (including meletin, isorhamnetin, and kaempferol). Both the United States Pharmacopoeia and the European Pharmacopoeia define as standardized only extracts that contain the active components of Ginkgo in a certain and defined content. In particular, the standardized extracts should contain 5-7% triterpene lactones, 22-27% flavonoids, and less than 5 ppm of ginkgolic acids, which are toxic ingredients of Ginkgo. \[[@B42]\]. Most toxicological, pharmacological, and clinical investigations have focused on the neuroprotective value of two main standardized extracts labeled EGb761 and LI 1370 \[[@B43]--[@B45]\]. The EGb761 extract consists of 24% flavone glycosides (mainly quercetin, kaempferol, and isorhamnetin) and 6% terpene lactones (2.8-3.4% ginkgolides A, B, and C and 2.6-3.2% bilobalide), while the extract LI 1370 is composed of 25% ginkgo flavone glycosides as well as 6% terpenoids. Several terpene lactones (ginkgolides and bilobalide) show substantial mitochondria-protecting properties, while the flavonoid fraction seems to play an important role in the free radical scavenging properties \[[@B46]\]. In the following parts, only the effects of standardized GBE will be discussed. ### 2.1.1. Mechanisms of Action Based on Preclinical Evidence {#sec2.1.1} 2.1.1.1. Direct Effects of GBE on Mitochondria. Several findings demonstrate the mitochondria-modulating effect of GBE, mainly in cellular and animal models of AD. In particular, GBE has been shown to attenuate effectively mitochondrial dysfunction through several mechanisms of action, such as antioxidant effect and free radical scavenging properties, with all the evidence leading to this conclusion having been reviewed extensively \[[@B35], [@B47]--[@B49]\]. In vitro, GBE was shown to ameliorate mitochondrial function by improving MMP and ATP levels at a low concentration of 0.01 mg/ml in pheochromocytoma cells (PC12) cells \[[@B46]\]. In amyloid precursor protein- (APP-) transfected human neuroblastoma cells, an AD cellular model with increased Aβ generation, GBE improved respiration of mitochondria, stimulated mitochondrial biogenesis, and increased ATP production \[[@B50]\]. Mitochondria-related modes of action of GBE are summarized in [Figure 2](#fig2){ref-type="fig"}. 2.1.1.2. Effects of GBE on Oxidative Stress, Aβ, and Tau Toxicity Related to Damage of Mitochondria. Aβ plaque deposition is one of the main hallmarks of AD. The overexpression of both Aβ itself and its precursor protein, the amyloid precursor protein (APP), has been used to create cellular and animal models of AD. GBE has been shown to be effective in reducing both the deposition of Aβ and its toxicity. In detail, the prooxidant Aβ~25-35~ peptide treatment was shown to decrease complex I and IV activities and to increase the level of reactive oxygen/reactive nitrogen species (ROS/RNS) in SH-SY5SY cells \[[@B51]\]. Thus, pretreatment with GBE was able to reduce the Aβ-related increase in ROS/RNS levels as well as to ameliorate the complex I and IV activities \[[@B51]\]. GBE protected against Aβ~1-42~ oligomer-induced neurotoxicity and cell damage with an indirect effect on SH-SY5Y neuroblastoma cells by improving Hsp70 protein expression and subsequently by activating the Akt (protein kinase B) pathways as well as ER stress \[[@B52]\]. GBE also attenuated Aβ~1-42~ oligomer-induced cell damage and protected against Aβ toxicity and oxidative stress \[[@B53], [@B54]\], as well as apoptosis \[[@B52]\]. GBE was also able to reduce Aβ production \[[@B55]\]. In terms of animal models, a chronic treatment with GBE improved cognitive defects in a transgenic mouse model of AD (Tg2576), a model that overexpresses a mutant form of APP \[[@B53]\]. GBE was also shown to decrease Aβ oligomers and to significantly increase neuronal proliferation in the hippocampus of young (6 months) and old (22 months) mice in a double transgenic mouse model (TgAPP/PS1) \[[@B54]\]. A chronic daily treatment with GBE for 6 months improved the cognitive function and alleviated amyloid plaque deposition in two-month-old APP/PS1 mice. Of note, GBE treatment seems to decrease the level of insoluble Aβ, while the soluble content of Aβ was unchanged \[[@B56]\]. GBE reduced the hyperphosphorylation of tau at AD-specific Ser262, Ser404, Ser396, and Thr231 sites, rescued the activity of tau phosphatase PP2Ac and kinase GSK3β, and reduced the oxidative stress in the hippocampus and prefrontal cortex on a hyperhomocysteinemia-treated rat model of AD. Memory lesions were also restored, and the expression of synapse-associated protein PSD95 and synapsin-1 protein was upregulated \[[@B57]\]. 2.1.1.3. Effects of GBE on Neuroplasticity Pathways. GBE exerts its beneficial effects not only by acting on the Akt pathway, as aforementioned, but also by acting on the cyclic AMP response element-binding protein (CREB) \[[@B54], [@B58], [@B59]\]. CREB phosphorylation induces transcriptional activation which results in the expression of BDNF, and therefore, in synaptic plasticity and cognitive enhancement. Conversely, lack of CREB phosphorylation is a pathological ailment of neurodegenerative diseases such as AD \[[@B60]\]. In detail, the administration of GBE restored CREB phosphorylation in the hippocampus of TgAPP/PS1 mice \[[@B54]\]. Quercetin and bilobalide are the major constituents that have contributed to GBE-induced neurogenesis \[[@B58]\]. Both constituents promoted dendritic processes in hippocampal neurons and restored Aβ oligomer-induced synaptic loss, as well as restored CREB phosphorylation \[[@B58]\]. Ginkgo flavonols quercetin and kaempferol have been shown to stimulate BDNF and phosphorylation of CREB in neurons isolated from double transgenic AD mouse (TgAPPswe/PS1e9) \[[@B59]\]. Recently, our team could confirm the neurite outgrowth stimulating effects of GBE in a 3D cell culture model ([Figure 4](#fig4){ref-type="fig"}). ### 2.1.2. Clinical Evidence {#sec2.1.2} Apart from the preclinical studies, the extract has been largely investigated in clinical trials in a range of age-associated cognitive conditions from SMI and MCI to dementia and AD. GBE has been suggested for both the symptomatic treatment and prevention of those cognitive decline-related diseases. The standardized GBE is considered a phytopharmacon, and the dose of 240 mg/day is recommended as the most effective by the guidelines for biological treatment of dementias \[[@B12]\]. There are 9 categories (A, B, C, C1, C2, C3, D, E, and F) and 5 grades (1-5) of pharmaceuticals used for AD and other dementias according to the level of existing clinical evidence and the occurrence of side effects, respectively. GBE belongs to category B of the level of evidence (limited positive evidence from controlled studies) and to grade 3 \[[@B12]\]. Here, we are going to highlight evidence on the extract\'s efficacy on subgroups of age-associated cognitive conditions in an ascending severity order ([Table 1](#tab1){ref-type="table"}). 2.1.2.1. Patients with SMI and MCI. Three randomized, double-blind, placebo-controlled, parallel-group trials were conducted for patients with memory complaints, one in SMI and two in MCI patients. In total, data from 61 SMI and 460 MCI patients were evaluated. One trial conducted in healthy aged patients with SMI showed that GBE enhanced cognitive flexibility without changes in brain activation and that it mildly increased prefrontal dopamine \[[@B61]\]. Two trials showed that GBE ameliorated neuropsychiatric symptoms (NPS) and cognitive ability in patients with MCI \[[@B62]\] as well as improved cognitive functioning and aspects of quality of life in subjects with very mild cognitive impairment \[[@B63]\]. 2.1.2.2. Patients with Dementia. GBE has been found particularly efficacious in demented people with neuropsychiatric symptoms (NPS) \[[@B64], [@B65]\]. In total, 3 original papers, 1 systemic review, 6 meta-analyses, and 1 Cochrane analysis involving 14974 demented patients were evaluated. In detail, the pooled analyses of 4 published trials in a systemic review, involving outpatients with mild to moderate dementia and BPSD, demonstrated the efficacy of GBE at a daily dose of 240 mg \[[@B66]\]. Six meta-analyses (3 trials included in these meta-analyses were conducted in 1997 \[[@B67]--[@B69]\]) of 32 controlled, randomized, double-bind clinical trials and one bivariate meta-analysis of 6 trials come to the conclusion that GBE is efficacious and well tolerated in patients with a diagnosis of AD, VaD, or mixed dementia in three typical domains of assessment in dementia, i.e., cognition, activities of daily living (ADL), and clinical global judgment \[[@B65], [@B70]--[@B74]\]. However, there are also the studies with inconclusive or contrasting results to the efficacious effect of GBE in demented subjects \[[@B75]--[@B77]\]. 2.1.2.3. Patients with Specific Dementia Type: AD and Vascular Dementia. In total, data from 1 original paper, 1 review, and 3 systematic reviews and meta-analyses involving 6880 patients with AD and VaD were evaluated. In detail, in an original paper, low doses of GBE administered to patients with vascular cognitive impairment in a randomized, double-blind, placebo-controlled trial showed significant deceleration of cognitive decline versus placebo only in one (Clinical Global Impression) of the four tests conducted in the trial \[[@B78]\]. The systematic reviews and meta-analyses (3 trials included in these meta-analyses were conducted in 1997 \[[@B67]--[@B69]\]) concluded that GBE exerts potentially beneficial effects on the improvement of activities of daily living, cognitive function, and on global clinical assessment in patients with MCI or AD, in mainly the AD type of dementia and in aged people with VaD having NPS \[[@B79]--[@B82]\]. 2.1.2.4. Prevention. The preventive effect of GBE was reported in 14812 patients in three original papers and one systematic review and meta-analysis. In contrast, there are 4 studies that do not support the efficacious effect of GBE in preventing the onset of AD in either healthy aged or aged with MCI people \[[@B83]--[@B86]\]. The outcome for the efficacy of GBE in preventing the onset of AD in healthy individuals varies among different studies. However, there is also high variability in the design of the studies in terms of GBE doses, duration of the treatment, sample size, statistical tools, and compliance with the medication. Therefore, there is space for criticism regarding the methodological design of studies and the interpretation of the outcome. There are two large studies which form good examples of scepticism towards their negative outcome: the GEM study and the GuidAge study \[[@B83], [@B84], [@B87]\]. The GEM study was conducted in healthy old people (80 years old or more) and found no efficacy of GBE in preventing the onset of AD. In this study, the compliance of subjects with the treatment was nonadequate, yet this parameter was not taken into account in the interpretation of the results. In the GuidAge study, the conversion rate from memory complaints to dementia was examined in aged people with memory complaints and no difference was found between GBE and placebo. However, the statistical power for the analysis of hazards was found low. The selection of suitable statistical methods to take into account increasing hazards overtime is crucial for meaningful results and increased significance \[[@B35]\]. Based on the included studies, GBE has been reported in only a few studies that show no effect. The majority of the recent studies demonstrated that the treatment with doses up to 240 mg/day was safe, well-tolerated, and efficacious against age-related disorders. In summary, GBE has been proven more effective in patients with cognitive impairment at baseline than preventing the onset of cognitive impairment in healthy aged subjects. As mentioned before (see Introduction), mitochondrial dysfunction is more profound in cognitive disorders than in normal aging. Similarly, GBE shows increasing promising effects with increasing cognitive impairment. This, again, represents an indicator that GBE exerts its effects clinically by acting on mitochondria \[[@B35]\]. Thus, we can conclude that GBE can potentially improve mitochondrial dysfunction across the aging spectrum. 2.2. Resveratrol {#sec2.2} ---------------- Resveratrol, known as a polyphenol from white hellebore (Veratrum grandiflorum), was discovered by Takaoka (1939) as a component of several dietary sources such as berries, peanuts, and red grape skin or wine. Siemann and Creasy discovered that resveratrol is present at high concentration in red wine \[[@B88]\]. Resveratrol has been reported to possess several benefits, including antitumor, antioxidant, antiaging, anti-inflammatory, cardioprotective, and neuroprotective properties. This polyphenol has emerged as a novel natural agent in the prevention and possible therapy of AD \[[@B89]\]. ### 2.2.1. Mechanisms of Action Based on Preclinical Evidence {#sec2.2.1} In vitro and in vivo, the direct molecular targets of resveratrol are not known in detail. However, there is evidence that resveratrol exerts a complex mode of actions through the protection of mitochondrial function and the activation of biogenesis, through its effect on certain signalling pathways, through its antioxidant effects, through the increase of Aβ clearance, and through the reduction of Aβ neurotoxicity \[[@B90]\] ([Figure 3](#fig3){ref-type="fig"}). 2.2.1.1. Direct Effects of Resveratrol on Mitochondria. Dietary supplementation with 0.2% (w/w) resveratrol suppressed the aging-associated decline in physical performance in senescence-accelerated mice (SAMP1) at 18 weeks of age by improving several mitochondrial functions such as the activity of respiratory enzymes, oxygen consumption, and mitochondrial biogenesis, as well as the activity of lipid-oxidizing enzymes \[[@B91]\]. In 18-month-old aged mice, resveratrol (15 mg/kg/day) and/or exercise for 4 weeks were able to counteract aging-associated oxidative damage targeting mitochondrial biogenesis and function by causing overexpression of peroxisome proliferator-activated receptor-gamma coactivator (PGC-1α) mRNA and by increasing citrate synthase enzyme activity \[[@B92]\]. Mitochondrial biogenesis is induced by resveratrol through SIRT1 activation and deacetylation of PCG-1α \[[@B90]\] ([Figure 3](#fig3){ref-type="fig"}). 2.2.1.2. Effects of Resveratrol on Oxidative Stress. Damaged mitochondria activate ROS production during oxidative stress which is involved in apoptosis \[[@B93]\]. ROS may damage the mitochondrial and cellular proteins and nucleic acids, causing lipid peroxidation and resulting in the loss of membrane integrity \[[@B94]\] ([Figure 3](#fig3){ref-type="fig"}). Resveratrol also protects mitochondria by increasing the expression of the ROS-inactivating enzymes glutathione peroxidase 1 (GPx1) and superoxide dismutase 1 (SOD1) and by reducing the expression of the ROS-producing enzyme NADPH oxidase 4 (Nox4) \[[@B93], [@B95]\] ([Figure 3](#fig3){ref-type="fig"}). In line with this, resveratrol rescued Aβ-treated human neural stem cells (hNSCs) from oxidative stress by increasing the mRNA of antioxidant enzyme genes such as SOD-1, nuclear factor erythroid 2-related factor 2 (NRF-2), Gpx1, catalase, and heme oxigenase 1 (HO-1) \[[@B96]\]. In addition, resveratrol exerted antioxidant properties and attenuated oxidative damage by decreasing iNOS and COX-2 levels \[[@B93]\]. 2.2.1.3. Effects of Resveratrol on Aβ Toxicity Related to Damage of Mitochondria. Thanks to its natural antioxidant properties and/or by sirtuin1 (SIRT1) activation, resveratrol shows a neuroprotective effect because it counteracts Aβ toxicity. In more details, resveratrol increases the clearance of Aβ through the activation of AMPK \[[@B90]\] ([Figure 3](#fig3){ref-type="fig"}). This natural molecule plays an important role in reducing Aβ neurotoxicity by phosphorylating protein kinase C delta (PKC-δ) \[[@B90]\] ([Figure 3](#fig3){ref-type="fig"}). Resveratrol also influences the Aβ-induced apoptotic signalling pathway through SIRT1 activation, including inhibiting the expression of caspase protein 3 (caspase-3), apoptotic regulator Bax, Forkhead box O (FOXO), and tumor protein p53, through blocking the activation of c-Jun N-terminal kinase (JNK) and restoring the decrease of B-cell lymphoma 2 (Bcl-2) expression, as well as through inhibiting the increase of the nuclear factor kappa-light-chain-enhancer of activated B cell (NF-κB) DNA binding \[[@B90]\] ([Figure 3](#fig3){ref-type="fig"}). Resveratrol (20 μM) protected PC12 cells against neurotoxicity caused by Aβ~25-35~ by provoking autophagy which was proven dependent on the tyrosyl tRNA synthetase-poly(ADP-ribose) polymerase 1 (TyrRS-PARP1) and SIRT1 pathway (TyrRS-PARP1-SIRT1 pathway) \[[@B97]\]. A very low concentration of resveratrol (0.2 mg/l) significantly attenuated Aβ neuropathology and AD-type deterioration of spatial memory function in Tg2576 mice compared to control \[[@B98]\]. In a transgenic mouse model of AD (Tg19959), dietary supplementation with resveratrol (300 mg/kg) decreased amyloid plaque formation \[[@B93]\]. In order to translate the animal doses into ones that are relevant in humans, a scaling factor of 0.08 is used to calculate the human equivalent dose (<http://www.fda.gov/cber/gdlns/dose.htm>). For resveratrol, this is about 24 mg/kg or 1.68 g per day for a 70 kg individual \[[@B93]\]. Resveratrol is also known to act as a phytoestrogen (this mode of action of resveratrol is discussed in more detail in Phytoestrogens). 2.2.1.4. Effects of Resveratrol on Metabolic and Signalling Pathways. Resveratrol has been suggested to regulate cellular processes by activating key metabolic proteins such as SIRT1, 5′ adenosine monophosphate-activated protein kinase (AMPK), and peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PCG-1α) \[[@B99]--[@B101]\]. Sirtuins and nicotinamide adenine dinucleotide- (NAD^+^-) dependent protein deacetylases are described as novel therapeutic agents against neurodegenerative disease pathogenesis \[[@B102]\]. In fact, the essential neuroprotective effect of resveratrol is based on the action of SIRT1 and AMPK and on the phosphorylation/acetylation status of PGC-1α that consequently activates the mitochondrial biogenesis leading to the improvement of the mitochondrial activity \[[@B103]\] ([Figure 3](#fig3){ref-type="fig"}). In a study using Aβ-treated hNSCs, the neuroprotective effect of (10 μM) resveratrol was demonstrated by the activation of the AMPK-dependent pathway by rescuing the expression levels of inhibitory kappa B kinase (IKK) and by restoring iNOS and COX-2 levels \[[@B104]\]. In the inducible p25 transgenic mouse model of tauopathy and AD, resveratrol-mediated (5 μg/μl) SIRT1 activation reduced learning impairment and hippocampal neurodegeneration \[[@B105]\]. The JAK/ERK/STAT signalling pathway (janus kinases, extracellular signal-regulated kinases, and signal transducers and activators of transcription) is implicated in cell survival, proliferation, and differentiation, while the dysregulation of the JAK/STAT pathway in neurodegenerative disorders contributes to neuronal loss, cognitive impairment, and brain damage \[[@B96]\]. Treatment with 20 mg/kg resveratrol exerted a neuroprotective effect via the JAK/ERK/STAT signalling pathway in a rat model of ischemia-reperfusion injury. In detail, resveratrol attenuated the increase in phosphorylation of JAK, ERK, STAT, and JNK caused by ischemia-reperfusion \[[@B96]\] ([Figure 3](#fig3){ref-type="fig"}). ### 2.2.2. Clinical Evidence {#sec2.2.2} Only eight clinical trials and four ongoing trials on resveratrol aim at evaluating the effects of this compound on cognitive function in humans \[[@B106]\] ([Table 2](#tab2){ref-type="table"}). Efficacy results of resveratrol are based only on one clinical trial in MCI and one in AD patients. 2.2.2.1. Young and Old Healthy Subjects. Witte et al. conducted a study to evaluate the effect of resveratrol (200 mg/day) supplementation in a formulation with quercetin 320 mg in 23 healthy overweight older individuals versus placebo during 26 weeks. They showed that resveratrol supplementation is able to improve memory performances and glucose metabolism and is able to increase hippocampal functional connectivity in older adults for the maintenance of brain health during aging \[[@B107]\]. No effect on cognitive function was detected in young healthy people \[[@B94], [@B95]\]. 2.2.2.2. Patients with Cognitive Decline and MCI. Lee et al. examined the effects of grape consumption (which contains resveratrol) on cognitive function and metabolism in the brain of patients with mild cognitive decline and demonstrated a protective effect of the grape extract against pathologic metabolic decline \[[@B108]\]. In a more recent 14-week study carried out on 80 postmenopausal women aged 45-85 years, it was proven that a regular consumption of a modest dose of resveratrol (75 mg twice daily) is able to enhance cerebrovascular function and cognition and to reduce their heightened risk of accelerated cognitive decline \[[@B109]\]. Clinical studies are underway to explore the beneficial effect of resveratrol on MCI. In the ongoing trials, one four-month resveratrol supplementation study in phase 1 aims at evaluating the efficacy and safety of bioactive dietary preparation (BDPP) at low, moderate, and high doses in treating mild cognitive impairment on 48 MCI subjects (55-85 years) \[[@B110]\]. The purpose of another study in phase 4 is to test the effect of a six-month administration of resveratrol on brain functions in MCI subjects (50-80 years) (National Institutes of Health, ClinicalTrials.gov) \[[@B111]\]. In a randomized, double-blind interventional study, resveratrol intake (200 mg/day, 26 weeks) reduced glycated hemoglobin A1c, preserved hippocampus volume, and improved hippocampus resting-state functional connectivity (RSFC) in 40 well-characterized patients with MCI (21 females, 50-80 years) \[[@B112]\]. 2.2.2.3. Patients with Moderate AD and Dementia. Class II evidence provided by the study of Turner et al. on patients with AD showed that resveratrol (500 mg/day to 2 g/day, 52 weeks) is well-tolerated, safe, and able to decrease Aβ~40~ levels in cerebrospinal fluid (CSF) and plasma but had no significant effects on cognitive score \[[@B113]\]. Recently, a phase 2 study was conducted investigating the effect of resveratrol (500 mg) in individuals with mild to moderate AD confirming its tolerability and safety as well as its modulation of AD biomarker pathways \[[@B114]\]. Currently, an ongoing study in phase 3 tests the effect of resveratrol supplementation (215 mg/day for 52 weeks) on cognitive and global functioning in mild-to-moderate AD patients (50-90 years) \[[@B115]\]. A second ongoing study in phase 3 aims at evaluating the effect of resveratrol combined with glucose and malate in slowing down the progression of AD after 12 months in mild-to-moderate AD (50-90 year old patients) \[[@B116]\]. On the basis of the results from the very few clinical trials in MCI and AD, no conclusion about the efficacy of resveratrol on cognition can be drawn at the current time, but promising trials are underway. 2.3. Neurosteroids {#sec2.3} ------------------ Neurosteroids offer therapeutic opportunities through their pleiotropic effects on the nervous system. They are a subcategory of steroids synthetized de novo from cholesterol in the central nervous system independently of supply by peripheral steroidogenic glands \[[@B117], [@B118]\] and accumulate within the brain in neurons or glial cells \[[@B119], [@B120]\]. Neurosteroids are derived from cholesterol which is translocated from the outside to the inside of mitochondria via the translocator protein (TSPO). In the inner mitochondrial membrane, cholesterol is then converted by the cytochrome cholesterol side-chain cleavage enzyme (P450scc) to pregnenolone, the precursor of all the neurosteroids \[[@B121]\]. In particular, pregnenolone and allopregnanolone play an essential role in aging, in the performance of memory, and in physiopathology. Indeed, the age-related drop of neurosteroids gives rise to neuronal degeneration and dysfunction in human and animal models owing to the loss of neurosteroid neuroregenerative and protective effects \[[@B122], [@B123]\]. Allopregnanolone is used in several studies as a plasmatic biomarker for AD because of its reduced level in the plasma of demented patients \[[@B122]\]. It is known to be a regenerative agent in the brain \[[@B124]\]. Several neurosteroids were quantified and were found decreased in postmortem brains of aged nondemented controls and aged AD patients. The transgenic mice model of AD (APPswe+PSEN1Δ9 mice) presents a decreased ability to form allopregnanolone in the hippocampus \[[@B125]\]. ### 2.3.1. Allopregnanolone {#sec2.3.1} 2.3.1.1. Mechanisms of Action Based on Preclinical Evidence. 2.3.1.1.1. Direct Effects of Allopregnanolone on Mitochondria. In control and APP/Aβ SH-SY5Y cells, allopregnanolone improved basal respiration and glycolysis as well as increased the bioenergetic activity and ATP production \[[@B126]\]. In APP-transfected cells, a pretreatment with allopregnanolone exerted a neuroprotective effect against oxidative stress-induced cell death via the amelioration of the cellular and mitochondrial energy, the reduction of ROS, and the improvement of mitochondrial respiration \[[@B126]\]. Thereby, it exerted its beneficial effect by improving the mitochondrial redox environment, such as MnSOD activity and mitochondrial ROS levels \[[@B127]\]. Moreover, allopregnanolone increased ATP levels and respiration in mouse primary cortical neurons \[[@B127]\]. In addition, in vitro, allopregnanolone potentiated mitochondrial respiration in both adult neural stem cells (NSCs), neurons, and mixed glia \[[@B128]\]. In vivo, allopregnanolone was able to restore the ovarectomized- (OVX-) induced decrease in mitochondrial respiration in both non-Tg and 3xTgAD mice \[[@B128]\]. Moreover, allopregnanolone also improved the activity of bioenergetic enzymes such as pyruvate dehydrogenase (PDH) and α-ketoglutarate dehydrogenase (αKGDH) \[[@B128]\]. 2.3.1.1.2. Effects of Allopregnanolone on Aβ Toxicity Related to Damage of Mitochondria. In a recent study, allopregnanolone was shown to exert an increased neuroprotective activity against Aβ~42~-induced cell death in neural stem cells \[[@B129]\] ([Figure 5](#fig5){ref-type="fig"}). In vivo, the natural neurosteroid allopregnanolone appears to be a promising therapeutic tool for the development of neurogenic and/or neuroprotective strategies, but diverse points have to be taken into account, including the dosing regimen, the treatment regimen, bioavailability, solubility, route of administration, and sex differences. Acute single administration of allopregnanolone promoted neurogenesis in the subgranular zone (SGZ) in the triple transgenic mouse model of AD (3xTgAD) at 3 months of age prior to the appearance of AD \[[@B71]\]. Allopregnanolone reversed memory and learning deficits in these mice. Chen et al. showed that allopregnanolone administration (once/week for 6 months) decreased Aβ generation and promoted survival of newly generated neurons in the brain of 3xTgAD \[[@B130]\]. They also demonstrated that allopregnanolone increased oligodendrocyte myelin markers and ameliorated cholesterol homeostasis and clearance from the brain by increasing the expression of PXR and Liver-X-receptor (LXR). Singh et al. reported that allopregnanolone is able to restore cognitive performance in the preplaque phase of AD as well as memory and learning in aging 3xTgAD mice \[[@B131]\]. All these studies demonstrated the neuroprotective effects of allopregnanolone against the Aβ toxicity in 3xTgAD mice and also its capacity to stimulate rodent and human neural progenitor cell proliferation and to compensate the cell loss \[[@B130], [@B132]\]. Continuous infusions of allopregnenanolone were antiregenerative, while intermittent administration promoted repair and renewal in a mouse model of AD \[[@B124]\]. The mode of action of allopregnanolone is summarized in [Figure 5](#fig5){ref-type="fig"}. 2.3.1.2. Clinical Evidence. Currently, there is only one phase I ongoing clinical trial testing the safety and the tolerability of allopregnanolone in patients with mild cognitive impairment and early AD \[[@B133]\] ([Table 3](#tab3){ref-type="table"}). The primary aim of this phase 1 study is to evaluate the maximally tolerated dose after intravenous injection of allopregnanolone (2, 4, or 6 mg, once per week for 12 weeks). Thus, no clinical evidence is currently available. The natural neurosteroid allopregnanolone appears to be a promising therapeutic tool with specific regard to its neurogenic properties besides its mitochondria-directed effects. However, more trials are urgently needed to prove that. 2.4. Phytoestrogens {#sec2.4} ------------------- Phytoestrogens are the most bioactive molecules of soy and present structural similarity to the 17β-estradiol, which is the major circulating estrogen. Specific estrogen receptors have been shown to localize in mitochondria in the frontal lobe and the hippocampus of men and women suggesting a role of estrogen in controlling cognitive functions and memory processes via energy supply \[[@B134]\]. Estrogen plays a neuroprotective role during the aging process, especially through its beneficial impact upon mitochondrial metabolism by increasing glucose utilization by cells as well as by enhancing ETC activity, by stabilizing the MMP, and by preventing ROS production and calcium-induced excitotoxicity \[[@B135]\]. Moreover, females live longer than males and this can be attributed in part to the antioxidant effect of estrogen and the upregulation of life longevity-related genes \[[@B19], [@B136]\]. The phytoestrogens are characterized by their ability to bind to estrogen receptor α and estrogen receptor β and to exert similar responses to endogenous estrogens \[[@B137]\]. Isoflavones are a subclass of phytoestrogens and are contained abundantly in soy and soybeans. Soy presents estrogenic effects attributed to genistein, daidzein, and glycitein. The most potent isoflavone is genistein, while daidzein and glycitein present an affinity to the estrogen receptor, 100-500 times lower than genistein \[[@B138]\]. Estrogen receptors are localized in the important brain areas, including the prefrontal cortex and the hippocampus that are also known to be vulnerable to age-related decline \[[@B139]--[@B142]\]. ### 2.4.1. Mechanisms of Action Based on Preclinical Evidence {#sec2.4.1} 2.4.1.1. Effects of Phytoestrogens on Aβ and Tau Toxicity and Cognitive Performance Related to Damage of Mitochondria. One of the most important phytoestrogens is resveratrol, an estrogen receptor agonist/antagonist. In particular, resveratrol acts on estrogen receptor β, whose activation is known to play a major role in cognitive processes, leading to the improvement of cognitive impairment in AD \[[@B143]\]. The soybean is a source of vegetable proteins and contains also other functional ingredients including phytoestrogens. The isoflavones genistein and daidzein have been shown to present protective effects against tau protein phosphorylation \[[@B144]\]. Animal models confirmed the neuroprotective effects of phytoestrogens. Genistein, the most active molecule of soy isoflavones, improved Aβ-induced cell death and reduced neuronal loss in rats \[[@B145]--[@B147]\]. In OVX female rats, dietary supplementation of soy phytoestrogens (0.4 g/kg or 1.6 g/kg) or 17β-estradiol (0.15 g/kg) for 12 weeks has been shown to increase the expression of brain neurotrophic factors such as BDNF and tropomyosin receptor kinase B (TrkB) and, as a result, to ameliorate hippocampal learning \[[@B148]\]. In normal and OVX transgenic AD mice, a selection of phytoestrogens in combination, composed of genistein, daidzein, and equol, has been shown to improve spatial working memory performance and to reduce mortality, as well as to delay neuropathological changes associated with AD \[[@B149]\]. 2.4.1.2. Effects of Phytoestrogens on Oxidative Stress. The phytoestrogens are also known for their neuroprotective antioxidant effects in neuronal cell models after exposure to neurotoxic substances \[[@B150]--[@B152]\]. Phytoestrogens are able to reduce ROS within a cell and to protect from cellular damage \[[@B153]\]. In aged mice, soybean supplementation has been shown to prevent cognitive deficits by decreasing free radical generation, by enhancing scavenging of free radicals, and by increasing GSH levels \[[@B154]\]. Compared to estrogen itself, less evidence is provided for the direct effects of phytoestrogens on mitochondria, but antioxidant properties were demonstrated \[[@B155]--[@B158]\]. The molecular effects of phytoestrogens are summarized in [Figure 6](#fig6){ref-type="fig"}. ### 2.4.2. Clinical Evidence {#sec2.4.2} Until today, no clinical trials in MCI and AD were performed. Thus, currently there is no clinical evidence. 2.4.2.1. Healthy and Postmenopausal Women. Among five randomized controlled trials, four recent studies reported the beneficial effect of phytoestrogens on cognitive function in healthy individuals ([Table 4](#tab4){ref-type="table"}). In a study with young healthy adults of both sexes, a high soya or a low soya diet for 10 weeks had a beneficial effect and showed significant improvements in short-term and long-term memory as well as in mental flexibility \[[@B159]\]. In another cross-over design study, the administration of 4 capsules/day containing soya isoflavones during 6 weeks improved the spatial working memory of men aged 30-80 years \[[@B160]\]. In postmenopausal women, 6 months of treatment duration with isoflavone supplementation provoked better learning, mental flexibility, and increased attention, as well as caused improvement in mood and lower depressive symptoms \[[@B161]\]. In a small mixed gender sample of older adults, soy supplementation ameliorated the visuospatial memory and the construction of verbal fluency and speeded dexterity \[[@B162]\]. All these studies demonstrated that phytoestrogens may affect human cognition. However, no clinical trials of phytoestrogens are known for the prevention or the treatment of AD. Inconclusive findings have also been reported from randomized controlled trials and observational studies in humans. In fact, these discrepant data could have several possible reasons. Investigation in European cohorts showed that a low dietary consumption of phytoestrogens had a significant effect on the improvement of the quality of life but no effect on cognition \[[@B163]\]. Mediating variables in the characteristics of the study population such as gender, age, ethnicity, and menopausal status appears to play an important role \[[@B164]\]. Phytoestrogens have been shown to have time-limited positive effects on cognition. These findings are in line with estrogen treatment which also exerts an initially positive short-term effect on cognition and a reversion after a long-term continuous use in aged women \[[@B164]\]. Globally, the effects of phytoestrogens can be dependent upon a window of opportunity for treatment and can affect males differentially than females due to the diminished presence of ER-mediated protective mechanisms and the tyrosine kinase activity with a potentially deleterious outcome of the supplements \[[@B165]\]. An age-dependent effect of phytoestrogen supplements is suggested in postmenopausal women \[[@B165]\]. In males, the findings are equivocal and sparse, and more investigations are needed to determine whether the effects will be deleterious or beneficial \[[@B165]\]. 3. Conclusion {#sec3} ============= In this article, the efficacy of standardized Ginkgo biloba extract, resveratrol, allopregnanolone, and phytoestrogens in combatting age-related cognitive decline has been reviewed. The mechanisms of action as well as preclinical and clinical evidence for each of those entities have been discussed. The four entities share common mechanisms of action but also diverse ones. In terms of the main AD features, Aβ and tau, all four categories were able to reduce the Aβ accumulation but only GBE and phytoestrogens seem to reduce tau hyperphosphorylation. Similarly (and quite predictably due to their phenolic character), all four act as antioxidants either by reducing ROS and oxidative stress (GBE, phytoestrogens, and allopregnanolone) or by enhancing the activity of antioxidant enzymes such as SOD and GPx1 (GBE, resveratrol, and phytoestrogens) and by reducing lipid peroxidation (GBE) and prooxidant enzymes such as Nox4 (resveratrol). GBE, resveratrol, and allopregnanolone target mitochondria by enhancing their functions (activities of complexes, oxidative phosphorylation, oxygen consumption, respiration, mitochondrial membrane potential, and ATP production), while in addition to this, GBE and resveratrol promote mitochondrial biogenesis. This is particularly important since mitochondria play a pivotal role in synaptic plasticity that is reduced in pathological states in the brain. However, there are also some differences in the mechanisms of action of the four discussed substances and mainly in the pathways through which they exert their beneficial effects. Based on our review of the literature, GBE rescues the Aβ neurotoxicity through the activation of the Akt pathway and through phosphorylation of CREB. Neurotrophic factors such as BDNF are stimulated both by GBE and by phytoestrogens. Resveratrol leads to Aβ clearance, enhancement of mitochondrial biogenesis and metabolism, and reduction of inflammation and ROS mainly through the activation of SIRT 1 and AMPK pathways as well as through the deacetylation of PGC-1α and the modulation of the JAK/ERK/STAT pathway. Phytoestrogens act as ER receptor modulators. Resveratrol can additionally act as a phytoestrogen and bind to the ERβ receptor. In terms of in vitro assays, it should be taken into account that the extract and the substances should be tested in meaningful, physiologically relevant concentrations and not in irrationally high ones. Regarding clinical trials, there is a different level of evidence for the four phytochemicals. Standardized GBE, resveratrol, allopregnanolone, and phytoestrogens appear in a descending order according to the level of existing clinical evidence. According to the World Federation of Societies of Biological Psychiatry (WFSBP) Guidelines, GBE has been classified in category B and grade 3 in terms of the outcome of existing studies. Therefore, there is sufficient and good clinical evidence for the efficacy of GBE. There is increasing and promising clinical evidence for resveratrol, but more studies of larger sample size are definitively needed. Lastly, there are no clinical trials indicating the beneficial effect of allopregnanolone and phytoestrogen in age-related cognitive decline disorders. There is only promising evidence from preclinical data regarding allopregnanolone and phytoestrogen. Notably, the four entities follow the same descending order regarding the existing level of clinical evidence and their mitochondria-improving properties. All in all, the effect on mitochondria goes hand in hand with the clinical effect and this highlights one more time the importance of these organelles not only in the pathogenesis of AD but also in aging in general. AE has received grant/research support from Schwabe and Vifor Pharma. 17β-estradiol: : Estradiol 2A PP2Ac: : Catalytic subunit of protein phosphatase 3xTgAD: : Triple transgenic mouse model of AD AD: : Alzheimer\'s disease Akt: : Protein kinase B AMPK: : 5′ adenosine monophosphate-activated protein kinase APP: : Amyloid precursor protein ATP: : Adenosine triphosphate Aβ: : beta-Amyloid protein αKGDH: : α-Ketoglutarate dehydrogenase Bax: : Apoptotic regulator BBB: : Blood-brain barrier Bcl-2: : Anti-B-cell lymphoma 2 BDNF: : Brain-derived neurotrophic factor BDPP: : Bioactive dietary preparation Ca^2+^: : Calcium CMRglc: : Cerebral metabolic rates of glucose COX-2: : Cyclooxygenase-2 CREB: : Cyclic AMP response element-binding protein CSF: : Cerebrospinal fluid CTRL: : Untreated SH-SY5Y cells DAT: : Dopamine transporters E2: : Estrogen ER: : Endoplasmic reticulum ERT: : Estrogen replacement therapies ETC: : Electron transport chain FAD: : Familial Alzheimer\'s disease FOXO: : Forkhead box O GBE: : Gingko biloba extract GPx1: : Glutathione peroxidase 1 GSK3β: : Glycogen synthase kinase 3 beta hNSCs: : Human neural stem cells HO-1: : Heme oxigenase 1 I~A~: : Transient potassium channel IBO: : Ibotenic acid IKK: : Inhibitory kappa B kinase IMM: : Inner mitochondrial membrane IMR-32: : Human neuroblastoma cells iNOS: : Nitric oxide synthase JAK/ERK/STAT: : Janus kinases/extracellular signal-regulated kinases/signal transducers and activators of transcription JNK: : c-Jun N-terminal kinase LXR: : Liver-X-receptor MCI: : Mild cognitive impairment MDA: : Malondialdehyde MMP: : Mitochondrial membrane potential MnSOD: : Manganese superoxide dismutase mPTP: : Mitochondrial permeability transition pore MRI: : Magnetic resonance imaging MTDLs: : Multitarget-directed ligands mtDNA: : Mitochondrial DNA NF-κB: : Nuclear factor kappa-light-chain-enhancer of activated B cells Nox4: : Nicotinamide adenine dinucleotide phosphate (NADPH) oxidase 4 NPS: : Neuropsychiatric symptoms NRF-2: : Nuclear factor erythroid 2-related factor 2 NRs: : Nuclear receptors OVX: : Ovarectomized OXPHOS: : Oxidative phosphorylation p53: : Tumor protein P450scc: : Cytochrome cholesterol side-chain cleavage enzyme PAF-AH-1: : Platelet-activating factor-acetylhydrolase-1 PCG-1α: : Peroxisome proliferator-activated receptor γ coactivator-1α PC12: : Pheochromocytoma cells PDH: : Pyruvate dehydrogenase PKC-δ: : Protein kinase C delta PPARγ: : Peroxisome proliferator-activated receptor gamma PSD95: : Synapse-associated protein PXR: : Pregnane xenobiotic receptor RNS: : Reactive nitrogen species ROS: : Reactive oxygen species RSFC: : Resting-state functional connectivity SAD: : Sporadic Alzheimer\'s disease SAMP1: : Senescence-accelerated mice SGZ: : Subgranular zone SH-SY5Y: : Human neuroblastoma cells SIRT1: : Sirtuin1 SMI: : Subjective memory impairment SOD: : Superoxide dismutase SOD1: : Superoxide dismutase 1 TH: : Tyrosine hydroxylase TNF-α: : Tumor necrosis factor alpha TrkA: : Tropomyosin receptor kinase A TrkB: : Tropomyosin receptor kinase B TSPO: : The translocator protein TyrRS-PARP1: : Tyrosyl tRNA synthetase-poly(ADP-ribose) polymerase 1 VaD: : Vascular dementia YY-1224: : A terpene trilactone-enhanced GBE. Disclosure ========== AE has served as a consultant or on advisory boards for Vifor Pharma and Schwabe. Conflicts of Interest ===================== With the relevance to this review, there is no direct conflict of interest to declare. ![Common mitochondria-related targets of natural substances in neuroprotection. In AD, the precursor of amyloid protein APP is cleaved sequentially by β- and γ-secretases leading to the production of Aβ peptides, their aggregation, and the formation of extracellular plaques. Different Aβ species exist, but Aβ~1-42~ is one of the most abundant and is the one that is mainly deposited in the brain due to its hydrophobic and fibrillogenic nature. AD is associated with electron transport chain (ETC) impairments leading to decreased ATP levels and basal respiration, with a decrease of antioxidant defenses and an increase of ROS production by complex I and complex III (orange dashed arrows). Globally, Gingko biloba, resveratrol, and phytoestrogens have been shown to protect against cell death in AD through a common mechanism of action by reducing abnormal aggregation of Aβ, amyloid beta (Aβ) toxicity, oxidative stress, mitochondrial impairments leading to neuronal dysfunction, and apoptosis. Gingko biloba, resveratrol, and phytoestrogens are suggested to exert a beneficial effect in AD affected neurons, but their specific mechanisms of mitochondrial interaction are not fully described yet. ↓: AD-related decrease. The green circle indicates the common mitochondria-related targets of GBE, resveratrol, phytoestrogen, and allopregnanolone.](OMCL2019-9695412.001){#fig1} ![The effects of Aβ, hyperphosphorylated tau, and standardized Ginkgo biloba extract (GBE) on mitochondrial function in AD. It has been shown that mitochondrial dysfunction is a key feature in AD and plays a pivotal role on the onset of the disease. While defining the chronologically first hallmark of the disease can be puzzling, there is evidence about mitochondrial dysfunction being the first hallmark at the early stages of AD with Aβ occurring as a result. Aβ has been shown to cause a decline in OXPHOS, taking place at the ETC, which leads to defective complexes IV and V and decreased ATP production. Faulty OXPHOS function results in the production of ROS which, when in excess, cannot be counterbalanced by the antioxidant enzymes like GSH-Px and SOD. ROS can cause membrane lipid peroxidation and instable MMP. Hyperphosphorylated tau inhibits complex I activity. However, GBE has been proven to reduce Aβ aggregation and tau hyperphosphorylation and to enhance OXPHOS, activities of complexes, and ATP levels, as well as to restore MMP. ROS and consequently lipid peroxidation are reduced due to GBE, while the extract has the ability to enhance SOD and GSH-Px activity and also induce mitochondrial biogenesis. ↓: represents increase; ⟂: represents inhibition.](OMCL2019-9695412.002){#fig2} ![Neuroprotective effects of resveratrol in AD. The precursor of amyloid protein APP is cleaved sequentially by β- and γ-secretases leading to the production of Aβ and their aggregation. Resveratrol increases the clearance of Aβ peptides through the activation of AMPK. Resveratrol plays an important role in the neuroprotective properties as it reduces Aβ neurotoxicity by phosphorylating PKC-δ. Damaged mitochondria generate ROS which are implicated in apoptosis. iNOS and COX-2 also enhance the production of ROS. Resveratrol exerts antioxidant properties and attenuates oxidative damage by decreasing iNOS and COX-2 levels. Resveratrol also protects mitochondria by increasing the expression of ROS-inactivating enzymes GPx1 as well as SOD1 and by reducing the expression of the ROS-producing enzyme Nox4. Resveratrol also influences the Aβ-induced apoptotic signalling pathway by inhibiting the expression of caspace-3, Bax, FOXO, and p53 by blocking the activation of JNK and by restoring the decrease of Bcl-2 expression, as well as by inhibiting the increase of NF-κB DNA binding. Mitochondrial biogenesis is induced by resveratrol through SIRT1 activation and deacetylation of PGC-1α. Resveratrol was also able to protect hippocampal neurons by alleviating cognitive impairment and reducing neuronal loss via modulating the janus kinases, extracellular signal-regulated kinases, and signal transducers, as well as the signalling pathway of the activators of transcription (JAK/ERK/STAT).](OMCL2019-9695412.003){#fig3} ![Standardized Ginkgo biloba extract (GBE) LI 1370 (Vifor SA, Switzerland) (100 μg/ml) increased neurite outgrowth of SH-SY5Y neuroblastoma cells after 3 days of treatment in 3D cell culture. Pictures were taken using a cell imaging multimode reader Cytation3 (Biotek Instruments Inc., X20 in black and white) after immunostaining (IMS, βIII-tubuline/Alexa488). Compared to the untreated SH-SY5Y cells (CTRL, (a)), 100 μg/ml of GBE (b) was efficient in increasing the formation of neurites.](OMCL2019-9695412.004){#fig4} ![Neuroprotective effects of allopregnanolone (AP) in AD. AP has been proven to reduce Aβ aggregation-induced cell death. It exerts a neuroprotective effect against oxidative stress-induced cell death via the improvement of the cellular and mitochondrial energy by enhancing the OXPHOS and ATP levels. AP ameliorates the mitochondrial redox environment by decreasing ROS and by improving the activity of the enzyme MnSOD. AP also has beneficial effects on bioenergetic enzymes such as PDH and αKGDH implicated in the TCA cycle. AP ameliorates cholesterol homeostasis and clearance for the biosynthesis of neurosteroids by raising the expression of PXR and LXR. AP promotes repair and renewal of neurons leading to restored cognitive performances in AD.](OMCL2019-9695412.005){#fig5} ![Modulation of mitochondrial function by estrogen and phytoestrogen. Less evidence is provided for the direct effects of phytoestrogen on mitochondria compared to estrogen, but antioxidant properties were demonstrated.](OMCL2019-9695412.006){#fig6} ###### Clinical trials on the effects of GBE. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Study design GBE dose/preparation Duration Subjects Purpose Main results References --------------------------------------------------------------------- ------------------------------------------------------- ---------------------------------- ---------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------ ------------------------------------------------------------------------------------------------------------ Patients with memory complaints, SMI, and MCI R, DB, PC 240 mg of GBE once daily or placebo 56 ± 4 days (61)\ Test the effect of GBE on cognitive functions associated with prefrontal dopamine GBE caused a mild increase in prefrontal dopamine; there were indications for enhanced cognitive flexibility and for ameliorated response inhibition results Beck et al., 2016 \[[@B61]\] Healthy aged patients with subjective memory decline (SMI) R, DB, PC 240 mg of GBE once daily or placebo 12 weeks (300)\ Evaluate the effects of GBE on cognition and quality of life in patients with very mild cognitive impairment GBE improved the cognitive ability and quality of life of patients Grass-Kapanke, 2011 \[[@B63]\] Patients 45-65 years old with very mild cognitive impairment (MCI) R, PC, DB, MC 240 mg of GBE once daily or placebo 24 weeks (160)\ Test the effect of GBE on NPS and cognition in patients with MCI GBE improved NPS and cognition; the extract was safe and well tolerated Gavrilova et al., 2014 \[[@B62]\] Patients with MCI Patients with dementia R, DB, PC 240 mg of GBE once daily 22 weeks (400)\ Test the efficacy of GBE on NPS of dementia GBE statistically superior to placebo in ameliorating NPS (e.g., irritability, apathy, and anxiety) Scripnikov et al., 2007 \[[@B64]\] Demented patients with NPS Systematic review 240 mg of GBE once daily 22 weeks (1628)\ Demonstrate efficacy of GBE in dementia with BPSD Improvements of quality of life, cognition, and BPSD activities of daily living clinical global impression Von Gunten et al., 2016 \[[@B66]\]\ Demented patients with behavioural and psychological symptoms (BPSD) (\[[@B12], [@B166]--[@B168]\]) Meta-analysis and systematic review Different dosages of GBE Not available Demented patients Test the efficacy of GBE in ameliorating symptoms of demented patients GBE improved cognitive function and activities of everyday life in patients with dementia Brondino et al., 2013 \[[@B72]\]\ (\[[@B12], [@B67]--[@B69], [@B79], [@B166], [@B169], [@B170]\]) Meta-analysis of randomized placebo controlled trials 120 mg or 240 mg of GBE per day or placebo 22-26 weeks (2684)\ Evaluate evidence for efficacy of GBE in dementia Confirmation of efficacy of GBE and good tolerability Gauthier and Schlaefke, 2014 \[[@B70]\]\ Demented patients (\[[@B12], [@B69], [@B166], [@B167], [@B169], [@B171]\]) Systematic review and meta-analysis of randomized controlled trials 240 mg of GBE once daily 22-26 weeks (2561)\ Evaluate the clinical efficacy and adverse effects of GBE in dementia and cognitive decline GBE was found more effective than placebo in decelerating cognition deficits and in improving daily life activities and NPS in dementia Tan et al., 2015 \[[@B65]\]\ Demented patients (\[[@B12], [@B62], [@B69], [@B167], [@B169]--[@B174]\]) Meta-analysis of randomized controlled clinical trials 240 mg/day 22 or 24 weeks Old patients aged over 60 years Effects of GBE on anxiety, dementia, and depression in aging patients Improvements in dementia, anxiety, and depression Kasper, 2015 \[[@B73]\] (\[[@B12], [@B166]--[@B168]\]) Meta-analysis of randomized controlled trials 240 mg of GBE once daily 22 or 24 weeks (1628)\ Test the effects of GBE on BPSD of demented patients Significant superiority of GBE to placebo in improving BPSD and therefore caregiver experience Savaskan et al., 2017 \[[@B74]\]\ Demented patients with behavioural and psychological symptoms (BPSD) (\[[@B12], [@B166]--[@B168]\]) Bivariate meta-analysis Different dosages of GBE Approximately 6 months Demented patients Evaluate baseline risk on the treatment effect and assess the efficacy of GBE on cognitive symptoms of dementia GBE was effective at improving cognitive functions in dementia after 6 months of treatment Wang et al., 2010 \[[@B71]\]\ (\[[@B12], [@B67], [@B69], [@B166], [@B169], [@B170]\]) R, DB, PC, PG, MC 160 mg or 240 mg of GBE daily 24 weeks (214)\ To assess the efficacy of GBE in aged demented patients or patients with age-related memory loss No beneficial effect of GBE for demented or age-related memory-impaired patients Van Dongen, 2000 \[[@B75]\] Patients with dementia or age-related memory loss R, DB, PC, PG 120 mg of GBE daily 6 months 176 mildly to moderately demented patients Assess the efficacy and safety of GBE for treating dementia in early stages GBE not beneficial in mild to moderate dementia after a 6-month treatment McCarney et al., 2008 \[[@B76]\] Cochrane analysis of R, DB, PC trials Different GBE doses ranging from low to high Different treatment periods Aging with dementia or cognitive impairment Assess the efficacy and safety of GBE in dementia and cognitive impairment GBE displays unreliable and inconsistent evidence in being beneficial for demented people Birks and Evans, 2009 \[[@B77]\] Patients with AD and vascular dementia R, DB, PC 120 mg of GBE, 60 mg of GBE, or placebo 6 months (90)\ Evaluate the efficacy and safety of GBE in vascular demented patients GBE slowed down the cognitive deterioration in vascular demented patients, effect shown in only one of the four neuropsychological tests Demarin et al., 2017 \[[@B78]\] Patients with vascular dementia (VaD) Review of R, PC 120 mg of GBE twice daily or 240 mg of GBE once daily 22 or 24 weeks (1294)\ Test the efficacy of GBE in older patients with AD/vascular dementia with NPS Confirmation of efficacy of GBE and good tolerability Ihl, 2013 \[[@B79]\]\ Demented patients (AD or VaD) with NPS (\[[@B12], [@B166], [@B167], [@B175]\]) Systematic review and meta-analysis GBE extract 12-52 weeks (2372)\ Evaluate the effects of GBE in AD and vascular and mixed dementias Superiority of GBE to placebo in improving everyday life activities in mainly the AD type of dementia Weinmann et al., 2010 \[[@B80]\]\ Patients with AD or vascular or mixed dementia (\[[@B67]--[@B69], [@B166], [@B169], [@B173], [@B175]\]) Systematic review and meta-analysis 240 mg and 120 mg of GBE daily 24 weeks Patients with MCI or AD Assess the effectiveness and safety of GBE in treating MCI and AD There is an indication for the beneficial effect of GBE in MCI and AD but the results were inconsistent Yang et al., 2016 \[[@B81]\] (AD: \[[@B67], [@B68], [@B169], [@B170], [@B174]--[@B176]\]; MCI: \[[@B62]\]) Systematic review of randomized controlled trials 240 mg of GBE daily Period ≥ 16 weeks Patients with mildly to moderately severe and severe AD Assess the beneficial effect of GBE in AD Evidence of beneficial effects of GBE in amelioration cognition, every day activities, and psychopathological symptoms but great heterogeneity among the results Janssen et al., 2010 \[[@B82]\]\ (\[[@B67], [@B69], [@B166], [@B169]\]) Prevention R, DB, PC, PG 120 mg of GBE twice daily 5 years Adults 70 years or older with occasional memory problems Efficacy of long-term use of GBE for the prevention of AD in aging with memory complaints GBE did not reduce the incidence of AD compared to placebo GuidAge study, Vellas et al., 2012 \[[@B83]\] R, DB, PC 120 mg of GBE twice daily Every 6 months from 2000 to 2008 (3069)\ Test whether GBE delays or prevents global or domain-specific cognitive impairment in aging GBE did not prevent cognitive decline in aging Snitz et al., 2009 \[[@B84]\] Healthy old people or people with MCI aged 72 to 96 years R, DB, PC 120 mg of GBE twice daily 5 years (3000)\ Assess the ability of GBE in the prevention of dementia in normal aging or those with MCI GBE does not prevent dementia GEM study, DeKosky et al., 2006 \[[@B87]\] Healthy subjects aging over 80 years old Systematic review and meta-analysis 240 mg of GBE daily Not available Nondemented patients aged 70 years or older Evaluate the efficacy of GBE for the prevention of dementia in nondemented adults GBE is not able to prevent the development of dementia Charemboon and Jaisin, 2015 \[[@B86]\] (\[[@B83], [@B85]\]) ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- SMI, subjective memory impairment; MCI, mild cognitive impairments; AD, Alzheimer\'s disease; VaD, vascular dementia; R, randomized; DB, double blind; PC, placebo controlled; MC, multicenter; PG, parallel group; BPSD, behavioural psychological symptoms; VCI: vascular cognitive impairment. The number of patients involved in the trials is indicated in parentheses. ###### Clinical trials on the effects of resveratrol. Ongoing trials are italicized. ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Study design Resveratrol dose/preparation Duration Subjects Purpose Main results References ---------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------- ---------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------ Young and aged healthy individuals R, DB, PC, CO Trans-resveratrol from Biotivia Bioceuticals 250 mg or 500 mg 21 days (24)\ Ability to increase cerebral blood flow and modulate mental function Increase in cerebral flow, no effect in cognitive function Kennedy et al., 2010 \[[@B94]\] 18-25 years healthy R, DB, PC, CO Trans-resveratrol 250 mg/day or trans-resveratrol 250 mg/day with 20 mg piperine 21 days (23)\ Effect of piperine on the efficacy and bioavailability of resveratrol Piperine enhances the effect of resveratrol on cerebral blood flow but no effect on bioavailability and cognition Wightman et al., 2014 \[[@B95]\] Healthy subjects aged 19-34 years Study in older adults 200 mg of resveratrol per day 26 weeks (46)\ Test whether resveratrol would improve memory performance in older adults Resveratrol ameliorates memory performance in combination with improved glucose metabolism and increased hippocampal functional connectivity in healthy overweight old people Witte et al., 2014 \[[@B107]\] Healthy overweight subjects aged 50-75 years Patients with cognitive decline and postmenopausal women R, DB, PC 72 g of active grape formulation 6 months $10$ Adults with mild cognitive decline with mean age of 72.2 years Evaluate the effects of grapes on regional cerebral metabolism Grapes could possess a protective effect against early pathologic metabolic decline Lee et al., 2017 \[[@B108]\] R, PC, intervention trial 75 mg twice daily of trans-resveratrol 14 weeks $80$ Postmenopausal women between 45 and 85 years old Test the effects of resveratrol on cognition, mood, and cerebrovascular function in postmenopausal women Resveratrol was well tolerated and able to improve cognition which was related to the improvement of cerebrovascular function. Mood was improved but not significantly. Evans et al., 2017 \[[@B109]\] Patients with MCI R, DB, interventional study 200 mg of resveratrol per day 26 weeks (40)\ Assess if resveratrol improves long-term glucose control, resting-state functional connectivity of the hippocampus, and memory function in patients with MCI Resveratrol supplementation decreased glycated hemoglobin A1c, preserved hippocampus volume, and improved hippocampus RSFC in patients with MCI Koebe et al., 2017 \[[@B112]\] Old patients with MCI R, DB\ Bioactive dietary polyphenol preparation (BDPP) at low, moderate, and high doses 4 months (48)\ Safety and efficacy in treating mild cognitive impairment --- [NCT02502253](https://clinicaltrials.gov/ct2/show/NCT02502253) \[[@B110]\] Phase 1 55-85 years MCI R, DB, PC\ Resveratrol or omega-3 supplementation or caloric restriction 6 months (330)\ Effects on brain function --- [NCT01219244](https://clinicaltrials.gov/ct2/show/NCT01219244) \[[@B111]\] Phase 4 50-80 years MCI Patients with mild to moderate AD R, DB, PC, MC\ Resveratrol 500 mg/day with escalation by 500 mg increments ending with 2 g/day 52 weeks (119)\ Assess efficacy and safety No effect on cognitive score, decrease of CSF and plasma Aβ40 levels Turner et al., 2015 \[[@B113]\] Phase 2 Over 49 years mild to moderate AD R, DB, PC Phase 2 Resveratrol 500 mg daily (orally) with a dose elevation by 500 mg every 13 weeks until a final dose of 1000 mg twice daily was reached for the final 13 weeks. 52 weeks (119)\ Evaluation of safety and tolerability of resveratrol and its effects on AD biomarkers and also on clinical outcomes Resveratrol was well tolerated and safe, it was detected in the cerebrospinal fluid (nM), it changed the AD biomarker paths, it modified the CNS immune response, and it maintained the BBB integrity; however, more research is needed Sawda et al., 2017 \[[@B114]\] Adults older than 49 years old with a diagnosis of mild to moderate dementia due to AD R, DB, PC\ Longevinex brand resveratrol supplement (resveratrol 250 mg/day) 52 weeks (50)\ Effects on cognitive and global functioning --- [NCT00743743](https://clinicaltrials.gov/ct2/show/NCT00743743) \[[@B115]\] Phase 3 50-90 years mild to moderate AD on standard therapy R, DB, PC\ Resveratrol with malate and glucose 12 months (27)\ Ability to slow the progression of AD --- [NCT00678431](https://clinicaltrials.gov/ct2/show/NCT00678431) \[[@B116]\] Phase 3 50-90 years mild to moderate AD ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ MCI, mild cognitive impairment; AD, Alzheimer\'s disease; R, randomized; DB, double blind; PC, placebo controlled; CO, cross over; MC, multicenter; CSF, cerebrospinal fluid. The number of patients involved in the trials is indicated in parentheses. ###### Ongoing clinical trial on the effects of allopregnanolone in MCI and mild AD. --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Study design Allopregnanolone dose/preparation Duration Subjects Purpose Main results References ----------------------------- ------------------------------------------------------------------------------------------------------------------- ---------- -------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------- --------------- ------------------------- R, DB, parallel assignment\ Allopregnanolone 2, 4, or 6 mg intravenous injection once per week or placebo intravenous injection once per week 12 weeks $8$ For each dose group, 55 years and older, both genders\ Determine the maximally tolerated dose, safety and tolerability, pharmacokinetic profile, and effects on cognitive function Not available NCT02221622 \[[@B133]\] Phase 1 MCI or mild AD\ (6) Randomized to AP\ (2) Randomized to placebo --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- The number of patients involved in the trials is indicated in parentheses. ###### Clinical trials on the effects of phytoestrogens. ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Study design Phytoestrogens dose/preparation Duration Subjects Purpose Main results References ---------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------ ---------- --------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------- Healthy individuals and postmenopausal women Randomized control trial High soya (100 mg total isoflavones/day) or a low soya (0.5 mg total isoflavones/day) diet 10 weeks (27)\ Effects on memory, attention, and frontal lobe function Improvements in short-term memory, long-term memory and mental flexibility File et al., 2015 \[[@B159]\] Young healthy adults (15 men, 12 women) DB, CO, PC 4 capsules/day containing soya isoflavones (116 mg isoflavone equivalents/day: 68 mg daidzein, 12 mg genistein, and 36 mg glycitin) or placebo 6 weeks (34)\ Effects on cognitive function Improvements of spatial working memory but no effect on auditory and episodic memory and executive function and visual-spatial processing Thorp et al., 2009 \[[@B160]\] Men aged 30-80 years 18\ Isoflavone supplementation 60 mg/day or placebo 6 months (78)\ Effects of soy isoflavones on mood and cognitive function in postmenopausal women Improvements in mental flexibility, attention, mood, and lower depressive symptoms Casini et al., 2006 \[[@B161]\] R, DB, CO, PC Postmenopausal women (mean age 49.5 years) R, DB, PC 100 mg/day soy isoflavones (glycoside weight) or matching placebo tablets 6 months Older nondemented men and women (age 62-89 years) Examination of safety, feasibility, and cognitive efficacy of soy isoflavone administration Improvements of visual-spatial memory and construction of verbal fluency and speeded dexterity Gleason et al., 2009 \[[@B162]\] R, DB, PC 20 g of soy protein containing 160 mg of total isoflavones 12 weeks (93)\ Effect of a high-dose isoflavones on cognition, quality of life, lipoproteins, and androgen status in postmenopausal women Significant improvement in the quality of life versus placebo. No significant effects in cognition. The testosterone and HDL levels were significantly lower at the end of the study. Basaria et al., 2009 \[[@B163]\] Healthy postmenopausal women (mean age 56 years) ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- The number of patients involved in the trials is indicated in parentheses. [^1]: Guest Editor: Nicola Amodio	{ "pile_set_name": "PubMed Central" }
Never would we have imagined that our third issue of 2020 would be printed in the middle of a pandemic. COVID-19, which has already infected more than 20 million people worldwide and is projected to kill a million by the end of the year, has changed the world in ways that we are only beginning to comprehend. Population scientists are playing an important role in the fight against COVID-19, from analyzing the disease's death toll to studying inequalities in infection rates and severe health outcomes along key demographic characteristics, such as age, gender, and race. The main factor limiting our efforts is data availability. Beyond specific hospital settings, access remains limited to individual-level information on those who have tested positive for COVID-19. Particularly alarming is the lack of knowledge about disease progression in population-based samples of children. This information is needed as millions of children across the world, including Canada, return to school. Research about COVID-19 has been growing at an incredible speed, oftentimes outpacing journals' capacity to publish. Nonetheless, it is more important than ever to ensure timely publication of results that could be important in our fight against the disease. For this reason, Canadian Studies in Population is issuing a special call for papers about COVID-19. All details can be found on the journal's website. We apologize for delays that Canadian Studies in Population is experiencing because of the pandemic, and we would like to ensure you that we are working hard to resolve these issues. Thank you to all our authors for their patience and to our reviewers for their irreplaceable commitment during these difficult times. Stay safe everyone and wear a mask. Simona Bignami-Van Assche, Editor-in-Chief Sophia Chae, Deputy Editor	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-nutrients-11-01865} =============== Nowadays, prevention of primary diseases and maintenance of health in the population, through a healthy nutrition, has become of great interest, in order to maintain the sustainability of healthcare systems worldwide \[[@B1-nutrients-11-01865]\]. In this sense, aged people have a considerable public health importance due to the cost management in treating older adult patients with multiple co-morbid conditions \[[@B2-nutrients-11-01865]\]. Abnormal lipid metabolism or dyslipidemia, non-alcoholic fatty liver disease (NAFLD), obesity, and insulin resistance are common in adult and aged people and increase the risk for other liver disorders, diabetes, and cardiovascular diseases, which remains the major cause of worldwide morbidity and mortality in the elderly population, despite scientific and therapeutic advances \[[@B3-nutrients-11-01865]\]. Although, postprandial lipemic behavior varies according to different age stages, the mechanism underlying this effect is still unknown \[[@B4-nutrients-11-01865]\]. Until recently, excessive postprandial dyslipidemia after fat load have been attributed to the increased levels of triglyceride-rich lipoprotein (TRL) of intestinal origin \[[@B4-nutrients-11-01865],[@B5-nutrients-11-01865]\]. Currently, it has been clearly demonstrated that VLDL (very low density lipoproteins) remnants are the major that are TRL increased postprandially, after a fat meal \[[@B6-nutrients-11-01865],[@B7-nutrients-11-01865],[@B8-nutrients-11-01865],[@B9-nutrients-11-01865]\]. Nevertheless, the extent and duration of postprandial dyslipidemia in aged organisms have also been associated with changes in the absorptive processes, with reduced activity of lipoprotein lipases (LPL) and a decreased rate of clearance of TRL remnants by receptor-mediated processes \[[@B5-nutrients-11-01865],[@B10-nutrients-11-01865]\], although the available data are still inconclusive. In the setting of NAFLD, postprandial dyslipidemia have been attributed to the state of insulin resistance, which increases the rate of peripheral lipolysis, and the rate of TG synthesis and VLDL secretion by the liver \[[@B11-nutrients-11-01865],[@B12-nutrients-11-01865]\]. Increased hepatic secretion and impaired clearance of VLDL are associated with high plasma concentrations of TG and low density lipoproteins (LDL), two risk markers of the metabolic syndrome, T2DM, and cardiovascular diseases \[[@B13-nutrients-11-01865]\]. Moreover, insulin resistance elevates plasma levels of NEFA. In this condition, the insulin-stimulated glucose uptake in skeletal muscles is reduced and the production of fatty acid derivates, such as long-chain acyl-CoA, DAG and ceramides, which promote oxidative stress, inflammation, and impaired insulin signaling, are increased, resulting in postprandial hyperglycemia \[[@B14-nutrients-11-01865]\]. Conversely, a recent report proposed that postprandial dyslipidemia is the main cause of insulin resistance in the obese state \[[@B15-nutrients-11-01865]\]. In this sense, excessive VLDL remnants in the postprandial state favors adipocyte hypertrophy and inflammation of visceral adipose tissue (vWAT), the prominent causes of insulin resistance and the syndromes associated with insulin resistance, such as NAFLD \[[@B15-nutrients-11-01865]\]. However, a significant gap of knowledge exists between the underlying cellular and molecular mechanisms leading to postprandial dyslipidemia and insulin resistance in aged organisms. In the present work, we aim to investigate the progression of postprandial dyslipidemia in a rat model of insulin resistance associated with ageing. The study was implemented in mature middle-aged (7 months) and middle-old-aged (24 months) Wistar rats, which exhibit increased adiposity and an altered lipid profile when fasting together with liver steatosis and fibrosis, but avoid the onset of diabetes mellitus \[[@B16-nutrients-11-01865],[@B17-nutrients-11-01865],[@B18-nutrients-11-01865]\]. Interestingly, in 24-month-old rats, the liver transcription factor ChREBP localizes at a nuclear fraction, even in the fasted state \[[@B16-nutrients-11-01865]\]. In addition, it has been reported that ChREBP overexpression in mice is able to integrally induce the lipogenic and esterification program in the liver, increasing the hepatic content of monounsaturated fatty acids, without alterations in insulin sensitivity, despite liver steatosis \[[@B19-nutrients-11-01865]\]. Hence, we hypothesized that ChREBP play a pivotal role in the development of fasting and postprandial hypertriglyceridemia in Wistar rats, during ageing. Moreover, because food restriction (FR) improved the systemic insulin sensitivity in 7-month-old but not in 24-month-old Wistar rats \[[@B17-nutrients-11-01865]\], we speculate that moderate chronic FR for 3 months would modify hepatic ChREBP expression and avoid postprandial dyslipidemia, at least in 7-month-old mature middle-aged rats. Since the most functional assessment of postprandial lipid metabolism involves ingestion of a high-fat meal, we performed an OLLT, using extra-virgin olive oil, in order to evaluate the effect of ageing and chronic FR in the time course of postprandial triglyceridemia, cholesterolemia, glycemia, glucagonemia, and insulinemia, in 7-month-old and 24-month-old Wistar rats. Furthermore, to determine the metabolic impact of the acute fat-overload on ChREBP activity, we determined the expression pattern of ChREBP isoforms in liver and adipose tissues. Additionally, we examined the expression of pivotal genes involved in fatty acid and lipoprotein uptake, lipogenesis, lipolysis, fat oxidation, and thermogenesis in these tissues, as well as the expression of the ChREBP target gene FGF21, because recent reports demonstrated that postprandial dyslipidemia and fatty liver might modulate postprandial FGF21 levels in humans and mice \[[@B20-nutrients-11-01865],[@B21-nutrients-11-01865]\]. The data presented herein indicate that postprandial hypertriglyceridemia, progressively increases with ageing and could be associated with the conservation of hepatic ChREBP activity and fatty acid esterification program. Interestingly, moderate chronic FR avoids excessive postprandial hypertriglyceridemia in both 7-month-old food-restricted (7mFR) and 24-month-old food-restricted (24mFR) Wistar rats, supporting the notion that postprandial TG levels firstly depend on the degree of adiposity. Nevertheless, we speculated that FR could improve dyslipidemia in 7mFR and 24mFR rats, through different mechanisms. Finally, using the OLLT we showed that an acute fat load induced molecular signatures of browning in vWAT from old rats, including PRDM16 and the thermogenic gene UCP-1, particularly in old FR rats. Thus, as has been suggested elsewhere \[[@B22-nutrients-11-01865]\], these would favor the use of vWAT for ectopic fat deposition in 24-month-old Wistar rats, avoiding the onset of diabetes mellitus. 2. Materials and Methods {#sec2-nutrients-11-01865} ======================== 2.1. Animals {#sec2dot1-nutrients-11-01865} ------------ The experiments were performed in male 3-, 7-, and 24-month-old Wistar rats from our in-house colony (Centre of Molecular Biology, Madrid, Spain). The maximal life span of the male Wistar rats is about 32--34 months, while their mean life span is about 24 months \[[@B23-nutrients-11-01865]\]. Thus, the 24-month-old rats used in the present study are middle-old-aged animals. In addition, according to the previous studies \[[@B24-nutrients-11-01865]\] and our previous observations \[[@B16-nutrients-11-01865],[@B17-nutrients-11-01865],[@B18-nutrients-11-01865]\], the 24-month-old rats are not at high risk of mortality and do not present apparent signs of frailty. Nevertheless, 24-month-old rats showed higher intracellular accumulation of lipofuscin than 3-month-old or 7-month-old Wistar rats (see [Supplementary Figure S1](#app1-nutrients-11-01865){ref-type="app"} for further details) and showed an increased amount of activated stellate cells in the liver \[[@B18-nutrients-11-01865]\], both of which are markers of cellular senescence. Animals were housed in climate-controlled quarters with a 12-h light cycle. All rats in this study (fed ad libitum or food-restricted; 8 to 10 animals per group), were fed with a standard chow diet (2014 Teklad Global 14% Protein Rodent Maintenance Diet) (see [Supplementary Table S1](#app1-nutrients-11-01865){ref-type="app"} for further details) from Harlan Laboratories and water. Animals were handled according to the European Union laws (2010/63/EU) and following the Spanish regulations (RD 53/2013) for the use of laboratory animals. The experimental protocols were approved by the institutional committee of bioethics. All efforts were made to minimize animal suffering and to reduce the number of animals used. 2.2. Food Restriction (FR) Protocol {#sec2dot2-nutrients-11-01865} ----------------------------------- 4- and 21-month-old rats housed individually were fed once daily over 3 months, with an amount of chow equivalent to 75%--80% of their normal food intake, until they reached a reduction of 20% of their body weight, as previously described \[[@B17-nutrients-11-01865],[@B25-nutrients-11-01865]\]. After 1 month from the start of nutritional restriction, the rats showed a body weight equivalent to ≈80%--85% of ad-libitum-fed aged-mates. After this, they were weighed weekly and the amount of restricted food provided was individually adjusted ([Table 1](#nutrients-11-01865-t001){ref-type="table"}), in order to maintain their body weight for two additional months. Food restricted rats were used at the age of 7 and 24 months (7mFR and 24mFR) ([Scheme 1](#nutrients-11-01865-sch001){ref-type="scheme"}). This FR neither compromised the nutritional status of the animals, nor produced fasting hypoglycemia as we have previously confirmed by OGTT and by the euglycemic-hyperinsulinemic clamp technique, as well as several fasting measurements of glucose levels in the food-restricted animals \[[@B17-nutrients-11-01865],[@B18-nutrients-11-01865]\]. 2.3. Oral Lipid Loading Test (OLLT) {#sec2dot3-nutrients-11-01865} ----------------------------------- The OLLT was performed in 7- and 24-month-old Wistar rats fed ad libitum (AL) or food-restricted (FR) (n = 10 in each of the four experimental groups). Animals were fasted for 16 h (including the overnight period) before the test. A bolus (1 mL/Kg of body weight) of extra-virgin olive oil (aceites Malagón S.L., Ciudad Real, Spain) was orally administered, using a long flexible silicone Abbocath Catheter connected to a sterile polypropylene syringe. Rats remained in their cages without food for the 4 h duration of the study. Blood aliquots from the tail vein at the following times 0, 30, 60, 90, 120, 180, and 240 min were taken. After 240 min, the rats were anesthetized by CO~2~ inhalation, and sacrificed by decapitation. Additionally, 3-month-old AL (3mAL), and 7-month-old and 24-month-old rats fed AL or FR (control groups) were sacrificed before the OLLT after 16 h of fasting. Liver and vWAT were rapidly excised and weighted. The collected tissues were snap-frozen in liquid N~2~ and stored at −80 °C. Serum samples were obtained by blood centrifugation. 2.4. Serum and Tissue Measurements of Hormones and Metabolites {#sec2dot4-nutrients-11-01865} -------------------------------------------------------------- Serum glucose was determined enzymatically, using the AmplexRed Glucose/Glucose Oxidase Assay Kit (Molecular Probes, Inc., Eugene, OR, USA). Serum TG, total-cholesterol, and HDL (high-density lipoprotein) cholesterol levels were measured, respectively, using an enzymatic kit from Stanbio Laboratory (Boerne, TX, USA). The total serum ketones body (KB) and nonesterified fatty acid (NEFA) levels were quantified using an enzymatic kit from Wako Chemicals (Neuss, Germany). Glycerol was quantified by the method of Eggstein and Kuhlmann \[[@B26-nutrients-11-01865]\]. Serum levels of FGF21 were determined by a rat-specific ELISA kit from R&D Systems (Minneapolis, MN, USA). Serum insulin and glucagon levels were measured using rat-specific ELISA kits from Mercodia (Uppsala, Sweden) and Wako Chemicals (Neuss, Germany), respectively. Liver and vWAT content of the TG was measured, as previously described \[[@B16-nutrients-11-01865],[@B25-nutrients-11-01865]\]. 2.5. Real-Time RT-PCR {#sec2dot5-nutrients-11-01865} --------------------- Total RNA was isolated from 100 mg of liver and vWAT using the Trizol reagent (Invitrogen). The cDNA was synthesized from 1.5 mg of DNase-treated RNA. Relative quantitation of peroxisome proliferator-activated receptor alpha (Ppara, Rn00566193_m1), peroxisome proliferator-activated receptor gamma coactivator 1-alpha (Pgc1a, Rn00580241_m1), aquaporin 7 (Aqp7, Rn00569727_m1), fatty acid transporter CD36 (Cd36, Rn00580728_m1), lipoprotein lipase (Lpl, Rn00561482_m1), low density lipoprotein receptors (Ldlr, Rn005998442_m1), LDL receptor related protein 1 (Lrp1, Rn01503901_m1), perilipin-1 (Plin1, Rn00558672_m1), diacylglycerol O-Acyltransferase 2 (Dgat2, Rn01506781_m1), microsomal triglyceride transfer protein (Mttp, Rn01522974_m1), carnitine palmitoyl transferase 1a (Cpt1a, Rn00676501_g1), carnitine palmitoyl transferase 1b (Cpt1b, Rn00566142_m1), fatty acid synthase (Fas, Rn01463556_g1), stearoyl CoA desaturase-1 (Scd1, Rn00594894_g1), medium-chain acyl-CoA dehydrogenase (Acadm, Rn00566390_m1), long-chain acyl-CoA dehydrogenase (Acadl, Rn00563121_m1), elongation of very long chain fatty acids 6 (Elovl6, Rn00592815_m1), phosphoenolpyruvate carboxykinase 1 (Pck1, Rn01529009_g1), glucose-6-phosphatase (Glu6Pase, Rn00598561_m1), fibroblast growth factor 21 (Fgf21, Rn00590706_m1), glucose transporters 1 (Glut1, Rn01417099_m1), 2 (Glut2, Rn00563565_m1), and 4 (Glut4, Rn00562597_m1), peroxisome proliferator activated receptor gamma (Pparg, Rn00440945_m1), tumor necrosis factor alpha (Tnfα, Rn99999017_m1), PRD1-BF1-RIZ1 homologous domain-containing 16 (Prdm16, Rn01516224_m1), beta-3 adrenergic receptor (Adrβ3 Rn00565393_m1), and uncoupling protein 1 (Ucp1, Rn00562126_m1) mRNA expression analysis from individual mRNA samples of rats of the four experimental groups (n = 8--10 rats per group) were performed on an ABI PRISM 7500 using Pre-Developed TaqMan Assay Reagents (PE Applied Biosystem, Foster City, CA, USA). The relative amount of mRNA in each sample was normalized to that of the reference control 18 S rRNA, using the comparative (2^−ΔCT^) method, according to the manufacturer's instructions. Carbohydrate-response element-binding protein isoforms (α and β isoforms) were measured using specific primers from Sigma Aldrich: ChREBP-α: 5′-AGGCTCAAGCATTCGAAGAG-3′ and 5′-TGCATCGATCACAGGTCATT-3′; ChREBP-β: 5′-CTTGTCCCGGCATAGCAAC-3′ and 5′-TCTGCAGATCGCGCGGAG-3′. Quantitative PCR was performed using FAST SYBR GREEN^®^ Master Mix (Applied Biosystems) with normalization of gene expression levels to 18 s. 2.6. Calculations and Statistical Analysis {#sec2dot6-nutrients-11-01865} ------------------------------------------ Visceral adiposity was the sum of the weight in grams of epidydimal and retroperitoneal fat pads. Insulin Resistance (IR) was evaluated by the homeostasis model assessment method (HOMA-IR), as follows: \[fasting insulin (μU/mL) × fasting glucose (mmol/liter)\]/22.5 as described \[[@B27-nutrients-11-01865]\]. The total response during the OLLT was measured as the areas under the curve (AUCs), which were calculated using the trapezoid rule. Statistical analysis was performed using the GraphPad Prism version 8.2 for Windows (GraphPad Software). Significant differences among groups (3mAL, 7mAL, 7mFR, 24mAL, and 24mFR) were determined by one-way, two-way, or three-way ANOVA, followed by Tukey's post hoc test. For statistical information about the effects of ageing, diet, OLLT, and their interactions, data have been included in all figures (see [Supplementary information](#app1-nutrients-11-01865){ref-type="app"} to [Figure 1](#nutrients-11-01865-f001){ref-type="fig"}, [Figure 2](#nutrients-11-01865-f002){ref-type="fig"}, [Figure 3](#nutrients-11-01865-f003){ref-type="fig"}, [Figure 4](#nutrients-11-01865-f004){ref-type="fig"}, [Figure 5](#nutrients-11-01865-f005){ref-type="fig"}, [Figure 6](#nutrients-11-01865-f006){ref-type="fig"}, [Figure 7](#nutrients-11-01865-f007){ref-type="fig"}, [Figure 8](#nutrients-11-01865-f008){ref-type="fig"}, [Figure 9](#nutrients-11-01865-f009){ref-type="fig"} and [Figure 10](#nutrients-11-01865-f010){ref-type="fig"} for further details). For simple comparisons between two groups (7mAL versus 24mAL; 7mAL versus 7mFR; 24mAL versus 24mFR, CONTROL versus OLLT), an unpaired Student's t test was performed. Statistical significance was set at p ≤ 0.05. 3. Results {#sec3-nutrients-11-01865} ========== 3.1. Physiological and Metabolic Parameters in the Fasting State {#sec3dot1-nutrients-11-01865} ---------------------------------------------------------------- We have previously shown that 7-month-old and 24-month-old Wistar rats fed AL are characterized by increased adiposity and overall insulin resistance \[[@B17-nutrients-11-01865]\]. In addition, we have also reported an increase in leptinemia, triglyceridemia, as well as the presence of fatty liver and hypertrophic adipocytes in these rats (16, 18, 25, and 28). Thus, in agreement with our previous observations, mature middle-aged (7-month-old ad libitum (7mAL)) and middle-old aged (24-month-old ad libitum (24mAL)) Wistar rats exhibit increased body weight, visceral adiposity, and leptinemia when compared to the young Wistar rats (3mAL) ([Table 1](#nutrients-11-01865-t001){ref-type="table"}). Moreover, the HOMA-IR values indicated that the state of insulin resistance was significantly higher in 24mAL than in 7mAL rats ([Table 1](#nutrients-11-01865-t001){ref-type="table"}). According to our previous observations \[[@B17-nutrients-11-01865],[@B25-nutrients-11-01865]\], moderate chronic FR reduced body weight and adiposity in 7-month-old and in 24-month-old rats ([Table 1](#nutrients-11-01865-t001){ref-type="table"}). Nevertheless, moderate chronic FR prevented the increase in HOMA-IR in 7mFR but not in 24mFR old rats ([Table 1](#nutrients-11-01865-t001){ref-type="table"}), confirming the efficiency of the FR at an earlier age \[[@B17-nutrients-11-01865]\]. Then, we studied the effect of FR in postprandial lipemic response, during an OLLT performed in rats at two different age stages, 7- and 24-month-old. In order to reduce the impact of the type of meal, a bolus of extra-virgin olive oil (1 mL/Kg of body weight) was given. 3.2. Effect of Food Restriction in Serum Glucose, Insulin, and Glucagon Profiles during the OLLT {#sec3dot2-nutrients-11-01865} ------------------------------------------------------------------------------------------------ Before the OLTT ([Scheme 1](#nutrients-11-01865-sch001){ref-type="scheme"}), there were no significant differences observed in the fasting serum levels of glucose and insulin between the four groups of rats used in the OLLT study ([Table 2](#nutrients-11-01865-t002){ref-type="table"}). Nevertheless, basal glucagonemia was markedly increased in the 24mAL rats ([Table 2](#nutrients-11-01865-t002){ref-type="table"}). During the OLLT, there were no change in the postprandial levels of glucose or insulin in 7mAL and 7mFR old rats ([Figure 1](#nutrients-11-01865-f001){ref-type="fig"}A, B). Nevertheless, the maximum glucose response to fat intake was very different in 24mAL rats, because after the fat bolus infusion, these rats exhibited a marked peak of hyperglycemia ([Figure 1](#nutrients-11-01865-f001){ref-type="fig"}A). In addition, a peak of hyperinsulinemia was also detected in the 24mAL rats, one hour after fat ingestion ([Figure 1](#nutrients-11-01865-f001){ref-type="fig"}B) although the value returned to basal levels at the end of the experiment. Thus, in response to the OLLT, the area under the curves (AUCs) for glucose and insulin were significantly higher in the 24mAL than in the 7mAL rats, and were reduced by FR in the older rats ([Figure 1](#nutrients-11-01865-f001){ref-type="fig"}A, B). Most interestingly, after fat intake, 7mAL rats showed the maximum postprandial response of glucagon ([Figure 1](#nutrients-11-01865-f001){ref-type="fig"}C). This response begins immediately and reached the maximal levels of 60 min after fat ingestion ([Figure 1](#nutrients-11-01865-f001){ref-type="fig"}C). Postprandial levels of glucagon were kept higher until 180 min, returning to baseline values at the end of the experiment ([Figure 1](#nutrients-11-01865-f001){ref-type="fig"}C). On the other hand, the postprandial levels of glucagon progressively decreased after fat intake in 24-month-old rats fed AL or FR, even though they were lower in 24mFR than in 24mAL rats. Additionally, the AUC data from insulin and glucagon showed inverse responses in these rats ([Figure 1](#nutrients-11-01865-f001){ref-type="fig"}B,C). Finally, in the postprandial state, the 7mAL rats achieved both a higher peak in serum glucagon levels as well as a higher overall glucagon concentration than the older rats, in response to the OLLT ([Figure 1](#nutrients-11-01865-f001){ref-type="fig"}C). 3.3. Serum Lipid Profiles during the OLLT {#sec3dot3-nutrients-11-01865} ----------------------------------------- As expected, serum lipid profile was altered at baseline in the 24-month-old rats compared to the 7-month-old rats ([Table 2](#nutrients-11-01865-t002){ref-type="table"}). FR impaired the increase in TG and NEFA, as well as the glycerol levels in both groups of -7 and 24-month-old rats ([Table 2](#nutrients-11-01865-t002){ref-type="table"}). On the other hand, confirming our previous observations \[[@B14-nutrients-11-01865]\], the 24mAL rats showed the lowest values of KB, compared to the 7mAL rats, at the fasting state ([Table 2](#nutrients-11-01865-t002){ref-type="table"}); being similar to those reached in food-restricted rats ([Table 2](#nutrients-11-01865-t002){ref-type="table"}). Next, the effects of the OLLT on the serum TG levels are shown in [Figure 2](#nutrients-11-01865-f002){ref-type="fig"}A. In the four groups of animals studied, the postprandial peak for TG was observed 180 min after the OLLT, although both the maximum and the total response of TG significantly increased further, at older ages, and decreased with FR ([Figure 2](#nutrients-11-01865-f002){ref-type="fig"}A). The postprandial NEFA, glycerol, and KB responses to the OLLT were similar between groups and started to peak at 90 min, reaching the maximal levels between 100--180 min after fat ingestion ([Figure 2](#nutrients-11-01865-f002){ref-type="fig"}B--D). There was a small increase in the NEFA concentration and their levels were close to the respective basal values at 240 min in the four groups of rats studied ([Figure 2](#nutrients-11-01865-f002){ref-type="fig"}B). In parallel, KB increased progressively and their levels remained relatively elevated in the four groups of rats, 4 h after the OLLT, compared to their respective basal levels ([Figure 2](#nutrients-11-01865-f002){ref-type="fig"}D). Moreover, 4 h after the OLLT, the glycerol concentration returned to their basal levels, only in the 24mAL and 24mFR rats ([Figure 2](#nutrients-11-01865-f002){ref-type="fig"}C). In summary, 7mAL and 24mAL showed the maximum rise and the total increase in TG, NEFA, glycerol, and KB levels in the non-fasting state, which are reflected in their respective AUC values ([Figure 2](#nutrients-11-01865-f002){ref-type="fig"}B--D). On the other hand, total cholesterol levels in the fasted state were higher in the 24mAL rats compared to the 7mAL rats ([Table 2](#nutrients-11-01865-t002){ref-type="table"}), whereas the fasting levels of HDL, the main cholesterol transport lipoprotein in rats \[[@B28-nutrients-11-01865]\], decreased in the 24-month-old rats. Moreover, FR induced significant changes in the HDL-cholesterol levels in the 7mFR but not in the 24mFR rats, compared to their AL littermates ([Table 2](#nutrients-11-01865-t002){ref-type="table"}). After the OLLT, the postprandial behavior of total cholesterol concentration did not change significantly in the four groups of rats ([Figure 3](#nutrients-11-01865-f003){ref-type="fig"}A). However, the total response in cholesterol levels to the fat-rich meal was significantly lower in the 7mFR rats than in the rest of the groups studied ([Figure 3](#nutrients-11-01865-f003){ref-type="fig"}A). In the case of the HDL-cholesterol, we observed a marked drop in the HDL-cholesterol concentration in the 7mFR rats, 60 min after fat ingestion ([Figure 3](#nutrients-11-01865-f003){ref-type="fig"}B). Nevertheless, the total response was similar between 7mAL and 7mFR rats and was significantly higher than the total response observed in the 24-month-old rats ([Figure 3](#nutrients-11-01865-f003){ref-type="fig"}B). 3.4. Changes in the Metabolic Capacity of Liver after the OLLT {#sec3dot4-nutrients-11-01865} -------------------------------------------------------------- Taking into account that delayed lipid absorption processes significantly affects the magnitude of postprandial dyslipidemia in ageing; we decided to study the expression of genes involved in the management of lipids and glucose disposal in liver and vWAT, through quantitative RT-PCR, at baseline and 4 h after the fat load. At baseline, the capacity to remove chylomicrons (CMs) and VLDL-remnants from the blood (LRP-1) was reduced in the 24-month-old rats, when compared to the 7-month-old rats ([Figure 4](#nutrients-11-01865-f004){ref-type="fig"}A). In addition, the expression of PPARα, PGC-1α ([Figure 4](#nutrients-11-01865-f004){ref-type="fig"}B,C), and those genes involved in beta-oxidation of CPT-1a, MCAD, and LCAD ([Figure 4](#nutrients-11-01865-f004){ref-type="fig"}D--F) were lower in the 24-month-old than in the 7-month-old rats. We observed the same trend for the genes involved in fatty acid synthesis and elongation (FAS, Scd-1, ELOVL6) ([Figure 5](#nutrients-11-01865-f005){ref-type="fig"}A--C), and those related to glycerol and TG synthesis/VLDL formation and secretion (DGAT2, MTTP, Pck1) ([Figure 5](#nutrients-11-01865-f005){ref-type="fig"}D--F), indicating that hepatic lipoprotein and lipid metabolism progressively decays from mature middle-age, in Wistar rats. On the other hand, after the OLLT, the mRNA levels of genes involved in fatty acid and cholesterol uptake from circulation as CD36 and LDLR were markedly increased in the four groups of animal studied ([Figure 4](#nutrients-11-01865-f004){ref-type="fig"}G,H), although the major changes were observed in the 24mFR rats. Additionally, the MCAD mRNA levels were not altered, except in the 7mFR rats, where it was found to be reduced ([Figure 4](#nutrients-11-01865-f004){ref-type="fig"}E), whereas the LCAD levels were reduced in all studied rats ([Figure 4](#nutrients-11-01865-f004){ref-type="fig"}F). Of note, the reduction in the hepatic oxidative response to a high fat load did not depend on the adiposity. Next, we measured the mRNA expression levels of the major genes involved in fatty acid esterification/VLDL formation, in response to the OLLT. As observed in [Figure 5](#nutrients-11-01865-f005){ref-type="fig"}A, all groups of rats exhibited enhanced expression of hepatic FAS. Interestingly, the mRNA levels of the Scd-1, DGAT2, and MTTP increased 5-, 10-, and 40-fold, respectively, in the 24mFR rats, compared to their basal levels ([Figure 5](#nutrients-11-01865-f005){ref-type="fig"}B,D,E), suggesting an increased capacity for fatty acid esterification in these rats. In line with the gene expression analysis, we found that the hepatic TG levels were markedly increased in the 24mFR rats after the acute fat load ([Figure 6](#nutrients-11-01865-f006){ref-type="fig"}A). Finally, the mRNA levels of the two genes related to gluconeogenesis, Pck1, and Glu6Pase, were increased ([Figure 5](#nutrients-11-01865-f005){ref-type="fig"}F,G), whereas the expression levels of Glut2, which is involved in glucose uptake in the liver, were decreased after the acute lipid load in all groups of rats, compared to their basal levels ([Figure 5](#nutrients-11-01865-f005){ref-type="fig"}H). Thus, acute fat ingestion decreased glucose utilization capacity by the liver, while inducing transcription of genes that code for enzymes and factors involved in lipoprotein metabolism, fatty acid esterification, and oxidation, as well as hepatic glucose production. 3.5. Changes in the Metabolic Capacity of vWAT after the OLLT {#sec3dot5-nutrients-11-01865} ------------------------------------------------------------- Before the OLLT, the mRNA levels of several genes related to white fat differentiation (PPARγ), hydrolysis of TG in TG-rich lipoproteins, lipoprotein-remnants clearance, and free fatty acids uptake from circulation (LPL, LDLR, and CD36), were significantly lower in the 24mAL and FR rats, when compared to the 7mAL or FR rats ([Figure 7](#nutrients-11-01865-f007){ref-type="fig"}A--D). In addition, the expression of genes involved in lipogenic capacity, triglyceride mobilization and glucose uptake (Plin1, AQP7, Glut4, Glut1) ([Figure 7](#nutrients-11-01865-f007){ref-type="fig"}E--H), as well as those involved in mitochondrial fatty acid oxidation (CPT-1b, MCAD, and LCAD) ([Figure 8](#nutrients-11-01865-f008){ref-type="fig"}A--C), were also decreased in 24mAL and FR rats when compared to the 7mAL or FR rats. These data confirmed that 24 m old rats have lower expression of metabolic genes in vWAT than 7 m old rats, sustaining that basal expression of genes involved in lipid metabolism deteriorates progressively during aging in vWAT independently of obesity. Interestingly, fat meal increased the mRNA levels of PPARγ, LPL, LDLR, CD36, Plin1, AQP7, Glut4, and Glut1 ([Figure 7](#nutrients-11-01865-f007){ref-type="fig"}A--H) by more than 4-fold and those of CPT-1b, MCAD, and LCAD, in the 24-month-old rats fed AL or FR ([Figure 8](#nutrients-11-01865-f008){ref-type="fig"}A--C). Additionally, the mRNA levels of the inflammation marker TNFα in the vWAT were also increased 4 h after the OLLT in the 7mAL, 24mAL, and 24mFR rats ([Figure 8](#nutrients-11-01865-f008){ref-type="fig"}D), paralleled to the visceral adiposity index reported in [Table 1](#nutrients-11-01865-t001){ref-type="table"}. To investigate whether the lipid overload induced brown fat selective genes in the vWAT, we analyzed gene expression of the PDRM16 and UCP1. Our results showed that PDRM16 and UCP1 mRNA levels increased significantly in the vWAT from 24-month-olds but not in the 7-month-old rats ([Figure 8](#nutrients-11-01865-f008){ref-type="fig"}F,H). These results suggest that an oral fat load markedly induced the expression of genes involved in the browning program in 24-month-olds but not in 7-month-old rats, although their metabolic consequences need to be elucidated. Interestingly, in response to fat load, the expression of PPARγ ([Figure 7](#nutrients-11-01865-f007){ref-type="fig"}A), LDLR, Glut4, Glut1 ([Figure 7](#nutrients-11-01865-f007){ref-type="fig"}C,G,H), as well as CPT-1b, LCAD, TNFα, PGC-1α, PDRM16, and UCP1 ([Figure 8](#nutrients-11-01865-f008){ref-type="fig"}A,C--F,H) was not induced in the vWAT from the chronic 7mFR-old rats, suggesting a decreased thermogenic capacity, following the FR in the 7-month-old rats. Nevertheless, the TG storage capacity was significantly increased in these rats ([Figure 6](#nutrients-11-01865-f006){ref-type="fig"}B), without any sign of inflammation ([Figure 8](#nutrients-11-01865-f008){ref-type="fig"}D). On the other hand, the TG storage capacity was also significantly increased in the 7mAL, 24mAL, and 24mFR rats ([Figure 6](#nutrients-11-01865-f006){ref-type="fig"}B). 3.6. Changes in Hepatic and vWAT ChREBP Activity after the OLLT {#sec3dot6-nutrients-11-01865} --------------------------------------------------------------- ChREBP, in liver, promotes lipogenesis and fatty acid esterification \[[@B19-nutrients-11-01865],[@B29-nutrients-11-01865]\], whereas in adipocytes, regulates the expression of PPARγ target genes involved in thermogenesis and promotes insulin sensitivity \[[@B29-nutrients-11-01865]\]. It is known that ChREBPα is translocated from the cytosol to the nucleus upon glucose stimulation, to induce ChREBPβ transcription, which remains localized in the nucleus and has much more potent trans-activity than ChREBPα \[[@B29-nutrients-11-01865]\]. In this work, and in agreement with our previous observations reporting that the liver of middle-old age Wistar rats exhibit higher levels of total ChREBP in the nuclear fraction in both fasted and fed state, independently of the levels of the starvation signal ketone bodies \[[@B16-nutrients-11-01865]\], we show that the hepatic mRNA levels of both ChREBP isoforms are higher in 24 m than in 7 m old Wistar rats ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}A,B). Interestingly, both ChREBP isoforms were significantly upregulated after 4 h of fat intake in the four investigated groups ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}A,B), indicating that in the liver, ChREBP activity remain higher in the postprandial state despite the lack of carbohydrates in the meal. Of note, ChREBP-β expression increased 25-fold in 24mFR rats ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}B). Then, we suspect that the increases in hepatic triglyceride content in 24mFR rats after the OLLT were due to increased ChREBP-β-mediated hepatic TG synthesis/esterification in these rats ([Figure 6](#nutrients-11-01865-f006){ref-type="fig"}A). In contrast to the liver, at baseline, the mRNA levels of both ChREBP isoforms were decreased in vWAT with ageing in Wistar rats ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}C,D). In response to fat load, ChREBP-α mRNA levels increased whereas those of ChREBP-β decreased in vWAT ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}C,D). Next, we determined the expression levels of FGF21, a direct target gene of ChREBP, in liver and vWAT, at baseline and after the OLLT. At baseline, hepatic FGF21 mRNA levels were significantly reduced in 7mFR and 24mFR rats ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}E), while in vWAT the mRNA levels of FGF21 changed with age and adiposity ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}F). Fat ingestion increased hepatic mRNA levels of FGF21 in the four groups of rats independently of age and nutritional status ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}E). In vWAT, the intake of fat increased the mRNA levels of FGF21 in 7mAL, 24mAL and 24mFR rats ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}F), rats characterized with the highest values of visceral adiposity ([Table 1](#nutrients-11-01865-t001){ref-type="table"}). This latter finding parallel the increase observed in LDLR, Glut4 and Glut1 mRNA levels in these animals ([Figure 7](#nutrients-11-01865-f007){ref-type="fig"}C,G,H). Conversely, the effects of fat ingestion on the expression of FGF21 ([Figure 9](#nutrients-11-01865-f009){ref-type="fig"}F) and the genes mentioned above ([Figure 7](#nutrients-11-01865-f007){ref-type="fig"}C,G,H) were blunted in vWAT from 7mFR rats, the rats with the lowest values of visceral adiposity ([Table 1](#nutrients-11-01865-t001){ref-type="table"}). 3.7. Changes in the Basal and Postprandial Serum Levels of FGF21 after the OLLT {#sec3dot7-nutrients-11-01865} ------------------------------------------------------------------------------- The data shown in [Figure 10](#nutrients-11-01865-f010){ref-type="fig"} indicate that in Wistar rats the circulating levels of FGF21 decrease after mature middle-age. Interestingly, baseline serum FGF21 levels in 7mFR rats, with the lowest values of adiposity ([Table 1](#nutrients-11-01865-t001){ref-type="table"}), were particularly high when compared to 7mAL, 24mAL and 24mFR rats, but decreased dramatically after acute fat ingestion ([Figure 10](#nutrients-11-01865-f010){ref-type="fig"}). Nevertheless, in rats with higher visceral adiposity index (7mAL, 24mAL, 24mFR) ([Table 1](#nutrients-11-01865-t001){ref-type="table"}), serum FGF21 levels increased 4 h after fat consumption ([Figure 10](#nutrients-11-01865-f010){ref-type="fig"}). In order to understand the effect of age on the circulating levels of FGF21 we measured it in young 3-month old rats fed AL. As shown in [Figure 10](#nutrients-11-01865-f010){ref-type="fig"}, serum FGF21 levels were low at 3-month, increase progressively until mature middle age and then fall slightly with aging. 4. Discussion {#sec4-nutrients-11-01865} ============= Elevated concentration of circulating lipids after a fat-rich meal is a feature of lipid profile in adult and older individuals, which becomes relevant in the development of obesity and its associated metabolic disorders like type 2 diabetes and liver and cardiovascular diseases. Our previous observations demonstrated that ageing in Wistar rats is associated with increased visceral adiposity, elevated levels of ChREBP and increased lipid deposition in the liver \[[@B16-nutrients-11-01865],[@B17-nutrients-11-01865],[@B18-nutrients-11-01865],[@B30-nutrients-11-01865]\]. In this work, we studied the effects of ageing and moderate FR on the development of dyslipidemia in the postprandial state, as well as the transcriptional response of liver and vWAT to the acute fat intake in mature middle-aged and middle-old-aged Wistar rats. Previously, an oral glucose tolerance test (OGTT) and the euglycemic-hyperinsulinemic clamp technique revealed that 24-month-old Wistar rats need more insulin than 3- and 7-month-old rats, to maintain the serum glucose concentration at euglycemic levels, reflecting peripheral insulin resistance (impaired glucose uptake by adipose tissues and muscle) without diabetes \[[@B17-nutrients-11-01865]\]. Herein, the OLLT confirmed these previous observation and brought out another relevant aspect that concerns α-cells in both 7mAL and 24mAL rats, because impaired suppression of glucagon in response to meal challenge is an indicator of insulin resistance in α-cells, as has been reported in diabetic and prediabetic humans \[[@B31-nutrients-11-01865],[@B32-nutrients-11-01865]\]. In this sense, derangements of α-cells at early age might precede the development of dyslipidemia with ageing in Wistar rats, as suggested in early and recent studies regarding the role of glucagon on lipid metabolism \[[@B33-nutrients-11-01865]\]. Moreover, our results clearly showed that postprandial dyslipidemia, mainly hypertriglyceridemia, progressively increased with ageing in Wistar rats, and as suggested by gene expression analysis, it goes in parallel with the progressive decrease in the hepatic capacity to oxidize fatty acids and the progressive decay of the adipose tissue lipogenic capacity, in a situation of metabolic inflammation. To the best of our knowledge, we found for the first time that hepatic expression of ChREBP-β isoform increased with age and after an acute fat intake in AL or in chronic FR rats. Additionally, the OLLT revealed a coordinated upregulation of ChREBP-β expression and their target genes CD36, LDLR, Glu6Pase, FAS, Scd-1, DGAT2, MTTP, and FGF21 in the liver. Thus, although glucose is the main inductor of ChREBP expression and activity in the liver \[[@B34-nutrients-11-01865]\], unexpectedly and despite the lack of glucose in meals, during a fat load, we found a significant induction of ChREBP expression in the liver, in response to an extra-virgin olive oil load. Previous studies have shown that liver X receptors (LXR) regulate the hepatic expression and transcriptional activity of ChREBP, SREBP-1c, and some lipogenic genes, such as FAS, Scd1, and ELOVL6 \[[@B35-nutrients-11-01865],[@B36-nutrients-11-01865]\]. Moreover, it has recently been reported that oleic acid, the most abundant component of the extra-virgin olive oil, promotes LXR-dependent hepatic lipogenesis in young mice, increasing the expression and nuclear levels of SREBP-1c, and to a lesser extent, that of ChREBP \[[@B37-nutrients-11-01865]\]. Despite this, in 24-month-old rats characterized by insulin resistance and high hepatic TG content, we have previously shown that the total and nuclear levels of ChREBP were markedly up-regulated, while hepatic expression of SREBP-1c and the nuclear levels of its mature form were not altered compared to 3-month-old rats \[[@B16-nutrients-11-01865]\]. Thus, according to the data reported herein we suggest that ChREBP expression in the liver, with ageing, could also be modulated by fats, through LXR. Nevertheless, the elucidation of the underlying mechanisms of ChREBP induced by olive oil, with ageing, requires further investigation. Alternatively, our data suggest that the absence of glucose, during OLLT, might explain the decrease of hepatic GLUT2 expression, and as reported, in liver-specific GLUT2 knockout mice \[[@B38-nutrients-11-01865]\], this would not affect the normal glucose output capacity, since gene expression of Glu6Pase was raised and normoglycemia was maintained in the 7mAL/7mFR rats and in the 24mFR rats, and hyperglycemia occurred in the 24mAL rats. Interestingly, in response to the acute fat load, the hepatic ChREBP-β expression increased in all rats studied and their lipogenic target genes markedly increased in the 24mFR rats. Thus, it was tempting to speculate that after the OLLT, intrahepatic carbohydrate metabolites increased and might have signaled the induction of glucose production in a setting of high-fat--zero-carbohydrate meal. Accordingly, these carbohydrate metabolites or the newly synthesized glucose could activate ChREBP and the lipogenic genes that redirect carbons from glucose to glycolysis and lipogenesis, linking endogenous glucose production with lipogenesis, in line with data obtained from high fructose feeding mice and in liver-specific GLUT2 knockout mice \[[@B29-nutrients-11-01865],[@B38-nutrients-11-01865]\]. Conversely, in the 24mAL rats, the muscle and adipose tissues are resistant to insulin, as we previously demonstrated \[[@B17-nutrients-11-01865]\], thus, liver exported glucose and the glycemia rise after the OLLT. Collectively, our results suggest that ChREBP-β might have a role in gluconeogenesis and fatty acid synthesis/esterification, as in mice fed with fructose or with a high fat diet \[[@B29-nutrients-11-01865],[@B39-nutrients-11-01865]\], reinforcing its role in the development of hepatic steatosis and hypertriglyceridemia, in ageing. In addition, an acute fat load markedly downregulated the ChREBP-β expression in vWAT, which could account for a decreased insulin sensitivity in these animals \[[@B40-nutrients-11-01865]\]. The employment of the OLLT allowed us to infer that moderate chronic food restriction avoids postprandial hypertriglyceridemia in mature middle-aged and also in middle-old-aged Wistar rats, supporting the notion that postprandial TG levels, first of all, depend on the degree of adiposity. Nevertheless, our gene expression data also suggest that FR improves dyslipidemia in 7mFR and 24mFR rats through different mechanisms related to the control of inflammation, insulin responsiveness, and metabolic functionality of adipocytes from vWAT, in mature middle-aged or accumulation of TG in the liver, and induction of vWAT lipogenesis, and thermogenesis in the middle-old aged Wistar rats, respectively ([Figure 11](#nutrients-11-01865-f011){ref-type="fig"}). However, the protein modifications associated with the process of ageing \[[@B41-nutrients-11-01865]\] unrelated to gene expression should be taken into consideration before a final conclusion might be drawn. Additionally, our data showed an increased basal expression of FGF21 in vWAT and its circulating levels in the 7mFR rats, before the OLLT, suggesting that FGF21 in an endocrine/paracrine manner, could coordinate the induction of the lipogenic program of white adipocytes from rats, when adiposity is markedly reduced \[[@B42-nutrients-11-01865],[@B43-nutrients-11-01865]\]. In addition, FGF21 might influence glucagon secretion in 7mFR rats, since administration of FGF21 decreased the serum levels of glucagon and insulin in ob/ob mice \[[@B44-nutrients-11-01865]\]. As expected \[[@B45-nutrients-11-01865],[@B46-nutrients-11-01865]\], acute fat load upregulates PPARα in the liver of rats with increased adiposity (7mAL, 24mAL and 24mFR rats) and, consequently, the postprandial levels of KB and FGF21 rise in these animals, as a sensor of fat overconsumption. With regards to FGF21, our study confirmed the findings that FGF21 is induced in obesity \[[@B20-nutrients-11-01865]\] through nutritional deprivation \[[@B45-nutrients-11-01865],[@B46-nutrients-11-01865]\] and through a high-fat diet, in nonhuman primates \[[@B47-nutrients-11-01865],[@B48-nutrients-11-01865]\]. Fewer data exist that have shown the impact of ageing, per se, on FGF-21, relative to the increased levels with ageing but decreased levels in centenarians \[[@B49-nutrients-11-01865],[@B50-nutrients-11-01865]\]. In this sense, our results showed that the FGF21 levels in Wistar rats increased progressively in mature middle-age and declined later with aging. Treatment with FGF21 mimetics, increases thermogenic genes and induces beigeing of white adipose tissue in hamster with increased adiposity, but not in animals with decreased adiposity \[[@B51-nutrients-11-01865]\]. Despite the controvert relationship between FGF21 and UCP1 \[[@B52-nutrients-11-01865]\], our results suggest that both age and adiposity are implicated. In this sense, it has been demonstrated that FGF21 stimulate lipid disposal and thermogenesis in brown adipose tissue in obesity, while it increases lipoprotein disposal into white adipose tissues in lean mice \[[@B41-nutrients-11-01865]\]. Furthermore, although the brown adipose tissue function decreases with ageing \[[@B53-nutrients-11-01865],[@B54-nutrients-11-01865]\], it is well-known that adipocytes expressing UCP1 are present in white fat of mice and rats, and UCP1 expression is increased after cold exposure or treatment with a beta-adrenoceptor agonist in white adipose tissue, as in typical brown fat \[[@B55-nutrients-11-01865]\]. In addition, it has recently been reported that olive oil consumption in mice is associated with upregulation of thermogenesis \[[@B56-nutrients-11-01865]\], thus, we propose that the vWAT lipogenic and thermogenic program was markedly induced in old rats, upon an acute high dose of extra-virgin olive oil, suggesting an adaptation against the excess of energy intake, as has been previously reported \[[@B57-nutrients-11-01865]\]. Of note, the reduction of postprandial triglyceridemia and glycemia in the 24mFR rats, could also be associated with the dramatic upregulation of PRMD16, FGF21, and UCP1 in vWAT. Then, in contrast to what was observed in subcutaneous adipose tissue in which UCP1 expression and brown-like adipose tissue function were reduced with aging \[[@B54-nutrients-11-01865]\], our data suggest the activation of the browning program in the vWAT of older rats. Accordingly, we hypothesized that liver from 24mFR rats could produce a molecular switcher, which determines browning of white adipocytes and overstimulates postprandial thermogenesis to dissipate energy as heat. Similar findings were reported in adenovirus-ChREBP-β-infected mice, although herein we observed that circulating levels of FGF21 did not increase in the 24mFR rats as in adenovirus-ChREBP-β-infected mice \[[@B58-nutrients-11-01865]\]. Hence, our findings support the notion that FGF21 acts in an autocrine/paracrine manner in vWAT from aged rats, overstimulating the postprandial thermogenic adaptive responses, as described \[[@B59-nutrients-11-01865]\]. Finally, based on the analysis about the effects of age, food intake (AL or FR), and their interactions in response to the OLLT, our data support that postprandial lipidemia and expression of genes in liver and vWAT are influenced by age and caloric intake. Additionally, the results of the present work reporting the induction of a thermogenic gene program in vWAT from old Wistar rats could be taken into account among the metabolic adaptations that prevents the aged Wistar rats from becoming overtly obese and diabetic. Expectedly, FR enhanced the increased thermogenic gene response to nutrients observed in the old rats, consistent with its positive effects on this tissue. Probably, the fatty acid composition of extra-virgin olive oil is a major cause of these thermogenic responses but this requires further investigation. 5. Conclusions {#sec5-nutrients-11-01865} ============== In summary, our results provide evidence that ageing and overconsumption of fat in aged rats acutely upregulates hepatic ChREBP-β expression and induces transcriptional events in liver and vWAT involved in the regulation of lipid metabolism and postprandial lipidemia. Moreover, moderate chronic FR avoids postprandial hypertriglyceridemia in both mature middle-aged and in middle-old-aged Wistar rats, supporting the notion that postprandial TG levels, first of all, depend on the degree of adiposity. However, the molecular mechanisms that influence the lipid profile in the non-fasting state are quite different between age, as well as the possible intervention of glucagon and FGF21 in the control of dyslipidemia. Although these results have been obtained in a limited low number of tested animals, and not in humans, they could contribute to a better understanding of the effects of caloric restriction and its ability to modify fasting and postprandial dyslipidemia with ageing. This work was supported by the Research Grants PCI08-0136 from Junta de Comunidades de Castilla-La Mancha \[JCCM\] and BFU2012-39705-C03-01 from Ministerio de Ciencia e Innovación. We thank Sergio Moreno for the excellent technical assistance. We are also indebted to Fernández-Agulló and Horrillo for their help in hepatic lipofuscin analysis. The following are available online at <https://www.mdpi.com/2072-6643/11/8/1865/s1>, Figure S1: Lipofuscin accumulation in hepatic senescent cells with ageing, Table S1: Macronutrients of rodent maintenance diet 2014 Teklad Global 14% protein. Supplementary information to figure 1--10: For statistical information about the effects of ageing, diet, OLLT, and their interactions. ###### Click here for additional data file. N.G. and A.A. supervised the experiments, analyzed the data, and wrote the manuscript. A.F., L.M., C.P., V.L., A.J.d.S., B.R., and M.R. developed the study design. C.P., L.M., B.R., and A.J.d.S., reviewed the manuscript and contributed to the discussion. Research Grants PCI08-0136 from Junta de Comunidades de Castilla-La Mancha (JCCM), and BFU2012-39705-C03-01 from Ministerio de Ciencia e Innovación, Spain. The authors declare no conflict of interests. OLLT---oral lipid loading test, TG---triglycerides, TRL---triglyceride-rich-lipoprotein, NEFA---non esterified fatty acids, NAFLD---non-alcoholic fatty liver disease, VLDL---very low density lipoprotein, ChREBP---carbohydrate response element binding protein, FGF21---fibroblast growth factor 21, KB---ketone bodies, OGTT---oral glucose tolerance test, DAG---diacylglycerol, vWAT---visceral white adipose tissue, T2DM---type 2 diabetes mellitus, and LXR---liver X receptors. Figures, Scheme and Tables ========================== ![Experimental design.](nutrients-11-01865-sch001){#nutrients-11-01865-sch001} ![Evolution of serum glucose (A), insulin (B), and glucagon (C) levels in 7- and 24-month-old rats fed ad libitum (AL) or food-restricted (FR) after an oral lipid loading test (OLLT). Blood aliquots from the tail vein were taken at 0, 30, 60, 90, 120, 180, and 240 min after the OLLT. Serum samples obtained by blood centrifugation at each time were pooled and used for the measurement of hormones and metabolites, made in duplicates. Results are the mean ± SEM of 8--10 rats per group. Overall changes in glucose, insulin and glucagon levels during the oral lipid loading test were calculated as the area under the curve (AUC) above the basal level. Unpaired Student's t-test, \\ p ≤ 0.0025, \\\\ p ≤ 0.0001 compared to 7-month-old rats under the same feeding protocol; + p ≤ 0.01, ++ p ≤ 0.0025, ++++ p ≤ 0.0001 compared to the age-matched ad-libitum-fed rats.](nutrients-11-01865-g001){#nutrients-11-01865-f001} ![Evolution of serum triacylglycerol (A), nonesterified fatty acid (NEFA) (B), glycerol (C), and ketone body (D) levels in 7- and 24-month-old rats fed ad libitum (AL) or food restricted (FR), after an oral lipid loading test (OLLT). Blood aliquots from the tail vein were taken at 0, 30, 60, 90, 120, 180, and 240 min after the OLLT. Serum samples obtained by blood centrifugation at each time were pooled and used for the measurement of hormones and metabolites, conducted in duplicates. Results are the mean ± SEM of 8--10 rats per group. Overall changes in TG, NEFA, Glycerol, and ketone body levels during the oral lipid loading test were calculated as the area under the curve (AUC) above the basal level. Unpaired Student's t-test, \\\\ p ≤ 0.0001 compared to 7-month-old rats under the same feeding protocol; ++ p ≤ 0.0025, ++++ p ≤ 0.0001 compared to age-matched ad libitum fed rats.](nutrients-11-01865-g002){#nutrients-11-01865-f002} ![Evolution of serum total cholesterol (A) and HDL-cholesterol (B) levels in the 7- and 24-month-old rats fed ad libitum (AL) or food restricted (FR), after an oral lipid loading test (OLLT). Blood aliquots from the tail vein were taken at 0, 30, 60, 90, 120, 180, and 240 min after the OLLT. The serum samples obtained by blood centrifugation at each time were pooled and used for the measurement of hormones and metabolites, conducted in duplicates. Results are the mean ± SEM of 8--10 rats per group. Overall changes in the total cholesterol and HDL-cholesterol levels during the oral lipid loading test were calculated as the area under the curve (AUC) above the basal level. Unpaired Student's t-test, \\\* p ≤ 0.001 compared to the 7-month-old rats under the same feeding protocol; ++++ p ≤ 0.0001 compared to the age-matched ad-libitum-fed rats.](nutrients-11-01865-g003){#nutrients-11-01865-f003} ![mRNA relative levels of LRP-1 (A), PPARα (B), PGC-1α (C), CPT-1a (D), MCAD (E), LCAD (F), CD36 (G), and LDL-R (H), in the liver of 7- and 24-month-old rats fed ad libitum (AL) or food-restricted (FR), before (CONTROL) and after an oral lipid loading test (OLLT). Results are the mean ± SEM of 8--10 rats per group, conducted in duplicates. Unpaired Student's t-test, \* p \< 0.05 compared to 7-month-old rats under the same feeding protocol; + p \< 0.05, ++ p \< 0.0001 compared to the age-matched ad-libitum-fed rats; \# p \< 0.05, \#\# p \< 0.0001 compared to basal levels before a fat load of age-matched rats under the same feeding protocol.](nutrients-11-01865-g004){#nutrients-11-01865-f004} ![mRNA relative levels of FAS (A), Scd-1 (B), ELOVL6 (C), DGAT2 (D), MTTP (E), Pck1 (F), G6Pase (G), GLUT2 (H) in the liver in 7 and 24-month old rats fed ad libitum (AL) or food-restricted (FR), before (CONTROL) and after an oral lipid loading test (OLLT). Results are the mean ± SEM of 8--10 rats per group, conducted in duplicates. Unpaired Student's t-test, \* p \< 0.05 compared to the 7-month-old rats under the same feeding protocol; + p \< 0.05, ++ p \< 0.0001 compared to the age-matched ad-libitum-fed rats; \# p \< 0.05, \#\# p \< 0.0001 compared to the basal levels before a fat load of the age-matched rats under the same feeding protocol.](nutrients-11-01865-g005){#nutrients-11-01865-f005} ![Content of triglycerides (TG) in liver (A) and visceral adipose tissue (B) in 7- and 24-month-old rats fed ad libitum (AL) or food restricted (FR), before (CONTROL) and after an oral lipid loading test (OLLT). Results are the mean ± SEM of 8--10 rats per group, conducted in duplicates. Unpaired Student's t-test, \* p \< 0.05 compared to the 7-month-old rats under the same feeding protocol; + p \< 0.05, ++ p \< 0.0001 compared to the age-matched ad-libitum-fed rats; \# p \< 0.05, \#\# p \< 0.0001 compared to the basal levels before a fat load of the age-matched rats under the same feeding protocol.](nutrients-11-01865-g006){#nutrients-11-01865-f006} ###### mRNA relative levels of PPARγ (A), LPL (B), LDL-R (C), CD36 (D), Plin-1 E), AQP7 (F), GLUT4 (G), and GLUT1 (H) in the visceral adipose tissue in the 7- and 24-month-old rats fed ad libitum (AL) or food-restricted (FR), before (CONTROL) and after an oral lipid loading test (OLLT). Results are the mean ± SEM of 8--10 rats per group, conducted in duplicates. Unpaired Student's t-test, \* p \< 0.05 compared to the 7-month-old rats under the same feeding protocol; + p \< 0.05, ++ p \< 0.0001 compared to the age-matched ad-libitum-fed rats; \#\# p \< 0.0001 compared to the basal levels before a fat load of the age-matched rats under the same feeding protocol. ![](nutrients-11-01865-g007a) ![](nutrients-11-01865-g007b) ###### mRNA relative levels of CPT-1b (A), MCAD (B), LCAD (C), TNF-α (D), PGC-1α (E), PRDM16 (F), Adrβ3 (G), and UCP-1 (H) in the visceral adipose tissue in the 7- and 24-month-old rats fed ad libitum (AL) or food-restricted (FR), before (CONTROL) and after an oral lipid loading test (OLLT). Results are the mean ± SEM of 8--10 rats per group, conducted in duplicates. Unpaired Student's t-test, \* p \< 0.05 compared to the 7-month-old rats under the same feeding protocol; + p \< 0.05, ++ p \< 0.0001 compared to the age-matched ad-libitum-fed rats; \# p \< 0.05, \#\# p \< 0.0001 compared to the basal levels before a fat load of the age-matched rats under the same feeding protocol. ![](nutrients-11-01865-g008a) ![](nutrients-11-01865-g008b) ![mRNA relative levels of α- and β-ChREBP isoforms in liver (A,B), and visceral adipose tissue (C,D), in 7- and 24-month-old rats fed ad libitum (AL) or food restricted (FR), before (CONTROL) and after an oral lipid loading test (OLLT); mRNA relative levels of FGF21 in liver (E) and visceral adipose tissue (F). Results are the mean ± SEM of 8--10 rats per group, made in duplicate. \* p \< 0.05 compared to 7-month-old rats under the same feeding protocol; + p \< 0.05, ++ p \< 0.0001 compared to age-matched ad libitum fed rats; \#\# p \< 0.0001 compared to basal levels before a fat load of age-matched rats under the same feeding protocol.](nutrients-11-01865-g009){#nutrients-11-01865-f009} ![Serum levels of the FGF21 in the 3-, 7-, and 24-month-old rats fed ad libitum (AL) or food-restricted (FR), before (CONTROL) and after an oral lipid loading test (OLLT). Results are the mean ± SEM of 8--10 rats per group, conducted in duplicates. Unpaired Student's t-test, \* p \< 0.05 compared to the 3-month-old rats under the same feeding protocol; ++ p \< 0.0001 compared to the age-matched ad-libitum-fed rats; \# p \< 0.05, \#\# p \< 0.0001 compared to the basal levels before a fat load of the age-matched rats under the same feeding protocol.](nutrients-11-01865-g010){#nutrients-11-01865-f010} ![Scheme summarizing the suggested mechanisms about the effects of ageing and moderate chronic food restriction on postprandial dyslipidemia after an oral lipid loading test (OLLT). (A) The progression of postprandial dyslipidemia with ageing, in Wistar rats, is associated to increased hepatic ChREBPβ activity and adiposity, as suggested by the gene expression analysis through increased gluconeogenesis and FA esterification in the liver. (B). Moderate chronic FR ameliorates postprandial dyslipidemia in middle-aged (7-month-old) and middle-old-aged (24-month-old) Wistar rats, through different mechanisms. In the 7-month-old rats, FR stimulated the transcriptional response leading to an increase in TG accumulation in vWAT, while in the 24-month-old rats, FR stimulated TG accumulation in both liver and in vWAT, and the activation of the browning program in vWAT.](nutrients-11-01865-g011){#nutrients-11-01865-f011} nutrients-11-01865-t001_Table 1 ###### General characteristics of the ageing model. Wistar Rats 3mAL 7mAL 7mFR 24mAL 24mFR ------------------------------ ----------- --------------- ---------------- ------------------ --------------- Body weight (g) 325 ± 17 425 ± 7 ^&^ 382 ± 4 ^+^ 646 ±19 ^&,^\* 490 ±12 ^&,+^ Mean daily food intake (g) 22 ± 1.2 22 ± 0.4 17 ± 0.5 ^+^ 23 ± 0.7 18 ± 0.3 ^+^ Visceral adiposity index (%) 2.1 ± 0.2 5.7 ± 0.3 ^&^ 1.2 ± 0.04 ^+^ 7.0 ± 0.5 ^&,^\* 1.9 ± 0.1 ^+^ Fasting serum leptin (ng/mL) 4.7 ± 0.6 7.5 ± 0.9 ^&^ 2.1 ± 0.1 ^+^ 16.3 ± 2 ^&,^\* 7.1 ± 1.2 ^+^ HOMA-IR 7.5 ± 1.5 7.8 ± 1.9 2.4 ± 0.5 ^+^ 10.1 ± 1 ^&,^\* 7.3 ± 1.2 Data were obtained from 8--10 rats per group and represent the mean ± SEM of separate determinations. The data were evaluated by one-way ANOVA followed by Tukey's post hoc test. ^&^ p \< 0.05 compared to 3-month-old rats; \* p \< 0.05 compared to 7-month-old rats under the same feeding protocol; ^+^ p \< 0.05 compared to the age-matched ad-libitum-fed rats. nutrients-11-01865-t002_Table 2 ###### Baseline levels of hormones and metabolites before the oral lipid loading test (OLLT). Wistar Rats 7mAL 7mFR 24mAL 24mFR -------------------------------- ------------- ------------------ ------------------ ------------------ Fasting serum glucose (mg/dL) 119 ± 5 128 ± 7 131 ± 5 112 ± 22 Fasting serum insulin (ng/mL) 2.0 ± 0.1 1.4 ± 0.9 2.8 ± 0.1 2.1 ± 1 Fasting serum glucagon (pg/mL) 671 ± 29 574 ± 5 1328 ± 62 \* 571 ± 8 ^+^ Fasting serum TAG (mg/dl) 156 ± 9 98 ± 14 ^+^ 171 ± 8 \* 103 ± 4 ^+^ Fasting serum NEFA (mmol/L) 1.27 ± 0.1 0.84 ± 0.002 ^+^ 1.08 ± 0.03 \* 0.87 ± 0.06 ^+^ Fasting serum glycerol (mg/dl) 47 ± 4 36 ± 0.9 ^+^ 39 ± 3 \* 24 ± 7 ^+^ Fasting serum KB (mM) 0.18 ± 0.01 0.06 ± 0.01 ^+^ 0.081 ± 0.008 \* 0.06 ± 0.002 ^+^ Fasting serum Cho (mg/dL) 101 ± 2 91 ± 19 120 ± 13 \* 107 ± 16 Fasting serum cHDL(mg/dL) 32.7 ± 1 53.0 ± 14 ^+^ 15.8 ± 1 \* 16.2 ± 1 \* Data were obtained from 8--10 rats per group and represent mean ± SEM of biochemical determinations made in duplicates. The data were evaluated by one-way ANOVA, followed by Tukey's post-hoc test. \* p \< 0.05 compared to 7-month-old rats under the same feeding protocol; ^+^ p \< 0.05 compared to the age-matched ad libitum fed rats. [^1]: These authors contributed equally to this work.	{ "pile_set_name": "PubMed Central" }
![](transedinobsoc81967-0071){#sp1 .43} ![](transedinobsoc81967-0072){#sp2 .44} ![](transedinobsoc81967-0073){#sp3 .45} ![](transedinobsoc81967-0074){#sp4 .46} ![](transedinobsoc81967-0075){#sp5 .47} ![](transedinobsoc81967-0076){#sp6 .48} ![](transedinobsoc81967-0077){#sp7 .49} ![](transedinobsoc81967-0078){#sp8 .50} ![](transedinobsoc81967-0079){#sp9 .51} ![](transedinobsoc81967-0080){#sp10 .52} ![](transedinobsoc81967-0081){#sp11 .53} ![](transedinobsoc81967-0082){#sp12 .54} ![](transedinobsoc81967-0083){#sp13 .55} ![](transedinobsoc81967-0084){#sp14 .56} ![](transedinobsoc81967-0085){#sp15 .57} ![](transedinobsoc81967-0086){#sp16 .58} ![](transedinobsoc81967-0087){#sp17 .59} ![](transedinobsoc81967-0088){#sp18 .60} ![](transedinobsoc81967-0089){#sp19 .61} ![](transedinobsoc81967-0090){#sp20 .62} ![](transedinobsoc81967-0091){#sp21 .63} ![](transedinobsoc81967-0092){#sp22 .64} ![](transedinobsoc81967-0093){#sp23 .65} ![](transedinobsoc81967-0094){#sp24 .66} ![](transedinobsoc81967-0095){#sp25 .67} ![](transedinobsoc81967-0096){#sp26 .68} [^1]: Read at a Meeting of the Edinburgh Obstetrical Society, 10th January 1934.	{ "pile_set_name": "PubMed Central" }
In 1971, "next to nothing was known about the organization of membrane proteins," says S. Jonathan Singer (University of California, San Diego, CA). Singer had proposed that there were two kinds of membrane proteins--- integral and peripheral---but the idea was, at the time, largely speculative (Singer, 1971). It was a collaborative study between Singer, his then graduate student Garth Nicolson, and Vincent Marchesi of the National Institutes of Health that provided strong evidence for the existence of peripheral proteins (Nicolson et al., 1971a). Earlier work by Marchesi and Steers (1968) had shown that the protein spectrin was associated with the membranes of red blood cells. It could be isolated by mild treatments and behaved like a water-soluble protein. Some researchers thought spectrin was typical of membrane proteins in general. Singer\'s model, however, proposed that integral membrane proteins, which passed through the membrane, would be insoluble in water. In contrast, proteins like spectrin belonged to a distinct category. "I thought that spectrin would be peripheral to the membrane and attached to specific integral proteins where they stuck out from the membrane into the cytoplasm," recalls Singer. To investigate this idea, Singer wanted to see where spectrin was located. He used his own ferritin-conjugated antibody technique (Singer, 1959) and got the antibodies into the red cell ghosts by fixing the ghosts while they had holes in their membranes from incubation in hypotonic medium (Seeman, 1967). Electron micrscopy (EM) analysis then localized the electron-dense ferritin-conjugated anti-spectrin antibodies specifically to the inner surface of the cell membrane (Nicolson et al., 1971a). Marchesi and Palade (1967) had first speculated that a kind of protein skeleton underneath the cell membrane provided mechanical strength. "We found very high concentrations of spectrin under the membrane. So we also thought it might be involved in what is now referred to as the membrane skeleton," says Singer. FigureSpectrin is localized to the inner membrane surface of red cell ghosts.SINGER Several years later these predictions were borne out. It is now known that spectrin is the most abundant peripheral membrane protein in red blood cells and the principal component of a protein meshwork, or membrane skeleton, that underlies the cell membrane. This membrane skeleton, which contains other proteins, including actin (Tilney and Detmers, 1975), restricts the lateral mobility of membrane-penetrating protein molecules (Nicolson et al., 1971b; Elgsaeter and Branton, 1973). It also maintains the structural integrity and biconcave shape of the red blood cell membrane, and explains why membranes are not mechanochemically identical to lipid bilayers (Evans, 1973). Peripheral proteins also exist beneath the plasma membrane of many nucleated cells, but these proteins form a discontinuous network under the membrane, do not include spectrin, and generally allow most integral proteins to diffuse globally in the membrane (Frye and Edidin, 1970). LB [^1]: <laura.bonetta@comcast.net>	{ "pile_set_name": "PubMed Central" }
Introduction ============ Semantic information conveyed by language is reflected in the brain response ([@bib5], [@bib6]; [@bib49], [@bib50]; [@bib41]). Specific brain activation patterns reveal fine-grained differences between semantic categories such as actions and objects ([@bib56]; [@bib42]), tools and animals ([@bib43]; [@bib7]), as well as color and form ([@bib44]; [@bib53]). Functional magnetic resonance imaging (fMRI) activation along the motor strip distinguishes between words and sentences that refer to actions involving the face, arms, or legs ([@bib52]; [@bib27]; [@bib65]; [@bib2]; [@bib32]). This "semantic somatotopy" has provided a major argument supporting the idea that semantic mechanisms are grounded in action--perception systems of the brain ([@bib50]; [@bib6]; [@bib25]). A range of behavioral, transcranial magnetic stimulation (TMS), and neuropsychological studies also supports this view ([@bib26]; [@bib46]; [@bib15]; [@bib54]; [@bib3]; [@bib10]; [@bib70]; [@bib9]; [@bib11]; [@bib45]). Previous research on the grounding of semantics in action--perception circuits has, however, suffered from a major shortcoming. Only concrete meaning of single words ("kick" vs. "pick") and sentences ("she kicks the ball" vs. "she picks the pen") has appropriately been examined. Here we ask whether semantic somatotopy in the motor system persists during processing of idiomatic sentence meaning (e.g., "she kicks the habit"). In a previous study, [@bib2] examined the pattern of brain activation during reading of metaphorical sentences including action words (e.g., "biting off more that you can chew") compared with action-related literal sentences (e.g. "biting the peach"). Although their results showed somatotopic activity in the premotor cortex for literal stimuli, they failed to reveal any significant motor activation for metaphorical sentences. This absence of effects may however be explained by methodological issues such as the limited number of stimuli (5 sentences per condition, repeated 8 times each). The aim of the present study was to more suitably test and clarify whether the motor system comes into play during comprehension of figurative action-related language. If the grounding of semantics in sensory--motor processes is a universal feature of the human cognitive system ([@bib24]; [@bib37]; [@bib26]; [@bib36]), the prediction is that action--perception information should influence semantic brain activation to sentences, even if their meaning is highly abstract. To test this, we chose to look at idioms that include words referring to actions performed using the arm and leg (e.g., "He grasped the idea" and "He kicked the habit," respectively) and examined fMRI activation in the motor areas related to upper and lower limbs as the dependent variable. The study also allowed us to address more general issues in cognitive science. According to compositional theories of semantics ([@bib17]; [@bib16]; [@bib66]), the meaning of abstract sentences is computed from the meaning of words included in these sentences and from combinatorial information. Similar to a semantic grounding perspective ([@bib24]; [@bib6]), compositional theories would therefore receive support from semantic somatotopy to idioms that include action-related words. In this case, it would be argued that the meaning of the constituent words influences fMRI activation patterns at the sentence level ([@bib23]). Alternatively, abstract idioms could be stored separate from their constituent words as whole units and could be retrieved similar to the way long words are accessed in long-term memory ([@bib8]; [@bib21]). In such a "lexicalist" approach to idioms, no action word-related semantic activation and therefore no semantic somatotopy should emerge in the fMRI signature of idioms. In language comprehension, early lexico-semantic processing of single words ([@bib51] for a review) may be followed, after a delay, by the understanding of sentence meaning ([@bib4]). In support of this view, metabolic activity related to semantic integration at the sentence level was found to be maximal at about 6--8 s after sentence completion ([@bib31]) or even later ([@bib62]). In the present study, we chose to monitor brain activation in 2 time-windows that might reflect different temporal steps during comprehension of idioms. Cortical activity was examined at the onset of the critical word of the sentences ("He grasped the IDEA"), which disambiguated the sentences as either idiomatic or literal (early analysis window), and 3 s after its end (late analysis window). The prediction was that (1) differences in cortical activation between literal and idiomatic sentences would emerge in both time-windows and that (2) semantic contribution of action words would become evident as semantic somatotopy at the sentence level, that is, in the late time-window. Materials and Methods ===================== Participants ------------ Eighteen healthy right-handed native English speakers (8 females) participated in the study. They had normal or corrected-to-normal vision and no history of neurological or psychiatric disorder. The mean age of the volunteers was 24.3 years (SD = 6.3). They were paid for their participation. Ethical approval was obtained from the Cambridge Local Research Ethics Committee. Materials --------- Seventy-six pairs of idiomatic and literal English sentences were used in this experiment. In each condition, half of the sentences included an arm-related action word (e.g., "John grasped the idea" and "John grasped the object") and the other half contained a leg-related action word (e.g., "Pablo kicked the habit," which means "to stop doing something that is difficult to stop doing," and "Pablo kicked the ball"). Four experimental conditions were thus compared: arm-related idiomatic sentences (n = 38), arm-related literal sentences (n = 38), leg-related idiomatic sentences (n = 38), and leg-related literal sentences (n = 38). Sentence length varied from 3 to 7 words. The critical words of the sentences (e.g. "idea" and "object"), which disambiguated the sentences as either idiomatic or literal, were matched using the CELEX lexical database for relevant psycholinguistic variables, including word frequency, lemma frequency, length in letters, number of syllables, bigram frequency, trigram frequency, and number of orthographic neighbors. Arm- and leg-related action words were matched along the same variables (see [Table 1](#tbl1){ref-type="table"}). The 2 types of sentences were also matched for syntactic structure (i.e., only the critical words differed between idiomatic and literal conditions) and cloze probability. The latter parameter was defined as the number of occurrences, on <http://www.google.co.uk>, of the critical verb phrase of the sentences (e.g., "grasped the idea"; 53 917 ± 12 571 for idioms vs. 37 892 ± 13 856 for literal sentences, P \> 0.05). ###### Mean values of word frequency, lemma frequency, length in letters, number of syllables, bigram frequency, trigram frequency, and number of orthographic neighbors are reported for idiomatic and literal critical words of the sentences, and for arm- and leg-related action verbs Critical words Action words ------------ ---------------- -------------- ------------- -------- -------- ------------- WORD FQ 83.8 80.8 P = 0.835 15.25 22.58 P = 0.391 LEMMA FQ 143.7 117.9 P = 0.251 93.75 113.33 P = 0.683 LETT 5.55 5.51 P = 0.891 4.67 4.17 P = 0.193 SYLL 1.59 1.53 P = 0.579 1 1 P = 0.1 BIGR 39 137 40 133 P = 0.616 32 984 23 052 P = 0.176 TRIG 4811 5126 P = 0.554 4859 1892 P = 0.122 ORTH NEIGH 5.34 5.35 P = 0.989 7.08 7.67 P = 0.783 Note: P values for ANOVAs (by items) are reported. WORD FQ = word frequency (per million); LEMMA FQ = lemma frequency (p/m); LETT = length in letters; SYLL = number of syllables; BIGR = bigram frequency (p/m); TRIG = trigram frequency (p/m); ORTH NEIGH = number of orthographic neighbors. Seventy-six baseline stimuli, consisting of strings of meaningless hash-marks matched in length with the sentences (e.g., "\#\# \#\#\#\#\#\#\# \#\#\# \#\#\#\#"), were also constructed (see [@bib27] for similar methods). Finally, 6 literal sentences (different from and not related to the experimental stimuli; e.g., "John opened the door") were used as probe sentences in a simple motor response task. Procedure --------- The main experiment was run in 2 blocks, each block consisting of 117 trials (76 experimental trials, 38 baseline trials and 3 probe trials). Sentences were presented word by word, each for 500 ms (stimulus onset asynchrony, SOA = 500 ms), in lower-case letters at the center of a computer screen. The SOA between critical words of 2 consecutive sentences was fixed (6.6 s) and the intersentence interval (ISI) (i.e., time interval between the offset of a sentence and the onset of the next sentence), during which a fixation cross remained on the screen, varied between 2.6 and 5.1 s (mean = 4.04 s, SE = 0.05; [Fig. 1](#fig1){ref-type="fig"}). Participants were given the following instructions: "After display of a fixation cross at the center of the screen, sentences will be presented word by word. Please read words silently but attentively. Sequences of symbols will also be displayed, please look at them attentively." They were told to attend to the meaning of each sentence and to be prepared to respond to test questions probing their comprehension. To this end, they had to answer simple yes/no questions about probe sentences, interspersed between critical sentences, by pressing a button on a 2-button response box either with their left index or middle finger. For instance, after reading "John opened the door", they had to answer "no" to the question "Did John open the fridge?". Note that subjects did not know which sentences were probes, that is, they had to expect questions after any sentence. Stimuli were presented in a randomized order by means of E-Prime software (Psychology Software Tools, 2001) and viewed via a back-projection screen located in front of the scanner and a mirror placed on the head coil. ![Design of the experiment. Each trial was composed of 10 displays/screens, here represented by gray boxes, where the consecutive stimuli---fixation cross "+" and words making sentences---appeared each for 500 ms. Two examples of arm- and leg-action--related sentences are given. SOA between 2 consecutive critical words (indicated in bold) was 6.6 s. A fixed delay of 2.1 s, where a fixation cross remained on the screen, was inserted between 2 consecutive trials, so that the ISI varied between 2.6 and 5.1 s. The oblique axis on the right illustrates the temporal sequence of the trials and gives the onset of the corresponding stimulus (in milliseconds). TR of the EPI sequence is also represented (TR = 2 s).](cercorbhn217f01_ht){#fig1} Subsequent to the main experiment, participants were asked to perform a motor localizer task. The localizer scans always followed the sentence experiment to avoid any attentional bias toward action-related aspects of the stimuli. Instructions on which extremity to move (right or left index finger, right or left foot) were presented visually on the computer screen. Instructions remained on the screen for 20 s each and were repeated 4 times in pseudorandomized order (see [@bib27]). Imaging Methods --------------- Subjects were scanned in a 3-T Siemens Tim Trio magnetic resonance system using a head coil. Echo-planar imaging (EPI) sequence parameters were time repetition (TR) = 2 s, time echo = 30 ms and flip angle = 78°. The functional images consisted of 32 slices covering the whole brain (slice thickness 3 mm, interslice distance 0.75 mm, in-plane resolution 3 × 3 mm). Imaging data were processed using SPM5 software (Wellcome Department of Cognitive Neurology, London, UK). Images were corrected for slice timing and then realigned to the first image using sinc interpolation. Any nonbrain parts were removed from the T1-weighted structural images by using a surface-model approach ("skull-stripping"; [@bib63]). The EPI images were coregistered to these skull-stripped structural T~1~ images by using a mutual information coregistration procedure ([@bib40]). The structural MRI was normalized to the 152-subject T~1~ template of the Montreal Neurological Institute (MNI). The resulting transformation parameters were applied to the coregistered EPI images. During the spatial normalization process, images were resampled with a spatial resolution of 2 × 2 × 2 mm^3^. Finally, all normalized images were spatially smoothed with a 10-mm full-width half-maximum Gaussian kernel, and single-subject statistical contrasts were computed by using the general linear model, including 3 orthogonal basis functions (canonical Haemodynamic Response Function \[HRF\], its time derivative and dispersion as implemented in SPM5; [@bib19]). Only the estimate for the canonical HRF was used for the second level statistics, which is a measure for the amplitude of the brain response. Low-frequency noise was removed with a high-pass filter (time constant 128 s). We modeled the 4 experimental conditions (arm idiomatic, leg idiomatic, arm literal, and leg literal sentences) with the onset of the HRF response time-locked to the onset of the critical words of the sentences (i.e., early analysis window) and to a point delayed by 3 s from their offset (i.e., late analysis window). The probe sentences were modeled as separate events, though the corresponding data were not analyzed. Results are presented for both analysis windows. Group data were analyzed with a random-effects analysis. For visual display, Figures report results at P = 0.001, uncorrected. Tables report activations that passed the threshold of P = 0.001, uncorrected; activations that survived false discovery rate (FDR) correction ([@bib20]) at P \< 0.05 are also indicated. Stereotaxic coordinates for voxels with maximal Z values within activation clusters are reported in the MNI standard space (which resembles very closely the standardized space of [@bib64]; see [@bib13]). Several statistical analyses aimed at testing different hypotheses were performed. First, to assess whether reading of idioms and literal sentences activated a common cortical network and/or possibly additional selective brain areas, we carried out an analysis with 7 regions of interest (ROIs; "General ROI analysis"). On the basis of random-effects analysis in the early analysis window, we defined 7 ROIs activated by all sentences compared with the baseline (see [Table 2](#tbl2){ref-type="table"}; left perisylvian language areas---i.e., inferior frontal gyrus \[IFG\] Brodmann area \[BA\] 45, middle temporal gyrus \[MTG\] BA 22, angular gyrus \[AG\] BA 39, fusiform gyrus \[FG\] BA 37, and temporal pole \[TP\] BA 38---dorso-lateral prefrontal cortex BA 9 \[known as being involved in idiom processing, see [@bib38]\] and right cerebellum). This definition was done by using MARSBAR software utility ([@bib14]). For each subject and each of these 7 ROIs, average parameter estimates over voxels were calculated for spheres of radius 10 mm. This was done in both early and late analysis windows. Cortical activity in the 7 ROIs was compared between the early and late windows using a 4-way ANOVA with the design Time-Window (early vs. late) × ROI × Idiomaticity (idiomatic vs. literal) × Body Part (arm vs. leg). To further analyze potential interactions, time-windows were then examined separately with a 3-way ANOVA (ROI × Idiomaticity × Body Part). Significant effects are reported in the text only if they survived Greenhouse--Geisser correction. ###### Coordinates and statistics for activation peaks produced by all sentences (idiomatic and literal) versus the baseline (hash-marks strings) in the (a) early and (b) late analysis windows ----------------------------- ![](cercorbhn217fx1_ht.jpg) ----------------------------- Second, we directly tested the hypothesis that both idioms and literal sentences that include action words activate the motor cortex somatotopically. To this aim, we performed an analysis including the 2 ROIs selected from the motor localizer task ("motor localizer ROI analysis"). Because idiomatic and literal sentences elicited mainly left-lateralized activity, we selected left-hemispheric ROIs from the right finger and right foot conditions only. Right finger movements yielded activity in the left postcentral gyrus (BA 2: −50 −22 44, t(17) = 9.54), whereas activity in a left dorsal area on the midline was observed during execution of right foot movements (BA 6: −4 −26 72, t(17) = 16.34). Note that these postcentral activations (for which maximal t-values were obtained) may be related to somatosensory self-stimulation during motor performance (see also [@bib27]). Activity in these regions was compared between the early and late analysis windows with a 4-way ANOVA (Time-Window × ROI × Idiomaticity × Body Part). Values were then subjected to a 3-way ANOVA (ROI × Idiomaticity × Body Part) where the 2 time-windows were analyzed separately. Because ROIs from the motor localizer ended up in postcentral cortex, we subsequently performed an additional statistical analysis with a priori selected ROIs along the central sulcus and precentral gyrus ("motor strip ROI analysis"; as in [@bib55]). A chain of 9 spheres with a radius of 10 mm was aligned along the central sulcus of the standard MNI brain between vertical z-coordinates 25 and 76 mm. An additional line of 9 regions just 1 cm anterior to the central ROIs was defined in the same way in the precentral gyrus. These central and precentral regions were selected a priori as belonging to the motor strip ([@bib47]; see also [@bib55] for similar methods). The subdivision of precentral and central cortex resulted in an array of 2 × 9 = 18 regions, for each of which activation values were obtained for each condition (idiomatic vs. literal, and arm- vs. leg-action relatedness) and subject. A 4-way ANOVA with the design Time-Window (early vs. late) × Dorsality (9 regions, inferior to superior) × Idiomaticity × Body Part was used to compare cortical activity along the motor strip between both analysis windows. Activity was then analyzed in each of these windows with an additional 4-way ANOVA (Frontality \[precentral vs. central\] × Dorsality × Idiomaticity × Body Part). Results ======= Behavioral Results ------------------ To ensure that the 18 participants were attentive to the silent reading task, they were asked to answer yes/no questions about probe sentences by pressing one of 2 buttons with left index or middle finger. Mean error rate was small (8.3%, SE = 2.45), indicating that they paid attention to the sentences. fMRI General Activation ----------------------- In both early and late analysis windows, comparison of all sentences (literal and idiomatic) to the baseline (hash-mark strings) revealed left-lateralized activation in core language areas, that is, the IFG and the MTG ([Table 2](#tbl2){ref-type="table"} and [Fig. 2](#fig2){ref-type="fig"}). Activity was also observed in left FG, left AG, left TP, left dorsolateral prefrontal cortex (DLPC), and right cerebellum. Regions in the primary motor and premotor cortex were further activated during reading of action-related sentences. As can be seen from [Figure 2](#fig2){ref-type="fig"}, this activity along the motor strip tended to be more distributed and to extend further in dorsal motor areas (z-coordinates \> 50 mm) in the late analysis window. ![Cortical activation during silent reading of all sentences (idiomatic and literal) relative to the baseline (hash-mark strings) in the (a) early and (b) late analysis windows (P \< 0.001, uncorrected). Results are rendered on a standard brain surface. Top panel: lateral view of the brain, Bottom panel: top view. Note the greater activation of precentral areas in the late window.](cercorbhn217f02_4c){#fig2} Idiomatic versus Literal Sentence Processing -------------------------------------------- The results broken down for the idiomatic and literal sentences, compared with the baseline, are presented in [Figure 3](#fig3){ref-type="fig"}a,b, respectively. A common network of cortical activity was observed for both conditions in both analysis windows, with the idioms eliciting overall more distributed activation. This network included core perisylvian language areas as well as the right cerebellum ([Table 3](#tbl3){ref-type="table"}). Importantly, the precentral and middle frontal gyri including the premotor and motor cortex were activated when both literal and idiomatic sentences were being processed. ###### Coordinates and statistics for activation peaks produced by literal and idiomatic sentences in the (a) early and (b) late analysis windows, compared with the baseline (hash-marks strings) Brain region MNI t(17) --------------------------------- ------- ---- ----- --------- ------- ------- $a$ Early analysis window Literal sentences IFG BA 45 LH −50 20 16 7.49 BA 47 LH −48 30 2 6.37 Middle frontal gyrus BA 6 LH −40 −2 48 8.71 BA 9 LH −38 10 24 4.81 Superior frontal gyrus BA 6 LH −4 10 54 7.27 BA 6 LH −6 12 70 4.29 MTG BA 21 LH −54 −6 −14 7.08 BA 39 LH −56 −58 12 6.82 Superior temporal gyrus BA 22 LH −64 −46 8 7.2 BA 38 LH −52 12 −18 5.61 FG BA 37 LH −42 40 −16 7.96 Cerebellum RH 12 −78 −34 8.46 Idiomatic sentences IFG BA 45 LH −48 30 4 9.52 BA 47 LH −36 32 −16 6.68 BA 46 LH −52 26 12 8.74 BA 9 LH −36 10 24 5.97 Medial frontal gyrus BA 6 LH −6 −14 70 3.89 Superior frontal gyrus BA 6 LH −4 12 58 7.38 BA 9 LH −8 50 30 5.9 MTG BA 21 LH −54 −6 −16 8.37 BA 39 LH −56 −60 12 7.02 Superior temporal gyrus BA 22 LH −64 −46 8 7.78 BA 38 LH −50 12 −22 7.5 Precental gyrus BA 6 LH −42 −4 48 7.59 BA 6 LH −38 0 32 7.42 Cerebellum RH 12 −78 −34 7.72 Idiomatic \> literal IFG BA 45 LH −48 28 6\* 5.61 BA 44 LH −58 16 10\* 4.78 $b$ Late analysis window Literal sentences IFG BA 45 LH −48 18 12 5.74 BA 47 LH −50 20 0 6.29 Middle frontal gyrus BA 47 LH −48 36 −6 4.98 BA 6 LH −40 4 48 7.63 BA 46 LH −50 24 26 5.85 MTG BA 22 LH −52 −40 2 6.91 BA 21 LH −60 −48 6 5.98 Superior temporal gyrus BA 22 LH −58 −54 12 5.48 BA 21 LH −50 −22 −6 5.12 Cerebellum RH −16 −80 −34 6.44 Idiomatic sentences IFG BA 45 LH −48 22 12 10.49 Middle frontal gyrus BA 47 LH −44 36 −4 10.68 BA 46 LH −52 24 4 11.65 BA 6 LH −34 −4 54 7.34 Medial frontal gyrus BA 8 LH −6 42 40 5.18 Superior frontal gyrus BA 6 LH −12 18 60 6.93 BA 9 LH −12 52 28 4.93 MTG BA 21 LH −52 −28 −4 6.78 BA 22 LH −56 −40 2 8.87 BA 39 LH −58 −58 20 11.01 FG BA 37 LH −46 −36 −10 6.79 Precental gyrus BA 6 RH 16 −18 68 8.17 Cerebellum RH 16 −82 −30 10.19 Idiomatic \> literal IFG BA 45 LH −44 30 2\* 4.84 Middle frontal gyrus BA 9 LH −56 20 26\* 4.37 MTG BA21 LH −62 −56 6\* 4.69 Cerebellum LH 20 −82 −32\* 5.13 Note: Activation peaks for the contrast idiomatic vs. literal stimuli are also reported. The corresponding BAs are indicated in the second column of the table. All clusters passed the significance threshold at P \< 0.001, uncorrected. Activations that did not survive FDR correction at P \< 0.05 are indexed by asterisks. MNI coordinates along with t-values are given for the maximally activated voxel in each local cluster. All clusters passed the significance threshold at P \< 0.001, uncorrected. Activations that did not survive FDR correction at P \< 0.05 are indexed by asterisks. ![Cortical activation during silent reading of (a) idioms and (b) literal sentences (P \< 0.001, uncorrected), compared with the baseline (hash-marks strings), in the early (top panel) and late analysis windows (bottom panel). Results are rendered on a standard brain surface. Specific activations for the direct contrast between idioms and literal sentences are reported for both windows in (c). The inset in (c) highlights the specific activation observed in the right cerebellum for idioms, compared with literal sentences, in the late window (bottom panel).](cercorbhn217f03_4c){#fig3} Direct comparisons between the 2 activation conditions showed that literal sentences failed to elicit stronger activation than idioms in any brain area. In contrast, stronger activation to idioms than to literal sentences was seen in IFG (pars triangularis of Broca\'s area) in both early and late analysis windows, in the pars opercularis of Broca\'s area in the early window only, and in MTG, right cerebellum and DLPC in the late window only ([Fig. 3](#fig3){ref-type="fig"}c and [Table 3](#tbl3){ref-type="table"}). General ROI Analysis -------------------- To test activity dynamics related to Time-Window, Idiomaticity, and Body Part reference of action verbs, along with their possible interactive effects, activity in 7 ROIs was compared between arm/leg-related idiomatic/literal sentences in the early and late analysis windows ("General ROI analysis," see Imaging Methods). A 4-way ANOVA (Time-Window × ROI × Idiomaticity × Body Part) revealed a significant main effect of Time-Window (F~1,17~ = 18.09, P = 0.001) and a significant Time-Window × ROI interaction (F~1,17~ = 12.31, P \< 0.001), indicating that cortical activity was weaker in the late than in the early analysis window in a range of areas (IFG, MTG, FG, TP, and AG, P values \< 0.003), whereas in other regions (cerebellum and DLPC), no significant change was observed (i.e., prolonged activation). To further analyze these complex interactions, the analysis windows were examined with separate 3-way ANOVAs (see Imaging Methods). Results, which are presented in [Figure 4](#fig4){ref-type="fig"}, first revealed a significant main effect of ROI (early window: F~1,17~ = 17.24, P = 0.001; late window: F~1,17~ = 11.65, P = 0.001), indicating that cortical activity was particularly strong in left perisylvian areas (IFG, MTG, and AG). A significant main effect of Idiomaticity (early: F~1,17~ = 11.92; P = 0.003; late: F~1,17~ = 16.42, P = 0.001) and a significant ROI × Idiomaticity interaction (in the early window only: F~1,17~ = 4.46, P = 0.002) also emerged, suggesting stronger activity for idioms than for literal sentences, especially in IFG, MTG, TP, and AG in the early analysis window (P values \< 0.025). In the late analysis window, the right cerebellum was strongly activated also showing enhancement of activity to idioms (P = 0.001). There was a significant ROI × Body Part interaction (early: F~1,17~ = 3.17, P = 0.02; late: F~1,17~ = 3.56, P = 0.01), documenting that activity in the 7 defined ROIs depended on the body part reference of the action verbs. In the early analysis window, arm-related sentences generally activated more strongly the defined ROIs than leg-related sentences (P values \< 0.035; with exceptions however in the IFG and DLPC). In the late window, activity was stronger for arm sentences than for leg sentences in the AG (P = 0.017). ![Mean parameter estimates (in arbitrary units) for the 7 ROIs in the 4 experimental conditions (arm idiomatic, leg idiomatic, arm literal, and leg literal sentences) in the (a) early and (b) late analysis windows. Error bars are reported. LDLPC, left DLPC; LIFG, left IFG; LTP, left TP; LMTG, left MTG; LAG, left AG; LFG, left FG; RCrbllm, right cerebellum.](cercorbhn217f04_4c){#fig4} Semantic Somatotopy ------------------- Inspection of [Figure 5](#fig5){ref-type="fig"} shows that in the critical comparison between idioms including arm- and leg-related action words, different activation patterns were obtained. Already in the early analysis window, only idioms that include leg words led to spreading of activity to leg areas in dorsocentral cortex compared with the baseline (right hemisphere \[RH\]: 10 −24 60, t(17) = 7.1; left hemisphere \[LH\]: −22 −26 60, t(17) = 5.3). This effect seemed even more pronounced in the late window (LH: −18 −28 72, t(17) = 7.19; RH: 22 −26 70, t(17) = 9; LH: −12 −20 72, t(17) = 7.71; RH: 20 −12 68, t(17) = 8.59). A 4-way ANOVA (Time-Window × ROI × Idiomaticity × Body Part) including the 2 ROIs selected from the motor localizer experiment ("motor localizer ROI analysis," see Imaging Methods) revealed a significant interaction between Time-Window, ROI, and Body Part (F~1,17~ = 7.6, P = 0.013). This establishes that cortical activity in motor areas was modulated by the body part relationship of words included in the sentences differently for the 2 time-windows. ![Somatotopic activation for idioms including arm- (in red) and leg-related action words (in blue), compared with the baseline (hash-mark strings), in the early (left panel) and late analysis windows (right panel; P \< 0.001, uncorrected). Results are rendered on a standard brain surface.](cercorbhn217f05_4c){#fig5} To analyze this triple interaction, activity in finger and foot areas was examined separately for early and late analysis windows (see Imaging Methods). Whereas the early analysis window failed to show any ROI × Body Part interaction, the late window revealed a significant interaction of these 2 factors (F~1,17~ = 7.42, P = 0.014). This result shows that sentences including leg-related action words elicited stronger activity in the left foot dorsal area, whereas arm-related sentences recruited more strongly the left finger lateral area. Note that there was overlap, but not absolute congruency, between movement and action-sentence-related focal activations in the motor system, an observation which is in very good agreement with previous work on single action words ([@bib27]; [@bib32]) where also overlap, but not exact congruency, between the brain loci processing movements and action-related language was found ([@bib27] reported the highest t-value to arm words at z = 48 mm but that for finger movements at z = 60 mm, 12 mm dorsal to it). A significant ROI × Idiomaticity interaction (F~1,17~ = 7.99, P = 0.012) finally documented that idioms produced greater activation than literal sentences particularly in the left foot dorsal area. Note again that the ROIs from the motor localizer were however located in the postcentral cortex, possibly due to somatosensory self-stimulation during movement execution (for discussion, see also [@bib27]). To further examine semantic somatotopy to arm- and leg-related sentences, we carried out an additional analysis for both analysis windows in regions along the motor strip ("motor strip ROI analysis," see Imaging Methods). A 4-way ANOVA (Time-Window × Dorsality × Idiomaticity × Body Part) showed a significant interaction between Time-Window, Dorsality, and Body Part (F~8,136~ = 4.09, P \< 0.001), documenting that the body part reference of action words modulated cortical activity in motor areas differently for the 2 time-windows. Given that no significant indication of somatotopy was found in the early analysis window with the "motor localizer ROI analysis" (see above), the "motor strip ROI analysis" including the chain of 2 × 9 regions along the motor strip was performed for the late window only. A 4-way ANOVA (Frontality × Dorsality × Idiomaticity × Body Part) revealed a significant interaction between the Dorsality and Body Part factors (F~8,136~ = 4.40, P \< 0.001). As can be seen in [Figure 6](#fig6){ref-type="fig"}, in dorsal ROIs (z-coordinate ∼75 mm), leg sentences elicited stronger activation than arm sentences, and the opposite pattern, relatively stronger arm-sentence activation, was seen in lateral ROIs (25 mm ≤ z ≤ 50 mm). Semantic somatotopy (Dorsality × Body Part interaction) was also found in separate analyses for central (P = 0.002) and precentral regions (P = 0.05), suggesting that both motor and premotor cortex contributed to this effect. A 2-way ANOVA (ROI × Body Part), including a dorsal motor ROI (computed over the 3 regions with 73 ≤ z ≤ 76 mm) and a lateral motor ROI (computed over the 3 regions with 35 ≤ z ≤ 52 mm), confirmed this result by showing a significant ROI × Body Part interaction (F~1,17~ = 6.55, P = 0.02). Finally, additional analyses on local differences revealed stronger leg- than arm-sentence responses in the most dorsal ROI (z = 75 mm; P = 0.033), whereas stronger activity for arm sentences was found in a more lateral ROI (z = 44 mm; P = 0.037). These regions are in good agreement with those reported by [@bib27] who found maximal activation probabilities (t-values) in precentral gyrus for arm words at z = 48 mm and for leg words at z = 64 mm. ![Semantic somatotopy for literal and idiomatic sentences along the motor strip in the late analysis window. Bar graphs show mean parameter estimates (in arbitrary units) for the 9 ROIs aligned along the central sulcus and the precentral gyrus that are reported for sentences including arm- (in red) and leg-related action words (in blue). For each graph, the /x/ and /z/ coordinates are indicated at the top right (x, z). The locations of the ROIs are reported (yellow circles) on a coronal slice of the brain. Somatotopic activations elicited during finger (in red) and foot movements (in blue) during the localizer experiment are also shown.](cercorbhn217f06_4c){#fig6} A significant Frontality × Idiomaticity × Body Part interaction also emerged (F~8,136~ = 23.61, P \< 0.001), suggesting that activity along the central sulcus and in the precentral gyrus was differentially modulated by the idiomatic nature of the sentences as well as by the body part reference of the component action words. This interaction, which was not influenced by body part representations (superior vs. lateral ROIs), was due to generally enhanced blood oxygenation level--dependent (BOLD) signals to idioms in precentral cortex and an additional tendency for leg-idiomatic sentences to more strongly activate central areas. Discussion ========== Silent reading of sentences including action words activated a range of left perisylvian fronto-temporal areas with a well-known role in language processing, along with the FG and the right cerebellum. Idioms activated most of these areas more strongly than literal sentences, both in early and late analysis windows. Activity patterns critically depended on the body part reference of action-related words embedded into both idioms and literal sentences. Semantic somatotopy with stronger dorsal motor cortex activation for "leg-action" idioms ("He kicked the habit") and relatively stronger lateral motor cortex activation for "arm-action" idioms ("He grasped the idea") was evident, especially when the modeled metabolic response was adjusted to a time period after critical word ending, to capture the metabolic indexes of sentence-level meaning processing ([Figs 5](#fig5){ref-type="fig"} and [6](#fig6){ref-type="fig"}). These results establish for the first time the differential involvement of motor and premotor cortex in idiom processing and support theories that view abstract semantics as grounded in action-perception systems ([@bib50]; [@bib6]; [@bib25]; [@bib41]). Materials putting a particularly heavy burden on the language system are known to activate the left fronto-temporal language network more strongly and in a more widespread fashion than relatively simple language stimuli. For instance, it is well known that the N400 brain response is enlarged to sentences including semantically unexpected constituent words ([@bib34]; [@bib67]), the main generators of this effect being localized in the posterior perisylvian cortex ([@bib68]). [@bib59] presented sentences with unexpected ambiguities and found enhanced and more distributed fronto-temporal fMRI activation relative to unambiguous control sentences. [@bib38] probed literal and idiomatic sentences and found left fronto-temporal activation also extending into anterior inferior frontal cortex, anterior temporal cortex and AG (see also [@bib58]; [@bib39]; [@bib69] for fMRI studies on idiom and metaphor processing). Our present activity enhancements to idioms at left inferior frontal and middle temporal sites are consistent with this pattern of results. They may, however, be best explained as an index of increased workload on the language system rather than as specific brain signature of idiom processing. In the late analysis window examined here, we found additional idiom-related activation enhancement in the cerebellum and in the middle frontal gyrus extending into frontocentral motor and premotor cortex. The stronger cerebellar activity to idioms extends previous findings on the role of this structure in perceptual and language processing ([@bib35]; [@bib12]; [@bib18]; [@bib1]). The co-occurrence of cerebellar and motor cortex increased activation to idiomatic sentences further suggests a consorted role of these structures in motor cognition ([@bib36]) brought about by abstract action-related language. For both idiomatic and literal sentences, we observed differences that reflected the meaning of their constituent arm- and leg-related action words. This influence became evident in a range of analyses, most notably with the significant ROI × Body Part interaction in the analysis of motor and premotor cortex activation ("motor strip ROI analysis"; [Fig. 6](#fig6){ref-type="fig"}). This interaction did not involve the "Idiomaticity" factor, thus documenting that the well-known semantic somatotopy found for concrete action words and sentences ([@bib52]; [@bib27]; [@bib65]; [@bib2]) can be replicated for abstract sentences including action words. Activation of frontocentral motor and premotor areas was relatively weak at the onset of critical words but was strong after their offset ([Fig. 5](#fig5){ref-type="fig"}). If the fMRI brain response reflecting sentence meaning is delayed relative to that of single words ([@bib31]), this late activation of motor areas can be linked to the sentence processing stage. Any contribution of individual action words would have been expected to arise at action word onset (on average 1.2 s before critical word onset and thus 4.7 s before the late analysis window) and to decrease with time. Our results therefore suggest that the orchestration of abstract meaning in the human brain is not solely explained by the activation of unspecific semantic centers in fronto-temporal cortex, but that it involves late complementary activations in the sensory--motor system. These referentially grounded activations may play a specific functional role in the composition of sentence meaning. However, further work using techniques such as TMS and neuropsychological studies in brain-damaged patients are necessary to draw firm conclusions on functional contributions of the motor system to idiom comprehension. The present results support a compositional perspective on semantic processing postulating that idiom meaning is computed from the semantics of constituent words and from combinatorial information. Semantic somatotopy to idioms indeed suggests that meaning aspects of words included in these sentences are being re-accessed and combined in the relatively late construction of sentence meaning. Access to concrete referential aspects of constituent words, as it regularly occurs in language comprehension, appears not as an irrelevant by-product but rather as an important step in the comprehension process, which may play a role in the comprehension of figurative language too ([@bib22]; [@bib23]; [@bib66]). In the context of the present study, one may argue that the activation of dorsal and lateral motor and premotor cortex was related to the processing of leg- and arm-related words per se and not to the comprehension of sentence meaning. We should remind the reader, however, that somatotopic semantic grounding of constituent arm/leg words in lateral/dorsal frontocentral cortex, respectively, was relatively weak at action word onset and also at presentation of critical words ([Fig. 5](#fig5){ref-type="fig"}). It became pronounced about 3 s after sentence ending, suggesting its specificity to a late stage of sentence processing. At such a late stage, it would be extremely unlikely that one particular word from the several ones included in the sentence is still processed in depth in isolation and dominates the brain response. Rather, it appears plausible that semantic integration at the sentence level underlies metabolic changes, which might occur especially late for highly abstract sentences. This pattern of results is therefore consistent with a gradual emergence of semantic somatotopy in the processing of idiom meaning and is not explained by word-related activation. Previous electrophysiological studies have shown instant spreading of activity to motor regions during action word recognition (\<200 ms; [@bib56]; [@bib28]; [@bib57]; [@bib33]; [@bib11]; for a review, see [@bib29]). Despite the apparent discrepancy between these results and those of the present study, the reader should however be reminded that 1) previous neurophysiological work focused on single words while we here used complex sentences, where action words had to be integrated into their context; 2) as opposed to fMRI, electrophysiological techniques offer high temporal resolution that allows precise tracking of the time-course of brain activation during cognitive processes. Future research, using electroencephalography and/or magnetoencephalography for instance, is therefore necessary to address the question of when grasping ideas activates the motor system. Comparing the present results with previous fMRI studies of brain activation to action-related sentences ([@bib65]; [@bib2]), it is still noteworthy that BOLD signal changes occurred relatively late in the present study. This indicates that, if idioms appear in the context of literal sentences, as they do in normal language use, semantic and any postunderstanding processes may be delayed relative to an experimental context where only literal sentences are presented. We note however that early brain reflections of semantic processing of word pairs ([@bib61]; [@bib30]) have recently been reported early-on (100--150 ms) and even the first effects of sentence-level semantics have been found between 100--200 ms ([@bib60]; [@bib48]). It may therefore be suggested that the idiomaticity of sentence stimuli in the present experiment critically contributed to the lateness of the hemodynamic brain response. To further explore the temporal structure of idiom comprehension, we would like to re-emphasize the need for future neurophysiological work. The late effects we observed here also raise the issue that motor activity could be epiphenomenal with respect to sentence comprehension or could reflect motor imagery after semantic access. It is indeed possible that postunderstanding processes ([@bib26]), following semantic sentence-level analysis, are reflected (e.g., imagining a picture or a scene matching the sentence content). If this is true, semantic somatotopy of this secondary process triggered by sentence meaning would still argue against an abstract symbolic perspective. Indeed, in such an abstract symbolic framework, the meaning of the phrase "grasp an idea" is semantically unrelated to grasping. The only way to account for such secondary somatotopic activation would be through semantic somatotopy in sentence meaning analysis and consequent somatotopy of the secondary (imagery or the like) process. Models assuming storage of idiom meanings as whole units unrelated to the meaning of their constituent words do not provide an explanation of the observed differences between idioms that include arm and leg action words. To provide such an explanation, 2 assumptions are necessary: 1) Idiom meaning must be computed on-line from the meaning of constituent words, and 2) Semantic aspects grounded in action--perception knowledge must play a critical role in the composition process yielding idiom meaning ([@bib23]). In this sense, the present results support both semantic compositionality and the grounding of figurative/abstract language in concrete sensory--motor information and in their corresponding specific brain circuits. Motor systems of the brain, including motor and premotor cortex, and the motor cognitions they process ([@bib36]) appear to be central for understanding idioms. When "grasping ideas," the motor system is engaged in a specific manner. Funding ======= Medical Research Council (UK; U1055.04.003.00001.01, U1055.04.003.00003.01) and by the European Community under the "New and Emerging Science and Technologies" Programme (NEST-2005-PATH-HUM contract 043374, NESTCOM) to F.P. V.B. was supported by a postdoctoral fellowship from the Fyssen Foundation. Funding to pay the Open Access publication charges for this article was provided by the Medical Research Council (UK). We thank Elisabeth Fonteneau and Kambiz Tavabi for helpful comments on a previous version of this paper, as well as Matthew H. Davis for his help with the SPM analysis. We also thank 4 anonymous Referees for their efforts and help in improving this work. Conflict of Interest: None declared.	{ "pile_set_name": "PubMed Central" }
Background ========== The process of linking genes to disease phenotypes is rapidly gaining momentum since the first disease-causing gene was identified 25 years ago \[[@B1]\]. Alternative approaches adopted in the past to identify disease genes are the candidate gene approach, where likely suspects are prioritised and screened on a genome-wide basis; and linkage analysis where specific loci are determined systematically using family studies. The two approaches have been synthesized into a pipeline by completion of the Human Genome Project; and further enabled by the increased availability of high-throughput experimental data and the development of sophisticated bioinformatics tools. In addition there have been efforts in the bioinformatics community to systematize and automate candidate gene prediction. Automated prediction systems provide geneticists with a reduced list of genes estimated to have a high probability of involvement in the disease phenotype by sifting through hundreds to thousands of genes. Ultimately, these tools aim to give the researcher the best possible guidance in honing in on the gene culprits for further biological confirmation. Since their introduction in the early 2000s, the predictive powers of automated candidate gene prediction systems have improved, largely due to increases in biological systems knowledge and more effective algorithms. Candidate gene prediction systems vary in their approach and the data sources they draw on in generating predictions. These are summarised in Figure [1](#F1){ref-type="fig"} and Table [1](#T1){ref-type="table"}. Comparing the performance of these systems can be difficult because of the use of custom benchmark test sets by individual groups. Typically, benchmarking data is derived from genotype-phenotype information from the Online Mendelian Inheritance in Man (OMIM) database \[[@B2]\], but groups have used varying subsets of diseases. Several groups have tried to use standard benchmark sets \[[@B3]-[@B5]\], but these efforts have been limited. In addition, it is difficult to predict whether benchmarks which predominately contain data on well characterised diseases with Mendelian transmission patterns (i.e. dominant, recessive, X-linked) resulting from mutations in single genes \[[@B6]\] will be effective in predicting genes involved in less well characterised diseases, or in complex diseases. ###### Automated Candidate Gene Prediction Systems ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Semi-Automated Systems GeneSeekeris a semi-automated web-server tool which selects positional candidates based on expression and phenotypic data from human and mouse. Queries must be formulated by the end-user using Boolean expressions \[[@B13],[@B33]\]. ♠ ◇ Systems Biology Techniques Prioritizeruses pathway and interaction data from KEGG \[[@B17],[@B34]\], Reactome \[[@B35]\], and HPRD \[[@B36]\]. Interactions are also predicted using a Bayesian technique based on GO keywords \[[@B23]\] and other databases \[[@B5]\]. In GentrepidCommon Pathway Scanning (CPS), pathways are associated with phenotypes using either known disease genes, or by searching for enrichment of pathways across multiple disease intervals associated with the phenotype \[[@B4]\]. ♠◇ Oti et aluse protein-protein interaction data from HPRD \[[@B36]\], Y2H \[[@B37],[@B38]\], and PCP \[[@B39],[@B40]\] giving coverage of 10 894 human genes \[[@B24]\]. Genotype-Phenotype Mapping Methods G2D\[[@B32]\] uses biomedical literature to associate pathological conditions with GO terms \[[@B23]\]. Candidate genes are identified by homology to GO-annotated disease-associated genes. ♠◇ GentrepidCommon Module Profiling (CMP) searches for enrichment of particular domains in gene clusters associated with a particular phenotype. Domains are extracted either from known disease genes or by comparison of multiple disease intervals \[[@B4]\]. ♠◇ POCUSsearches for over-representation of functional annotation among multiple loci associated with the same disease. Functional annotation is based on keywords from InterPro domains \[[@B22]\] and GO \[[@B23]\]. No a prioriknowledge of the phenotype is required \[[@B3]\]. ♠ Techniques based on a bipartite distribution of \"disease\" and \"non-disease\" genes The eVOCsystem uses text mining of biomedical literature to associate a phenotype with anatomy terms and links these with human expression data to produce a ranked list of disease genes. The classifier is a machine-learning technique, based on a bipartite training set of 17 known \"disease genes\" and 306 \"non-disease genes\" \[[@B30]\]. ♠ DGP (Disease Gene Prediction)is a web tool which selects genes based on protein sequence properties. The properties analysed by DGP include protein length, degree of sequence conservation, the extent of phylogenetic relationship and paralogy patterns \[[@B31],[@B41]\]. ♠ PROSPECTR(PRiOrization by Sequence and Phylogenetic Extent of CandidaTe Regions) uses an alternating decision tree to discriminate \"disease genes\" from \"non-disease genes\" using a classifier based on sequence features such as gene length, protein length, and similarity of homologs in other species \[[@B12]\]. ♠ Hybrid techniques SUSPECTScombines a genotype-phenotype mapping method based on disease-gene-associated keywords from InterPro and GO, and expression libraries, with the PROSPECTRBoolean classifier. Disease genes are prioritized \[[@B21]\]. ♠ ◇ ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ♠ Assessed here, ◇ Webserver. ![Data sources and approaches used in automated candidate gene prediction methods. (A):Most systems draw on at least two types of data. SUSPECTS\[[@B21]\] (not shown) uses keywords from InterPro \[[@B22]\] and GO \[[@B23]\], co-expression data, and also incorporates the PROSPECTRmodule \[[@B12]\] (shown on right). (B): Upper leftGene clustering approaches associate a gene cluster with a phenotype via a group member. For example, Systems Biology approaches \[[@B4],[@B5],[@B24]\] group genes whose protein products interact; and link them to a phenotype using a group-member gene associated with the phenotype. Systems Biology methods assume oligogenic diseases are associated with disruption in proteins that participate in a common complex or pathway \[[@B25]\]. Other gene clustering systems look for enrichment of keywords or domains associated with particular phenotypes and suggest candidate genes with similar properties. These systems are based on the principle that candidate genes have similar functions to disease genes already determined \[[@B26]-[@B28]\]. Upper rightPhenotype clustering approaches such as that of Freudenberg & Propping \[[@B29]\] group related phenotypes into superphenotypes. Lower leftMost of the Machine Learning approaches do not use phenotype information and are based on the concept that the genome consists of a bipartite distribution of genes: those which cause diseases, and those that do not. By analysing these two gene sets with respect to discriminating variables, a profile for \"non-disease genes\" and \"disease genes\" is produced which enables training of a classifier. A novel gene submitted to the classifier is flagged as either \"disease-causing\" or \"non-disease causing\". Systems include eVOC\[[@B30]\], PROSPECTR\[[@B12]\], SUSPECTS\[[@B21]\] and DGP\[[@B31]\]. Finally G2D, lower right, is a transitive method that maps phenotypes to genes \[[@B32]\] by interfacing literature- and keyword-based ontologies.](1471-2105-10-S1-S69-1){#F1} A recent effort by Tiffin and colleagues \[[@B7]\] to identify candidate disease genes for the complex disease type II diabetes (T2D) and the related obesity trait predicted 12 genes in previously implicated chromosomal regions. The study also allowed a limited comparison of seven candidate gene prediction systems. Since that time two genome-wide association studies (GWAs) on T2D undertaken by the Wellcome Trust Case Control Consortium (WTCCC) and the Genetics Replication and Meta-analysis Consortium (DIAGRAM) have been published \[[@B8],[@B9]\]. GWAs are a powerful tool for identifying genetic variants linked to complex diseases because they are more sensitive than linkage studies to small to moderate effect size contributions from polygenic and oligogenic diseases. The data from these GWAs allow the assessment of the predictions made by Tiffin et al., as well as evaluation of the effectiveness of predictions made by the individual automated candidate gene prediction systems used in their study and our system, Gentrepid\[[@B4]\]. We assessed the candidate gene predictions systems\' ability to select robustly supported genes from the GWAs and used them to filter noisy data from statistically less well supported genes to select favoured candidates. Results ======= Predictions ----------- All methods were given the starting set of 9556 genes mapped to chromosomal intervals implicated in T2D as assessed by Tiffin et al., except for POCUSwhich was run against a search space of 562 genes. The POCUSmethod was confined to the smaller search space because \"poorly defined susceptibility regions or regions with questionable association with the disease are obscured by background noise\" \[[@B7]\]. The number of candidate gene predictions made by the eight methods varied from two to 3093. POCUSgenerated the smallest number of candidates but neither of the two predictions matched genes in either the highly significant (HS) or medium-to-highly significant (MHWD) data sets. Other candidate gene prediction methods made considerably more predictions. The largest numbers of predictions were made by G2D(3,093 candidate gene predictions) and eVOC(2,496 predictions). These comprise almost one third and one quarter of the search space respectively. Thus neither of these methods prune the search space particularly well. Excluding POCUS, the least number of predictions was made by Gentrepidcomprising 502 genes in known-disease-gene mode. Accuracy of predictions ----------------------- To assess the accuracy of the predictions, all eight systems were compared with genes found in previously-implicated intervals strongly linked to T2D by the GWAs. Figure [2](#F2){ref-type="fig"} shows the comparative performance of seven of these methods in selecting the 11 genes in the HS GWA data set. Several metrics were calculated to assess accuracy. No metrics were calculated for POCUSas neither of its two predictions matched genes in either the HS or MHWD data sets. ![Comparison of methods against the HS (left) and MHWD (right) T2D gene data sets. Top: Relative Enrichment Ratios. Bottom: Comparisons based on Sensitivity and Specificity.](1471-2105-10-S1-S69-2){#F2} The Enrichment Ratio is a general measure of the system\'s ability to accurately prune the search space. Enrichment Ratios ranged from 1 to 5 for the seven remaining prediction systems. The highest Enrichments Ratios were obtained by Gentrepidand GeneSeeker. These results were robust when the upper and lower (not shown) 95% confidence interval limits were taken into account. The lowest Enrichment Ratios were associated with the Machine Learning methods. This is not surprising, as the classifiers are trained to distinguish \"disease genes\" from \"non-disease genes\" and are ignorant of any concept of phenotype. The Specificity of a system measures its ability to reject genes not associated with the phenotype. Specificity scores among all seven methods ranged from 0.68 to 0.99, with a median of 0.92. As a group, the Machine Learning methods were poorer at rejection. G2Dalso performed poorly on this metric, but this result is slightly misleading because it does not take into account G2D\'s prioritization method which will be discussed later. The Sensitivity is a measure of a system\'s ability to find the disease genes in the search space. A caveat here is not all of the GWA predictions are currently confirmed. G2Dis by far the standout performer in Sensitivity, with eVOCranked second. However, as can be seen from the other metrics, this result is obtained at the expense of Specificity for both systems. Gentrepid\'sSensitivity is on par with most of the Machine Learning methods but with higher Specificity. The high Specificity reflects the high quality of the data in the underlying databases. The lower Sensitivity is due to incompleteness of these databases with respect to all human genes. Figure [2](#F2){ref-type="fig"} shows the comparative performance of methods when assessed against the 61 T2D associated genes with moderate to strong SNP signals (MHWD) in the Tiffin chromosomal intervals. The MHWD data set is not as statistically well supported as the HS set, and would be expected to contain some genes associated with T2D and others that are false positives. Perhaps the most interesting metric to look at here is the Sensitivity which should fall compared to the values for the HS set because of the lower signal to noise ratio in the MHWD set. All the systems except one, SUSPECTS, passed this negative test. More importantly, application of the systems to this noisy genetic data allows selection of a subset of candidates on the basis of molecular data (see below). The results shown for Gentrepidin Figure [2](#F2){ref-type="fig"} are for the known-disease-gene mode. In ab initiomode, Gentrepid\'s CPS method identified 506 pathways containing a total of 1980 candidate gene predictions. This resulted in Enrichment Ratios of 3.3 and 2.1 when the HS and MHWD full gene sets were considered (Table [2](#T2){ref-type="table"}). ###### Gentrepid ab initioresults Predictions Reference list ER L95% U95% S L95% U95% -------------------------- ------------------- ----- ------ ------ ------ ------ ------ CPS rank 8+ pathways HS 3.3 1.1 9.4 0.45 0.21 0.72 CPS rank 8+ pathways HS -- annotated 7.2 2.1 25 1.00 0.57 1.00 CPS rank 8+ pathways MHWD 2.1 1.3 3.6 0.30 0.20 0.42 CPS rank 8+ pathways MHWD -- annotated 6.8 3.6 13 0.95 0.75 0.99 CPS interactions top 50% HS 4.1 1.2 15 0.27 0.10 0.57 CPS interactions top 50% HS -- annotated 9.0 2.2 37 0.60 0.23 0.88 CPS interactions top 50% MHWD 1.7 0.79 3.8 0.11 0.06 0.22 CPS interactions top 50% MHWD -- annotated 8.1 3.2 20 0.54 0.29 0.77 CMP top 10% HS 2.2 0.3 17 0.1 0.02 0.38 CMP top 10% MHWD 2.0 0.8 4.8 0.1 0.08 0.18 Abbreviations in Table: ER -- Enrichment Ratio, L95% -- Lower 95% confidence limit, U95% -- Upper 95% confidence limit, S -- Sensitivity In ab initiomode, the CMP method generated 527 predictions by limiting the selection to the top 10% most probable genes. This resulted in correct prediction of one gene from the HS set and five from the MHWD set, yielding a Enrichment Ratio of 2.2 when applied to the HS and 2.0 for the MHWD gene data sets. It is also interesting to note the effect of lack of annotation on these results. Only five of 11 genes in the HS dataset, and 19 of 61 genes in the MHWD set contained KEGG or BioCarta pathway annotations.When we included only genes containing pathway information from the gene datasets (designated \'annotated\' in Table [2](#T2){ref-type="table"}) we observed Enrichment Ratios of 7.2 against the HS and 6.8 against the MHWD pathway-annotated sets. Sensitivities also improved by a factor of 2 for the HS dataset. By extrapolation, if all genes were pathway annotated, we could expect approximately two- to three-fold improvement in Enrichment and Sensitivity scores. Prioritization -------------- Although the metrics discussed provide useful measures of a candidate gene prediction system\'s performance, another criterion of importance to geneticists is the system\'s ability to prioritize predictions. Although several methods claim the ability to prioritize (Table [1](#T1){ref-type="table"}), only G2Dprovided prioritized predictions in Tiffin et al. \[[@B7]\]. Hence only G2Dand Gentrepidwill be discussed here. Because Gentrepidonly made 502 predictions in toto, we took the top 502 predictions made by G2Dand recalculated the Enrichment Ratio, Sensitivity and Specificity for this restricted set of favoured predictions. ER and Specificity significantly improved to 3.1 and 0.95 such that G2Dsurpassed Gentrepid\'s gross ER and almost equalled Gentrepidin Specificity. The improvement in these two metrics came at the expense of Sensitivity which was reduced to 0.16, but the G2Dsystem still managed to maintain its lead on this metric. In G2D\'s prioritization system, a GO-metric is calculated for each gene in the search space based on how well its GO profile fits the GO profile of the disease genes inferred from MeSH terms. An R-score is calculated for each gene by normalizing against the number of genes in the genome with better GO-metrics for the phenotype. Genes with R-scores closer to zero are better fits to the phenotype. GentrepidCPS ranks genes by the number of loci in the search space involved in a particular pathway. In ab initiomode, of the 53 intervals searched, the top pathway, focal adhesion, was represented in 35 of these. All five of the HS dataset genes were represented by pathways found in at least eight intervals. Pathways implicated in at least eight intervals constituted the top 40% of the 506 pathways containing 1749 candidates. In these pathways, Gentrepididentified all five pathway annotated genes from the HS dataset, and 18 of the 19 pathway annotated genes from the MHWD gene data set. Other figures for CMP are given in Table [2](#T2){ref-type="table"}. CMP ab initiolooks for protein domains enriched in the search space compared to the genome by taking a census of domains in the search space and the genome. Two expectation values are calculated to estimate the frequency of occurrence of genes with domains of interest based on a random combination of these domains e~a~and the rarest domain e~b~\[[@B4]\]. Figure [3](#F3){ref-type="fig"} shows the data ranked on a χ^2^test based on e~a~was most effective in prioritizing the HS data. This reflects our experience of phenotypes with genotypes encoded by multi-domain proteins, as would be expected for diseases associated with signaling. For metabolic diseases associated with single-domain proteins, e~b~may be a better measure. ![MHWD dataset filtered against prioritized automatic candidate gene predictions. Genes in bold are robustly supported genes from the GWA studies (HS set).](1471-2105-10-S1-S69-3){#F3} Although the G2Dprioritization system appears more sensitive than the coarse-grained prioritization of Gentrepid, the performance of both systems was roughly equivalent against the HS set. Both systems were moderately successful in prioritizing the HS data. For example, of the seven genes in the HS dataset predicted by G2D, four were ranked in the top 15% by G2D\'s prioritization method (bold in Figure [3](#F3){ref-type="fig"}). Significant work needs to be done to improve the prioritization schemes of both G2Dand Gentrepid. Finally, we used the candidate gene predictions systems to filter the less statistically-well-supported MHWD data set (MHWD -- HS): effectively adding more power to the GWA study. Prioritized predictions are the unbolded genes in Figure [3](#F3){ref-type="fig"}. Additional unprioritized predictions made for the MHWD dataset using the other candidate predictions systems are given as supplementary data in Additional file [1](#S1){ref-type="supplementary-material"}. Discussion ========== Candidate disease gene prediction is a rapidly moving area of bioinformatics research with the potential to deliver great benefits to human health. By assisting geneticists to use existing biological information to investigate disease loci obtained by linkage analysis and association studies, disease genes can be identified more rapidly. The need for good applications in the area of candidate gene prediction is becoming increasing important as the proliferation of SNP-based association studies produces valuable genetic information in need of analysis. The biggest problem facing candidate gene prediction today is the accuracy and completeness of the underlying databases. Failure to make a prediction is mostly due to incomplete data coverage. For example, 65% of human proteins have GO terms but only 25% of these are manually annotated. Systems drawing on GO terms like G2Dare potentially able to make predictions for 65% of genes but only around one third of these are likely to be accurate. Systems Biology methods like GentrepidCPS are reliant on pathway and protein-protein interaction data. One of the databases CPS draws on is OPHID \[[@B10]\], one of the most complete protein-protein interaction datasets, containing over 48 000 interactions. However these 48,000 interactions are estimated to be only 13% of the complete human interactome \[[@B11]\]. Completeness of the underlying data clearly impacts the Sensitivity of the GentrepidCPS method. As time goes on this constraint will ease as these databases are further populated. In the meantime, we have shown that the use of independent biological data to make complementary candidate gene predictions is one way to ameliorate the problem of incomplete data coverage (see Figure [3](#F3){ref-type="fig"}) \[[@B4]\]. In addition to the predictions made by the individual candidate gene prediction systems in Tiffin et al., a set of nine \"winners\" were chosen using a consensus approach \[[@B7]\]. These nine candidate genes were independently predicted by six of the seven prediction systems studied. A larger consensus set, chosen by five of the seven methods, contained 94 genes \[[@B7]\]. None of the genes in either of these consensus lists matched any of the genes in the HS and MHWD gene sets. Even if we compile a third tier of consensus genes from any four of the seven methods (269 genes) only one gene (VEGF) fell within the HS data set and only three genes (CHN2, B4GALT5, VEGF) matched the MHWD data set. Clearly the consensus approach is not working and it is easy to see why when the underlying databases are considered (Figure [1A](#F1){ref-type="fig"}). Candidate gene prediction systems that use an independent data set, not drawn upon by most of the other methods, will be penalized. Possibly the only benefit of a consensus approach is to give the user a false sense of accuracy when confronted with noisy data. Clearly much work still remains to improve the sensitivity and specificity of candidate gene prediction methods but some general conclusions are possible. Machine Learning methods were not as effective as other methods. Most of the Machine Learning approaches do not use phenotype information and are based on the concept that the genome consists of a bipartite distribution of genes: those which cause diseases, and those that do not. The evidence supporting this assumption is limited \[[@B12]\]. We believe the concept that there is a difference between \"disease genes\" and \"nondisease genes\" is intrinsically flawed and no such Boolean classification exists. We hypothesize that the ability of these methods to predict disease genes in test sets is based on selection effects in the data: possibly rare, highly penetrant monogenic diseases, such as those involved in metabolic syndromes, are over-represented among known disease genes because they have been easier targets to identify. Although these systems were not as effective as the other candidate gene prediction systems, their performance was not greatly different. However, we believe that unlike systems which attempt to map genotype to phenotype, Machine Learning systems based on the disease gene/non-disease gene concept will not improve as more biological data becomes available. Conclusion ========== Candidate gene prediction systems have typically been benchmarked on well characterized oligogenic phenotypes. GeneSeeker \[[@B13]\] produced a 10-fold enrichment using a data set consisting of eight diseases. Gentrepid\'s combined methods \[[@B4]\] produced an Enrichment Ratio of 13 when 29 diseases with a total of 170 known disease genes were used. For 29 diseases with 163 genes, POCUS \[[@B3]\] reported Enrichment Ratios between 12 to 42-fold, depending on the size of the intervals in the search space. The PRIORITIZER \[[@B5]\] method yielded a 2.8-fold enrichment using a data set consisting of 96 heritable disorders. In summary, Enrichment Ratios of 3 to 13 have been reported in benchmarks, but a substantial part of the data used for these studies has been limited to oligogenic phenotypes, where several different genes may cause the disease, but a single mutation in each case or family has a large effect. Some doubts have been raised about the ability of systems to predict candidates for complex polygenic diseases such as T2D where multiple genes interact to create a permissive gene pool for disease genesis. The candidate gene prediction systems did prune the genome in favour of moderately to highly significant SNPs identified by the GWAs under semi-blind testing on a complex polygenic disease. Enrichment Ratios calculated in this study suggest that most of the oligogenic benchmarks have been reasonably good predictors of system performance. Methods ======= Benchmark datasets ------------------ Eight candidate gene prediction systems were assessed on their ability to predict genes involved in T2D by comparison against genes implicated by recent GWAs. Two data sets of T2D-implicated genes were used as the benchmark: a Highly Significant gene set (HS) of 21 genes and a Moderate to Highly significant gene set derived from the WTCCC and DIAGRAM studies (MHWD) of 172 genes \[[@B8],[@B9]\]. The HS gene set contained 11 genes which mapped to the chromosomal regions investigated by Tiffin et al.(hereafter Tiffin intervals) \[[@B7]\]. Genes associated with 706 moderately significant SNPs with a frequentist additive p-value of \<0.001, good clustering and intact NCBI build 36 reference ids were taken from WTCCC T2D data \[[@B8]\]. SNPs positioned between the 5\' UTR and 3\' UTR of a known gene structure, 1000 bases upstream of a 5\' UTR or 1000 bases downstream of 3\' UTR of a known gene were considered to implicate the gene in T2D disease susceptibility. This moderately significant list was combined with the genes from the HS data set to generate the MHWD dataset, yielding 172 genes genome wide of which 61 genes mapped to the Tiffin intervals \[[@B7]\]. Data sources for predictions ---------------------------- The search space available to all eight automated candidate gene prediction systems consisted of 9556 genes in 53 chromosomal loci assessed by Tiffin et al. to be involved in T2D by various linkage and association studies. We matched 96.5% of all Ensembl gene entries \[[@B14]\] provided to NCBI Entrez ids. All remaining genes were unable to be matched due either to the Ensembl entry having an unknown gene symbol label or because the entry was ambiguous or associated with a redundant gene symbol name entry. Ensembl entries and NCBI id matching was carried out at four levels, in order: approved symbol name, previous symbol names, Uniprot/SwissProt Accession and RefSeq Ids. Data conversion keys for matching between databases were acquired from BioMart \[[@B15]\]. Predictions made by seven candidate gene prediction methods were also obtained from Tiffin et al. \[[@B7]\]. Nine disease-implicated genes were available to the systems as seeds (PPARG, GYS1, IRS1, INS, KCNJ11, ABCC8, SLC2A1, PPARGC1, CAPN10). Two of the genes -- PPARG and KCNJ11, are implicated by the highly significant SNPs detected by GWAs but are not in the Tiffin intervals and are thus not included in the search space or benchmark set. Candidate gene predictions -------------------------- The candidate gene predictions for seven of the systems are detailed elsewhere \[[@B7]\]. Briefly, GeneSeekerselected genes from the search space using a Boolean expression based on 14 keywords selected by an expert user \[[@B7]\]. PROSPECTR, DGPand eVOCare Boolean classifiers which require only the search space as input. G2Dand Gentrepidin ab initiomode, also only require the search space. POCUSpotentially only needs the search space as input, but this was restricted to the seven best supported intervals of the 53 available, as judged by the POCUSteam. SUSPECTSand Gentrepid, in known-disease-gene mode, used the nine known disease genes associated with the phenotype as seeds. SUSPECTSadditionally draws on predictions from the PROSPECTRBoolean classifier. Gentrepidpredictions are discussed in detail here for the first time. Gentrepidimplements two different modules to derive predictions: CPS -- a systems biology method; and CMP -- a method that associates phenotypes with particular domains. CPS and CMP can be used in two input modes: using known disease genes as a seed or using only the search space (ab initiomode). In known-disease-gene input mode, CPS searches all pathway and interaction data in BioCarta \[[@B16]\], KEGG \[[@B17]\] and I2D (formerly OPHID) \[[@B10]\] to extract all genes associated with the disease gene, and then filters this list against implicated loci. Genes are ranked based on the total number of genes implicated in the pathway. For example, if two known disease gene seeds and three genes in the loci being investigated are found in the same pathway, the pathway is given a rank of five against the phenotype. CMP parses the protein sequences of the known disease genes associated with the phenotype into domains using the Pfam library of Hidden Markov Models (HMMs) \[[@B18]\] and then retrieves any other genes with related domain content from the genome. A score between 0 and 1 is generated reflecting the candidate gene\'s similarity to a known disease gene \[[@B4]\]. The same nine disease genes and 53 chromosomal cytogenetic bands were used by Gentrepidas per Tiffin et al.. In ab initiomode, Gentrepidcan make predictions in the absence of known disease genes if two or more loci are provided as input. Gentrepid\'s CPS ab initiomethod is based on the premise that pathways whose genes are more prevalent within disease-implicated loci (chromosomal regions) compared to the entire genome have a higher probability of involvement in the pathoetiology of the disease phenotype of interest. Analogous to the known disease gene mode, pathways are ranked by the number of loci involved. The CMP ab initiomethod searches for enrichment of domains in the loci with respect to the genome and ranks genes based on the statistical significance of the domain enrichment (equations 2 and 4 in \[[@B4]\] where mnis replaced by Σ -- the total number of genes in the intervals examined). For each input mode, a final list of predictions is made by consolidating all predictions from both the CMP and CPS modules. Metrics for comparisons ----------------------- Systems were compared using three metrics: Enrichment Ratio, Sensitivity and Specificity. The Enrichment Ratio calculations were calculated as below \[[@B4]\]: $$EnrichmentRatio = \frac{TP/(TP + FP)}{{\sum{genes_{implicated}}}/{\sum{genes_{all}}}}$$ The denominator was obtained by dividing the number of T2D implicated genes by the total number of genes within all surveyed chromosomal regions. Sensitivity and Specificity were calculated as below: $$Sensitivity = \frac{TP}{(TP + FN)}$$ $$Specificity = \frac{TN}{(TN + FP)}$$ TP is the number of true positives, FP is the number of false positives, TN is the number of true negatives and FN is the number of false negatives. Sensitivity is the proportion of true positives among all disease genes in the chromosomal regions. Specificity is the proportion of true negatives among genes not associated with the disease in chromosomal regions. Confidence intervals were estimated using the method of Newcombe \[[@B19]\] implemented using the CIcalculator software \[[@B20]\]. List of abbreviations used ========================== T2D: Type II diabetes; WTCCC: Welcome Trust Case Control Consortium; DIAGRAM: Genetics Replication and Meta-analysis Consortium; SNP: Single nucleotide polymorphism; GWA: Genome wide association studies; HPRD: Human Protein Reference Database; BIND: Biomolecular Interaction Network Database; DGP: Disease Gene Prediction; PROSPECTR: PRiOrization by Sequence and Phylogenetic Extent of CandidaTe Regions; OMIM: Online Mendelian Inheritance in Man; OPHID: Online Predicted Human Interaction Database; KEGG: Kyoto Encyclopedia of Genes and Genomes; GO: Gene Ontology; MeSH: Medical Subject Headings; HS: Highly Significant gene set; MHWD: Moderate to Highly gene set derived from WTCCC and DIAGRAM studies; ER: Enrichment Ratio; S: Sensitivity; TP: true positives; FP: false positives; TN: true negatives; FN: false negatives. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= MW design, concept and oversight of study, and manuscript author. ET manuscript author, and data generation for study. JL implementation, construction and maintenance of database. SB additional data generation, figures and manuscript preparation. DF Genetics consultant. All authors read and approved the final manuscript. Supplementary Material ====================== ###### Additional file 1 Additional unprioritized MHWD matches. Additional unprioritized MHWD matches data. ###### Click here for file Acknowledgements ================ The authors wish to acknowledge funding from the Ronald Geoffrey Arnott Foundation. This article has been published as part of BMC BioinformaticsVolume 10 Supplement 1, 2009: Proceedings of The Seventh Asia Pacific Bioinformatics Conference (APBC) 2009. The full contents of the supplement are available online at <http://www.biomedcentral.com/1471-2105/10?issue=S1>	{ "pile_set_name": "PubMed Central" }
Biological context {#Sec1} ================== Cyclic nucleotides (cNMPs) are important secondary messenger molecules that mediate a multitude of processes by activating several different proteins in the signalling cascade. These proteins share a conserved regulatory protein unit referred to as cyclic nucleotide-binding domain (CNBD). Most of the CNBD harboring proteins are allosterically modulated. Upon binding cNMPs the regulatory CNBD renders the protein in an active conformation. One such family of proteins which is regulated by cNMPs is ion channels (Kaupp and Seifert [@CR8]). On binding cNMPs, the channels are activated resulting in an increase in membrane conductance. Although a considerable amount of information is available on the function of these channel proteins, the molecular events that relay ligand binding to channel gating are not well understood. A major limitation in our understanding of the channel function arises from the fact that ligand binding and channel activation are two separate events that are reciprocally coupled, i.e. ligand binding affects channel gating, and conversely, channel gating affects ligand binding. Until recently, it was believed that cNMP-gated channels are allosterically modulated and more than one ligand molecule is required to gate open the channel. However, the bacterial mlCNG channel displays distinctively different channel functioning property in being non-cooperative (Cukkemane et al. [@CR4]; Nimigean et al. [@CR10]). This provides a facile system in understanding the binding--gating relationship. In order to comprehend the molecular mechanisms of activation, a structural understanding of the channel and the CNBD is a prerequisite. The crystal structure (Clayton et al. [@CR3]) and the solution NMR (Schünke et al. [@CR12]) data of the CNBD from mlCNG are available. In line with other cAMP binding proteins (Berman et al. [@CR2]), the cNMP binding site of the CNBD comprises three α-helices and eight β-strands. The β-strands form a cavity resembling a flattened β-barrel like structure which is popularly referred to as β-jelly roll-like structure. The cNMP is bound within this cavity through a network of polar and hydrophobic interactions with a short 3~10~ helix and loop between strands β6 and β7. This region is commonly referred to as the phosphate binding cassette and is the most conserved region of the protein. Solid-state NMR (ssNMR) offers means to monitor structural or dynamical aspects of soluble proteins under a variety of preparatory conditions (see, e.g., Seidel et al. [@CR18]) and can be used to study membrane proteins (MPs) in a functional membrane setting. Previously, we showed that spectroscopic studies in MPs encompassing several hundred amino acids are facilitated by reference experiments using smaller protein constructs (Etzkorn et al. [@CR15], [@CR16]). Importantly, resonance assignments in the solid state may differ from values obtained in solution due to differential protein mobility or intermolecular interactions (Seidel et al. [@CR19]). Therefore we investigated the CNBD of mlCNG channel using solid-phase preparations paving the way for structural studies on the full-length channel. Methods and experiments {#Sec2} ======================= The CNBD (residues 214--355) was expressed as a GST-fusion construct in pGEX-2T vector here (Clayton et al. [@CR3]; Cukkemane et al. [@CR4]) in E.coli BL-21 (DE3) pLysE using standard ^15^N/^13^C labeled media. Overnight expression (20°C) of the transformed E.coli cells was induced at OD~600~ \~ 0.6 using 0.6 mM IPTG. The harvested cells were lysed in PBS buffer by sonication. The supernatant containing the GST-CNBD was separated from the cell pellet by centrifugation. The pellet was discarded and the supernatant was loaded on to a glutathione affinity column (Glutathione Sepharose-4B). After washing with 10 ml (per 1 ml column) of washing buffer, the CNBD part of the fusion protein was cleaved off by thrombin. 40 units of thrombin were added to 1 ml of Glutathione Sepharose-4B column to release \~10 mg of CNBD protein. Incubation was done overnight at 27°C. The CNBD was eluted by a washing step whereas the GST part remained bound to the column. The purified protein was dialysed against 10 mM Tris, pH 8.0, 250 μM cAMP and 0.01% Na-Azide at room temperature. The dialysate was flash-frozen and lyophilized in a 15 ml falcon tube (Greiner Bio-one). The lyophilized sample was rehydrated by placing the protein powder in a water bath for 3 days in an assembly that included a 250 ml conical flask with \~20 ml of water. The lyophilized sample was incubated in the flask, covered using parafilm at room temperature for 3 days during which the sample hydrated with a gel consistency. Around 10 mg of the rehydrated sample was packed in a volume of \~35 μl in a 3.2 mm zirconium rotor by centrifugation. All ssNMR experiments were conducted using a 3.2 mm triple-resonance (^1^H,^13^C,^15^N) probe head at static magnetic field of 16.4T corresponding to 700 MHz proton resonance frequency (Bruker Biospin). Assignments were obtained using a combination of 2D and 3D correlation experiments. These experiments included 2D \[^13^C,^13^C\] correlations obtained using proton-driven spin diffusion under weak coupling conditions (Seidel et al. [@CR13]) with spin diffusion times of 20 and 150 ms to encode intra-residue and sequential correlations, respectively. Furthermore, 2D \[^13^C,^13^C\] DQ-SQ and 3D \[^13^C,^13^C,^13^C\] DQ-SQ-SQ (Heise et al. [@CR7]) were performed. SPC5 (Hohwy et al. [@CR17]) was used for the generation of double quantum coherence with a contact time of 800 μs. 2D \[^15^N,^13^Cα/C′\] (Baldus et al. [@CR1]) spectra were recorded using SPECIFIC-CP time of 3 ms. 2D \[^15^N,^13^Cα--Cx\] and 2D \[^15^N,^13^C′--Cx\] were recorded with ^13^C--^13^C mixing times of 40 and 50 ms, respectively. All the above mentioned experiments were performed at 10.127 kHz MAS at −10°C with SPINAL 64 (Fung et al. [@CR6]) decoupling on ^1^H while acquiring the FID. The 3D \[^15^N,^13^C′--Cx\] experiment was performed at 17 kHz with 50 ms ^13^C--^13^C MIRROR recoupling (Scholz et al. [@CR11]) and XiX (Detken et al. [@CR5]) decoupling on ^1^H during acquisition. Chemical shifts were referenced to the upfield peaks of 31.48 and 38.9 ppm for ^13^C and ^15^N using Adamantane and tri-peptide AGG, respectively. Assignments and data deposition {#Sec3} =============================== Backbone and side-chain chemical shifts of the CNBD are presented in Table S1 and have been deposited in the BioMagResBank (<http://www.bmrb.wisc.edu>) under the accession number 18024. Figure [1](#Fig1){ref-type="fig"}A, B display results of a 2D \[^13^C--^13^C\] spin diffusion and a 3D NCOCA correlation experiment obtained on uniformly \[^13^C,^15^N\] labeled CNBD, respectively. Intra-residue and sequential correlations are indicated. Overall, the line-width observed in both ^13^C and ^15^N dimensions compared favorably to results obtained previously in microcrystalline proteins (Seidel et al. [@CR18]). Most signals assigned in the 2D spectrum reflect amino acids that reside in the core region rich in β strands (see Figure 1A/S1). Signals for residues in α-helices were resolved when using 3D ssNMR experiments. As an example, the strip plot shown in Fig. [1](#Fig1){ref-type="fig"}b contains correlations of residues located at the C-terminal helix of the protein. Resonances of five residues in the extreme N- and C-terminal parts of the CNBD were not identified. The Cα chemical shifts of residues of V224, Q228, L229, V23, A242 and E336 were not identified. Apart from those, the N chemical shifts of most of the Prolines and residues R225, Q237, V243, E267 and I337 were not assigned.Fig. 1Characterization of the CNBD by ssNMR using (A) 2D \[^13^C--^13^C\] and (B) 3D \[^15^N,^13^C′--Cx\] correlation spectra. Intra-residue and sequential correlations are indicated in a and b respectively. Most signals assigned in the 2D \[^13^C--^13^C\] spectrum reflect amino acids that reside in the core region rich in β-strands (see Fig. [2](#Fig2){ref-type="fig"}A, B). The signals for residues in α-helices were resolved in a 3D \[^15^N,^13^C′--Cx\] spectrum. The strip plot shown in B contains correlations of residues located at the C-terminal helix of the protein We obtained an assessment of the secondary structure and the overall fold of the CNBD using chemical shift differences (Fig. [2](#Fig2){ref-type="fig"}a). The backbone chemical shift of the CNBD is comparable to the results of the protein in solution (Figure S2 and S3D). Deviations between ssNMR and solution-state NMR assignments were found throughout the protein sequence (Figure S2) with largest differences generally detected towards the N and C terminus. Nevertheless, a secondary chemical shift analysis (Fig. [2](#Fig2){ref-type="fig"}B) revealed a backbone fold similar to that of previous crystal (Clayton et al. [@CR3]) and solution-NMR (Schünke et al. [@CR12]) studies of the cAMP-bound structure (Figures S3). Further analysis on the secondary structure was performed by predicting the dihedral angles ϕ and φ (Fig. [2](#Fig2){ref-type="fig"}B, C) with the program TALOS+ (Shen et al. [@CR14]). Comparison of the TALOS+ predicted torsion angles between the ssNMR results and solution NMR (Figure S3A and S3B) show a profile indicating similar protein secondary structure. Overall, the chemical shifts of the CNBD determined by ssNMR results indicate a well folded protein. We obtained assignments for 86% of backbone residues with missing assignments predominantly found at both protein termini (Cα shifts of V224, Q228, L229, V233, A242 and E336). These parameters are, together with amino-acid side chain assignments, given in table S1.Fig. 2The pictorial representation of the secondary structure elements of the CNBD is shown in the toppanel of A. Experimentally observed chemical shift parameter (δCα--δCβ) (Luca et al. [@CR9]) in A of the CNBD compared against the derived backbone torsion angles using the program TALOS+ (Shen et al. [@CR14]) in B and C Interestingly, we observed peak doublings in several regions of the protein (shown in Figure S3D), including loop residues connecting β1--β2, β2--β3, β7--β8 as well as the segment β6-PBC, and the N-terminal end of αC. Such spectroscopic polymorphism suggests the presence of multiple conformations in our ssNMR preparations even in the absence of the lipid-embedded transmembrane domain. Hence, our results on the 141 amino-acid CNBD will provide a valuable reference for correlating structural and dynamical aspects of the full-length channel to ligand binding in mlCNG in a membrane setting, and also aid in developing ssNMR methodology. Electronic supplementary material {#AppESM1} ================================= {#SecESM1} Supplementary material 1 (DOC 728 kb) This work was funded by NWO (grant 700.26.121) and the Seventh Framework Programme \[FP7/2007-2013\] under grant agreement n° \[211800\]. MW is supported by a FEBS long-term postdoctoral fellowship. We thank Dr. H. Wienk for helpful discussions. Open Access {#d29e556} =========== This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#s1} ============ Inflammatory breast cancer (IBC) is an uncommon (5% of all cases) but highly aggressive form of breast cancer \[[@R1]\]. IBC is clinically characterized by a rapidly growing edema and erythema of the breast involving at least 1/3 of the skin breast, and may include at the pathological level tumor emboli in the dermal lymphatics \[[@R2]\]. Even though significant improvements were achieved with multimodal treatment, the 5-year survival rate is only ∼ 40%, and the prognosis of IBC remains worse than that of non-inflammatory locally advanced breast cancer \[[@R1]\]. Biologically, IBC remains poorly understood but includes a high vascularity and an increased micro-vessel density consistent with a high expression of angiogenic factors, including the vascular endothelial growth factor (VEGFA) \[[@R2]\]. This led to the evaluation of anti-angiogenic therapies in this disease. In the recent open-label, single-arm, multicenter phase II BEVERLY-2 study, neoadjuvant and adjuvant bevacizumab was added to trastuzumab-based neoadjuvant chemotherapy and adjuvant trastuzumab in patients with nonmetastatic HER2-positive IBC leading to a high pathological complete response (pCR) rate of more than 60% \[[@R3]\] and promising 3-year disease-free survival (DFS) and overall survival (OS) rates of 68 and 90%, respectively. However, in the absence of a randomized design, the specific impact of bevacizumab was not established. In addition, the use of bevacizumab in breast cancer was recently challenged. Thus, biomarkers able to predict bevacizumab benefit in breast cancer, including HER2-positive IBC, would be of a crucial interest. Ideal biomarkers should be easily measurable, on multiple points upon treatment, and standardized \[[@R4]\]. Baseline levels and/or variation of numerous intratumor or circulating candidate prebiomarkers have been explored \[[@R5]\]. In-situ prebiomarkers such as VEGFA, VEGF-R2 or CA9, as well as circulating prebiomarkers such as VEGFA, VEGFR1, ICAM1, IL6, IL8 or circulating tumor cells count \[[@R6]\] were reported to predict bevacizumab benefit, but this predictive value was generally weak and rarely confirmed \[[@R5]\]. Matrix metalloproteinases 2 (MMP2) and 9 (MMP9) belong to the MMP family, whose activity is implicated in proteolysis of extra-cellular matrices, regulation of cell adhesion and migration, processing of growth factors and cytokines, and liberation of angiogenic factors \[[@R7]\]. We recently reported the association between the baseline circulating MMP2 level and, to a lesser extent, the MMP9 level, and the response rate, PFS and OS of patients treated with bevacizumab for recurrent high-grade glioma \[[@R8]\]. A baseline of high and low plasma levels of MMP2 and MMP9 respectively were associated with a high response rate and a prolonged progression-free survival and OS in recurrent high-grade gliomas treated with bevacizumab. Moreover, no association was found with patient survival in a similar population treated with cytotoxic agents only without anti-angiogenic therapy, suggesting a specific predictive value of these biomarkers. BEVERLY-2 study included a prospective collection of serum samples before initiation of neoadjuvant chemotherapy and before surgical resection. In this study, our objective was to evaluate the prognostic impact of MMP2 and MMP9 serum levels, at baseline and during treatment, in patients with IBC treated with neoadjuvant bevacizumab, trastuzumab and chemotherapy. RESULTS {#s2} ======= Patient characteristics and baseline (b) MMP2 and MMP9 serum levels {#s2_1} ------------------------------------------------------------------- Serums were available for 45 of 52 patients included in the trial. Characteristics of this patients\' subset were similar to those of the entire population (Table [1](#T1){ref-type="table"}). In our population, pCR was observed in 73.3% (95% Confidence Interval \[CI\]: 59.0-84.0) of patients. After a median follow-up of 3 years, 5 patients died and 14 presented disease recurrence. Median DFS and OS were not reached. The 3-year DFS and OS rates were 67.8% (95% CI: 55.2-83.3) and 88.1% (95%CI: 78.8-98.5), respectively. ###### Patient characteristics of the 45 patients with available serum samples Factors N = 45 \% ------------------------------- ------------------ ----------- Age, median (range) 51,7 (37,0-67,9) Estrogen receptor-positive 16 35,6 Progesteron receptor-positive 7 15,6 Hormone receptors-positive 16 35,6 SBR grade 2 18 41,9 3 25 58,1 Pathological response pCR 33 73,3 No pCR 12 26,7 Median follow-up (95%CI) 35.6 months (35-36.2) 3-years DFS 73% 3-years OS 91% CR: complete response; DFS: disease-free survival; OS: overall survival; 95%CI: confidence interval of 95% At baseline, median bMMP2 and bMMP9 serum levels were 203.6 ng/ml (range, 116.2-338.9) and 629.6 ng/ml (range, 191-1189), respectively, and were inversely correlated (R= −0.498, p=0.001). bMMP2 and bMMP9 levels and correlation with histo-clinical features {#s2_2} ------------------------------------------------------------------- We searched for correlations between the bMMP2 and bMMP9 serum levels and histo-clinical features. As shown in [Supplementary Table 1](#SD1){ref-type="supplementary-material"}, no significant correlation existed with patient\'s age, HR status and SBR grade. bMMP2 and bMMP9 levels were not associated with the count of baseline CTC and CEC, analyzed as continuous or qualitative variables. bMMP2 and bMMP9 serum levels and correlation with clinical outcome {#s2_3} ------------------------------------------------------------------ Neither bMMP2 nor bMMP9 in continuous values were correlated to pCR (p=0.277 and p=0.104, respectively). So, for each MMP, subjects were divided into two groups (high and low bMMP) using the median value as cutoff. As binary variables, bMMP9 (p=0.208) and bMMP2 (p=0.146) were not correlated to pCR. By contrast, correlations existed with survival. In univariate analyses (Table [2](#T2){ref-type="table"}), high bMMP2 was associated with longer DFS (p=0.001) and OS (p=0.032), whereas high bMMP9 was associated with worse OS (p=0.022) and tended to be associated with worse DFS (p=0.071) (Figure [1](#F1){ref-type="fig"}). Because of the low number of death events, multivariate analysis was only performed for DFS and included SBR grade (p=0.057). Both bMMP2 (p=0.003, HR: 0.115) and bMMP9 (p=0.041, HR: 3.511) remained significantly and oppositely correlated to DFS. Thus, 3-year DFS rates were 52.4% versus 91% for patients with low versus high bMMP2 serum level respectively, while DFS rates were 86.4% versus 56.5% for patients with low and high bMMP9. ###### Univariate and multivariate (DFS only) analyses of MMP2 and MMP9 serum level at baseline and other potential prognostic factors ------------------------------------------------------------------------------------------------------------------ DFS OS ----------------------------------------------------------- ------------- -------------------------- ------------- MMP2 (cutoff: median) *0,001* *0,006\ 0,032* HR: 0.115 (0.025-0.533) MMP9 (cutoff: median) *0,071* *0,042\ 0,022* HR: 3,511 (1,044-11,806) HR-positive 0,244 0,519 ER-positive 0,244 0,519 PR-positive 0,404 0,628 pCR 0,005 0,041 SBR grade (2 versus 3)[\](#tfn_001){ref-type="table-fn"} 0,057* 0,418 Age (cutoff: median) 0,685 0,726 ------------------------------------------------------------------------------------------------------------------ adjusted by SBR grade Multivariate analysis not allowed by the number of events. ![Disease-free survival (DFS) and overall survival (OS) according to baseline MMP2 A, C. and MMP9 B, D. serum levels.](oncotarget-07-18531-g001){#F1} As continuous variables, bMMP2 (p=0.021), but not bMMP9 (p=0.250) levels were correlated to DFS, whereas bMMP2 (p=0.098) and bMMP9 (p=0.068) levels presented a borderline significant trend for correlation to OS. Moreover, bMMP2 (p=0.049, AUC: 0.801) and bMMP9 (p=0.035, AUC: 0.793) levels were significantly correlated to the risk of death, while bMMP2 serum levels were correlated to the risk of recurrence (p=0.002, AUC: 0.805) (Figure [2](#F2){ref-type="fig"}). ![ROC analysis of the correlation between baseline MMP2 A, C. and MMP9 B, D. serum levels and the risks of death (A, B) and of recurrence (C, D).](oncotarget-07-18531-g002){#F2} Exploratory scoring systems and ratio analysis {#s2_4} ---------------------------------------------- Given the opposite prognostic value of bMMP2 and bMMP9 levels, we divided our patient population in four classes: high bMMP2 and low bMMP9 (N=16), high bMMP2 and high bMMP9 (N=6), low bMMP2 and low bMMP9 (N=6), and low bMMP2 and high bMMP9 (N=15) serum levels. These four classes were significantly associated with DFS (p=0.015) and OS (p=0.033). All death events were observed in the worse subgroup defined by low bMMP2 and high bMMP9 serum levels (Figure [3](#F3){ref-type="fig"}). At the 3-year follow-up, DFS rates were 94% in the "high bMMP2/low bMMP9" class, 83% in the "high bMMP2/high bMMP9" class, 50% in the "low bMMP2/low bMMP9" class, and 43% in the "low bMMP2/high bMMP9" class, while OS rates were 100%, 100%, 100%, and 71%, respectively. ![Disease-free survival and overall survival according to the four classes based on baseline MMP2 and MMP 9 serum levels\ Class 1: high MMP2 and low MMP9 (N=16), class 2: high MMP2 and high MMP9 (N=6), class 3: low MMP2 and low MMP9 (N=6) and class 4: low MMP2 and high MMP9 (N=15).](oncotarget-07-18531-g003){#F3} Finally, we evaluate the prognostic impact of the MMP9/MMP2 ratio, which was associated both with OS (p=0.044) and DFS (p=0.053). MMP2 and MMP9 serum levels changes during treatment and outcome {#s2_5} --------------------------------------------------------------- Post-neoadjuvant chemotherapy pre-surgery (ps) samples were available for 31 patients: median psMMP2 and psMMP9 serum levels were 298.5 ng/ml (range, 195.7-552.9) and 319.9 ng/ml (range, 20.1-858.4), respectively. Under treatment, all patients experienced an increase in psMMP2 serum level (p\<0.001), while 87% of patients had a decrease in psMMP9 serum level (p\<0.0.01) (Figure [4](#F4){ref-type="fig"}). psMMP2 and psMMP9 serum levels tended to be inversely correlated (R=−0.308, p=0.092). By contrast, the magnitudes of changes of MMP2 and MMP9 serum levels were not correlated (p=0.991) and were not associated with DFS or OS. ![Changes of MMP2 and MMP9 serum levels under treatment, (mean, standard error of mean)](oncotarget-07-18531-g004){#F4} Tumor MMP2 and MMP9 mRNA expressions in IBC {#s2_6} ------------------------------------------- Since only a limited number of frozen tumor samples were available from BEVERLY-2 patients, we analyzed MMP2 and MMP9 mRNA expression in 137 clinical IBC samples profiled within the World IBC Consortium (E-MTAB-1006, E-MTAB-1547 and GSE22597). None of these patients received bevacizumab as part of their treatment. Their clinical characteristics are shown in [Supplementary Table 2](#SD1){ref-type="supplementary-material"}. Expression levels were heterogeneous across samples with a range of intensities over 3 decades in log10 scale ([Figure S2](#SD1){ref-type="supplementary-material"}). We searched for correlations (Fisher\'s exact test) between MMPs expression status ("high" versus "low" groups) and histo-clinical variables in IBC samples ([Table S3](#SD1){ref-type="supplementary-material"}). Regarding MMP2 expression, no correlation was found with patients\' age, histological type and grade, ER and PR, whereas high expression was associated with HER2-positive status (p=0.039). Regarding MMP9 expression, no correlation was found with age, histological type and HER2 status, whereas high expression was associated with higher grade (p=0.020), ER-negative status (p=0.002) and PR-negative status (p=0.0007). Neither MMP2 nor MMP9 RNA expression levels were correlated to pathological response and to metastasis-free survival ([Table S4](#SD2){ref-type="supplementary-material"}). DISCUSSION {#s3} ========== This analysis showed for the first time to our knowledge that, in the HER2-positive IBC patient population of the BEVERLY-2 trial, baseline MMP2 and MMP9 serum levels are associated with DFS and OS, are anti-correlated, and significantly changed during bevacizumab- and trastuzumab-based neo-adjuvant chemotherapy. These results are in accordance with the predictive value of these biomarkers for bevacizumab benefit, we recently reported in patients with recurrent high grade glioma \[[@R8]\]. Such predictive value was not found in glioma patients untreated with bevacizumab. Whether the same is true in HER2-positive IBC patients would ideally require a control group in a randomized design. To our knowledge, serum MMP2 and MMP9 dosages have not been reported in IBC, limiting access to some historical control. Despite that point, the similar results observed in glioma and HER2-positive IBC reinforce the interest of these biomarkers as predictive factors candidate for prediction of bevacizumab activity. Given the increasing use and cost of antiangiogenic agents, and the heterogeneity of their benefit, particularly for bevacizumab, biomarkers of efficacy that could drive their therapeutic use, remain an unmet need in oncology \[[@R4]\]. In breast cancer, the use of bevacizumab was associated with a weak and debated activity, leading to the withdrawal of the FDA authorization and highlighting the especial need for predictive biomarkers. In the particular context of HER2-positive IBC, this is a critically important issue. Indeed, the BEVERLY-2 study demonstrated a high anti-tumor activity in terms of pCR and survival outcomes, comparing very favorably with historical controls \[[@R6]\]. However, other attractive strategies in this patient population may include optimization of anti-HER2 targeting using double blockade combination such as trastuzumab-pertuzumab \[[@R13]\] or trastuzumab-lapatinib associations \[[@R14]\]. Thus, predictive biomarkers that could help selecting among these two innovative approaches might be particularly useful. Numerous studies explored potential predictive biomarkers for bevacizumab efficiency in breast cancer, but to date, none has been validated \[[@R5]\]. Hypertension and polymorphisms affecting the VEGFA pathway have shown some predictive value of bevacizumab benefit, although their limitations included lack of standardization, and inconsistent effect among studies. Tissue expressions of VEGFA, CD31, and PDGFRβ were associated with response, but their predictive value was not confirmed \[[@R15]\]. Moreover, baseline circulating biomarkers such as VEGFA, soluble VEGFR2, VCAM-1, and E-Selectin as well as circulating endothelial cells have been associated with outcome \[[@R16], [@R17]\]. However, their predictive value was inconsistent between studies. Recently, we reported that CTC count at baseline had an independent prognostic impact restricted to DFS in the BEVERLY-2 trial \[[@R6]\]. Interestingly, we did not find any correlation between baseline CTC count and MMP2 or MMP9 serum levels. We observed MMP serum levels variations during treatment: MMP2 increased for all patients and MMP9 decreased for most of them. Interestingly, we found similar results in our cohort of bevacizumab-treated glioma patients, with an increase of MMP2 and a decrease of MMP9 plasma levels during treatment before progression. In addition, in glioma patients, MMP2 plasma level significantly decreased at the time of progression. Variations of circulating markers were already reported under anti-angiogenic therapies, but their predictive values were limited \[[@R18]\]. Thus, MMP2 and MMP9 plasma levels could allow the initial patients\' selection, but also could help in the treatment monitoring. MMP2 and MMP9 belong to the matrix metalloproteinase (MMP) family, whose activity has been implicated in proteolysis of extra-cellular matrix. \[[@R7]\]. In breast cancer, MMP9 was reported to be highly expressed, in contrast to MMP2. Moreover, MMP2 seems to have no prognostic value for breast cancer patients while the prognostic impact of tissue expression or circulating level of MMP9 remains controversial \[[@R19], [@R20]\]. In this study, we have examined, for the first time in the literature, MMP2 and MMP9 mRNA tumor expression in a large population of IBC patients from the world IBC consortium but no relationship was found with pCR or survival. Of note, none of these patients had received bevacizumab in their initial neoadjuvant treatment. Thus, even though no data was available on corresponding serum levels, the lack of prognostic value observed for MMP2 and MMP9 tumor expression in this retrospective cohort may argue in favor of a predictive and specific impact of these markers for bevacizumab benefit. One intriguing result of our study was the opposite correlation and prognostic impact of MMP2 and MMP9. MMP2 has been involved in angiogenesis, associated to pericyte recruitment, vascular maturation and functionality via multiple effectors \[[@R21], [@R22]\]. Interestingly, in orthotopic mouse model of glioblastoma completely devoid of MMP2 activity, tumor vascularity appeared to be less functional with reduction of VEGFR2 expression in tumor vessels and decrease of pericyte coverage \[[@R21]\]. In another model of lung cancer, MMP2 was reported to be implicated in the VEGFA expression by the tumor cells, through the PI3K/Akt pathway, underlining the major role played by MMP2 in the angiogenic process \[[@R22]\]. By contrast, MMP9 seems implicated in another vascularization process, called vasculogenesis, based on the recruitment of circulating endothelial and myeloid precursors \[[@R23]\]. Thus, these distinct roles in the tumor vascularization, supported by the inverse correlation of their serum concentrations, could explain their opposite predictive effect for bevacizumab activity. Our study presented some limitations. The sample size of our population was small but related to the rarity of this disease. This limited number of patients and events could explain the borderline significance of our continuous analysis results. In this context, the exploratory results could be carefully interpreted taking into account these limitations. However, the anti-correlation of the survival impact of these two markers evaluated in serum, their changes under bevacizumab as well as their similar impact observed in glioma treated with bevacizumab reinforce their potential value. MATERIALS AND METHODS {#s4} ===================== Study design and patient population {#s4_1} ----------------------------------- BEVERLY-2 was a single-arm, multicenter, non-randomized phase II study. Study design, inclusion criteria and patient characteristics, as well as procedures, treatment ([Figure S1, supplementary methods](#SD1){ref-type="supplementary-material"}), efficacy and safety results have been previously published \[[@R3], [@R6]\]. All patients had a centrally reviewed HER-2 positive non metastatic IBC, defined as T4d (any N), or PEV2 or 3 according to the PEV (Poussée EVolutive) classification. Written informed consent was provided by patients before any screening procedure and was required for the translational research studies. The study was approved by the ethical board (Comité de Protection des Personnes Sud Méditerranée I) and registered (NCT00717405 and EUDRACT 2008-000783-16). Procedures and treatment {#s4_2} ------------------------ Patients received neoadjuvant chemotherapy consisting of 4 three-weekly cycles of FEC100 (500mg/m^2^ fluorouracil, 100 mg/m^2^ epirubicin, 500 mg/m^2^ cyclophosphamide) plus bevacizumab (15 mg/kg), followed by four 3-weekly cycles of docetaxel (100 mg/m^2^), bevacizumab (15 mg/kg), and trastuzumab (initially at a loading-dose of 8 mg/kg, and then a dose of 6 mg/kg). Surgical treatment consisted of mastectomy and axillary lymph node dissection. Bevacizumab was stopped 4 weeks prior to surgery and resumed (for a further 30 weeks) once the wound was healed entirely, during or after radiotherapy. Patients continued receiving trastuzumab (6 mg/kg) during the perioperative period which continued for another 30 weeks following surgery (42 weeks in total). Adjuvant radiotherapy was administered according to standard practice in combination with trastuzumab, and bevacizumab. Patients with hormone receptor (HR)-positive tumors received adjuvant hormone therapy. Blood sample collection and serum analysis {#s4_3} ------------------------------------------ Additional blood samples (2 × 7.5 mL) were taken from patients at different times ([supplementary methods](#SD1){ref-type="supplementary-material"}): (1) before the first cycle of neoadjuvant chemotherapy and bevacizumab (baseline); (2) before the first trastuzumab administration during the neoadjuvant period (cycle 5); (3) before the surgical procedure (cycle 8); (4) during the adjuvant period before the reintroduction of bevacizumab (post-operative assessment); and (5) at the final visit at the end of adjuvant treatment (one-year follow-up). After collection of the whole blood, samples were allowed to clot at room temperature during 15-30 minutes. Then, serum was obtained by centrifuging at 2,000 x g for 10 minutes in a refrigerated centrifuge. Serum was immediately distributed into 0.5 ml aliquots, stored, and transported at −80°C or lower. Serums were thawed immediately before analysis. In the present study, serums at baseline and before surgery were analyzed for levels of MMP2 and MMP9 using commercially available enzyme-linked immunosorbent (ELISA) assay kits (R&D Systems^®^). Samples were run in duplicate, and the average was recorded. Serum levels were analyzed as continuous values and as binary values using the median value as cutoff. Circulating tumor cells (CTC) and circulating endothelial cells (CEC) counts were measured in blood samples as previously published \[[@R6]\] using CellSearch. mRNA expression of MMP2 and MMP9 in IBC tumor samples {#s4_4} ----------------------------------------------------- MMP2 and MMP9 expressions have not been described in IBC tumor samples. Unfortunately, no frozen tumor sample from patients treated in the Beverly 2 trial was available for analysis. We thus analysed the gene expression dataset of the World IBC Consortium that included the whole-genome transcriptional profiles of 137 IBC profiled using Affymetrix DNA microarrays \[[@R9], [@R10]\]. The series and experiments have been described \[[@R9], [@R10]\], as well as the preprocessing of data. MMP2 and MMP9 mRNA expressions were measured by analyzing the 201069_at and 203936_at Affymetrix probe sets respectively, whose identity and specificity were verified using the NCBI program BLASTN 2.2.31+ and showed 100% accuracy. Expression was measured as discrete value by using the median expression level as cut-off: "high" expression was superior to the median, and "low" was inferior. Statistical analysis {#s4_5} -------------------- For evaluation of pCR, Sataloff criteria were used \[[@R3]\]. Central review of tissue blocks was required for each patient and pCR was defined as a total or near total treatment effect with loss of nodal involvement (Sataloff classification TA and NA or NB) \[[@R11]\]. DFS was defined as time from first administration of neoadjuvant treatment to local, regional or distant recurrence, contralateral breast cancer, second primary cancer (other than squamous or basal cell carcinoma of the skin, melanoma in situ, carcinoma in situ of the cervix, colon carcinoma in situ, or lobular carcinoma in situ of the breast) or death from any cause. OS was defined using death from any cause. We analyzed MMP2 and MMP9 serum levels data with descriptive analyses, in terms of the variation between time points and correlation with pCR and survivals. Categorical variables were presented as frequencies and percentages, continuous variables as median and range. Comparisons of categorical variables distribution were performed by χ2 or Fisher\'s exact test. For continuous variables, non-parametric Mann-Whitney U test was used. The Kaplan-Meier method was used to estimate survival distributions. Log-rank tests were used for univariate comparisons. Cox proportional hazards models were used for multivariate analyses and to estimate hazard ratios in survival regression models. Multivariate analysis included all variables with a p-value \< 0.10. Baseline and pre-surgical values of MMP2 and MMP9 serum levels were compared using with the Wilcoxon signed-rank test. Correlations were analyzed using the Spearman test. All reported p-values are two-sided, and p\<0.05 was considered statistically significant. All statistical analyses were performed by paswstatistics.version21^®^. We followed the reporting REcommendations for tumor MARKer prognostic studies (REMARK criteria) \[[@R12]\]. CONCLUSION {#s5} ========== This analysis performed in the BEVERLY-2 phase II trial identified MMP2 and MMP9 as candidate biomarkers correlated consistently to DFS and OS in patients with HER2-positive inflammatory breast cancer. The prediction of bevacizumab benefit versus a prognostic value of MMP2/MMP9 should be assessed ideally in controlled studies. Further studies are needed to reinvestigate the biological role of MMP2 and MMP9 and their interaction with the VEGFA pathway. SUPPLEMENTARY FIGURES AND TABLES {#s6} ================================ This work was supported by SIRIC program (INCa-DGOS-Inserm 6038). J.-P. Borg\'s lab is supported by La Ligue Nationale Contre le Cancer (Label Ligue 2013). The BEVERLY-2 trial was sponsored by Roche (France). This sponsor was not involved in the design of the present study neither in the data analysis, interpretation and in writing the report. CONFLICTS OF INTEREST J-YP, JG, J-MF, OC and AG have had consultant or advisory roles for and have received honoraria and research funding from Roche. TP has had consultant or advisory roles for Roche and Novartis and has received honoraria from Roche. CM has had consultant or advisory roles for Servier and Novartis and has received honoraria from Novartis. TB has had consultant or advisory roles for and has received research funding from Roche. All other authors declared that they have no conflicts of interest.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-jcm-07-00555} =============== One of the most exciting directions in medicine is based on stem cell research, which holds the promise of treatments and cures for various diseases and conditions that have so far frustrated traditional pharmaceutical approaches. Both stem cell research and clinical trials in regenerative medicine are presently dominated by pluripotent stem cells: embryonic stem cells (ESC) and induced pluripotent stem cells (iPSCs). However, these remarkable cells face important and unresolved hurdles, including the risk of generating teratomas, the truly arduous and inefficient processes of directed commitment to desired lineages and the limited regenerative capacity of derived lineages \[[@B1-jcm-07-00555],[@B2-jcm-07-00555],[@B3-jcm-07-00555],[@B4-jcm-07-00555],[@B5-jcm-07-00555],[@B6-jcm-07-00555]\]. The promise of pluripotent stem cells has largely overshadowed efforts to harness so-called "adult" or "somatic" stem cells intrinsic to regenerative tissues. Green and colleagues developed methods for cloning epidermal stem cells \[[@B7-jcm-07-00555]\] in their most immature and clonogenic form, which could be differentiated at will to yield a stratified squamous epithelium. The Green strategy has also been applied to capture stem cells from other stratified epithelia, including those of the corneal, thymic and airway epithelia \[[@B8-jcm-07-00555],[@B9-jcm-07-00555],[@B10-jcm-07-00555]\]. However, stem cells of columnar epithelial tissues, such as those of the gastrointestinal tract, liver and kidney, have resisted cloning in a manner that maintains their immaturity during proliferative expansion. Instead, many researchers maintain these columnar epithelia as regenerative "organoids" that contain a minor fraction of stem cells that drive their growth \[[@B11-jcm-07-00555],[@B12-jcm-07-00555],[@B13-jcm-07-00555]\]. While their potential for regenerative medicine is obvious \[[@B14-jcm-07-00555]\], organoids are labor-intensive, very slow to expand and mostly comprised of differentiated cells rather than regenerative stem cells. To overcome this significant barrier in adult stem cell biology, highly robust technologies have been developed to clone and propagate clonogenic, "ground-state" stem cells of the human intestine and colon \[[@B15-jcm-07-00555]\]. These cultured stem cells show remarkable stability in their genomic integrity and epigenetic commitment programs, show unlimited replicative expansion and maintain high clonogenicity, suggesting tremendous potential in disease modeling and regenerative medicine. 2. Technologies for Adult Epithelial Stem Cell Culturing {#sec2-jcm-07-00555} ======================================================== Since 1975, the 3T3J2 feeder system generated in Green's lab has been used widely to culture adult stem cells derived from p63-expressing stratified epithelium, including skin, thymus and lung \[[@B7-jcm-07-00555],[@B9-jcm-07-00555],[@B10-jcm-07-00555]\]. Recently, a new technology developed in our lab, using a combination of 3T3J2 feeder cells and a specialized medium, was employed to clone ground-state stem cells from the columnar epithelium \[[@B15-jcm-07-00555],[@B16-jcm-07-00555],[@B17-jcm-07-00555]\]. The common feature of these two feeder-dependent systems is their ability to keep highly clonogenic adult stem cells in culture so that stem cells can rapidly proliferate and reach sufficient numbers required for downstream applications, such as autologous transplantation, in a short time. In addition, the adult stem cells cultured in these feeder systems maintain the unipotent (e.g., skin) or multipotent differentiation ability (e.g., lung and intestine) and region-specificity despite long-term culturing \[[@B7-jcm-07-00555],[@B10-jcm-07-00555],[@B15-jcm-07-00555],[@B18-jcm-07-00555]\]. The J2 strain of 3T3 cells, originally developed by Green, was used in both of these systems and cleared regulatory hurdles for clinical use thirty years ago, without reports of adverse effects \[[@B19-jcm-07-00555],[@B20-jcm-07-00555],[@B21-jcm-07-00555],[@B22-jcm-07-00555],[@B23-jcm-07-00555],[@B24-jcm-07-00555]\]. Several clinical-grade cell culture methods have been developed under feeder-free conditions \[[@B20-jcm-07-00555],[@B21-jcm-07-00555],[@B22-jcm-07-00555]\]. However, methods based on the Rheinwald and Green culture protocol remain the gold standard for clinical applications \[[@B23-jcm-07-00555]\] due to the high clonogenicity and regenerative capacity. Due to the different signaling requirements of the maintenance of stemness in the stratified epithelium and columnar epithelium, the growth medium of Green's method and the method that we developed (herein Xian--McKeon method) are significantly distinct. In addition to the growth factors such as IGF and EGF that were included in the medium in Green's method, the Xian--McKeon lab developed a media containing novel combinations of growth factors and regulators of TGF-β, Wnt/β-catenin, EGF, IGF, and Notch pathways \[[@B25-jcm-07-00555],[@B26-jcm-07-00555]\] that supports the maintenance of human columnar epithelial stem cells, including intestinal stem cells, in a highly clonogenic, ground-state form. Importantly, single-cell derived, pedigree lines of human intestinal stem cells can be induced to differentiate into all cell types of the intestinal epithelium in air--liquid interface culture systems ([Figure 1](#jcm-07-00555-f001){ref-type="fig"}). Regio-specificity of adult stem cells has been demonstrated in both of these systems \[[@B10-jcm-07-00555],[@B15-jcm-07-00555]\]. In the Xian--McKeon method, the ground-state stem cells possessed robust epigenetic programs of commitment to regio-specific intestinal differentiation that are stable, despite more than a year of continuous propagation. This cell-autonomous regio-specificity of stem cells along the intestinal tract argues against a unitary "intestinal stem cell" or even one for each of the histologically recognized segments, and rather demonstrates a developmentally established spectrum of stem cells that ultimately maintain the histological and functional properties that define these segments. A heuristic deciphering of the commitment code from the regio-specific expression patterns described in Wang et al. \[[@B15-jcm-07-00555]\] will guide parallel efforts with iPSCs to achieve appropriate lineage fates. Despite the extensive studies on p63-expressing stem cells in the upper airways \[[@B27-jcm-07-00555]\] and our own findings of p63-expressing cells contributing to lung regeneration following acute respiratory distress syndrome in mice \[[@B10-jcm-07-00555],[@B18-jcm-07-00555]\], it was completely unclear whether we were dealing with one p63+ airway stem cell occupying different niches or many that were committed to regio-specific differentiation. Green's method was used to clone tracheal airway stem cells (TASCs) and distal airway stem cells (DASCs). These regio-specific airway stem cells demonstrated a clear distinction in their differentiation ability in various in vitro assays and in vivo transplantation experiments. Thus, despite the morphological and gene expression similarities between the upper airway stem cells and the distal airway stem cells, they clearly possess distinct gene expression profiles and cell fate programs. However, in the final analysis, the identification of distinct stem cells in the airways is of critical importance for any future scheme in regenerative medicine or the testing of drugs or biologics for their ability to rally such cells for regenerative endpoints. Human ESC and iPSC lines acquire, with successive passages, genomic structural variations, including some that confer a selective advantage \[[@B28-jcm-07-00555],[@B29-jcm-07-00555]\]. To assess the genomic stability of ground-state stem cells, we examined the copy number (CNV) and single nucleotide variation (SNV) in our cloned ground-state intestinal stem cell pedigrees and showed that ground-state stem cells sustain few genomic changes within the first 100 days of proliferative expansion \[[@B15-jcm-07-00555]\]. Due to their high clonogenicity, we estimated that more than one billion stem cells would be generated from a single stem cell following 60 days culturing in this system \[[@B17-jcm-07-00555]\]. Previous research suggested that epithelial cells require the loss of the pRB/p16^INK4a^ cell cycle control mechanisms, in addition to hTERT overexpression, to achieve immortality \[[@B30-jcm-07-00555]\]. However, the remarkable proliferative potential of these ground-state stem cells, despite the absence of RB or p16 mutations suggests that their apparent immortality is a normal and intrinsic property of adult stem cells, rather than an indication of pathology. 3. Ground-State Stem Cells versus Organoids {#sec3-jcm-07-00555} =========================================== Mammalian cells have been cultured in collagen and laminin-rich matrices as organoids in 3D culture since at least 1980 \[[@B12-jcm-07-00555],[@B31-jcm-07-00555],[@B32-jcm-07-00555],[@B33-jcm-07-00555],[@B34-jcm-07-00555],[@B35-jcm-07-00555]\]. An organoid is defined as a unit of function of a given organ that is able to reproduce, in vitro, a biological structure similar in architecture and function to its counterpart in vivo. There are multiple origins of organoids that include a fragment of tissue, a stem cell isolated in an adult organ, an embryonic stem cell, or an induced pluripotent stem cell. Importantly, organoids can also be generated from cells that have lost stem cell capacity---so-called "transit amplifying cells" with limited proliferative potential---indicating that organoids per se are not strictly assays for stem cells \[[@B36-jcm-07-00555]\]. Despite fundamental advances in stem cells afforded by organoid technology \[[@B37-jcm-07-00555]\], the organoid method does not support cloning the intestinal stem cells or any other adult stem cells in their highly immature, clonogenic state. We solved the problem of culturing populations of pure stem cells of the human gastrointestinal tract using a cocktail of factors impacting Wnt, Notch, and TGF-β signaling pathways among others and were able to clone ground-state stem cells from all regions of the intestine and colon from endoscopic biopsies \[[@B15-jcm-07-00555],[@B17-jcm-07-00555]\]. In other words, the stem cells of columnar epithelia can now be captured in the same ground-state that Green showed for the stem cells of stratified epithelia such as the epidermis. We found the following properties of these ground-state stem cells that make this technology particularly advantageous compared with the organoid method: (1) a typical 1 mm endoscopic biopsy yields 100--300 independent stem cell clones; (2) each of these clones can be independently propagated while maintaining genomic integrity; (3) these clones uniformly express stem cell markers and have a clonogenicity rate of 60--90%; (4) each of these clones retain epigenetic memory of the region from which they were derived; and (5) single clones can be differentiated to yield 3-D intestinal epithelia with regionally-appropriate cell types. It should be acknowledged that many of the features of these gastrointestinal stem cells were unknown prior to Wang et al. \[[@B15-jcm-07-00555]\], Yamamoto et al. \[[@B16-jcm-07-00555]\] and Duleba et al. \[[@B17-jcm-07-00555]\], including the stability of the epigenetic programs underlying commitment, despite months of continuous in vitro propagation, the fact that these stem cells possess all of the information required to assemble into a higher order, and that regio-specific 3-D epithelia are nearly identical to those found in the mature intestine in vivo. As stem cells comprise only a minor component of organoids, perhaps less than 1% \[[@B38-jcm-07-00555]\], the molecular features of the stem cells of columnar epithelia, such as the intestinal tract, have remained unclear ([Figure 2](#jcm-07-00555-f002){ref-type="fig"}). Therefore, the selective cloning and proliferative expansion of highly clonogenic, ground-state intestinal stem cells through the Xian--McKeon method offer a first glimpse into the molecular properties of these cells. Our recent success in adapting this technology to clone stem cells in precancerous lesions, such as Barrett's esophagus, addressed the long-standing confusion in the cellular origin of Barrett's esophagus. Furthermore, using this technology, we generated broad sets of patient-matched stem cells corresponding to all stages of precursor lesions both in-line and out-line with the progression to adenocarcinoma. We can anticipate that each of these clones can be highly annotated with genomics, gene expression and differentiation fate information, accessible through a dynamic biorepository and thus be valuable substrates for new investigations across the cancer biology community. 4. Biobanking of Ground-State Stem Cells and Personalized Regenerative Medicine {#sec4-jcm-07-00555} =============================================================================== The onset of adult-stem-cell-based regenerative medicine started in the 1980s. Green and colleagues demonstrated the first example of cell therapy using cultured stem cells. They showed that human epidermis could be grown in the laboratory and transplanted to patients to reconstitute a functional epidermis \[[@B39-jcm-07-00555],[@B40-jcm-07-00555]\]. Since then, transplantation of cultured epidermal stem cells has long been used to treat patients with burns, chronic wounds and stable vitiligo \[[@B41-jcm-07-00555]\]. This is a life-saving procedure for patients with large area of burns. Moreover, the long-term effectiveness and safety of genetically-modified epidermal stem cells in correcting the severe skin blistering disease, epidermolysis bullosa, has been shown clinically \[[@B42-jcm-07-00555]\]. In addition to skin, cultured stem cells from other epithelial tissues can be the source of stem-cell-based regenerative medicine. For example, a feature of lung regeneration that bodes well for regenerative medicine is that the underlying stem cell is highly clonogenic, which shows unlimited expansion capacity in vitro, and readily transplants to form functional alveoli in acutely damaged lungs \[[@B18-jcm-07-00555]\]. We showed that a single p63+/Krt5+ DASC can be cloned, expanded and transplanted via intratracheal delivery to acutely damaged lungs, where they selectively inhabit damaged regions and differentiate to form Clara cells and alveoli composed of type I and type II pneumocytes. Importantly, these same p63+/Krt5+ DASCs showed no incorporation in mice without prior acute lung injury, suggesting that the efficient regenerative properties of these cells are not marred by "off-target" incorporation. Lastly, DASCs are readily cloneable from simple bronchoscopic biopsies, from bronchopulmonary lavage, or from transmural biopsies, providing good sources of autonomous stem cells that can be expanded to hundreds of billions of cells in weeks. Taken together, the established properties of DASCs, including clonogenicity, expandability, and facility for accurate transplantation obviate many theoretical objections that could have limited their use in regenerative medicine for either acute or chronic lung diseases. It is also conceivable that the newfound ability to derive ground-state stem cells from columnar epithelium provides excellent sources for autologous transplantation to treat a wide range of disorders that current treatments are not able to help. For example, cultured ground-state intestinal stem cells may be of use to restore the intestinal epithelial functions following autologous transplantation in patients with severe forms of short bowel syndrome (SBS) \[[@B43-jcm-07-00555]\], congenital disorders \[[@B44-jcm-07-00555]\], or inflammatory bowel disease (IBD) \[[@B45-jcm-07-00555],[@B46-jcm-07-00555]\]. In conclusion, we now have the technologies for cloning and culturing adult stem cells from nearly all types of epithelial tissues. Given the tremendous success of the pioneering work by Green and colleagues on the use of cultured adult stem cells in regenerative medicine, we should anticipate that more scientists and clinicians will recognize the potential of adult stem cells, appreciate the potential of biobanking various types of adult stem cells from individuals of diverse HLA haplotypes, and make attempts to use them for stem-cell-based personalized regenerative medicine. We thank all the members in the Xian--McKeon laboratory for helpful discussions and support. We thank H. Green for advice and support. Conceptualization, W.X., M.V. and F.M.; Writing-Original Draft Preparation, W.X., M.D., and F.M; Writing-Review and Editing, W.X., Y.Y., M.V., M.D., and F.M.; Supervision, W.X. and F.M.; Funding Acquisition, W.X. and F.M. This work was supported by grants from the Cancer Prevention Research Institute of Texas (CPRIT; RR15014 to WX and RR15088 to FM), the National Institutes of Health (1R01DK115445-01A1 to WX, and U24CA228550 to FM), the US Dept. of Defense (W81XWH-17-1-0123 to WX), the University of Texas Presidential Award (to WX) and the American Gastroenterology Association Research and Development Pilot Award in Technology (to WX). W.X., F.M., M.D. and M.V. have filed a patent related to the technology used in the present work. ![Xian--McKeon method of cloning ground-state intestinal stem cells. (a) Right, Representative image of a single-cell derived ISC^GS^ colony. Left, Top view of in vitro intestinal epithelium generated from stem cells of a pedigree of ISC^GS^ differentiated in an air--liquid interface setting. (b) Schematic diagram of establishing pedigrees from ISC^GS^. Long-term culturing does not affect the capacity of ISC^GS^ to form single-cell derived pedigrees. (c) Clonogenicity assay revealing nearly unchanged number of Rhodamine red-stained colonies despite long-term culturing.](jcm-07-00555-g001){#jcm-07-00555-f001} ![Rapid expansion of a single cell to one billion cells in approximately 60 days using the Xian--McKeon method. In comparison, one cell can become 36 organoids in the organoid method.](jcm-07-00555-g002){#jcm-07-00555-f002}	{ "pile_set_name": "PubMed Central" }
Introduction ============ Sarcoidosis is an inflammatory granulomatous disease which primarily affects the lungs. The disease is characterized by an accumulation of CD4^+^lymphocytes and the formation of non-caseating epithelioid cell granulomas in the affected organs. The granuloma consists of highly differentiated mononuclear phagocytes (epithelioid cells and multinucleated giant cells) surrounded by lymphocytes \[[@B1]\]. The disease either resolves spontaneously or develops into a more chronic disease where the sarcoid granulomas develop fibrotic changes, which in the airways may lead to a progressive loss of lung function. Factors that influence granuloma formation and the development of fibrosis are not well understood in sarcoidosis \[[@B2]\]. Löfgren\'s syndrome is a form of sarcoidosis, which affects about 1/3 of Scandinavian sarcoidosis patients, and is characterized by an acute onset of disease with fever, bilateral lymphadenopathy, erythema nodosum and/or ankle arthritis \[[@B3]\]. Löfgren\'s syndrome is mostly associated with complete disease resolution, often within two years, without the need of any treatment while an insidious onset (non-Löfgren sarcoidosis) is accompanied with a higher risk of developing chronic disease with progressive fibrosis of the lungs. We have recently reported higher levels of nerve growth factor (NGF) in the airways of patients with sarcoidosis as compared to healthy subjects \[[@B4]\]. NGF, brain-derived neurotrophic factor (BDNF) and neurotrophin-3 (NT-3) belong to the family of neurotrophins, and are structurally and functionally related mediators. Neurotrophins are essential survival factors for nerve cells and are critical for the development of peripheral sensory neurons \[[@B5]\]. However, neurotrophins and their corresponding receptors are not only expressed within the nervous system, but are also present in non-neuronal cells and in the airways \[[@B6],[@B7]\]. Structural cells, like epithelial and smooth muscle cells \[[@B6]-[@B8]\], and immune cells, such as mast cells, eosinophils and lymphocytes \[[@B9]-[@B11]\], express neurotrophins as well as their receptors. NGF has been immunolocalized to fibrotic tissue in the lungs and found in elevated levels in sputum from patients with interstitial pulmonary fibrosis (IPF) \[[@B12]-[@B14]\]. Several studies have shown that neurotrophins have tissue healing properties, and are able to promote tissue remodelling in airway disease \[[@B8],[@B14],[@B15]\]. In addition, neurotrophins seem to have pro-inflammatory properties and mediate effects such as mast cell survival and degranulation \[[@B16]\], eosinophil chemotaxis \[[@B17]\] and lymphocyte activation \[[@B18],[@B19]\]. In this context, neurotrophins have been shown to play a role in pulmonary inflammation in asthma \[[@B20]\]. Increased levels of NGF, BDNF and NT-3 have been found in asthmatic airways and are closely linked to airway hyper responsiveness \[[@B6],[@B18],[@B21],[@B22]\]. While knowing that NGF is elevated in bronchoalveolar lavage fluid (BALF) of patients with pulmonary sarcoidosis, less is known about the neurotrophins BDNF and NT-3. Moreover the cellular sources of neurotrophins and the distribution of the corresponding neurotrophin receptors in the airways of these patients are poorly understood. The aim of the present study was to compare the concentrations of the neurotrophins NGF, BDNF and NT-3 in BALF of patients with newly diagnosed pulmonary sarcoidosis with those of healthy controls. Furthermore, the aim was to identify the localization of neurotrophins, and the corresponding neurotrophin receptors, within the sarcoid lung tissue. Methods ======= Subjects -------- This study included 41 patients with newly diagnosed sarcoidosis (4 current smokers, 10 ex-smokers, 27 never-smokers). All subjects had a typical clinical and radiographic picture compatible with the disease in addition to an elevated bronchoalveolar lavage (BAL) CD4/CD8 ratio and/or a biopsy showing non-caseating epithelioid cell granulomas. Diagnosis was established according to defined criteria set up by the World Association of Sarcoidosis and other Granulomatous Disorders (WASOG) \[[@B1]\]. Twentysix of the patients were diagnosed with Löfgren\'s syndrome \[[@B3]\]. Twentyseven never-smoking healthy volunteers with normal chest radiographs were included as healthy controls. No subject received corticosteroids at the time of BAL and blood (serum) sampling. Paired blood and BAL samples were obtained from 37 of the sarcoidosis patients and from all healthy subjects. Clinical characteristics are presented in Table [1](#T1){ref-type="table"}. ###### Clinical characteristics of study subjects included for bronchoalveolar lavage studies. Sarcoidosis patients ------------------------------------- --------------- ------------- ---------------------- ------------- Subjects, n 27 41 26 15 Sex, (M/F) 11/16 26/15 12/14 12/3 Age, yr 30 (24-39) 37 (31-41) 36 (29-41) 38 (32-42) Radiograph stage, (I/II/III/IV) \- 22/16/3/0 16/10/0/0 6/6/3/0 Pulmonary function tests, % VC 99 (94-106) 95 (83-104) 97 (84-105) 91 (79-101) FEV~1~ 99 (92-108)\* 91 (86-99) 90 (86-102) 95 (77-101) Data are presented as medians (interquartile ranges). M, male; F, female; VC, vital capacity; FEV~1~, forced expiratory volume in 1 s. ^\^; p \< 0.05 versus sarcoidosis patients For immunohistochemistry, biopsy specimens showing non-caseating epithelioid cell granuloma formation compatible with sarcoidosis were collected from 19 additional sarcoidosis patients, diagnosed according to the above defined criteria. In 17 cases the biopsies were transbronchial, in one patient lung tissue was sampled through video-assisted thoracoscopy, and in one case an intrathoracic lymph node biopsy was obtained during mediastinoscopy. Four of the patients were current smokers, 2 ex-smokers, and 13 were never-smokers. Seven of the patients had Löfgren\'s syndrome. Bronchoscopy, including BAL (see below) and biopsy sampling, were performed in all patients as they were referred to the lung clinic at Karolinska University Hospital in Stockholm, Sweden, for diagnostic purposes. The study was approved by the Regional Ethical Review Board in Stockholm (<http://www.epn.se>) (Dnr: 2005/1031-31) and in accordance with the Helsinki Declaration. All subjects gave their written informed consent. Fiberoptic bronchoscopy ----------------------- BAL of sarcoidosis patients and healthy subjects was performed as previously described \[[@B23]\]. Briefly, a flexible fiberoptic bronchoscope (Olympus Optical Co., Japan) was wedged into a middle-lobe bronchus and five aliquots of 50 ml sterile PBS solution were instilled and re-aspirated. Recovered BAL fluid (BALF) was separated into a cell- and debris free BALF, which was stored at -70°C until analyzed, and a cell fraction from which cytospin slides for differential cell counts were prepared and analyzed as previously described \[[@B23]\]. Biopsies were fixed in a buffered 10% formalin solution for 24 h and embedded in paraffin. Analysis of neurotrophins with ELISA ------------------------------------ Neurotrophins were quantified in BALF and serum by commercially available, two-site enzyme-linked immunosorbent assay (ELISA)-kits according to the manufacturer\'s instructions (Promega, USA) and as previously described \[[@B12],[@B21]\]. Detection limit was 4.7 pg/ml for NT-3 and 7.8 pg/ml for NGF and BDNF ELISA kits. All samples were analyzed in duplicates and serum samples were diluted in PBS before analysis (1:100 for NT-3 analysis and 1:500 for BDNF analysis). Immunohistochemistry -------------------- Serial 4 μm thick sections were mounted on slides and processed for immunohistochemistry. Sections were deparaffinized in xylene, stepwise rehydrated through graded ethanol, and antigen retrieval was achieved by boiling slides in 10 mM citrate buffer (pH 6.0) (for neurotrophins and neurotrophin receptors) or ethylenediamine tetraacetic acid (EDTA) buffer (pH 9.0) (for CD68) for 20 min in microwave oven. After cooling and washing in PBS, slides were incubated in 0.3% H~2~O~2~for 30 minutes to block endogenous peroxidase activity. After blocking with 5% goat serum or horse serum (for CD68) for 1 h at room temperature, slides were exposed to primary antibodies diluted in blocking buffer over night at 4°C. Primary antibodies and dilutions are presented in Table [2](#T2){ref-type="table"}. In control experiments, primary antibodies were omitted. Non-specific binding of anti-NGF, -BDNF, -NT-3, -TrkA, TrkC and TrkB was evaluated by incubating slides with the antibodies pre-adsorbed with the corresponding blocking peptides (ratio 1:5) (Santa Cruz Biotechnology Inc, Santa Cruz, CA, USA). After incubation, slides were washed and exposed to relevant biotinylated secondary antibodies (goat anti-rabbit or horse anti-mouse) (1:300) (Vector Laboratories, Burlingame, CA, USA) for 1 h at room temp. The product of immune reaction was revealed using Vectastain^®^, Elite^®^, ABC Kit (Vector Laboratories) followed by SIGMA FAST™ 3,3 diaminobenzidine (Sigma-Aldrich, St. Louis, MO, USA). Sections were then counter-stained with Mayer\'s hematoxylin before they were dehydrated, mounted and viewed under light microscope (Leica DMLB) at a magnification of ×100, ×200 and/or ×400. ###### Anti-human antibodies used for immunohistochemical staining. Antibody Cat No Source Dilution Manufacturer ---------- -------- -------- ---------- ------------------------- NGF sc-548 rabbit 1:100 Santa Cruz Biotech. Inc BDNF sc-546 rabbit 1:100 Santa Cruz Biotech. Inc NT-3 sc-547 rabbit 1:100 Santa Cruz Biotech. Inc TrkA sc-118 rabbit 1:100 Santa Cruz Biotech. Inc TrkB sc-12 rabbit 1:100 Santa Cruz Biotech. Inc TrkC sc-117 rabbit 1:100 Santa Cruz Biotech. Inc CD68 M 0876 mouse 1:200 DakoCytomation Statistical analysis -------------------- Data are presented as medians (interquartile range). Mann-Whitney test, or Kruskal-Wallis test followed by Dunn\'s post test, were used for group comparisons. A p-value \< 0.05 was considered significant. Analyses were performed with Graphpad Prism 4.03 (Graphpad Software Inc., USA). Results ======= BAL analysis and differential cell counts ----------------------------------------- BAL recovery was higher in healthy subjects as compared to sarcoidosis patients (78; 68-79% vs. 68; 60-75%, p \< 0.05). BAL cell viability was similar in both healthy subjects and sarcoidosis patients (median: 95%). BAL differential cell counts are presented in Table [3](#T3){ref-type="table"}. In BALF, the total cell concentration as well as concentrations of macrophages, lymphocytes and neutrophils were significantly higher in sarcoidosis patients as compared to healthy subjects. As expected, the percentage of macrophages was lower and the percentage of lymphocytes was higher in sarcoidosis patients compared to healthy subjects. ###### Differential cell counts in bronchoalveolar lavage. Sarcoidosis patients ---------------------------- ---------------------- --------------- ---------------------- --------------- Total cell conc. \10^6^/L 81 (62-94) \\\* 201 (133-308) 188 (131-308) 252 (132-319) Macrophages % 94 (91-95) \\\* 75 (63-84) 77 (63-86) 69 (60-80) Macrophages \10^6^/L 75 (58-92) \\\ 143 (90-210) 141 (85-230) 165 (91-200) Lymphocytes % 5.2 (3.8-7.0) \\\* 23 (14-34) 22 (13-33) ^\#^ 29 (19-40) Lymphocytes \10^6^/L 4.4 (2.7-6.1) \\\ 45 (22-81) 41 (15-76) 54 (36-87) Eosinophils % 0 (0-0.2) 0.0 (0-0.7) 0.1 (0-0.6) 0.0 (0-1.0) Eosinophils \10^6^/L 0 (0-0.2) 0.0 (0-1.2) 0.2 (0-1.2) 0.0 (0-1.6) Neutrophils % 1.0 (0.4-1.4) 1.0 (0.6-1.8) 1.4 (0.5-2.1) ^\#^ 0.6 (0.6-1.0) Neutrophils \10^6^/L 0.7 (0.3-1.1)\\\* 2.3 (0.9-3.5) 2.6 (1.2-4.1) 1.5 (0.9-2.6) CD4/CD8 ratio ND 7.3 (4.2-11) 8.2 (4.9-14) 6.9 (3.8-9.8) Data are presented as medians (interquartile ranges). ND: not determined. *\\\; p \< 0.001 versus sarcoidosis patients, ^\#^; p \< 0.05 versus non-Löfgren. Neurotrophin levels in BALF --------------------------- Significantly elevated concentration of NT-3 was found in BALF from sarcoidosis patients as compared to healthy subjects (Figure [1A](#F1){ref-type="fig"}). When sub-grouping the sarcoidosis patients, significantly higher levels of NT-3 were found in BALF from patients with non-Löfgren sarcoidosis compared to patients with Löfgren\'s syndrome (Figure [1B](#F1){ref-type="fig"}). In addition, higher NT-3 levels were associated with more advanced disease stage (Figure [1C](#F1){ref-type="fig"}). In line with our previous report \[[@B4]\], NGF was significantly elevated in sarcoidosis patients (14.8; 6.1-22.6 pg/ml) as compared to healthy subjects (4.7; 2.0-17.0 pg/ml) (p \< 0.01). In the present material, NGF showed no significant association with Löfgren\'s syndrome or disease stage. BDNF in BALF was below the detection limit of the ELISA kit. ![Neurotrophin-3 (NT-3) levels are increased in sarcoidosis*. NT-3 protein levels in bronchoalveolar lavage fluid (BALF) from healthy subjects and patients with sarcoidosis (A), and in subgroups of sarcoidosis patients divided according to Löfgren\'s syndrome or not (B) and radiographic stage (C). Horizontal bars indicate median values. \: p \< 0.05, \\: p \< 0.01.](1465-9921-11-156-1){#F1} Neurotrophin levels in serum ---------------------------- No significant differences were found in NT-3 or BDNF concentrations between patients and controls or between subgroups of sarcoidosis patients. The concentration of NT-3 was approximately 3000 times higher in serum (41; 22-79 ng/ml) compared to BALF (13.0; 10.0-19.3 pg/ml). BDNF concentration in serum was 16; 12-24 ng/ml. NGF levels in serum were not determined as the levels have previously been reported by us to be below detection limit \[[@B4]\]. Neurotrophin and neurotrophin receptor expression in sarcoid lung tissue ------------------------------------------------------------------------ Serial sections from 18 lung biopsies and one lymph node biopsy were analyzed for NGF, BDNF, NT-3, TrkA, TrkB and TrkC, respectively, by immunohistochemistry. Figures [2](#F2){ref-type="fig"}, [3](#F3){ref-type="fig"}, [4](#F4){ref-type="fig"} and [5](#F5){ref-type="fig"} show representative immunostainings. ![Lung biopsy sections of sarcoidosis patients. The presence of non-caseating epithelioid cell granulomas (Gr) with a surrounding layer of lymphocytes (hematoxylin-eosin staining, ×10) is shown in A). CD68 immunostaining in alveolar macrophages (AM), epithelioid cells (Ec) (×10) is shown in B) and CD68 immunostaining in a multinucleated Giant cell (Gc) and epithelioid cells (Ec) (×40) is shown in C).](1465-9921-11-156-2){#F2} The lung biopsies from sarcoidosis patients contained typical non-necrotizing granulomas composed of epithelioid cells and multinucleated giant cells and a surrounding layer of lymphocytes (Figure [2A](#F2){ref-type="fig"}). Strong immunoreactivity for CD68 (commonly used as a marker for monocyte and macrophage-derived cells) was found in macrophage-like cells, epithelioid cells and multinucleated giant cells (Figures [2B](#F2){ref-type="fig"} and [2C](#F2){ref-type="fig"}). Marked NGF, BDNF and NT-3 immunoreactivity was observed in the granulomas and was localized to epithelioid cells and giant cells within the granulomas (Figures [3A, B](#F3){ref-type="fig"} and [3C](#F3){ref-type="fig"}). No or less immunoreactivity for NGF, BDNF or NT-3 was found within fibrotic tissue around granulomas (Figures [3A, B](#F3){ref-type="fig"} and [3C](#F3){ref-type="fig"}). Analysing neurotrophin receptor immunoreactivity in the tissue sections, marked TrkA, TrkB and TrkC immunoreactivity was observed within the granulomas (Figures [3D, E](#F3){ref-type="fig"} and [3F](#F3){ref-type="fig"}). Also immunoreactivity for the neurotrophins and their receptors was found in inflammatory cells surrounding the granulomas (Figure [3A](#F3){ref-type="fig"} and not shown). ![Neurotrophin and neurotrophin receptor immunostaining in sarcoid lung granulomas. Immunostainings for NGF (A), NT-3 (B), BDNF (C), TrkA (D), TrkC (E) and TrkB (F) on lung biopsy sections from sarcoidosis patients (×40). Positive immunostaining was localized to epitheioid cells and Giant cells within the granulomas. F: fibrosis; Ic: infiltrating inflammatory cells. Non-specific (NS) immunostaining is shown in (G) and was obtained by exposing the sections to the antibodies preabsorbed with the corresponding blocking peptide (×20).](1465-9921-11-156-3){#F3} Sarcoid granulomas in the mediastinal lymph node also showed positive immunoreactivity for NGF, BDNF, NT-3, TrkA, TrkB and TrkC, localized to the granulomas and the surrounding lymphoid tissue (Figures [4A, B, C, D, E](#F4){ref-type="fig"} and [4F](#F4){ref-type="fig"}). ![Neurotrophin and neurotrophin receptor immunostaining in sarcoid lymph node granulomas. Immunostainings for NGF (A), NT-3 (B), BDNF (C), TrkA (D), TrkC (E) and TrkB (F) on sections of a mediastinal lymph node from a patient with sarcoidosis (×20). Positive immunostaining was localized to the granulomas. Non-specific (NS) immunostaining is shown in G) and was obtained by exposing the sections to the antibodies preabsorbed with the corresponding blocking peptide.](1465-9921-11-156-4){#F4} Two of the lung biopsies contained ciliated bronchial epithelium and submucosa. Marked immunoreactivity for NGF, BDNF, NT-3, TrkA and TrkB, and weaker immunoreactivity for TrkC was found in the epithelium (Figures [5A, B, C, D, E](#F5){ref-type="fig"} and [5F](#F5){ref-type="fig"}). Smooth muscle cells in the submucosa showed immunoreactivity for NGF, NT-3 and TrkA (Figures [5A, B](#F5){ref-type="fig"} and [5D](#F5){ref-type="fig"}) and infiltrating inflammatory cells in the submucosa showed positive immunostaining for NGF, BDNF, NT-3, TrkA and TrkB (Figures [5A, B, C, D](#F5){ref-type="fig"} and [5F](#F5){ref-type="fig"}). ![Neurotrophin and neurotrophin receptor immunostaining in sarcoid lung tissue. Immunostainings for NGF (A, G), NT-3 (B, H), BDNF (C, I), TrkA (D, J), TrkC (E, K) and TrkB (F, L) on lung biopsy sections from sarcoidosis patients (×20). Ep: epithelium, Sm: smooth muscle, Ic: infiltrating inflammatory cells and AM: alveolar macrophages. Non-specific (NS) immunostainings are shown in (M & N) and were obtained by exposing the sections to the antibodies preabsorbed with the corresponding blocking peptide.](1465-9921-11-156-5){#F5} Marked immunoreactivity for NGF, BDNF, NT-3 as well as for TrkA, TrkB and TrkC was observed in alveolar macrophages (Figures [5G, H, I, J, K](#F5){ref-type="fig"} and [5L](#F5){ref-type="fig"}). The wall of the alveoli displayed positive immunoreactivity for NT-3, BDNF and TrkB (Figures [5H, I](#F5){ref-type="fig"} and [5L](#F5){ref-type="fig"}) and weaker immunoreactivity for NGF and TrkA (Figures [5G](#F5){ref-type="fig"} and [5J](#F5){ref-type="fig"}). TrkC immunoreactivity was not observed within the alveolar wall (Figure [5K](#F5){ref-type="fig"}). Granulomas from patients with Löfgren\'s syndrome and non-Löfgren sarcoidosis showed similar localization and strength of immunoreactivity for the different neurotrophins and corresponding receptors (not shown). Discussion ========== The present study provides evidence for increased levels of the neurotrophins NT-3 and NGF locally in the airways of patients with pulmonary sarcoidosis as compared to healthy individuals. This study also describes, for the first time, neurotrophin and neurotrophin receptor expression localization in sarcoid granulomas within the lungs and lymph nodes of patients with pulmonary sarcoidosis. We have previously reported elevation of NGF in BALF of sarcoidosis patients \[[@B4]\], and this study confirms those results, and extends them by including the analysis of the neurotrophins NT-3 and BDNF in both BALF and serum, and by identifying cellular sources of, and targets for, NGF, BDNF and NT-3 in the lungs of sarcoidosis patients. This study shows that not only the NGF protein, but also NT-3 protein, was elevated in BALF from sarcoidosis patients as compared to healthy subjects, while BDNF protein levels were undetectable in both healthy subjects and sarcoidosis patients. Sarcoidosis patients presenting Löfgren\'s syndrome have an acute disease onset and often a favorable prognosis with high rate of spontaneous disease resolution. In contrast, patients with non-Löfgren sarcoidosis run a higher risk of developing chronic disease with lung fibrosis. In this study we sub-grouped the sarcoidosis patients into those with Löfgren\'s syndrome and those with non-Löfgren sarcoidosis, and we could show that the levels of NT-3 were significantly higher in BALF in the non-Löfgren sarcoidosis group. Differences between subgroups of sarcoidosis patients was not seen for NGF in this study, but has been indicated in one of our previous studies, including a larger study population \[[@B4]\]. Chest radiographic staging (stage 0-IV) is a measurement of lung involvement in sarcoidosis and reflects disease severity, where stage 0 describes no visible intrathoracic findings and stage IV, being the most advanced stage, is characterized by pulmonary fibrosis. When sub-grouping the patients according to their radiographic stage, we found higher NT-3 levels in BALF in patients with more advances disease stage. Taken together, these results suggest an association between NT-3 and disease severity and prognosis and it could be speculated on whether NT-3 could serve as an early clinical marker for disease activity and progression in sarcoidosis. In contrast to some other studies \[[@B24]\], BDNF was not possible to detect in BALF in our study. A plausible reason for this would be the difference in age of study participants, different exposure to environmental factors that may influence neurotrophin production and the fact that we used higher lavage volumes. The latter would dilute any acellular component to a larger degree. Apart from our study, also other studies have shown that the content of BDNF in BALF is near or below detection limit of the used ELISA kit \[[@B21]\]. In serum, concentrations of BDNF and NT-3 of healthy and sarcoidosis patients did not differ between healthy subjects and patients. Previous studies by us and others have shown that NGF concentration in serum is low or under the detection limit \[[@B4],[@B25]\]. This supports the concept of a local enhancement and a possible local origin of neurotrophins in pulmonary sarcoidosis. To elucidate the possible cellular sources of neurotrophins in the airways in sarcoidosis we performed immunohistochemistry on lung biopsy sections from distal airways of sarcoidosis patients. An intense immunostaining for NGF, BDNF and NT-3 was localized to granulomas and specifically found in epithelioid cells and giant cells of the granulomas. Epithelioid cells are macrophage-derived, highly differentiated cells with secretory functions and giant cells are multinucleated cells resulting from fusion of epithelioid cells. To our knowledge, this is the first report on neurotrophin expression in these cells and we suggest that they are possible cellular sources of the enhanced NGF and NT-3 levels detected in the bronchoalveolar lavage fluid. In our study we also had access to one pulmonary draining lymph node obtained with mediastinoscopy from a patient with pulmonary sarcoidosis. The lymphoid tissue exhibited marked granuloma formations, which were immunopositive for NGF, BDNF and NT-3. Previous reports have shown that infection-induced hepatic- and brain granulomas produce NGF, and that enhanced levels of NGF can be detected in granulomatous tissue \[[@B26]-[@B28]\]. Also an infectious cause has been suggested in the pathogenesis of sarcoidosis and in sarcoid granuloma formation \[[@B1]\]. NGF, BDNF and NT-3 immunostainings were also detected in structural- and inflammatory cells in sarcoid lung biopsies, as supported by previous studies in healthy and asthmatic airways \[[@B6],[@B7]\]. Interestingly, in vitrostudies on airway structural cells, such as epithelial cells, fibroblasts and smooth muscle cells, have shown that these cells produce neurotrophins constitutively and that the production is enhanced under inflammatory conditions \[[@B29]-[@B31]\]. In addition, Ricci and co-workers have demonstrated that immune cells, such as alveolar macrophages and T-lymphocytes, retrieved from BAL from sarcoidosis patients, express NGF, NT-3 and BDNF to a larger degree than BAL cells from healthy subjects \[[@B32]\]. We confirm and extend these results by showing positive NGF, BDNF and NT-3 immunostaining in alveolar macrophages present within the lung parenchyma of sarcoidosis patients. It is well known that T-lymphocytes and macrophages infiltrate the lungs in sarcoidosis patients, and therefore it may be postulated that an increased number of neurotrophin-expressing inflammatory cells, and specifically alveolar macrophages, in sarcoidosis patients contribute to enhanced levels of neurotrophins in BALF of these patients. Thus, we have identified multiple sources of neurotrophins in sarcoidosis airways, where the granulomas seem to be a unique source. To elucidate the cellular targets for NGF, BDNF and NT-3 within the airways in sarcoidosis, we studied the presence of the corresponding neurotrophin receptors, TrkA, TrkB and TrkC, in lung biopsies. TrkA, TrkB and TrkC belong to the protein tyrosine kinase (Trk) family of receptors, which bind neurotrophins with high affinity. While TrkA is the primary receptor for NGF \[[@B33]\], TrkB is the primary receptor for BDNF \[[@B34]\] and TrkC is the primary receptor for NT-3 \[[@B35]\]. We found that the sarcoid lung granulomas were immunoreactive for both TrkA, TrkB and TrkC, indicating that neurotrophins, which are also produced within the granulomas, are able to function in an autocrine and/or paracrine manner in the granuloma microenvironment in lung tissue in sarcoidosis. Similarly, granulomas in the mediastinal lymph node were positive for the two neurotrophin receptors arguing for a possible local role of neurotrophins also in lymphoid tissue. In addition, we demonstrated TrkA, TrkB and TrkC immunoreactivity in structural cells in sarcoid biopsies, in line with previous studies in healthy airways \[[@B7],[@B8],[@B36]\]. As described previously in both healthy and sarcoid airways, we also observe neurotrophin receptor immunoreactivity in alveolar macrophages \[[@B32],[@B37]\]. Taken together, the current findings suggest that, besides the granulomas, also structural- and inflammatory cells are possible targets for the neurotrophins in the airways in sarcoidosis. Despite the lack of detectable levels of BDNF in BALF, we could detect both BDNF and its receptor in the granulomas and airway cells in sarcoidosis by immunohistochemistry in a similar fashion as for NGF and NT-3. This indicates that besides NGF and NT-3, BDNF could be a messenger molecule of relevance in pulmonary sarcoidosis. Further studies are required to support this hypothesis. The functional roles of neurotrophins in inflammatory conditions of the airways are considered to be multiple and NGF is often referred to as an inflammatory mediator. Elevations of neurotrophins have previously been described to be linked to pulmonary inflammatory diseases, including asthma and interstitial pulmonary fibrosis (IPF). In asthma, neurotrophins have been described to be elevated in BALF as compared to healthy subjects \[[@B6],[@B21]\] and to enhance airway inflammation and airway hyperreactivity \[[@B22],[@B38]-[@B40]\]. Importantly, airway hyperreactivity is not a specific feature for asthma only, but is also present in patients with sarcoidosis \[[@B41]\]. In addition, a role for neurotrophins in wound healing and fibrosis has been suggested \[[@B42]\]. Thus, neurotrophin expression has been linked to airway tissue remodelling, shown to be immunolocalized to fibrotic tissue in patients with IPF \[[@B12],[@B14]\], and found in increased levels in sputum of these patients \[[@B13]\]. Furthermore, neurotrophins have been shown to modulate fibroblast migration and pro-fibrotic phenotype \[[@B14],[@B15]\]. Interestingly, we found an association of higher NT-3 levels in subgroups of sarcoidosis patients associated with a higher risk of developing chronic disease and fibrosis. Whether neurotrophin expression in sarcoid granulomas may promote the persistence of the granuloma and/or promote the development of fibrosis needs to be further investigated. In conclusion, the present study describes that the neurotrophins NGF, BDNF and NT-3 are expressed in sarcoid granulomas in the airways and that enhanced levels of NGF and NT-3 are found in bronchoalveolar lavage fluid of patients with pulmonary sarcoidosis as compared to healthy individuals. The findings of immunoreactivity for TrkA, TrkB and TrkC, the high-affinity receptors for NGF, BDNF and NT-3, respectively, within granulomas, structural- and inflammatory cells, suggest that these are possible cellular targets for the neurotrophins in sarcoid airways. Taken together, this study supports the concept that the neurotrophins are involved airway inflammation, granuloma biology and fibrosis in inflammatory pulmonary diseases. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= CD performed experiments and data analysis, participated in study planning and wrote the manuscript. JG participated in the design of the study, patient recruitment, data analysis and critically reviewed the manuscript. MK performed immunohistochemical stainings and analysis. HH-O participated in study planning, patient recruitment and material collection. GE participated in study planning, patient material collection and data analysis. AE participated in study planning, patient recruitment, patient material collection and critically reviewed the manuscript. COH conceived of the study and its design, did data analysis and manuscript writing. All authors read and approved the final manuscript. Acknowledgements ================ This work was supported by the Swedish Research Council, Swedish Heart Lung Foundation, King Oscar II Jubilee Foundation, Stockholm County Council, Swedish Asthma and Allergy Association, Torsten och Ragnar Söderberg\'s Foundations, Osher Center for Integrative Medicine Karolinska Institutet, Centre for Allergy Research Karolinska Institutet, and Karolinska Institutet. COH was supported by the Swedish Research Council and Karolinska Institutet, and CD by the Swedish Heart Lung Foundation. The authors are grateful for the assistance of H Blomqvist, M Dahl, B Dahlberg, B Engvall and G de Forest.	{ "pile_set_name": "PubMed Central" }
When we return from a conference, the return trip often seems shorter than the initial trip. This is odd, since the distance traveled and the actual time spent traveling are usually identical. Here, we investigated the existence of a return trip effect and explored two possible causes of this phenomenon. If a return trip effect exists, one possible cause might be an increase in familiarity and predictability. Previous research has shown that novel tasks are often remembered as taking longer than they actually do, while familiar tasks are remembered as taking less time (Avni-Babad & Ritov, [@CR1]; Boltz, Kupperman, & Dunne, [@CR6]; Hinds, [@CR11]; Roy & Christenfeld, [@CR14]). Similarly, tasks that are unpredictable are remembered as taking longer than similar coherent or predictable tasks (Boltz, [@CR3], [@CR4], [@CR5]). The unpredictability of the initial leg of the trip may make it so that it is remembered as being overly long. Conversely, the return trip is remembered as being short because it is now more familiar and predictable. Alternatively, the return trip effect could be due to a violation of expectations. People often predict that tasks will take less time than they actually do (Buehler, Griffin, & Ross, [@CR7]; Kahneman & Tversky, [@CR12]; Roy, Christenfeld, & McKenzie, [@CR15]). In general, people are fairly inaccurate when estimating how long things either have taken in the past (Block & Zakay, [@CR2]; Burt & Kemp, [@CR8]; Poynter, [@CR13]; Wallace & Rabin, [@CR17]) or will take in the future (Buehler et al. [@CR7]; Roy, Christenfeld & McKenzie [@CR15]). It may be that people have an expectation for a trip that is overly short, leading to a violation of this expectation when they take the trip. Hence, the initial trip takes longer than expected. For the return trip, the expectation is likely to be based on the experience of the (disappointingly long) initial trip. This leads to an upward adjustment in expectations for the return trip that is, happily, unmet. The return trip, therefore, feels shorter than the initial trip. Here, we try to establish whether or not there is a return trip effect and whether this might be due to a shift in familiarity or expectation. Before the studies were undertaken, a pretest was conducted to test whether people actually experience this phenomenon in their daily life. Fifty-six students (43 females, M ~age~ = 19 years) indicated whether they "sometimes feel that the return trip seems shorter than the initial trip" and whether they sometimes feel that it seems longer, on scales from 1 (never) to 7 (always). In comparison with the initial trip, participants more often reported that the return trip felt shorter (M = 5.23, SD = 1.36) than that it felt longer (M = 3.59, SD = 1.80), paired-sample t(55) = 4.21, p \< .001, r ^2^ = .24. When given a forced choice about whether the initial trip or the return trip tended to feel shorter, participants were almost 3 times as likely to indicate that the return trip tended to feel shorter (73% chose the return trip vs. 27% for the initial trip). Similar to our own experiences, participants felt that a return trip effect occurs frequently. The return trip effect was further examined in a field study ([Study 1](#Sec1){ref-type="sec"}), a field experiment ([Study 2](#Sec2){ref-type="sec"}), and a controlled lab experiment ([Study 3](#Sec5){ref-type="sec"}). Study 1 {#Sec1} ======= In [Study 1](#Sec1){ref-type="sec"}, the return trip effect was tested in a real-world situation, and possible explanations for the phenomenon were explored. Sixty-nine participants (57 females, M ~age~ = 33 years) who had just returned from a day trip by bus from either a housekeeping fair (a large event where people received free goods related to housekeeping) or the Efteling (a large theme park) indicated which part of the trip felt longer. We took care to include only participants for whom the initial and return trips actually took the same time (which varied between 25 and 120 min)[1](#Fn1){ref-type="fn"} and who were awake during both trips. After the return trip ended, the participants indicated how long they felt that the return trip had taken, as compared with the initial trip, on an 11-point scale from -5 (a lot shorter) to +5 (a lot longer). Participants exhibited a return trip effect, with the return trip feeling like it took less time than the initial trip (M = -0.55, SD = 2.16, with an average response significantly lower than 0), t(68) = 2.12, p = .04, d = 0.25. To examine the influence of trip familiarity/predictability on the return trip effect, participants also indicated the extent to which they recognized certain waypoints on the return trip that they had seen on the initial trip on a scale from 1 (none) to 9 (a lot) (M = 5.32, SD = 2.15). A regression analysis showed that how much they recognized along the way (as a measure of familiarity) was unrelated to the return trip effect, β = .15, t(67) = 1.21, p = .23; recognizing more waypoints during the return-trip did not influence the perception that the return trip took less time than had the initial trip. To examine the effect of expectations on the return trip effect, participants indicated whether the initial trip took shorter or longer than they had expected on an 11-point scale from -5 (much shorter) to +5 (much longer). On average, participants indicated that the initial trip took longer than they had expected (M = 0.59, SD = 2.27, which differs from 0), t(68) = 2.17, p = .03, d = 0.26. Furthermore, this violation of expectations predicted the return trip effect: The more travelers thought that the initial trip took longer than they had expected, the more they felt that the return trip took less time than the initial trip had, β = -.27, t(67) = 2.25, p = .02. Study 2 {#Sec2} ======= [Study 2](#Sec2){ref-type="sec"} attempted to replicate the findings of [Study 1](#Sec1){ref-type="sec"} using a different task. In [Study 1](#Sec1){ref-type="sec"}, participants' familiarity with the return route, as measured by the number of waypoints recognized, did not influence the return trip effect. However, this measure might not have accurately or fully captured their familiarly with the return route. Therefore, the possible role of familiarity in the return trip effect was examined in [Study 2](#Sec2){ref-type="sec"} by having participants return either by the same or by a different, equidistant route. Method {#Sec3} ------ Ninety-three 1st-year students (67 females, M ~age~ = 19 years) at a get-to-know-each-other event at the start of the academic year traveled by bike from a base camp (where they had spent one night) to a nearby forest to play some games for about 2 h. Participants traveled in small groups of 5--10 participants and followed someone who knew the way (who was not included in the sample). Participants were randomly assigned to travel along one of two equally long routes on the initial leg of the trip (route 1 = 9.23 km, route 2 = 9.19 km), both of which, in a pretest, took 35 min. A separate group of participants (a control group) traveled the initial leg via route 1 (n = 10) or route 2 (n = 8). They were asked at the end of the initial trip how long it had felt to them in minutes. They confirmed that both routes felt equally long (M ~route1~ = 41.5 min, M ~route2~ = 42.5 min), t(16) = 0.24, p = .83, d = 0.10. For the main group of participants, we manipulated whether they returned by the same route as that initially traveled (n = 32) or by the other one (n = 50).[2](#Fn2){ref-type="fn"} After participants arrived back at base camp, they filled out a questionnaire measuring their time perceptions of the initial and return trips. All estimates were given after both trips had been completed so that participants on each leg of the trip were unaware that they would be estimating duration (knowing that an estimation will be required can change time perception; Block & Zakay, [@CR2]). Results {#Sec4} ------- We conducted a 2 × 2 mixed-model Type III ANOVA with same route versus different route as a between-subjects variable and time estimate (in minutes) of initial trip versus return trip as a within-subjects variable. Participants exhibited a clear return trip effect, estimating that the initial trip took longer (44.3 min) than the return trip did (36.9 min), F(1, 80) = 19.02, p \< .001, η ~p~^2^ = .19. Note that the estimates for how long the initial trip took (M = 44.3 min) did not differ from the estimate of the control group (M = 41.9 min), who made a time estimate directly after the initial trip, t(98) = 0.78, p = .44, d = 0.16. Even though participants who completed the full experiment gave estimates for the initial leg after a long delay, their estimates did not differ from those of the control group, who estimated duration directly after completing the trip. As further evidence for the return trip effect, how long the return trip felt to participants (M = 36.9 min) was shorter than the estimate the control group made for how long the initial trip felt (M = 41.9 min), t(98) = 2.06, p = .04, d = 0.42. Varying the return route allowed us to directly examine the role of familiarity in the return trip effect. First, there was an unexpected main effect of condition showing that participants who took a different route on the way back made longer estimates for both the initial trip and the return trip (42.3 min on average) than did participants who returned by the same route (38.2 min), F(1, 80) = 5.34, p = .02, η ~p~^2^ = .06. Importantly, however, the magnitude of the return trip effect was not significantly different for participants who returned either by the same or by a different route, F(1, 80) = 1.77, p = .19, η ~p~^2^ = .02. While a lack of familiarity might have caused participants who returned by novel routes to give longer estimates overall, familiarity with the route did not explain the return trip effect. As in the previous experiment, participants indicated whether or not the initial trip took longer than they had expected on the scale from -5 (a lot shorter than expected) to +5 (a lot longer than expected). Participants thought that the initial trip took longer than expected (M = 1.46, SD = 1.64, which differed from 0), t(89) = 8.19, p \< .001, d = 0.89. More important, this violation of expectations again predicted the return trip effect: The more participants thought that the initial trip took longer than expected, the shorter they felt the return trip took, as compared with the initial trip (as measured by the difference in minutes of the estimated duration of the initial and the return trips), β = -.33, t(80) = 3.12, p = .003. Study 3 {#Sec5} ======= In a final study, the return trip effect was tested in a more controlled setting. To do so, participants watched a video of someone traveling by bike from her home to a friend's house and back again at a later time. We made the video so that the initial and return trip were of exactly the same length, both in time and in distance traveled. Therefore, for example, different speeds of travel could not influence the time estimates (as found by Cohen & Cooper, [@CR9]; Cohen, Hansel, & Sylvester, [@CR10]). To further test the hypothesis that the return trip effect is caused by a violation of initial expectations, after which the return trip feels less bad in comparison, a condition was added where the initial expectations were manipulated. If, as indicated by participants in the previous studies, a feeling that the initial trip took much longer than expected contributes to the return trip effect, making participants expect an overly long initial trip should lessen or eliminate the return trip effect. Finally, order of estimates was manipulated to make sure that when the estimates were given could not account for the return trip effect. Method {#Sec6} ------ One hundred thirty-nine participants (94 females, M ~age~ = 21 years) took part in a series of studies of which ours was part. The basic setup was that participants were seated behind a computer and were told that they would see a video of a student who traveled by bike from her house to that of a friend. The video was shot from the viewpoint of the person riding the bike (the rider could not be seen). The bike ride took exactly 7 min, during which a distance of 2.25 km was traveled. After this video had been seen, other studies followed for about 10 min. Next, the participants again watched a video of the student riding a bike, but now described as returning from her friend's house to her home (with exactly the same time length). The main dependent variables were again how long participants felt the initial trip and the return trip took. Five conditions were created:Basic return trip effect. After seeing the return trip, participants were asked to indicate how long they felt that the return trip had taken (on a sliding scale that could range between 1 and 20 min). After answering this question, they indicated how long they had felt that the initial trip had lasted, on the same scale.Question order control. This condition was exactly the same as the basic return trip effect condition, except that the order of the estimates was varied. Participants first indicated how long the initial trip had felt and then how long the return trip had felt.Question timing control. This procedure was the same as that of the basic return trip effect condition, but participants indicated how long the initial trip felt directly after viewing the initial trip. After they saw the return trip, they indicated how long they felt the return trip had taken.Different return trip. The procedure was exactly the same as that of the basic return trip effect, but the video of the return trip showed a different (but equally long) route on the return trip.Expectancy manipulation before initial trip. The procedure was exactly the same as that of the basic return trip effect, but before the participants saw the initial trip, they were told that they would first read what a previous participant had written about his experience after watching the video. They were presented with a handwritten statement, which was scanned in and presented on the screen. It first read the instruction to that other participant: "Please write down what your experience was (what you thought or felt) when you watched the video of the student riding her bike." It was answered with: "phewwww, that video took a lot longer than I expected." Results {#Sec7} ------- We conducted a mixed-model Type III ANOVA with time estimates of the initial and return trips as within-subjects variables and the five conditions described above as the between-subjects variable. Across conditions, a clear return trip effect existed. Participants felt that the initial trip took a lot longer (M = 9.12 min, SD = 3.54) than the return-trip did (M = 7.35, SD = 2.59), F(1, 134) = 70.35, p \< .001, η ~p~^2^ = .34, even though both took exactly 7 min. An interaction effect indicated that differences existed between conditions, F(4, 134) = 2.97, p = .022, η ~p~^2^ = .08 (see Table [1](#Tab1){ref-type="table"} for full results). In the three basic control conditions (where only the question order or timing was manipulated), a significant return trip effect existed. Note that an analysis of the difference between the estimates of the initial and the return trips showed that asking for a time estimate of the initial trip directly after that trip had been seen reduced the return trip effect, as compared with the basic return trip effect condition (p = .061) and the question order control condition (p = .042). The latter conditions did not differ (p = .972). This demonstrates that the order in which questions about the time estimates of the initial or return trip are asked does not eliminate the return trip effect, although asking for a time estimate directly after the initial trip does seem to reduce it somewhat.Table 1Return trip effect per condition in [Study 3](#Sec5){ref-type="sec"}Initial Trip (minutes)Return Trip (minutes)Return Trip EffectM(SD)M(SD)Paired t(28)pr ^2^Basic return trip effect9.54(2.93)7.14(2.21)5.01\< .001.47Question order control9.36(4.02)6.86(2.24)5.52\< .001.52Question timing control9.21(3.99)8.07(3.13)2.50.019.18Different return trip9.39(3.68)7.29(2.58)4.54\< .001.42Expectancy manipulation before initial trip8.04(2.94)7.37(2.72)1.34.192.06 A conservative way of testing the return trip effect is to compare the time estimates of the participants who estimated the duration of the initial trip directly after seeing it (in the question timing control condition, M = 9.21 min, SD = 3.99) with those of the participants who made an estimate of the return trip directly after seeing the return trip (in the basic return trip effect condition, M = 7.14 min, SD = 2.21). This between-subjects analysis also strongly confirms the existence of the return trip effect, t(54) = 2.40, p = .020, d = 0.64. This shows that the return trip effect is not only a within-subjects phenomenon, and it rules out the possibility that the pattern exists because people might have a lay theory regarding a return trip that they wish to confirm. Given that there was a return trip effect and question order did not appear to matter, we next examined the effect of familiarity with the return trip. One group of participants watched the rider travel back via a different (but equally long) route, while the remaining participants saw the rider return by the same initial route. As in the previous study, the return trip effect remained when participants returned by a different route: The different return trip group did not differ from the basic control (p = .667), the question order control (p = .554), or the question timing control (p = .147) group. Being familiar with the return trip is not necessary for the return trip effect to occur, nor does it seem to influence the strength of the effect. Recall that in the previous two studies, the more participants felt that the initial trip took longer than expected, the stronger they experienced the return trip effect. Therefore, in the expectancy manipulation condition, participants were led to believe that the initial trip would take a long time so that there would not be a violation of expectations. If the return trip effect is due to a violation of initial expectations for a short trip, lengthening participants' expectations for the initial trip should lessen or eliminate the return trip effect. Indeed, in this condition, no return trip effect existed, since the estimates of how long the initial and return trips felt did not differ (see Table [1](#Tab1){ref-type="table"}). Ironically, when a manipulation made participants expect a longer initial trip, they actually experienced the trip as taking less time, as compared with participants in the other conditions, as indicated by a planned contrast t(134) = 1.76, p = .08, d = 0.31. This is further support for the idea that one of the causes of the return trip effect is that people are generally disappointed in the initial trip, after which the return trip seems relatively short again. Conclusion {#Sec8} ========== These studies demonstrate that there is a return trip effect; a pretest shows that people regularly experience the return trip as being shorter than the initial trip, and the effect was found on a bus trip ([Study 1](#Sec1){ref-type="sec"}), a bicycle trip ([Study 2](#Sec2){ref-type="sec"}), and while a video of someone else traveling by bike was watched ([Study 3](#Sec5){ref-type="sec"}). The return trip effect is quite large: In [Study 2](#Sec2){ref-type="sec"}, the return trip was felt to be 17% shorter in duration than the initial trip; in [Study 3](#Sec5){ref-type="sec"}, it was felt to be 22% shorter (across the four conditions in which an effect was predicted). The results indicate that the return trip effect is not due to an increase in familiarity, since the return trip effect also exists when people travel a different but equally long trip back. Instead, the return trip effect is likely due to a violation of expectations. Participants felt that the initial trip took longer than they had expected. In response, they likely lengthened their expectations for the return trip. In comparison with this longer expected duration, the return trip felt short. The greater the participants' expectations were violated on the initial trip, the more they experienced the return trip effect (Studies 1 and 2). In [Study 3](#Sec5){ref-type="sec"}, where participants' expectations for the duration of the initial trip were increased via a manipulation, the return trip effect disappeared. It is possible that other causes exist for the return trip effect and that it is a multidetermined phenomenon. The main goal of the present research was to test whether a return trip effect exists and to test two possible explanations. It provides a starting point for subsequent research that examines the role of other aspects, such as motivation, valence of the trip and destination, and the effect of learning on the return trip effect---what strengthens it and what the boundary conditions are. Finding that a violation of expectations (at least partly) causes the return trip effect does allow us to make new predictions related to the return trip effect. One of our personal observations on the return trip effect is that it does not seem to occur for routes traveled frequently, such as commuting to work. It is likely that for these routes, the expectations become more accurate with repeated feedback (Roy, Mitten, & Christenfeld, [@CR16]), which attenuates the return trip effect. More generally, the findings on the return trip effect could also help us predict people's time estimates on other repeated tasks. For example, we predict that people who watch a movie for a second time are likely to perceive the second viewing as taking less time (especially if the first viewing seemed overly long). Indeed, one of the authors noted a similar effect when reading a story to his children the second time. These findings on the return trip effect can thus help us make new predictions on how people experience the duration of tasks unrelated to traveling as well. How long the trip took had no influence on any of our measures, all βs \< .15, ps \> .2. Eight participants were excluded from the analysis, since they got lost during the return trip. Three more were deleted because a multivariate outlier analysis indicated that their responses deviated strongly from the responses of the other participants. The authors thank Leanne Braber and Terri Seuntjens for their help in conducting the studies and Veolia and OAD for providing the opportunity to contact their travelers in [Study 1](#Sec1){ref-type="sec"}. Part of this research was made possible by a Rubicon grant (446-09-013) provided by the Netherlands Organization for Scientific Research (NWO). Open Access {#d29e999} =========== This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ The presence of dendritic cells (DC) in numerous human [@pone.0017515-Vicari1] and murine tumors [@pone.0017515-Chiodoni1], [@pone.0017515-Stoitzner1] is well established. The role of these tumor-infiltrating DC (TIDC) in the tumor-specific immune response, and their value as indicators of disease progression, are, however, unclear [@pone.0017515-Iwamoto1], [@pone.0017515-Sandel1]. A number of studies have shown that TIDC have poor tumor antigen presenting function in vitro and in vivo [@pone.0017515-Stoitzner1], [@pone.0017515-Chaux1]--[@pone.0017515-Herber1]. Tumors represent an immunosuppressive environment containing a range of inhibitory mechanisms such as decreased inflammatory cytokines, increased anti-inflammatory cytokines [@pone.0017515-Engleman1], [@pone.0017515-Marincola1] and increased Treg infiltration [@pone.0017515-Bui1], [@pone.0017515-Grauer1], which are likely to affect the function of local T cells [@pone.0017515-Chen1] as well as DC [@pone.0017515-Vicari2]. The term "regulatory T cells" (Treg) refers to a range of cells which express distinct phenotypes but share a common suppressive function [@pone.0017515-Shevach1]. Among these suppressive populations, CD4^+^CD25^+^ "natural" Treg (from now on referred to as "Treg") are prominent due to their essential role in the maintenance of self-tolerance. Treg require expression of the transcription factor Foxp3 for their normal development in the thymus, and are thought to require antigen specific activation by DC in order to acquire effector function in the periphery [@pone.0017515-Billiard1], [@pone.0017515-Shevach2]. Mice in which Treg function is defective develop severe autoimmunity that can be prevented by the transfer of CD4^+^CD25^+^ T cells [@pone.0017515-Asano1], [@pone.0017515-Kim1]. Treg suppress T cell proliferation and degranulation, inhibit CTL function, and may cause T cell death through production of the anti-inflammatory molecules adenosine, transforming growth factor (TGF)-β and interleukin (IL)-10, and inhibition of IL-2 transcription in T cells (reviewed in [@pone.0017515-Shevach2]). It has further been proposed that Treg may cause cytokine deprivation-induced apoptosis of target T cells, or even directly kill target cells using granzyme B and perforin [@pone.0017515-Shevach2], [@pone.0017515-Grossman1], [@pone.0017515-Grossman2]. Treg have been shown to inhibit production of inflammatory cytokines such as Interferon-γ [@pone.0017515-Oberle1], Tumor Necrosis Factor-α [@pone.0017515-vanMierlo1] and the cytolytic granule proteins perforin and granzymes [@pone.0017515-Yang1]. In addition to their suppressive effect on T cells, Treg may also suppress macrophages, natural killer (NK) cells, B cells [@pone.0017515-Shevach2] as well as DC. Studies in non-obese diabetic mice have shown that Treg can inhibit the expression of the DC activation markers CD40, CD80, CD86 and MHCII, both in vitro and in vivo [@pone.0017515-Shevach2], [@pone.0017515-Cederbom1], [@pone.0017515-Serra1], and interact directly with DC during immune responses [@pone.0017515-Tang1], decreasing the interaction time between effector T cells and DC [@pone.0017515-Tadokoro1]. A number of studies have also shown that Treg indirectly control DC homeostasis in vivo [@pone.0017515-Kim1], [@pone.0017515-DarrasseJeze1], [@pone.0017515-Liu1]. In tumor bearing mice, Treg have been shown to induce DC death in the lymph node (LN) [@pone.0017515-Boissonnas1], but little information is available on whether Treg may affect the number, phenotype or function of DC within the tumor context. Effects on DC antigen uptake and/or function might result in diminished T cell activation and effector differentiation within the tumor [@pone.0017515-Thompson1]. Effects on DC migration and/or function might also lead to decreased antigen presentation in the draining LN. In this paper we use a B16.OVA melanoma model to investigate and report the effects of Treg depletion on the antigen presenting function of TIDC in vitro and in vivo. Materials and Methods {#s2} ===================== Ethics statement {#s2a} ---------------- All experimental procedures were approved by the Victoria University Animal Ethics Committee (permits 2004R6M and 2007R4M) and carried out according to Institutional guidelines. Mice {#s2b} ---- All mice were bred at the Malaghan Institute of Medical Research Biomedical Research Unit. C57BL/6J mice were originally from Jackson Laboratories, Bar Harbor, ME, while CD45-congenic B6.SJL-Ptprc^a^Pep3^b^/BoyJArc (CD45.1) were from the Animal Resource Centre, Canning Vale, Western Australia. OTI and OTII mice expressing transgenic T cell receptors (TCR) specific for K^b^+OVA~257--264~ and I-A^b^+OVA~323--339~, respectively, were obtained with the permission of F. Carbone, Melbourne University, Australia. Foxp3GFP mice [@pone.0017515-Fontenot1] were obtained from Prof. A. Rudensky, University of Washington, USA; hemizygous males were used in all experiments. B6.129-Rag1^tm1^Mom mice (RAG1^−/−^) were from the Walter and Eliza Hall Institute, Melbourne, Australia. Mice were used when 6--8 weeks old and gender-matched within experiments. Media and reagents {#s2c} ------------------ Cells were cultured in Iscove\'s Modified Dulbecco\'s Medium (IMDM) supplemented with 5% Fetal Bovine Serum (FCS), 100 U/ml penicillin, 100 µg/ml streptomycin and 50 µM 2-mercapto-ethanol (all from Invitrogen, Auckland, NZ). Synthetic peptides were from Mimotopes Pty Ltd (Clayton, Victoria, Australia). Antibodies and flow cytometry {#s2d} ----------------------------- Monoclonal antibodies specific for murine CD11c, MHCII, F4/80, CD25 and CD3 were affinity-purified from hybridoma culture supernatants using protein G-Sepharose (Pharmacia Biotech, Uppsala, Sweden) and were left purified or conjugated to various fluorophores. Fluorescent antibodies specific for CD45, CD8α CD11b, CD11c, MHCII, CD86, CD40 and CD25 were from BD-Pharmingen (San Diego, USA). Anti-CD45 and anti-Foxp3 intracellular staining kits were from eBioscience (San Diego, USA). All reagents were used according to manufacturer\'s instructions. Live cells were identified by FSC and SSC properties. Treg depletion {#s2e} -------------- Mice were given two intraperitoneal (i.p.) injections of 100 µg purified PC61, 3 days apart [@pone.0017515-Prasad1]. Where applicable, tumor inoculation was carried out one day following the last PC61 injection. Treg depletion was estimated by flow cytometry of tail blood samples. Isolation of T cells {#s2f} -------------------- LN from OTI or OTII mice were pressed through a 70 µm cell strainer (BD Falcon, USA). CD8^+^ T cells were positively selected from OTI cell suspensions using anti-CD8α-MACS beads (Miltenyi Biotec, Germany) and magnetic sorting. Effector CD4^+^ T cells were first depleted of CD25^+^ cells by incubation with anti-CD25-PE followed by anti-PE MACS beads and negative magnetic selection, followed by positive selection with anti-CD4-MACS beads. OTI and OTII cell populations were routinely \>95% and ∼80% pure, respectively. Tumor experiments {#s2g} ----------------- The B16.OVA cell line [@pone.0017515-Lugade1] was kindly provided by Drs. Roslyn Kemp and Dick Dutton, Trudeau Institute, USA, while B16.F1 was from ATCC, Manassas, VA, USA. Mice were injected subcutaneously (s.c.) with 10^5^ tumor cells into the flank, and tumor growth was measured every 2 days using calipers. Tumors were removed, weighed, dissociated using tweezers and digested with 0.4 mg/ml Liberase CI (Roche Applied Science, Mannheim, Germany) and 0.1 mg/ml DNase I (Roche) for 30 min at 37°C. Digestion was stopped with 10 mM EDTA for 5 min at 37°C and suspensions pressed through 70 µm cell strainers. Isolation and quantification of TIDC {#s2h} ------------------------------------ Tumor cell suspensions were prepared, counted using trypan blue, and the frequency of live cells that were CD45^+^, CD11c^hi^ was determined by flow cytometry. For sorting experiments, leukocytes were enriched to ∼20% using anti-CD45-PE, anti-PE MACS beads, and magnetic selection. Cells were then incubated with anti-CD11c and electronically sorted to obtain a population that was \>96% pure. In vitro suppression assay {#s2i} ---------------------------- Tumor cell suspensions from Foxp3GFP mice were enriched for CD4^+^ cells using anti-CD4-MACS beads and magnetic selection. Cells were then incubated with anti-CD45-PE, and GFP^+^CD45^+^ cells were electronically sorted to approximately 98% CD45^+^GFP^+^. These Treg were cultured at differing ratios with a constant number of DC (2.4×10^3^/well), CD4^+^ CD25^−^ effector T cells (4×10^4^/well), and 1 µg/ml anti-CD3 for 3 days. ^3^H-thymidine (1 mCi/ml, Amersham, Aylesbury, UK) was added during the last 6 h of culture before harvesting on a Tomtec cell harvester (Orange, CT, USA) and counting on a Betacounter (Wallac, Turku, Finland) to determine the amount of proliferation. In vitro proliferation assays {#s2j} ------------------------------- TIDC were sorted and titrated in duplicate into 96 well U bottom plates containing 2×10^5^ purified OTI or OTII T cells in a total volume of 200 µL. After 3 days, 1 µCi ^3^H-thymidine was added to each well for 6 hours. Cells were harvested and counted as above. Carboxyfluorescein succinimidyl ester (CFSE) labeling {#s2k} ----------------------------------------------------- Single cell suspensions (5×10^6^ cells/ml) were incubated for 10 min at 37°C with 0.2 mM CFSE (Molecular Probes, Eugene, Oregon). The reaction was stopped by adding one volume of FBS. Cells were washed once with complete media and twice with PBS. In vivo proliferation assays {#s2l} ------------------------------ B6.SJL mice were inoculated with tumor and 13 days later were injected s.c. in the forearm with 2×10^5^ DC that were loaded with 1 uM OVA~257--264~ (SIINFEKL) or left untreated. One day later, mice were injected i.v. with 1.5×10^6^ OTI and 1.5×10^6^ OTII T cells labeled with CFSE. Tumor-draining and non-draining LN were removed 3 days after T cell transfer, and analyzed for T cell proliferation by flow cytometry. Results {#s3} ======= Tumor-derived CD25^+^ Foxp3^+^ Treg suppress T cell proliferation in vitro and are depleted by PC61 treatment {#s3a} --------------------------------------------------------------------------------------------------------------- To establish whether Treg were present in B16.OVA tumors, we used Foxp3GFP mice where natural Treg can be easily identified by Green Fluorescent Protein (GFP) expression ([Figure S1](#pone.0017515.s001){ref-type="supplementary-material"} and [@pone.0017515-Fontenot1]). In B16.OVA tumors, ∼14% of the CD4^+^ T cell population was Foxp3GFP^+^, as opposed to the ∼10% observed in the tumor-draining and non-draining LN ([Figure 1A and 1B](#pone-0017515-g001){ref-type="fig"}), and their frequency increased during tumor growth ([Figure 1C](#pone-0017515-g001){ref-type="fig"}). This Foxp3GFP^+^ population may also include "induced" Treg, but is unlikely to include Tr1 cells which are Foxp3GFP^−^ and CD25^−^ [@pone.0017515-Shevach1]. ![Tumor-infiltrating Foxp3^+^ Treg are suppressive in vitro and affect tumor growth.\ Foxp3GFP mice were treated with PC61 or left untreated, and injected with B16.OVA tumors s.c. Tissues were removed for analysis at different times after tumor challenge. (A) The frequencies of Treg in tumors from non-depleted mice, or mice depleted of Treg by PC61 treatment, were determined by flow cytometry. Each panel refers to an individual representative mouse. (B, C) Frequencies of Foxp3GFP^+^ cells in different tissues (B) and in tumors of different sizes (C), determined as shown in panel A. Each dot represents one mouse. Data are from 2 experiments, each with 4--5 mice per group, collected 14--17 days after tumor inoculation. Average values are shown by a horizontal line, p values were calculated using one-way ANOVA. (D) Mice were treated with PC61 (solid arrows) or left untreated, and injected with B16.OVA (arrow with broken line). Mice were bled over time to monitor the frequency of Foxp3GFP^+^ Treg within the peripheral CD4^+^ population. Average ± SEM for groups of 5 mice are shown. (E) CD4^+^ Foxp3GFP^+^ Treg were sorted from tumors and titrated into wells containing constant numbers of purified CD4^+^CD25^−^ effector T cells, DC, and anti-CD3. Proliferation was measured 3 days later. Bars represent average ± range for duplicate samples. (F) C57BL/6 mice were treated with PC61 as in D or left untreated, and inoculated with tumor. Average tumor sizes ± SEM are shown. Results are from one of 4 repeat experiments that gave similar results. Values of p (where \\<0.05 and \\\\<0.001) were calculated using a non-parametric one-way ANOVA with a Dunn\'s post-test.](pone.0017515.g001){#pone-0017515-g001} In tumor-bearing mice, PC61 treatment routinely achieved a 40--70% depletion of the Foxp3GFP^+^ population in blood, which was maintained throughout the experiment ([Figure S1A](#pone.0017515.s001){ref-type="supplementary-material"} and [Figure 1D](#pone-0017515-g001){ref-type="fig"}). The remaining Foxp3GFP^+^ population was essentially CD25^−^ until the very final stages of tumor growth ([Figure S1B](#pone.0017515.s001){ref-type="supplementary-material"}). The intratumoral Foxp3GFP^+^ population was still reduced in PC61-treated mice compared to untreated mice at day 17 ([Figure 1A and 1B](#pone-0017515-g001){ref-type="fig"}). As also shown by others [@pone.0017515-Shevach2], [@pone.0017515-Curiel1], tumor derived Foxp3GFP^+^ cells were able to suppress the proliferation of conventional T cells in vitro ([Figure 1E](#pone-0017515-g001){ref-type="fig"}). Treatment with PC61 in vivo resulted in a significant delay in tumor growth, and by day 20 tumors in Treg-depleted mice were only about one third of the size in the non-depleted group ([Figure 1F](#pone-0017515-g001){ref-type="fig"}). Together, these data suggest that Treg are present in B16.OVA tumors from an early stage, and may contribute to their rapid growth. Treg do not affect the frequency or phenotype of DC in B16.OVA tumors and in draining LN {#s3b} ---------------------------------------------------------------------------------------- Since Treg are present at relatively high frequencies in both tumor tissue and LN ([Figure 1B](#pone-0017515-g001){ref-type="fig"}), we hypothesized that one of their targets for suppression would be resident DC. TIDC were identified as CD45^+^ and CD11c^hi^, and comprised CD11b^hi^ and CD11b^int^ subpopulations ([Figure 2A](#pone-0017515-g002){ref-type="fig"}). Both subpopulations expressed CD40, CD86 and MHCII, although expression was highest on CD11b^hi^ DC. In addition to the CD11c^hi^ subpopulations, tumors also contained a population of cells expressing markers consistent with plasmacytoid DC [@pone.0017515-Stoitzner1]; these cells represented only a small proportion of TIDC and were not further examined. ![Treg depletion does not affect the frequency or phenotype of TIDC and DC in LN.\ C57BL/6 mice were treated with PC61 or left untreated, and injected with B16.OVA s.c. Tumors and LN were removed for analysis at different times after tumor challenge. (A) Gating strategy used to identify TIDC (CD45^+^, CD11c^hi^, CD11b^hi/int^) and assess expression of CD40, CD86 and MHCII. Fluorescence-minus-one controls (CD40) and isotype controls (CD86 and MHCII) are shown as grey filled histograms, while empty histograms show marker expression. Percentages of cells expressing the relevant markers are shown. (B) Frequencies of DC in tumors, expressed as % CD11c^hi^ cells in the CD45^+^ population. Each dot corresponds to one mouse. Data are from 3 experiments, each with 4--5 mice per group. (C) Numbers of DC per mg of tumor when tumors in both groups are of similar size (200--300 mg). Bars show the average number of DC+SE from 2 experiments each with 3--4 mice per group. The average tumor size in the two groups was similar. (D) Percentages of TIDC expressing the indicated maturation markers in Treg-depleted and non-depleted mice. Bars show the average+SEM for a compilation of 3 independent experiments each with 5 mice per group. (E) Expression of maturation markers on DC in untreated mice, and in Treg-depleted and non-depleted tumor-bearing mice. Both the tumor-draining and non-draining LN were examined. Averages+SEM from 3 independent experiments each with 5 mice per group are shown.](pone.0017515.g002){#pone-0017515-g002} As shown in [Figure 1F](#pone-0017515-g001){ref-type="fig"}, B16.OVA tumors grew more slowly in Treg-depleted mice. To obtain tumors of similar size, Treg-depleted and non-depleted mice were injected with tumor 1--2 days apart. When TIDC were compared between these two groups, no difference could be observed in DC frequency, number per milligram of tumor tissue, or phenotype ([Figures 2B, 2C, 2D](#pone-0017515-g002){ref-type="fig"}). Results shown in [Figures 2B, 2C, and 2D](#pone-0017515-g002){ref-type="fig"} refer to CD11b^hi^ cells, but similar results were obtained with the CD11b^int^ population (data not shown). CD40, CD86 and MHCII were similarly expressed in the TIDC from both treatment groups, both as percentage of positive cells ([Figure 2D](#pone-0017515-g002){ref-type="fig"}) and as Mean Fluorescence Intensity (MFI) in the total population (data not shown). The relative frequencies of CD11b^hi^ and CD11b^int^ TIDC subpopulations were also similar in untreated and Treg-depleted mice (data not shown). Treg depletion did not affect the number or maturation status of DC in the LN, nor was there a difference in maturation status between DC in the draining and non-draining LN, both when measured as percent positive cells or as MFI ([Figure 2E](#pone-0017515-g002){ref-type="fig"} and data not shown). Thus, Treg depletion does not appear to affect the numbers or phenotype of DC in the tumor, or DC phenotype in the tumor-draining LN. Treg depletion does not improve the ability of TIDC to induce proliferation of tumor-specific T cells in vitro {#s3c} ---------------------------------------------------------------------------------------------------------------- CD45^+^CD11c^+^ cells, including both CD11b^hi^ and CD11b^int^ DC subpopulations, were electronically sorted from tumor cell suspensions to a high purity (\>96%) and used to stimulate the proliferation of OVA-specific CD4^+^ OTII and CD8^+^ OTI T cells in vitro. To test the presentation of tumor antigen taken up by DC within the tumor context, no further antigen was added to the assay except in the positive controls. As previously reported [@pone.0017515-Stoitzner1], in the absence of added antigen, sorted DC were unable to stimulate proliferation of OVA-specific OTII cells ([Figure 3A](#pone-0017515-g003){ref-type="fig"}) and induced only minimal proliferation of OVA-specific OTI cells ([Figure 3B](#pone-0017515-g003){ref-type="fig"}). The low proliferation of OTI cells was not antigen-specific, as it was observed even when DC were prepared from B16.F1 tumors that do not express OVA [@pone.0017515-Stoitzner1]. Sorted DC were otherwise able to present synthetic peptide to OTI and OTII cells [@pone.0017515-Stoitzner1], or added OVA protein to OTII cells (data not shown), to induce proliferation in vitro. When TIDC were prepared from the tumors of Treg-depleted mice, no increase in the proliferation of OTI or OTII cells was observed ([Figure 3A and 3B](#pone-0017515-g003){ref-type="fig"}). ![Treg depletion does not increase the ability of TIDC to stimulate T cell proliferation ex vivo.\ C57BL/6 mice were treated with PC61 or left untreated, and injected with B16.OVA s.c. After 14--17 days CD45^+^ CD11c^hi^ TIDC were sorted and titrated in triplicate into cultures containing (A) OTII T cells, (B) OTI T cells or (C) BALB/c allogeneic T cells. As a positive control, specific peptide antigen was loaded on 10^3^ TIDC/well before co-culture with T cells where indicated. Proliferation was measured 3 days later. Each panel shows one of 3 independent experiments that gave similar results. Average ± SEM are shown. p was calculated using a two-way ANOVA test with a Bonferroni post-test.](pone.0017515.g003){#pone-0017515-g003} Although TIDC were unable to present OVA taken up within the tumor context, they appeared functional as they could stimulate the proliferation of specific T cells in the presence of peptide ([Figure 3A and 3B](#pone-0017515-g003){ref-type="fig"}), and proliferation of allogeneic BALB/c T cells in vitro ([Figure 3C](#pone-0017515-g003){ref-type="fig"}). In both cases, Treg depletion did not improve the response. Treg depletion does not improve the proliferation of tumor-specific OTI T cells in the tumor-draining LN {#s3d} -------------------------------------------------------------------------------------------------------- Since removing DC from the tumor environment might reverse the effects of Treg on DC in situ, we also examined the presentation of tumor-derived antigen using an in vivo assay. Presentation of tumor antigen in the draining LN is thought to reflect migration of DC from the tumor site, which is known to be defective [@pone.0017515-Vicari1] but might improve if the frequency of Treg in the tumor is decreased. Presentation in the LN may also reflect transfer of antigen from migratory to resident DC [@pone.0017515-Carbone1], which would bypass antigen presentation defects in tumor-derived DC. Naïve, CFSE-labeled OTI T cells were transferred into Treg-depleted or non-depleted tumor-bearing mice, and their proliferation in tumor-draining LN was compared 3 days later. A representative dot plot of OTI proliferation in tumor-bearing mice is shown in the top panel of [Figure 4A](#pone-0017515-g004){ref-type="fig"}. As a control, we used non tumor-bearing mice immunized with DC+OVA~257--264~; a representative dot plot of OTI proliferation in the LN draining the DC injection site is shown in the lower panel of [Figure 4A](#pone-0017515-g004){ref-type="fig"}. Little OTI T cell proliferation was observed in non tumor-bearing mice that received DC not loaded with OVA~257--264~ ([Figure 4B](#pone-0017515-g004){ref-type="fig"}). A high level of proliferation was detected in some of the tumor-bearing mice, which was antigen-specific [@pone.0017515-Stoitzner1], but always lower in extent than the proliferation observed in the positive controls ([Figure 4A](#pone-0017515-g004){ref-type="fig"}). This proliferation was not increased in magnitude ([Figure 4B](#pone-0017515-g004){ref-type="fig"}) or frequency ([Figure 4C](#pone-0017515-g004){ref-type="fig"}) by Treg depletion. Surprisingly, the proportion of mice where T cells proliferated was even decreased in Treg-depleted mice, however, this decrease was not statistically significant (p = 0.17 by a Fishers exact probability test) and further experiments are required to establish the reproducibility of this observation. High proliferation was observed in all mice injected with OVA~257--264~-loaded DC, indicating that the transferred T cells were functional and were not adversely affected by the presence of tumor, or by PC61 treatment ([Figure 4B](#pone-0017515-g004){ref-type="fig"}). ![Treg depletion does not affect the proliferation of tumor-specific T cells in vivo.\ C57BL/6 mice were treated with PC61 or left untreated, and injected with B16.OVA s.c. After 13--16 days each mouse was injected with 1.5×10^6^ naïve, CFSE-labeled OTI T cells. LN were removed 3 days later and OTI T cell proliferation was determined by flow cytometry. (A) Representative dot plots of proliferating OTI T cells in the draining LN of tumor-bearing mice (top panel) or non tumor-bearing mice immunized with OVA~257--264~ -loaded DC (bottom panel). The percent divided cells is shown. (B) Division of OTI T cells in LN draining the tumor or DC immunization site. Where both tumor and DC were given, the LN draining the immunization site was examined. Horizontal lines show the average percentages of divided cells. The graph is representative of 4 independent experiments with 5--10 mice per group per experiment. (C) Percentage of mice showing OTI T cell proliferation in the tumor-draining LN. The total number of mice (n) in each group is shown. Data are compiled from 3 separate experiments; ns, not significant by a Fisher\'s exact probability test.](pone.0017515.g004){#pone-0017515-g004} Similar experiments were carried out to examine the proliferation of CD4^+^ OTII cells. As previously reported using both low and highly immunogenic models, no CD4^+^ OTII T cell proliferation was detected in tumor-bearing mice [@pone.0017515-Stoitzner1], [@pone.0017515-Gerner1]. Additional experiments in PC61-treated mice showed that proliferation was not restored by Treg depletion (data not shown). The proliferation of OTI T cells is not increased in the tumor-draining LN of RAG1^−/−^ compared to C57BL/6 mice {#s3e} ---------------------------------------------------------------------------------------------------------------- To determine whether the residual Treg population seen in PC61-treated mice ([Figures 1A and 1B](#pone-0017515-g001){ref-type="fig"}) might be sufficient to affect DC function in tumor-bearing mice, we examined tumor-specific T cell proliferation in RAG1^−/−^ mice, which lack all T cells including Treg. TIDC were present in tumors from RAG1^−/−^ mice at frequencies similar to those in C57BL/6 tumors ([Figure S2A](#pone.0017515.s002){ref-type="supplementary-material"}) and expressed similar, although slightly less mature, activation phenotypes ([Figure S2B](#pone.0017515.s002){ref-type="supplementary-material"}). TIDC phenotype in RAG1^−/−^ mice was not significantly affected by the transfer of purified CD8^+^ T cells or CD4^+^CD25^+^ Treg one day before tumor inoculation (data not shown). To examine the effect of Treg on tumor-specific T cell proliferation in vivo, C57BL/6 and RAG1^−/−^ mice were inoculated with either B16.OVA or OVA-negative B16.F1 tumors. Both tumors grew at a similar rate in the two strains ([Figure 5A](#pone-0017515-g005){ref-type="fig"}) suggesting that in C57BL/6 hosts there was little spontaneous T cell response to the B16 tumors. Tumor-bearing mice received CFSE-labeled OTI and CD25-depleted OTII T cells and OVA-specific proliferation was compared in the tumor-draining ([Figure 5B](#pone-0017515-g005){ref-type="fig"}) and non-draining LN 3 days later. As shown in [Figure S3](#pone.0017515.s003){ref-type="supplementary-material"}, the OTII populations contained low frequencies of Foxp3^+^ T cells that were further decreased by CD25 depletion; about half of the remaining cells expressed the transgenic TCR and were presumably OVA-specific. ![The proliferation of tumor-specific OTI T cells in tumor-bearing C57BL/6 and RAG1^−/−^ hosts is similar.\ C57BL/6 and RAG1^−/−^ mice were injected with B16.OVA or B16.F1 s.c.. After 15 days each mouse was injected with 1.5×10^6^ naïve, CD8-enriched, CFSE-labeled OTI T cells and 1.5×10^6^ naïve, CD25-depleted, CD4-enriched, CFSE-labeled OTII T cells. LN were removed 3 days later and T cell proliferation was determined by flow cytometry. (A) Tumor size at the experimental endpoint (day 18); each dot corresponds to one mouse and the horizontal line shows the average tumor size. (B) Gating strategy used to identify the donor OTI T cell population in the draining LN. Antigen-specific division was determined on the basis of CFSE dilution as shown in the right-hand dot plots. (C) Antigen-specific OTI T cell division in the tumor-draining and non-draining LN of C57BL/6 and RAG1^−/−^ mice. Bars show the average percentage of cells that divided more than 3 times, +SEM. Data is from one experiment with 6--9 mice per group. p was calculated using two-way ANOVA with a Bonferroni post-test.](pone.0017515.g005){#pone-0017515-g005} It has been shown that in lymphopenic hosts OTI T cells undergo 2--3 rounds of homeostatic proliferation in the absence of antigen [@pone.0017515-Kieper1] therefore, to exclude homeostatic proliferation, only OTI cells that had divided more than 3 times were considered. This division was mostly antigen-specific, as it occurred in the LN draining B16.OVA tumors but not in those draining B16.F1. Specific OTI T cell proliferation was not increased in RAG1^−/−^ mice compared to C57BL/6 ([Figure 5C](#pone-0017515-g005){ref-type="fig"}), suggesting that Treg were not inhibiting OTI T cell division in C57BL/6 mice. Similarly, little or no division of OTII T cells was observed in C57BL/6 or RAG1^−/−^ mice injected with either B16.OVA or B16.F1 tumors (data not shown), again suggesting that lack of division was not due to the presence of Treg in C57BL/6 mice. Discussion {#s4} ========== The recruitment of Treg to tumors is an early event in tumor establishment [@pone.0017515-Qin1]. The negative correlation observed between the presence of Treg in tumors and survival outcome [@pone.0017515-Beyer1] suggests that these Treg are probably assisting in the evasion of the immune response. As also reported in other tumor models [@pone.0017515-Grauer1], [@pone.0017515-Curiel1] we found that even small B16.OVA melanoma tumors contained higher frequencies of CD25^+^ Foxp3^+^ Treg than blood, and that tumor-infiltrating Treg had suppressive activity in vitro. Somewhat surprisingly, we also found that depletion of Treg appeared to have no detectable effect on the frequency, phenotype or function of TIDC. A partial but functionally relevant depletion via injection of the anti-CD25 monoclonal antibody PC61, or a more profound depletion by the use of RAG1^−/−^ hosts, both failed to affect TIDC phenotype. Similarly, defective proliferation of tumor-specific CD4^+^ and CD8^+^ T cells in tumor-draining LN, which may reflect presentation of tumor antigen by DC migrating from the tumor to the LN, was not improved in these mice. Therefore, CD4^+^CD25^+^ Treg appeared to have no effect on the presentation of tumor antigen by DC in this tumor model, possibly suggesting that other regulatory populations that are not affected by PC61 treatment were involved. It has been shown that Treg suppress the responses of many cell types, using a variety of mechanisms including direct killing [@pone.0017515-Shevach2], [@pone.0017515-Grossman1], [@pone.0017515-Boissonnas1]. Using the B16.OVA melanoma model, we found no significant difference between the numbers, frequency or relative proportions of TIDC populations in Treg-depleted and non-depleted mice. The frequency and absolute number of DC in the LN were also similar (data not shown). These observations fail to support the possibility that in this tumor model CD4^+^CD25^+^ Treg inhibit DC function by directly or indirectly affecting their survival. Experiments using both in vitro and in vivo models have shown that, in the presence of Treg, DC express decreased levels of the maturation markers CD40, CD80, CD86 and MHCII (reviewed in [@pone.0017515-Shevach2]). The DC subpopulation/s that are involved in presentation of tumor antigen in B16.OVA melanoma tumors have not been precisely identified, however, in other tumor models multiple DC subpopulations can present tumor antigen in the draining LN [@pone.0017515-McDonnell1]. On this basis, we chose to include in our analysis all CD11c^hi^ DC subpopulations in these tissues. While DC in B16.OVA tumors and LN appeared relatively mature, there was no effect of Treg depletion on the proportion of TIDC subpopulations, or on any of the DC activation markers tested. Both TIDC and Treg localized mainly to the peripheral area of B16 tumors and could often be found in close proximity to each other (data not shown), suggesting that the two populations had the potential to interact within the tumor context. Additional experiments showed that Treg depletion also did not rescue or enhance the ability of TIDC to stimulate the proliferation of tumor-specific T cells in vitro, or the ability of DC in tumor-draining LN to stimulate T cell proliferation in vivo. Taken together, our results suggest that Treg had little or no effect on TIDC in this tumor model. Our inability to demonstrate an effect of PC61 treatment on DC function could not be explained by incomplete Treg depletion, as the same results were observed using a RAG1^−/−^ model in which Treg depletion was profound. Transfer of CD8^+^ OTI and CD25-depleted CD4^+^ OTII T cells into RAG1^−/−^ tumor-bearing mice 3 days before analysis provided a model in which significant conversion of CD4^+^ T cells into Treg was unlikely to occur, but CD4^+^ T cells could provide help to DC thereby rescuing potential defects in DC function due to lymphopenic conditions [@pone.0017515-Shreedhar1]. Even in this situation, the proliferation of OTI or OTII cells was not improved, suggesting the presentation of tumor antigen in the draining LN was similar in RAG1^−/−^ hosts and C57BL/6 mice. It is conceivable that the effects of Treg on DC might have been missed in our experiments, if these only occur early during tumor development, at time points where the limited amount of tumor tissue available makes the study of TIDC more difficult. We believe that this is unlikely to be the case, as we and others have shown that as the tumor increases in size, the frequency of Treg increases and the microenvironment becomes more suppressive. Under these increasingly hostile conditions, it is doubtful that suppression could be spontaneously overcome, and any effects that the Treg may have had on the DC at early timepoints are likely to still be evident at later stages of tumor growth. A recent paper used an OVA-expressing 3-methylcholanthrene (MCA)-induced tumor to report that tumor antigen-specific Treg could directly kill DC in the tumor draining LN [@pone.0017515-Boissonnas1]. A detailed comparison of those findings with ours is not possible, as the presence and localization of Treg and DC within tumors, and the ability of tumor-derived DC to present OVA to T cells, were not characterized [@pone.0017515-Boissonnas1]. The observation that Treg rapidly kill DC presenting tumor antigen in the LN might suggest that DC in MCA tumors do not come into contact with Treg, and can therefore survive unaffected until they reach the LN. A differential ability of TIDC from MCA or B16 tumors to present OVA antigen in the context of MHCII, and induce activation of CD4^+^ T cells and Treg, might also contribute to the different findings in these studies. It has been reported that antigen derived from normal tissues can be presented to self-reactive CD4^+^ T cells in the draining LN and elicit a response [@pone.0017515-Hoglund1], [@pone.0017515-Scheinecker1]. In tumors this is not always the case, and reports by other Authors and ourselves [@pone.0017515-Stoitzner1], [@pone.0017515-Gerner2]--[@pone.0017515-vanMierlo2] suggest that TIDC may be unable to induce productive activation of CD4^+^ T cells. A reduced ability of TIDC to activate conventional CD4^+^ T cells might also imply a reduced ability of these DC to activate Treg, as suggested by experiments showing that in inflamed tissue the same DC drive cytokine production by both CD4^+^ Teff and Treg [@pone.0017515-McLachlan1]. If this is the case, the results reported in this study might be explained by hypothesizing that Treg accumulate in the tumor, but fail to recognize the appropriate ligands in the context of DC to become activated to effector function. As it has been shown by us ([Figure 1F](#pone-0017515-g001){ref-type="fig"}) and by other Authors [@pone.0017515-Shimizu1] that Treg depletion delays the growth of B16 tumors ([Figure 1F](#pone-0017515-g001){ref-type="fig"}), and improves the anti-tumor efficacy of both prophylactic [@pone.0017515-Prasad1] and therapeutic [@pone.0017515-Matsushita1] DC immunotherapy, Treg must be suppressing anti-tumor immune responses by mechanisms other than acting on DC. For example, reports using intravital microscopy of mouse LN have demonstrated that Treg can directly act on T cells to suppress immune responses [@pone.0017515-Mempel1]. In our studies, Treg depletion did not improve CD4^+^ or CD8^+^ T cell proliferation, implying that any direct effect of Treg on T cells might have an impact on the quality of the T cell response, rather than the quantity. The defective DC function we and others observe in tumor-bearing mice [@pone.0017515-Stoitzner1], [@pone.0017515-Chaux1], [@pone.0017515-Vicari2], [@pone.0017515-PreynatSeauve1]--[@pone.0017515-Norian1] may then be attributable to other suppressive factors present in the hostile tumor environment, such as other regulatory T cell populations, myeloid-derived suppressor cells, or anti-inflammatory cytokines, which may act directly or indirectly on DC in tumor and/or draining LN. Indeed, a recent study by Herber et al [@pone.0017515-Herber1] showed that DC in tumor bearing hosts upregulate the scavenger receptor A resulting in increased lipid uptake, and impaired antigen presenting function. In conclusion, we show that CD4^+^CD25^+^ Treg appear to have little effect on the number, phenotype and function of TIDC in B16 melanoma, suggesting that the delayed tumor growth observed in Treg-depleted mice is unlikely to be due to improved DC function. The question of whether or not Treg affect DC in a tumor model is complex and highly relevant to designing improved therapeutic anti-tumor vaccines. This study suggests that, in order to be optimally effective, regulatory T cell depletion should be used in association with other forms of immunotherapy Supporting Information {#s5} ====================== ###### Identification of Treg in the blood of untreated and PC61 treated mice. Foxp3GFP mice were treated with PC61 or left untreated, and injected with B16.OVA tumors s.c. (A) The frequencies of Treg in blood from non-depleted mice, or mice depleted of Treg by PC61 treatment, were determined by flow cytometry. The lower panels show the Foxp3GFP^+^ populations as identified in the top panels. Each panel refers to an individual representative mouse. (B) Mice were treated with PC61 or left untreated, and tail bled over time to monitor the frequency of Tregs in blood by flow cytometry. CD25^+^Foxp3^+^ cells were gated as shown in the lower part of panel A. Average ± SEM for groups of 5 mice are shown. (EPS) ###### Click here for additional data file. ###### TIDC frequency and phenotype are similar in C57BL/6 and RAG1^−/−^ mice. C57BL/6 and RAG1^−/−^ mice were injected with B16.OVA tumors s.c. Tumors were removed 17 days later and analyzed by flow cytometry. (A) Frequencies of CD11c^hi^ TIDC in tumors were compared between C57BL/6 and RAG1^−/−^ mice. p was calculated using the Student\'s one tailed t test. (B) Frequencies of TIDC expressing the indicated maturation markers were compared between C57BL/6 and RAG1^−/−^ mice. Values of p were calculated using a two-way ANOVA test with a Bonferroni post-test. Data is from one experiment with 9--10 mice per group. (EPS) ###### Click here for additional data file. ###### Frequencies of Foxp3^+^ T cells in naïve OTI and OTII cell populations before transfer into RAG1^−/−^ hosts. (A) Frequencies of CD8^+^ T cells (top panels) and Foxp3^+^ Treg (lower panels) in the total OTI lymphocyte population were determined before and after enrichment for CD8^+^ cells. (B) Frequencies of CD4^+^ T cells (top panels) and Foxp3^+^ Treg (lower panels) in the total OTII lymphocyte population were determined before and after the cells were depleted of CD25^+^ cells and enriched for CD4^+^ cells. The frequency of OVA-specific (Valpha2^+^, Vbeta5.1,5.2^+^) Foxp3^+^ Treg in the CD4-enriched population is shown in the right lower panel. Foxp3 expression was determined by intracellular staining. (EPS) ###### Click here for additional data file. The authors wish to thank all colleagues at MIMR for useful discussions. The expert animal husbandry of Biomedical Research Unit staff is gratefully acknowledged. Competing Interests:The authors have declared that no competing interests exist. Funding:This work was funded by a research grant from the Health Research Council of NZ and Cancer Society of NZ to FR (HRC 04/249). HA was funded by a PhD scholarship from the Health Research Council of NZ and by the Wellington division of the Cancer Society. PS was supported by the Erwin Schroedinger Auslandsstipendium from the Austrian Science Fund (FWF-J2479). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. [^1]: Conceived and designed the experiments: HA PS FR. Performed the experiments: HA EH KP. Analyzed the data: HA EH KP. Wrote the manuscript: HA FR. [^2]: Current address: Center for Gene Therapy, Baylor College of Medicine, Houston, Texas, United States of America [^3]: Current address: Department of Dermatology and Venereology, Innsbruck Medical University, Innsbruck, Austria	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== Idiopathic inflammatory myopathies (IIM) are systemic autoimmune diseases characterized by myositis and extra muscular features \[[@CR1]\]. Lymphocytes and monocytes are acknowledged as fundamental effector cells of the pathogenic autoimmune response in IIM \[[@CR2], [@CR3]\], since mononuclear cells constitute the main component of the inflammatory infiltrate in muscle biopsies \[[@CR4]\]. Monocytes are the circulating precursors of macrophages and dendritic cells and are characterized by their ability to phagocytize, produce cytokines, present antigens \[[@CR5]--[@CR7]\] and their expression of a wide range of toll-like receptors (TLRs), especially TLR2 and TLR4 \[[@CR8], [@CR9]\]. In patients with dermatomyositis (DM), polymyositis (PM), immune mediated necrotizing myopathy (IMNM) and anti-synthetase syndrome (AS) macrophages and dendritic cells are prominent in muscle biopsies \[[@CR10]\], highlighting the relevance of monocytes in the immunopathology of IIM. Also, the relevance of TLRs in the pathogenesis of inflammatory myopathies has been demonstrated in animal models \[[@CR11]\] and muscle biopsies of these patients \[[@CR12]\]. In subjects with DM and PM, an enhanced expression of TLR2, TLR4 and TLR9 in the endomysial and perimysial inflammatory infiltrate \[[@CR13]\] as well as an overexpression of IFN-γ, IL12p40 and myeloid differentiation factor-88 (MyD88) has been shown in muscle biopsies \[[@CR14]\]. Furthermore, the expression of TLR4 correlated with the amount IFN-γ, IL-4, IL-17 and TNF-α in inflammatory cells invading the muscle \[[@CR13]\], underscoring the relevance of TLR2 and TLR4 as pro-inflammatory effectors in the pathogenesis of IIM. In recent years, different monocyte subsets have been acknowledged according to their expression of the LPS receptor (CD14) and the FcγRIII (CD16) and are classified as classical (CD14++/CD16−), intermediate (CD14++/CD16+) and non-classical (CD14+/CD16++) \[[@CR5], [@CR15]\]. These monocyte subsets are known to be genetically and functionally distinctive \[[@CR16]\], and an expansion of intermediate monocytes has been described in autoinflammatory and autoimmune diseases \[[@CR15]--[@CR18]\]. Notwithstanding the importance of monocytes and TLRs in the pathophysiology of IIM, studies describing the monocytes subsets and their expression of TLR4 and TLR2 in peripheral blood as biomarkers of disease activity are lacking. The aim of this study was to correlate the amount of the distinct monocyte subsets and their expression of TLR2 and TLR4 with the clinical features of patients with IIM. Methods {#Sec2} ======= Clinical evaluation of patients of IIM {#Sec3} -------------------------------------- We recruited 45 Mexican-mestizo adult patients with DM, clinically amyopathic dermatomyositis (CADM), juvenile dermatomyositis (JDM), AS and PM according to the ACR/EULAR, Bohan and Peter, Connor and Sontheimer criteria \[[@CR19]--[@CR22]\] who were followed-up in a tertiary care center (Instituto Nacional de Ciencias Medicas y Nutricion Salvador Zubiran) from 2016 to 2018 and 15 age and sex-adjusted healthy controls. We excluded patients with any kind of acute or chronic infection, pregnancy, puerperium and neoplasia. All healthy controls and patients signed an informed consent before inclusion and the protocol was approved by our institutional ethics committee (Ref. 2152) in compliance with the Helsinki declaration. The following disease activity and damage scales were evaluated by a certified Rheumatologist: manual muscle test 8 (MMT8), patient's and physician's global disease activity with a visual analogue scale (VAS), the cutaneous dermatomyositis disease area and severity index (CDASI), the myositis disease activity assessment tool (MYOACT and MITAX), and the myositis damage index (MDI) \[[@CR23]\]. We registered the type and dose of immunosuppressive therapy. Complete clinical response and remission were defined as the absence of muscular and extra-muscular disease activity for at least six months while taking immunosuppressive therapy or without treatment respectively \[[@CR24]\]. Also, we evaluated the presence of dilated, absent or mega capillaries, hemorrhage, thrombosis and neo-vascularization \[[@CR25]\] with a qualitative nailfold capillaroscopy, which was performed with a 500× capillaroscope. The presence of interstitial lung disease (ILD) as well as the pulmonary and echocardiographic parameters were registered from the medical charts. We assessed antinuclear antibodies with indirect immunofluorescence in all patients and the myositis specific and associated antibodies with the commercial membrane strip for the detection of antigens EUROLINE (Euroimmune AG, Luebeck, Germany). Multiparametric flow cytometry analysis {#Sec4} --------------------------------------- After peripheral blood mononuclear cells (PBMCs) isolation by density gradients with Lymphoprep (Stemcell Technologies, Vancouver, Canada), cells were re-suspended in RPMI with phenol red (Thermo Fisher scientific), washed in PBS with 5% FBS (fetal bovine serum) and stained with the following fluorescent labeled-antibodies: CD14-PerCP (Biolegend, San Diego, CA, USA, catalog number: 325632), CD16-BV605 (Biolegend, San Diego, CA, USA, catalog number: 302040), TLR4-APC (Biolegend, San Diego, CA, USA, catalog number: 312816) and TLR2-BV421 (BD, Biosciences, Franklin Lakes, NJ, USA, catalog number: 565350). The percentage of every monocyte subset was determined according to the International Union of Immunological Societies \[[@CR26]\]. The absolute numbers of each subset were calculated taking into account the number of total monocytes from a complete blood count taken at the time of the blood draw. Also, we evaluated the percentage TLR4+ and TLR2+ monocytes in every monocyte subset as well as their expression of TLR expressed as the mean fluorescence intensity (MFI) in arbitrary units (AU). The gating strategy is depicted in Fig. [1](#Fig1){ref-type="fig"}. The analysis was performed using the Flow-Jo v10 software. The serum levels of IL-17A, IFN-γ, TNF-α, IL-10, IL-6, IL-4, IL-2 were measured with the cytometric bead array (CBA) human Th1/Th2/Th17 cytokine kit (BD, Biosciences, Franklin Lakes, New Jersey, USA). The data were analyzed with the FCAP array software v3.0 (BD, Biosciences, Franklin Lakes, New Jersey, USA).Fig. 1Gating strategy for the monocyte subsets and TLR expression assessment. Classical monocytes were defined as CD14++, CD16−; intermediate monocytes as CD14++, CD16+ and non-classical monocytes as CD14+, CD16++. The expression of TLRs was assessed by measuring the mean fluorescence intensity (MFI) of TLR2 and TLR4 in every monocyte subset Statistical analysis {#Sec5} -------------------- We expressed quantitative variables as medians and interquartile ranges (IQR) or minimum and maximum (min--max). Differences between groups were assessed with Mann--Whitney U and the Kruskal--Wallis tests. Correlation between quantitative variables was assessed with Spearman Rho. The analysis was also performed adjusting by prednisone use. To evaluate the usefulness of monocyte subsets and their TLRs expression for the diagnosis of active inflammatory myopathies and to identify patients in complete clinical response we created receiving operating characteristic (ROC) curves, with sensitivity, specificity, area under the curve and 95% confidence intervals (95% CI). A P value \< 0.05 was considered as statistically significant. The statistical analysis was performed with the support of the SPSS v25 software (IBM Corp. Armonk, NY, USA). Results {#Sec6} ======= Clinical characteristics of patients with IIM {#Sec7} --------------------------------------------- Twenty-nine patients were women (64.4%). The median (IQR) of age was 46 (37--59) years. Thirty-two patients had DM (71.1%), 3 (6.7%) PM, 4 (8.9%) anti synthetase syndrome and 6 (13.3%) were adults with previous diagnosis of JDM. Regarding immunosuppressive therapy, 29 patients were receiving prednisone (64.4%), 20 (44.4%) methotrexate, 16 (35.6%) azathioprine, 14 (31.1%) hydroxychloroquine, 2 (4.4%) mycophenolate mofetil and 1 (2.2%) cyclophosphamide at the time of blood draw. Immunosuppressants combination was used in 18 patients (40%) and consisted exclusively in the simultaneous use of methotrexate and azathioprine. At the time of evaluation, 30 patients (66.7%) had cutaneous features, 26 (57.8%) had an abnormal nailfold capillaroscopy, 9 (20%) had interstitial lung disease, 7 (15.6%) calcinosis and 5 (11.1%) dysphagia. Nine patients (20%) were in complete clinical response and none of them in remission. The most frequent auto-antibody was anti-Mi2 (22.2%), followed by anti-Ro52 (8.9%). In Table [1](#Tab1){ref-type="table"}, we depict the main disease activity and damage parameters of the patients with IIM.Table 1Clinical, laboratory and cardio-pulmonary features of patients with idiopathic inflammatory myopathiesVariableMedian (min--max)Disease activity and damage Manual muscle test 8 (MMT8)144 (45--150) Visual analogue scale of physician's disease activity5 (0--10) Visual analogue scale of patient's disease activity5 (0--10) Cutaneous dermatomyositis disease area and severity index (CDASI) acute4 (0--76) Cutaneous dermatomyositis disease area and severity index (CDASI) chronic2 (0--21) Visual analogue scale of constitutional disease activity0 (0--10) Visual analogue scale of cutaneous disease activity0 (0--10) Visual analogue scale of pulmonary disease activity0 (0--10) Visual analogue scale of cardiovascular disease activity0 (0--10) Visual analogue scale of other disease activity0 (0--10) Visual analogue scale of extramuscular disease activity3 (0--10) Visual analogue scale of muscular disease activity0 (0--10) Visual analogue scale of global disease activity5 (0--10) Total myositis disease activity assessment visual analogue scales (MYOACT)1.4 (0--7.5) Total myositis intention to treat activity index (MITAX)0.85 (0--5.71) Visual analogue scale of muscular damage0 (0--10) Visual analogue scale of skeletal damage0 (0--9) Visual analogue scale of cutaneous damage1 (0--10) Visual analogue scale of gastrointestinal damage0 (0--10) Visual analogue scale of pulmonary damage0 (0--8) Visual analogue scale of cardiovascular damage0 (0--10) Visual analogue scale of vascular damage0 (0--5) Visual analogue scale of endocrine damage0 (0--10) Visual analogue scale of ocular damage0 (0--10) Visual analogue scale of infection damage0 (0--10) Visual analogue scale of malignancy damage0 (0--5) Visual analogue scale of other damage0 (0--10) Visual analogue scale of global damage5 (0--10) Damage extension0.04 (0--0.52) Damage severity0.045 (0--0.5) Extended damage0 (0--10) Health assessment questionnaire (HAQ)0 (0--3)Treatment Prednisone dose (mg/day)15 (2--100) Methotrexate dose (mg/week)20 (2.5--30) Azathioprine dose (mg/day)75 (50--175) Mycophenolate mofetil dose (g/day)1.5 (0.5--2.5) Anti-malarial dose (mg/day)200 (150--400)Laboratory and cardio-pulmonary features Creatine phosphokinase (U/L)170 (10--13,325) Aldolase (U/L)8.7 (6.2--131) Alanine aminotransferase (U/L)31 (5--435) Aspartate aminotransferase (U/L)30 (10--1441) Lactate dehydrogenase (U/L)226 (45--1243) C-reactive protein (mg/dL)0.26 (0.01--13.3) Erythrocyte sedimentation rate (mm/Hr)8 (1--56) Percentage of predicted forced vital capacity81 (45--108) Pulmonary artery systolic pressure (mmHg)32 (20--62) Percentage of left ventricle ejection fraction61 (37--76) Tricuspid annular plane systolic excursion (TAPSE)20 (11--26) Patients with IIM are characterized by a differential profile of circulating monocyte subsets {#Sec8} --------------------------------------------------------------------------------------------- As shown in Fig. [2](#Fig2){ref-type="fig"}, healthy donors had a higher percentage of classical monocytes in comparison to patients with IIM and active disease either using prednisone (90.5% (86.8--92.9%) vs 80.85% (45.78--86.63%), P = 0.004) or without prednisone treatment (90.5% (86.8--92.9%) vs 76.5% (32.9--88.8%), P = 0.035). Likewise, in comparison to healthy donors, IIM patients with active disease and prednisone treatment had a higher percentage of intermediate monocytes (13% (8.45--49.65%) vs 4.49% (3.5--6.71%), P = 0.014) as well as those with active disease without prednisone consumption (17.1% (6.4--54.7%) vs 4.49% (3.5--6.71%), P = 0.018). We did not find a difference in the monocyte subsets nor the TLRs expression in patients receiving azathioprine, methotrexate, cyclophosphamide, mycophenolate mofetil or antimalarials.Fig. 2Differential percentage of monocyte subsets and TLR4 expression in patients with inflammatory myopathies according to disease activity. a--c Patients with IIM and active disease have a lower percentage of classical monocytes (a) and higher amounts of intermediate monocytes (b). Non-classical monocytes were not different among the study groups (c). d--f All monocyte subsets from patients with IIM have a higher expression of TLR4 regardless of disease activity and prednisone treatment The absolute number of classical monocytes inversely correlated with the total myositis disease activity assessment visual analogue scales (MYOACT) (Rho = − 0.429, P = 0.006) and the myositis intention to treat activity index MITAX (Rho = − 0.355, P = 0.027). The percentage of intermediate monocytes was correlated with the VAS of patient's disease activity (Rho = 0.3, P = 0.05). Differential expression of TLR in monocyte subsets among patients with IIM and its relationship with clinical features and circulating cytokines {#Sec9} ------------------------------------------------------------------------------------------------------------------------------------------------ Classical monocytes from patients with active IIM and prednisone use had a higher expression of TLR4 in comparison to healthy donors (1780 AU (1448--2409 AU) vs 502 AU (412--832 AU), P = 0.001). The same results were obtained when we compared the expression of TLR4 in classical monocytes from patients with IIM in complete clinical response without prednisone treatment and healthy donors (1758 AU (1430--2348 AU) vs 502 AU (412--832 AU), P = 0.007). Expression of TLR4 in intermediate monocytes was higher in patients with active disease and prednisone use in comparison to healthy donors (2464 AU (1701--3105 AU) vs 569 AU (483--837 AU), P \< 0.0001) as well as in those with active disease without prednisone treatment (2039 AU (1273--2725 AU) vs 569 AU (483--837 AU), P = 0.049) and in patients with complete clinical response without prednisone treatment (2250 AU (1823--2647 AU) vs 569 AU (483--837 AU), P = 0.009). The expression of TLR4 in non-classical monocytes was higher in patients with active disease and prednisone use in comparison to healthy donors (1598 AU (954--2243 AU) vs 281 AU (237--353 AU), P \< 0.0001); as well as in those with active disease without prednisone treatment (1293 AU (1006--1896 AU) vs 281 AU (237--353 AU), P = 0.007) and those with complete clinical response without prednisone use (1243 AU (1118--1580 AU) vs 281 AU (237--353 AU), P = 0.023). Subjects with ILD had a higher percentage of TLR4+ non-classical monocytes (98.25% (97.45--98.93%) vs 95.9% (93.43--98.15%), P = 0.031). Interestingly, the expression of TLR4 was higher in classical (1778 AU (1419--2713 AU), P = 0.03) and intermediate (1315 AU (910.8--1506 AU), P = 0.054) monocytes among patients with anti-Mi2 antibodies (Fig. [3](#Fig3){ref-type="fig"}). There was a trend towards a significant positive correlation between the percentage of TLR4+ non-classical monocytes and c-reactive protein levels (Rho = 0.457, P = 0.075) and the VAS of pulmonary disease activity (0.52, P = 0.006). The absolute number of TLR4+ non-classical monocytes was correlated with lactate dehydrogenase (LDH) (Rho = 0.7, P = 0.002).Fig. 3Differential pool of monocyte subsets and TLR expression according to the clinical features. a Patients with IIM and dysphagia have a higher expression of TLR2 in intermediate monocytes. b Subjects with interstitial lung disease have a higher percentage of TLR4+ non-classical monocytes. c, d Patients with abnormal nailfold capillaroscopy have a higher percentage of TLR2+ classical monocytes (d) and those with mega capillaries a higher percentage of TLR2+ non-classical monocytes. e, f The expression of TLR4 was higher in classical (e) and intermediate (f) monocytes of patients with anti-Mi2 antibodies Patients with dysphagia had a higher expression of TLR2 in intermediate monocytes (13605 AU (11761--16570 AU) vs 10964 (8543 vs 12487 AU), P = 0.025). The percentage of TLR2+ classical monocytes was correlated with the VAS of gastrointestinal disease activity (Rho = 0.377, P = 0.033. Likewise, patients with abnormal nailfold capillaroscopy had a higher percentage of TLR2+ classical monocytes (98.6% (97--99.4%) vs 97% (94.6--98.5%), P = 0.047) and those with mega-capillaries had a trend towards a higher percentage of TLR2+ non-classical monocytes (98.5% (94.2--98.9%) vs 95.8% (94.1--97.3%), P = 0.06) (Fig. [3](#Fig3){ref-type="fig"}). Furthermore, we found that the serum concentration of IL-6 correlated with the absolute number of intermediate monocytes (Rho = 0.395, P = 0.034) and with the expression of TLR2 in all monocyte subsets (Fig. [4](#Fig4){ref-type="fig"}). The monocyte subsets and their TLR expression did not correlated with any other serum cytokine. We did not find differences in the monocyte subsets nor the expression of TLRs with respect to the patients' gender.Fig. 4Correlation between TLR expression and IL-6. a--c The serum levels of IL-6 correlated with the expression of TLR2 in every monocyte subset TLR4 expression in all monocyte subsets and the percentage of intermediate monocytes are associated with the diagnosis of inflammatory myopathies {#Sec10} ------------------------------------------------------------------------------------------------------------------------------------------------- To test the diagnostic capacity of the monocyte subsets and their TLRs expression, we created ROC curves for each parameter. As shown in Table [2](#Tab2){ref-type="table"} and Fig. [5](#Fig5){ref-type="fig"}, the expression of TLR4 in all monocyte subsets and the percentage of intermediate monocytes are related to the diagnosis of inflammatory myopathies with a high area under the curve, specificity and positive likelihood ratio (LR (+)). Also, a cutoff value of \< 4.20% for non-classical monocytes was associated with the diagnosis of complete clinical response with a LR (+) of 4.62, sensitivity (95% CI) of 51.43 (35.57--67.01), specificity (95% CI) of 88.89 (56.50--99.43), and area under the curve (95% CI) of 0.70 (0.55--0.86), P = 0.05.Table 2Association between monocyte subsets and TLRS expression and the diagnosis of idiopathic inflammatory myopathiesVariableCutoff valueArea under the curve (95% CI)Sensitivity (95% CI)Specificity (95% CI)Likelihood ratio (+)PMean fluorescence intensity of TLR4 in non-classical monocytes (arbitrary units)\> 733.50.94 (0.88--1)88.37 (75.52--94.93)93.33 (70.18--99.66)13.26\< 0.0001Mean fluorescence intensity of TLR4 in intermediate monocytes (arbitrary units)\> 20230.90 (0.80--1)54.76 (39.93--68.78)93.33 (70.18--99.66)8.21\< 0.0001Mean fluorescence intensity of TLR4 in classical monocytes (arbitrary units)\> 9670.88 (0.75--1)90.48 (77.93--96.23)86.67 (62.12--97.63)6.78\< 0.0001% of intermediate monocytes\> 20.60.83 (0.71--0.95)36.36 (23.78--51.13)93.3 (70.18--99.66)5.450.0001Fig. 5ROC curves for the % of intermediate monocytes and the expression of TLR4 in all monocyte subsets as diagnostic biomarkers in patients with IIM Discussion {#Sec11} ========== The main findings of this study are that patients with IIM have an expansion of circulating intermediate monocytes and that their monocytes subsets have a differential expression of TLR4 and TLR2, which correlate with serum IL-6, as well as with distinctive clinical features. Previously, it was shown that anti-TNF therapy is able to reduce the expression of TLRs in monocytes from patients with rheumatoid arthritis (RA) and spondylarthopathy \[[@CR27], [@CR28]\]. Nevertheless, most studies have demonstrated that immunosuppressive therapy does not have an effect in the proportion of monocyte subsets, as shown in patients with RA under methotrexate treatment \[[@CR29]\] and in patients with solid organ transplantation \[[@CR30]\]. This is in agreement to our results since we did not find a difference in the monocyte subsets nor the TLRs expression according to the immunosuppressive therapy. Nevertheless, prednisone treatment is known to augment the proportion of intermediate monocytes and to diminish the percentage of non-classical monocytes \[[@CR30]\], therefore, we decided to adjust our analysis for prednisone intake, confirming that our results are not a consequence of prednisone treatment. Furthermore, the differential pool of monocyte subsets was maintained in patients with IIM in complete clinical response. Previous studies have shown an increased gene expression of the TLR4 and IFN-γ signaling pathway in patients with inflammatory myopathies \[[@CR31]\]. Also, nuclear factor kappa B (NF-κB), tumor necrosis factor a (TNF-a), interleukin 1 a (IL-1a), interleukin 22 (IL-22), toll-like receptor 2 (TLR-2), toll-like receptor 4 (TLR-4), toll-like receptor 9 (TLR-9), interferon alpha (IFNA), interferon gamma (IFNG), and retinoic acid inducible gene 1 (RIG-1) are genetic risk factors involved in the pathogenesis of IIM \[[@CR32]\]. The constitutional overexpression of pro-inflammatory and TLR-related pathways may explain the differential pool of monocyte subsets and TLRs expression in patients with IIM in complete clinical response. Similar findings have been described in patients with familial Mediterranean fever, in whom increased expression of TLR2 in monocytes has been demonstrated, even during quiescent disease \[[@CR33]\]. Similar to our results, lower amounts of CD14++ monocytes have been described in patients with multiple sclerosis (MS) \[[@CR34]\] and juvenile idiopathic arthritis (JIA) with enthesitis \[[@CR35]\]. Besides, a higher percentage of CD16+ intermediate and non-classical monocytes with a pro-inflammatory phenotype has been described in patients with MS \[[@CR34]\], neuromyelitis optica \[[@CR36]\], RA \[[@CR18]\], SLE \[[@CR37]\], ANCA-vasculitis \[[@CR38]\], sarcoidosis \[[@CR39]\], IgA nephropathy \[[@CR40]\], JIA with enthesitis \[[@CR35]\], type 1 diabetes mellitus \[[@CR41]\], thromboembolism \[[@CR42]\], atherosclerosis and stroke \[[@CR43]\] which is according to our results. Also, we found that the absolute number of classical monocytes inversely correlated with the disease activity (MYOACT and MITAX), which is according with previous data in patients with RA, where there is a higher percentage of intermediate monocytes during disease activity and a higher proportion of classical monocytes during remission \[[@CR44]\]. Our data confirm that a differential proportion of monocytes is found in subjects with autoimmune pathologies, according to disease activity. Intermediate and non-classical monocytes have been described as proinflammatory \[[@CR45]\]. Intermediate monocytes possess phagocytic and pro-inflammatory features, since they secrete IL-1β and TNF-α \[[@CR37]\], IL-6 \[[@CR46]\] and express higher amounts of TLR 2, 4 and 5 than any other subset. Additionally they express CD80, CD86, HLA-DR and are able to differentiate to M1 macrophages, promote a Th17 response \[[@CR37]\] and to induce T-cells proliferation due to their higher expression of CD40 \[[@CR35]\]. In contrast, in animal models of muscle injury, non-classical monocytes are recruited in the muscle after tissue damage to promote its repair \[[@CR47]\]. Therefore, an expansion of intermediate monocytes may contribute to the pro-inflammatory environment in peripheral blood of patients with IIM, whilst the higher proportion of non-classical monocytes in these patients could be a reflection of muscle damage, since non-classical monocytes are known to respond to CX~3~CL1, which promotes their migration, survival and recruitment in tissues \[[@CR16]\]. Regarding TLR expression, a higher expression of TLR2 in monocytes has been described in patients with RA \[[@CR48]\], especially in CD16+ monocytes \[[@CR49]\] which is according to our results. In IIM patients with nailfold capillaroscopy abnormalities, we found a higher amount of classical and non-classical monocytes expressing TLR2. These data are according to the role of monocytes in endothelial damage in other autoimmune diseases \[[@CR50]\]. In the steady state, monocytes patrol the endothelium, but in patients with RA and SLE, it was shown that activated monocytes contribute to vascular damage \[[@CR50]\], which could explain the association between this monocyte subset and an abnormal nailfold capillaroscopy in subjects with IIM. Regarding the relationship between the distinctive monocyte subsets and their TLR expression with the clinical features of patients with IIM, we found a higher percentage of non-classical monocytes in patients with dysphagia and a higher proportion of TLR4+ non-classical monocytes in subjects with ILD. The higher proportion of non-classical monocytes in patients with dysphagia may reflect a more intense and persistent tissue damage, since it is known that this monocyte subset is recruited after muscle injury \[[@CR47]\]. According to our results, previous studies have shown that the deficiency of TLR4 decreases pulmonary inflammation and fibrosis in the bleomycin-induced lung injury \[[@CR51]\] supporting the relationship between TLR4 and interstitial lung disease found in our study. We found that the expression of TLR2 in all monocyte subsets correlated with serum IL-6. It is known that CD16 + monocytes expressing TLR2 secrete TNF-α, IL-1, IL-6, IL-8, IL-12p40, IL-1Ra and IL-10 after stimuli with lipotheicoic acid \[[@CR49]\] confirming the pro-inflammatory profile of these cells in response to pathogen associated molecular patterns (PAMPs). Regarding the specific and associated myositis antibodies, the monocytes from patients with anti Mi2 antibodies had a higher expression of TLR4+. The importance of TLR4 in DM is highlighted by the presence of TLR4+ cells in the perimysium of these patients \[[@CR13]\]. Also, patients with anti-Mi2 antibodies are characterized by intense myositis and an abundant inflammatory infiltrate in muscle biopsy \[[@CR52]\]. In this regard, TLR4 is a key mediator of the pathogenic autoimmune and inflammatory response in IIM. In the animal model of myositis induced by intramuscular immunization with histdyl-tRNA synthetase, the TLR4 deficiency suppress the isotype change of the pathogenic autoimmune humoral response in a reaction dependent on Toll/IL-1 receptor (TIR) domain-containing adaptor protein inducing IFN-β (TRIF) \[[@CR53]\]. Likewise, in the murine model of myositis induced by immunization with myosin binding protein (C-MBP) and the antibody fusion protein (MYBPC2-MBP), an over-expression of TLR4 and its ligand, the high mobility group box 1 (HMGB1), has been found in muscle biopsies, and it correlated with the expression of major histocompatibility complex I (MHC-I), a key histopathologic finding of inflammatory myopathies \[[@CR54]\]. Also, it is known that muscle cells express TLR4 and that their stimulation with HMGB-1 in vitro promotes muscle dysfunction and MHC-I expression \[[@CR55]\]. The murine models of myositis have demonstrated the importance of TLR2 and TLR4 in the induction of disease in IIM, since the deficiency of both TLRs \[[@CR56]\] or their signaling protein MyD88 completely abolish the disease phenotype \[[@CR56]\]. Although these encouraging results suggest a potential therapeutic role of the TLRs inhibition in IIM, it is well known that the TLR2, TLR4 and MyD88 deficiency implies a severe immunodeficiency \[[@CR57]\]. In this regard, different peptides have been created to inhibit the TLR4 signaling pathway in animal models of sepsis and mastitis \[[@CR58], [@CR59]\] with good results. Inhibition of TLR4 is a therapeutic candidate in autoimmune diseases including IIM. TLR4 has many ligands including heat shock proteins (HSP60, HSP70, gp96, HSP22), HMGB-1, beta-defensin and saturated free fatty acids \[[@CR60]\]. Therefore, the inhibition of TLR4 could diminish the inflammatory response secondary to muscle damage in IIM. Nevertheless, inhibition of the TLR4 signaling adaptor molecules such as MyD88, Toll/IL-1 receptor (TIR) domain-containing adaptor protein (TIRAP), TRIF, TIR domain-containing adaptor molecule (TICAM-1) and TRIF-related adaptor molecule (TRAM)/TICAM-2 would result in immunodeficiency since these adaptors are shared by other TLRs \[[@CR60]\]. TAK-242 is a TLR4 inhibitor able to suppress the constitutional activation of NF-kB secondary to overexpression of TLR4 \[[@CR60]\]. Furthermore, TAK-242 diminish the production of IL-1β, TNF-α and IL-6 and modulates the LPS-mediated secretion of these cytokines in human mononuclear cells without interfering with other TLRs \[[@CR61]\] or the TLR4 adaptor molecules \[[@CR60]\]. This relatively specific mechanism of action may overcome the infection predisposition as a limitation for the use of TLR4 inhibitors as a therapeutic target in IIM. Finally, our data suggest that the differential pool of monocyte subsets and their TLRs expression in peripheral blood, may be useful for the diagnosis of IIM and to detect patients in complete clinical response. Previous studies have demonstrated the usefulness of the expression of MHC of class I and II in muscle fibers to differentiate inflammatory myopathies from non-inflammatory myopathies and neurogenic conditions \[[@CR62]\]. We found that the expression of TLR4 in all monocyte subsets and the percentage of intermediate monocytes predict the diagnosis of inflammatory myopathies with a high area under the curve and specificity. It would be interesting to explore if these parameters are useful to differentiate inflammatory myopathies from non-inflammatory myopathies or neuropathies. Furthermore, we found a predictive capacity of the percentage of non-classical monocytes for the diagnosis of complete clinical response in patients with IIM. This is a crucial point to avoid excessive treatment in patients in whom persistent muscle weakness is secondary to muscle atrophy instead of disease activity. Our study has many limitations. First, it is a transversal study with a relatively small sample of Mexican-mestizo patients which may preclude us to find differences regarding the association between disease activity status, the monocyte subsets and the circulating cytokines prospectively. Also, the findings might be limited to the ethnicity of our patients. Nevertheless, it is the first study to address the monocyte subsets in peripheral blood and their relationship with the clinical characteristics and circulating cytokines of patients with IIM. Conclusion {#Sec12} ========== All monocyte subsets of patients with IIM have a higher expression of TLR4. The expression of TLR2 in monocytes is related to circulating IL-6 and the presence of dysphagia and ILD is related to a differential expression of TLR2 and TLR4 in CD16+ monocytes. The differential pool of monocyte subsets and their expression of TLR4 are associated with the diagnosis of IIM and the complete clinical response. Further prospective studies are needed to unveil if the circulating monocyte pool is changed after achieving complete clinical response and to unveil the inhibition of TLR4 as a potential therapeutic target in IIM. ACR : American College of Rheumatology ANCA : Anti-neutrophil cytoplasmic antibodies AS : Anti-synthetase syndrome CADM : Clinically amyopathic dermatomyositis CBA : Cytometric bead array CDASI : Dermatomyositis disease area and severity index DM : Dermatomyositis EULAR : European League Against Rheumatism HRS : Histidyl tRNA synthetase IFN : Interferon IIM : Idiopathic inflammatory myopathies ILD : Interstitial lung disease IMNM : Immune-mediated necrotizing myopathy IQR : Interquartile range JDM : Juvenile dermatomyositis JIA : Juvenile idiopathic arthritis LDH : Lactate dehydrogenase MDI : Myositis damage index MFI : Mean fluorescence intensity MITAX : Myositis intention to treat activity index MMT8 : Manual muscle test 8 MYOACT : Myositis Disease Activity Assessment Tool PBMC : Peripheral blood mononuclear cell PM : Polymyositis RA : Rheumatoid arthritis SLE : Systemic lupus erythematosus TLR : Toll-like receptors TNF : Tumor necrosis factor VAS : Visual analogue scale Publisher\'s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. We would like to acknowledge the technical support provided by the Red de Apoyo a la Investigación, (RAI-CIC UNAM). DG participated in the conceptualization, acquisition, design, investigation process, project administration, supervision, data curation, formal analysis, validation, and visualization of the work. JT participated in the investigation process, data curation, formal analysis, validation and visualization of the work, and writing the original draft. DC, DP, RV, CN, GJ participated in the investigation process and data curation. All authors read and approved the final manuscript. This work was supported by Consejo Nacional de Ciencia y Tecnología (CONACYT) and Secretaría de Educación Pública-Ciencia Básica 2017--2018 \[A1-S23262\] and by a research grant from the Fondo de Investigación del Colegio Mexicano de Reumatología 2018. All data generated or analysed during this study are included in this published article. All healthy controls and patients signed an informed consent before inclusion and the protocol was approved by our institutional ethics committee (Ref. 2152) in compliance with the Helsinki declaration. The images and data contained in this manuscript are not related to a single individual and are entirely unidentifiable. The authors declare that they do not have any conflict of interest.	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== Respiratory muscle weakness may develop in ventilated critically ill patients \[[@CR1]--[@CR4]\]. For instance, Jaber and colleagues \[[@CR1]\] demonstrated that after 5--6 days of controlled mechanical ventilation in intensive care unit (ICU) patients, force-generating capacity of the diaphragm was reduced by ±32 %. In ICU patients, impaired capacity of the respiratory muscles is accompanied by an increased load due to elevated elastic and resistive forces of the respiratory system. This imbalance in load and capacity plays an important role in the development of ventilatory failure, for instance during a weaning trial. Respiratory muscle weakness is associated with adverse clinical outcomes, including difficult weaning from mechanical ventilation, increased mortality, and increased risk of ICU/hospital readmission \[[@CR5]\]. It is reasonable to propose that strategies that aim to restore respiratory muscle function in these patients improve outcome. The aim of this review is to discuss (future) strategies that prevent the development of respiratory muscle weakness or restore respiratory muscle function in weak ICU patients. We will mainly focus on interventions that are most likely to be of clinical importance in the near future. Pathophysiology of respiratory muscle weakness in the critically ill {#Sec2} ==================================================================== Reduced force output of the respiratory muscles in the critically ill may result from injury at any point between the central respiratory centers and the contractile proteins of diaphragm muscle fibers \[[@CR6], [@CR7]\]. In the absence of sedatives, reduced central respiratory drive is unlikely to explain reduced force output of the respiratory muscles in ICU patients \[[@CR8]\]. Phrenic nerve neuropathy, as assessed by prolonged phrenic nerve conduction time, has been demonstrated in ICU patients, indicating that injury of the peripheral nerve may play a role in reduced force output \[[@CR9]\]. Contractile dysfunction of the respiratory muscles in ICU patients may result from the loss of muscle mass (atrophy) and/or dysfunction of the remaining contractile proteins. In a landmark paper, Levine and colleagues \[[@CR10]\] demonstrated the rapid development of diaphragm muscle atrophy in ventilated brain-dead patients. More recently, Hooijman and colleagues \[[@CR11]\] performed in-depth functional and structural analysis of diaphragm biopsies in critically ill patients on the ventilator. In that study, muscle fiber cross-sectional area was reduced by ±25 % after an average of 7 days of mechanical ventilation. Muscle atrophy is the final result of an imbalance between protein synthesis and degradation. Upregulation of several proteolytic pathways has been demonstrated in the respiratory muscles of ICU patients \[[@CR11]\]. For instance, key regulators of the ubiquitin-proteasome pathway are upregulated in the diaphragm of these patients \[[@CR10], [@CR11]\]. Other pathways such as lysosomal protein degradation and autophagy may play a role as well (Fig. [1](#Fig1){ref-type="fig"}) \[[@CR12], [@CR13]\]. In addition to enhanced proteolysis, decreased protein synthesis has been reported in the diaphragm of rodents subjected to controlled mechanical ventilation \[[@CR14], [@CR15]\]. Besides atrophy, diaphragm weakness may be the result of contractile protein dysfunction. Even when corrected for loss of protein, muscle fibers in ICU patients develop less force \[[@CR4]\]. Furthermore, the sensitivity of the contractile proteins for calcium is reduced \[[@CR4]\]. The pathophysiology of contractile protein dysfunction in these patients is incompletely understood, but animal models of mechanical ventilation and endotoxemia indicate that phosphorylation and oxidative modifications of the sarcomeric proteins and mitochondrial proteins play a role in dysfunction and injury \[[@CR16]--[@CR19]\]. For an extensive background on the pathophysiology of muscle dysfunction in the critically ill, we refer to a recent excellent review on this subject \[[@CR20]\].Fig. 1Proposed scheme of pathophysiologic pathways in the development of respiratory muscle weakness during critical illness. Oxidative stress \[[@CR89]\], inflammation \[[@CR71], [@CR74]\], increased nuclear factor (NF)-κB activity \[[@CR90]\], and mechanical unloading \[[@CR10], [@CR11]\] have been proposed to initiate respiratory muscle weakness. These initiators can result in contractile protein becoming dysfunctional \[[@CR4]\], decreased synthesis \[[@CR14], [@CR15]\], or muscular autophagy \[[@CR12]\]. Oxidative stress and inflammatory pathways can activate caspases and calpains \[[@CR89], [@CR91]\], thereby delivering substrates for the ubiquitin-proteasome \[[@CR10], [@CR11], [@CR92]\], which further degrades contractile proteins Evaluation of respiratory muscle function in ICU patients {#Sec3} ========================================================= We will briefly discuss the readily available techniques that are relevant and feasible in clinical practice. For a detailed overview we refer to other reviews \[[@CR21]--[@CR23]\]. Maximal inspiratory pressure (MIP) and maximal expiratory pressure (MEP) are used to evaluate global respiratory muscle strength and can be applied to selected ICU patients. MIP and MEP are measured using a handheld pressure device connected to the endotracheal tube or tracheostomy while the patient performs specific maneuvers. Although American Thoracic Society/European Respiratory Society guidelines advise a single inspiratory maneuver at residual volume in non-intubated patients \[[@CR24]\], the reliability of MIP measurement in ICU patients is improved when using a unidirectional expiratory valve connected to the endotracheal tube or tracheostomy \[[@CR25]\]. Alternatively, pressures can be assessed using the ventilator by performing an end-expiratory hold maneuver. A 20-s end-expiratory occlusion can be performed to obtain more reliable measurements in poorly cooperative patients \[[@CR26]\]. It should be acknowledged that assessment of MIP and MEP requires a cooperative patient. Today, most ICU ventilators provide automatic functions to provide some useful parameters to evaluate the diaphragm function. High values for MIP and MEP exclude clinically significant weakness, but low values are common and may also reflect poor technique or effort \[[@CR27]\]. Esophageal pressure (Pes) is an estimate of pleural pressure \[[@CR28]\] and can be used to calculate the amount of work performed by the respiratory muscles. Simultaneous recording of Pes and gastric pressure (Pga) allows the calculation of transdiaphragmatic pressure (Pdi = Pga − Pes), a specific measure of diaphragm contractility. The latter is useful for close monitoring and evaluation of diaphragm function in difficult-to-wean patients \[[@CR29]\]. However, acquisition, calculation, and interpretation of Pes, Pga, and their derived measures are rather complex and therefore not widely accepted in clinical practice. Ultrasonography is an increasingly popular tool for assessment of diaphragm function in ICU patients \[[@CR30]\]. In the subcostal view reduced caudal movement of the diaphragm during unassisted breathing is consistent with weakness and paradoxal movement indicates diaphragm paralysis \[[@CR31], [@CR32]\]. Diaphragm atrophy can be assessed by measuring diaphragm thickness in the midaxillary line at the level of the diaphragm dome \[[@CR32], [@CR33]\] and its thickening fraction during inspiration to asses contractile activity \[[@CR30]\]. Diaphragm atrophy can also be evaluated more precisely with computed tomography, although this is more cumbersome than echography \[[@CR34]\]. Using ultrasound, Goligher and colleagues \[[@CR30]\] recently demonstrated in ICU patients that diaphragm atrophy is associated with diaphragm dysfunction and, in a minority of these patients, diaphragm thickness surprisingly increased while on the ventilator, which was associated with dysfunction as well. Diaphragm electromyography (EMG) reflects the electrical activity and is the gold standard to assess neural respiratory drive. Diaphragm EMG can be recorded best using an esophageal catheter with multiple electrodes. With the introduction of neurally adjusted ventilatory assist (NAVA; Maquet, Solna, Sweden) \[[@CR35]\], the (processed) EMG signal can be obtained continuously in ICU patients. In these patients, diaphragm EMG may be used to monitor respiratory muscle unloading \[[@CR36]\] and patient--ventilator interaction \[[@CR37]\]. Modulation of contractile activity: disuse and inspiratory muscle training {#Sec4} ========================================================================== Prevention of disuse atrophy {#Sec5} ---------------------------- Like any other striated muscle, respiratory muscle mass is affected by contractile inactivity. In fact, the respiratory muscles appear more sensitive to the effects of disuse compared with other striated muscles \[[@CR10], [@CR11], [@CR34]\]. In humans, relatively brief periods of diaphragm disuse (\<3 days) due to controlled mechanical ventilation are associated with diaphragm muscle fiber atrophy \[[@CR10]\]. Animal studies have demonstrated that mechanical ventilation-induced diaphragm atrophy and dysfunction is less severe in assisted modes of mechanical ventilation \[[@CR38]\]. Recently, Goligher et al. \[[@CR30]\] used ultrasound techniques to demonstrate that low levels of diaphragm activity, resulting from high levels of ventilator support, are associated with diaphragm atrophy and dysfunction in ICU patients. On the other hand, administration of muscle relaxants for 48 h in patients with early severe acute respiratory distress syndrome (ARDS) resulted in earlier withdrawal from mechanical ventilation and did not adversely affect peripheral muscle function \[[@CR39]\]. Therefore, it appears that, under certain conditions (i.e., very severe ARDS), controlled mechanical ventilation is preferred to facilitate lung-protective ventilation and this beneficial effect outweighs the possible adverse effects on the respiratory muscles. Nevertheless, in general it is reasonable to limit the duration of controlled mechanical ventilation and prevent high levels of support under assisted ventilation in order to reduce the risk of disuse atrophy \[[@CR10], [@CR30], [@CR40]\]. It is important to recognize that ventilator pressure and flow waveforms are unreliable to confirm the presence of respiratory muscle activity \[[@CR21], [@CR22], [@CR41]\]. We recommend additional monitoring techniques as outlined above. Although the optimal level of activity for the respiratory muscles is unknown during mechanical ventilation, these monitoring techniques allow us to detect complete inactivity of the muscles due to over-assist. Inspiratory muscle training {#Sec6} --------------------------- In general, training can be instituted to enhance muscle endurance or strength. These types of training require different strategies and have distinct physiological responses. In healthy subjects, respiratory muscle activity is characterized by development of low pressure during the entire life span of a subject. The pressure generated by the inspiratory muscles is only ±5 cmH~2~O (5 % of maximum inspiratory pressure) to generate a tidal volume of 500 ml, when respiratory compliance is 100 ml/cmH~2~O. At first sight, training of the respiratory muscle should, therefore, be designed to improve endurance. Indeed, in patients who are difficult to wean from the ventilator, progressive weaning trials (T-tube or low pressure support) are frequently instituted as training stimulus. Although reasonable from a physiological perspective, it has never been proven that this strategy indeed improves respiratory muscle endurance. In very few circumstances is high inspiratory pressure required for prolonged periods of time and therefore strength training of the diaphragm seems of limited relevance. It has been demonstrated, however, that respiratory effort sensation depends on the maximal inspiratory pressure \[[@CR42]\]. In healthy subjects, pharmacological induction of inspiratory muscle weakness increases respiratory effort sensation for the same workload \[[@CR42]\], confirming the importance of adequate strength beyond that strictly required to generate tidal volume. It has been shown that inspiratory muscle strength training (IMST) improves whole body exercise performance, in particular in less fit subjects \[[@CR43]\]. Also, in patients with chronic obstructive pulmonary disease (COPD), IMST improves inspiratory muscle strength and total body exercise and reduces dyspnea sensation \[[@CR44]\]. Only three randomized studies have reported the effectiveness of IMST in ventilated ICU patients. In the trial by Cader et al. \[[@CR45]\], 41 ventilated ICU patients with respiratory muscle weakness were randomized between inspiratory threshold loading and no training intervention. Training consisted of an inspiratory load of 30 % maximum inspiratory pressure for 5 minutes, twice a day, 7 days a week throughout the weaning period. Maximum inspiratory pressure was significantly increased in the training group (15 ± 3 to 25 ± 4 cmH~2~O) but not in the control group (15 ± 2 to 18 ± 2 cmH~2~O). The study was underpowered for clinically relevant endpoints, although weaning time was reduced in the training group. In another study, Martin and colleagues \[[@CR46]\] randomized 69 ventilator-bound patients (mean duration of ventilation at inclusion ±44 days) to IMST or sham training added to endurance training. Strength training consisted of four sets of 6--10 breaths per day with 2 minutes rest between each set. Loading was individualized and set at a level that could just be tolerated by the patient. IMST significantly improved maximum inspiratory pressure (end of training 54 cmH~2~O in intervention group and 45 cmH~2~O in sham group, P \< 0.05). Also, successful weaning at day 28 after inclusion was more likely in the intervention group compared with sham (71 versus 47 %, P \< 0.05). No adverse events of IMST were reported in these two trials. Finally, Condessa et al. \[[@CR47]\] randomized ventilated patients to inspiratory strength training on top of usual care versus usual care only. Each training session consisted of five sets of loaded breaths (40 % maximum inspiratory pressure), twice a day, 7 days a week. In this study, IMST significantly increased maximum inspiratory pressure but did not affect weaning time. In conclusion, inspiratory muscle training is feasible and appears safe in patients with respiratory muscle weakness who are difficult to wean from the ventilator, which is supported by a recent systematic review \[[@CR48]\]. Studies in other patient categories, including COPD, indicate that IMST improves outcome. In our opinion, it is reasonable to add IMST to endurance training in stable, difficult-to-wean patients with confirmed respiratory muscle weakness. Future studies are needed to determine the optimal training protocol and appropriate timing for initiation of IMST. Antioxidants and nutrition {#Sec7} ========================== As outlined above, oxidative stress may play a role in the pathophysiology of ICU-acquired respiratory muscle weakness. Several experimental studies demonstrate that antioxidants attenuate the detrimental effects of controlled mechanical ventilation and/or systemic inflammation on respiratory muscle structure and function \[[@CR18], [@CR49]--[@CR51]\]. In healthy subjects, high dose N-acetylcysteine (150 mg/kg intravenously) attenuated diaphragm fatigue induced by an inspiratory resistive load \[[@CR52]\]. Today, no study has specifically evaluated the effect of antioxidants on respiratory muscle function in ICU patients; however, some indirect evidence is available. In an open trial, 595 critically ill surgery or trauma patients were randomized between antioxidant supplementation (alpha-tocopherol and ascorbic acid) and standard care or standard care only \[[@CR53]\]. Patients in the antioxidant group spent less time on the ventilator (3.7 versus 4.6 days, P \< 0.05). It should be noted that patients in this study were young (38 ± 15 years) and total ventilation time was rather short. Therefore, it is questionable whether the beneficial effects of antioxidants were the result of improved respiratory muscle function. Heyland et al. \[[@CR54]\] studied the effects of antioxidants and glutamine in a heterogeneous ICU population (N = 1223). Patients were divided into four groups: placebo, glutamine (0.35 g/kg/24 h of body weight intravenously), antioxidants (selenium, zinc, beta carotene, vitamin E, and vitamin C), or antioxidants plus glutamine. Again, respiratory muscle function was not specifically evaluated but no difference in duration of mechanical ventilation was observed among the four groups. Today, only one study has evaluated the effects of two nutritional strategies on skeletal muscle structure and function in critically ill patients \[[@CR55]\]. This study was a planned subanalysis of the EPaNIC trial that compared the effects of tolerating macronutrient deficiency versus early parenteral nutrition on skeletal muscle structure and function \[[@CR56]\]. In that study, patients were randomized between early (within 2 days of ICU admission) versus late (\<8 days after ICU admission) parenteral nutrition to prevent macronutrient deficiency. Skeletal muscle strength was assessed in 600 ICU patients using the Medical Research Council sum score. Weakness occurred less often in the late versus early parenteral nutrition group (34 versus 43 %, P = 0.030). Compared with healthy subjects, muscle fibers exhibited atrophy but were not significantly different between the early and late parenteral nutrition groups. However, markers for autophagosome formation were significantly higher in the late parenteral nutrition group. This indicates that autophagy plays an important role in protein turnover next to other effects to provide substrate for recycling. In conclusion, tolerating macronutrient deficiency in the first week after ICU admission is not associated with the development of muscle fiber atrophy but surprisingly appears to improve muscle contractility. Although the effect of high-dose antioxidant administration on respiratory muscle structure and function is encouraging in animal models, no data support the routine administration of antioxidants or other specific feeding strategies on respiratory muscle function in the critically ill ventilated patient. Improving respiratory muscle protein content: anabolics {#Sec8} ======================================================= Loss of muscle mass plays an important role in the development of ICU-acquired respiratory muscle weakness \[[@CR10], [@CR11]\]. Pharmacological interventions that restore protein balance seem a reasonable approach in these patients. First, inhibitors of proteolysis increase respiratory muscle mass in animal models with heart failure and under mechanical ventilation \[[@CR57], [@CR58]\] but toxicity limits the application of these agents for this specific indication in humans. Second, anabolic hormones have been used to enhance skeletal muscle mass under a variety of conditions (reviewed in \[[@CR59]\]). We focus on the effects of endogenous and exogenous anabolic hormones on the respiratory muscles. The most important endogenous anabolic hormones are growth hormone (GH), insulin-like growth factor-1 (IGF-1), insulin, and the anabolic steroid testosterone and its analogues. The effects of GH on skeletal muscle have been reviewed recently \[[@CR60]\]. GH enhances production of IGF-1 in the liver but has direct anabolic effects on skeletal muscle as well. The effect of administration of GH/IGF-1 or its exogenous analogues on respiratory muscle function in critically ill patients has not been studied. However, Takala et al. \[[@CR61]\] reported the effects of recombinant GH (Genotropin) administration in critically ill ICU patients relatively early during ICU stay in two placebo-controlled trials. Unexpectedly, both trials demonstrated significantly increased mortality in recombinant GH-treated patients (47 versus 25 % in the Finnish study and 61 versus 26 % in the multinational study). The mechanisms for increased mortality are incompletely understood, but modulation of immune function may play a role. Interestingly, Schols et al. \[[@CR62]\] reported the effects of the exogenous anabolic steroid nadrolone in patients with stable COPD during an 8-week pulmonary rehabilitation program. Patients (N = 217) were randomized between placebo, placebo with high caloric feeding, and nandrolone with high caloric feeding. They found that nandrolone together with high caloric feeding significantly improved inspiratory muscle strength. These findings were more or less confirmed in a later trial by the same group \[[@CR63]\]. Although case series report the use of anabolic steroids in difficult-to-wean patients \[[@CR64]\], no randomized studies have evaluated the safety and efficacy of anabolic steroids in ICU patients with respiratory muscle weakness. Accordingly, anabolic hormones should not be used in the early stage of ICU admission but may have a role in more chronic and stable ICU patients with respiratory muscle weakness and who are difficult to wean from the ventilator. Positive inotropes {#Sec9} ================== In addition to atrophy, dysfunction of the remaining muscle fibers has been demonstrated in critically ill patients \[[@CR4]\]. Accordingly, optimizing contractility using positive inotropes seems a reasonable approach in these patients. ß-Adrenoreceptor agonists indeed exhibit a direct positive inotropic effect on the diaphragm muscle in vitro by increasing intracellular calcium influx. The effects of β2 adrenoreceptor agonists on respiratory muscle function in vivo are, however, controversial. Albuterol (oral) did not affect fatigability of the diaphragm in healthy subjects. However, fenoterol (oral) delayed the development of diaphragm fatigue in healthy volunteers subjected to inspiratory loading. In mechanically ventilated COPD patients with respiratory failure, respiratory muscle function significantly improved after dopamine infusion, probably by augmentation of diaphragm blood flow and improved cardiac output. Currently the administration of ß-adrenoreceptor agonists cannot be recommended to improve respiratory muscle function in ICU patients. Muscle fibers isolated from the diaphragm of ICU patients display decreased maximal force-generating capacity, indicating intrinsic muscle weakness \[[@CR11]\]. In fast-twitch diaphragm fibers the reduction of sub-maximal force generation even exceeds the loss of maximal force-generating capacity \[[@CR4]\]. This implies that these fibers require more calcium to generate normal force levels, i.e., their sensitivity to calcium is reduced. Calcium sensitizers have been developed to treat similar pathology of cardiac muscle \[[@CR65], [@CR66]\]. Currently, levosimendan is the only calcium sensitizer approved for use in humans (\>50 countries worldwide). Experimental studies have shown that levosimendan improves calcium sensitivity of diaphragm muscle fibers from patients with COPD \[[@CR67]\]. Moreover, a recent double-blind, randomized, crossover study demonstrated that administration of levosimendan improved neuromechanical efficiency by \>20 % and prevented contractile fatigue during a diaphragm-loading task in healthy subjects \[[@CR68]\]. A randomized clinical trial (ClinicalTrials.gov identifier NCT01721434) is currently investigating whether levosimendan indeed facilitates liberation from the ventilator. In contrast to levosimendan, the effectiveness of other calcium sensitizers has so far only been studied in vitro. For example, exposure to EMD 57033 (a troponin activator) partially restored calcium sensitivity in diaphragm fibers isolated from piglets after 5 days of mechanical ventilation \[[@CR69]\]. Furthermore, in diaphragm fibers from critically ill patients, CK-2066260 completely restores calcium sensitivity \[[@CR4]\]. Taken together, calcium sensitizers might exert an energetically beneficial effect on diaphragm work \[[@CR67]\]. When less calcium is needed to maintain force generation, muscle work becomes more efficient \[[@CR68]\] and could improve respiratory muscle contractility in critically ill patients. However, further clinical trials are needed to prove the benefits of the calcium sensitizers in ICU patients with respiratory muscle weakness. Future developments {#Sec10} =================== Modulation of inflammation {#Sec11} -------------------------- Activation of pro-inflammatory pathways is associated with respiratory muscle weakness \[[@CR70]--[@CR73]\]. Therefore, modulation of the inflammatory response to combat respiratory muscle dysfunction has been a focus of interest in recent experimental studies \[[@CR71], [@CR74], [@CR75]\]. We demonstrated, for instance, that interleukin (IL)-6 plays an important role in the loss of contractile proteins in muscle fibers exposed to plasma from septic shock patients \[[@CR71]\]. However, diaphragm fiber atrophy due to disuse in brain death patients was not associated with upregulation of IL-6 \[[@CR10]\]. IL-10 is an interleukin with anti-inflammatory properties \[[@CR76]\]. In a murine model of Pseudomonas aeruginosa pneumonia, diaphragm dysfunction was attenuated after experimental IL-10 administration \[[@CR76]\]. During critical illness with subsequent inflammatory status, nuclear factor (NF)-κB is the key factor for transcription of several cytokines \[[@CR77]\]. Recently, evidence was found that inhibition of NF-κB in endotoxemic mice protects against diaphragm muscle weakness, probably due to decreased generation of pro-inflammatory cytokines \[[@CR75]\]. Proteolytic pathways can be activated through Toll-like receptor (TLR)-4, present in muscle plasma membrane \[[@CR74]\]. TLRs are essential receptors in recognizing microbes and initiating an inflammatory immune response \[[@CR78]\]. In TLR-4 knockout mice, loss of diaphragm contractile protein associated with controlled mechanical ventilation was attenuated compared with wild-type mice \[[@CR74]\]. In a large clinical trial the TLR-4 antagonist eritoran did not, however, improve outcome in patients with severe sepsis or septic shock (±80 % on mechanical ventilation) \[[@CR79]\]. Nevertheless, neither skeletal muscle function nor duration of mechanical ventilation was assessed in this trial. Traditionally, steroids are associated with myopathy, atrophy, and dysfunction of the respiratory muscles \[[@CR80]--[@CR82]\]. However, the final effects appear to be dose- and time-dependent, at least in experimental studies. For instance, Maes et al. \[[@CR83]\] demonstrated that "low-dose" (5 mg/kg) methylprednisolone exacerbated ventilator-induced diaphragm dysfunction in rats, whereas a high dose (30 mg/kg) protected against the deleterious effects of controlled mechanical ventilation on diaphragm function. In ICU patients the effect of steroids on muscle function are conflicting \[[@CR84]--[@CR86]\]. However, no study has prospectively evaluated the effects of corticosteroids on respiratory muscle function in ventilated ICU patients. In conclusion, despite the encouraging data that modulation of inflammation improves respiratory muscle function in animals, data in humans are scarce and, where present, disappointing. Modulation of proteolytic pathways {#Sec12} ---------------------------------- Since activation of proteolytic systems plays a key role in the development of respiratory muscle dysfunction during critical illness, several experimental studies have investigated the effect of specific inhibitors. For instance, in rats exposed to 24 h of mechanical ventilation, bortezomib treatment partially prevented the reduction of diaphragm force and atrophy \[[@CR58]\]. These small positive effects were probably mediated by the ability of bortezomib to indirectly reduce caspase-3 activity \[[@CR57]\]. Proteasomes are only able to process myofilaments that have been cleaved from the sarcomere by enzymes like caspases and calpains \[[@CR87]\]. Accordingly, inhibition of the proteasome alone is not expected to have substantial effects on muscle function as this would leave the muscle cell with only unprocessed, but cleaved, myofilaments. Furthermore, considering the basic housekeeping cell functions of the proteasome, it is no surprise that the clinical application of bortezomib is accompanied by serious toxic adverse events, such as cytopenia and peripheral neuropathy \[[@CR88]\]. Finally, although several compounds targeting proteolytic pathways upstream of the proteasome have a high potential to prevent the development of diaphragm weakness, this does not necessarily imply that these agents can also improve function of the weakened diaphragm. Nevertheless, modulation of the proteolytic system is a potentially interesting target to modulate loss of respiratory muscle function due to controlled mechanical ventilation. Conclusion {#Sec13} ========== Weakness of the respiratory muscles frequently develops in the ICU patient and is associated with adverse outcome, including prolonged mechanical ventilation. Despite the high incidence and clinical impact of ICU-acquired respiratory muscle dysfunction, no specific preventive or therapeutic interventions have been tested in large randomized controlled trials. Therefore, we should rely on interventions that seem reasonable from a physiological perspective or are supported by small clinical studies. As pointed out in Fig. [2](#Fig2){ref-type="fig"}, interventions could be subdivided into three categories: prevention of respiratory muscle dysfunction; therapeutic strategies that aim to improve respiratory muscle function; and so-called rescue interventions that should only be applied in exceptional cases and only after discussion with the patient or primary decision makers.Fig. 2Three groups of interventions to counteract respiratory muscle weakness during critical illness Preventive strategies should limit development of disuse atrophy and muscle damage. We suggest using techniques that monitor diaphragm muscle function \[[@CR21], [@CR22]\] to confirm a physiologically acceptable level of diaphragm contractility and allow the clinician to optimize ventilator settings in order to improve patient--ventilator interaction. Drugs with potential side effects on skeletal muscle, in particular corticosteroids and muscle relaxants, should be avoided when appropriate. Once ICU-acquired weakness has developed, a combined program of respiratory muscle endurance training and strength training should be considered. Endurance training can be instituted using progressive weaning trials and strength training by using a device for variable inspiratory threshold loading connected to the endotracheal tube \[[@CR46]\]. Use of respiratory muscle positive inotropes, in particular levosimendan, is the subject of a current randomized controlled trial (NCT01721434) and not currently recommended for difficult-to-wean patients. ARDS : acute respiratory distress syndrome COPD : chronic obstructive pulmonary disease EMG : electromyography GH : growth hormone ICU : intensive care unit IGF-1 : insulin-like growth factor-1 IL : interleukin IMST : inspiratory muscle strength training MEP : maximal expiratory pressure MIP : maximal inspiratory pressure NF : nuclear factor Pdi : transdiaphragmatic pressure Pes : esophageal pressure Pga : gastric pressure TLR : toll-like receptor Competing interests LMAH has received travel grants and speakers fees from Orion Pharma (Finland), Maquet Critical Care (Sweden), and Biomarin (USA). LMAH has received research grants paid to the institution from Bayer Pharma and Orion Pharma. The other authors declare that they have no competing interests. Authors' contributions WJMS, HWHvH, JD, LHR, and LMAH contributed to writing the manuscript. GJS and JGvdH contributed to revising the manuscript. All authors read and approved the final manuscript for publication.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Monteggia fractures are rare injuries of the forearm and were first described by Giovanni Battista Monteggia in 1814 as a fracture of the shaft of the ulna combined with an anterior dislocation of the radial head \[[@R1]\]. Monteggia injuries account for only 2--5% of all proximal forearm fractures \[[@R2]\]. In 1967, Bado introduced the concept of a Monteggia lesion and presented a classification into four types depending on the direction of the radial head's dislocation and the angulation of the fracture of the ulna \[[@R3]\], \[[@R4]\]. Type I describes a dislocation of the radial head in anterior, type II in posterior, and type III in lateral direction. Type IV is defined as a fracture of both bones of the forearm with dislocation of the radial head \[[@R4]\]. Furthermore, Bado described Monteggia equivalent injuries which differed in radiological appearance but possessed similar characteristics especially in the mechanism of injury and in its treatment. Type II injuries are most common (about 80%), followed by type I (about 15%), while type III and IV injuries are rare (all together 5%) \[[@R2]\]. Jupiter further classified the posterior Monteggia lesion (Bado type II) depending on the location and type of the ulnar fracture as well as the pattern of radial head injury \[[@R5]\], \[[@R6]\]. In type IIA, the fracture of the ulna involves the distal part of the olecranon and the coronoid process; in type IIB, the fracture involves the transition from metaphysis to diaphysis, distally to the coronoid process; in type IIC, a diaphyseal fracture is present and in type IID, the fracture extends from the olecranon to the proximal half of the ulna. Radial head fractures were classified in 4 types: type 0: no fracture; type 1: one part fracture; type 2: two part fracture; type 3: three part fracture; type C: comminuted fracture \[[@R4]\], \[[@R6]\]. The Bado and Jupiter classifications have been applied more frequently in the literature due to considerations of surgical practicality \[[@R3]\]. However, over the years additional injuries, such as radial head fractures, coronoid fractures, and combined radial head and coronoid fractures have added complexity to the classification of Monteggia-related injuries \[[@R3]\]. Not all of those injuries are included in the Bado and Jupiter classifications. Therefore the eponym of Monteggia fracture includes various patterns of complex fracture-dislocations of the proximal ulna and radius which are not well defined yet \[[@R6]\]. Other authors describe Monteggia lesions which include additional injuries such as radial head fractures (Mason, grade I-III) and coronoid fractures (Regan and Morrey I-III) as Monteggia-like lesions or Monteggia-equivalent injuries \[[@R2]\], \[[@R6]\], \[[@R7]\]. Nevertheless, according to Giannicola et al., in Monteggia-like pattern, six essential lesions can be identified and each of them must be recognized and treated: Ulnar fracture,Radio-humeral dislocation,Ulnohumeral dislocation,Proximal radio-ulnar dislocation,Radial fracture andDistal radio-ulnar joint\\interosseus membrane lesion. The various combination of these critical lesions can explain the complexity and variety of their treatment \[[@R6]\]. However, despite these confusing classifications and descriptions: Monteggia lesions with concomitant injuries to the radial head and/or the coronoid process do exist. Therefore, they should be considered separately. We were able to diagnose and treat a total of 10 adult patients with Monteggia-like lesions which could be evaluated clinically and radiologically one year after the injury on average. The objective of this retrospective study was to better define circumstances and patterns of injury and to document the short-term results of treatment with current fixation techniques. Materials and methods ===================== Ten adult patients were surgically treated for a Monteggia-like lesion between 2012 and 2014. Preoperatively, all patients received a CT-Scan of the injured elbow joint in order to rule out associated injuries and improve preoperative classification and planning. We defined a Monteggia-like lesion as a fracture of the proximal ulna distal to the end of the olecranon process with an associated dislocation of the radiocapitellar joint in combination with a radial head fracture alone or combined with a fracture of the coronoid process. All 10 patients met these criteria and their injuries were classified according to the systems of Bado and Jupiter et al. \[[@R4]\], \[[@R5]\]. Radial head fractures were classified according to Mason \[[@R8]\], fractures of the coronoid process according to Regan and Morrey \[[@R9]\], respectively. Demographics, injury patterns, and details of the surgical treatment were extracted from chart review. All 10 patients were available for follow-up examinations at an average of 12.3 months (range: 11--13 months) after the trauma. All patients were evaluated by the same experienced examiner blinded to the classification of injury. The average age was 52.4 years (range, 18--83 years). Six women and four men were included. In four patients the injury resulted from a motor-vehicle accident, in three from a casual fall from a standing height. The remaining three fractures were caused by a fall from a height of at least four meters. In four of the patients, the injury was part of a polytrauma including additional skeletal lesions in three, thoracic or abdominal trauma in two, and head injuries in two patients. The dominant arm was affected in eight patients. According to the classification of Bado two type I and eight type II fractures were seen. Two open fractures were classified grade I according to the classification of Gustilo and Anderson \[[@R10]\]. No patient had associated neurologic involvement. All eight Monteggia fractures with posterior dislocation of the radial head were further classified according to Jupiter et al. into two type IIa, four type IIb, two type IIc and two type IId fractures. Osteosynthesis of the ulna was performed using a proximally contoured 3.5 mm LCP (locking compression plate) or precontoured 3.5 mm LCP Olecranon Plate (both Synthes GmbH, Umkirch, Germany) applied to the posterior surface of the ulna in dynamic compression mode (Figure 1 [(Fig. 1)](#F1){ref-type="fig"}, Figure 2 [(Fig. 2)](#F2){ref-type="fig"}, Figure 3 [(Fig. 3)](#F3){ref-type="fig"}). All patients had a fracture of the radial head. A total of five were comminuted (Mason type III), three were classified as Mason type II and two as Mason type I. All patients with a type III radial head fracture received a cemented bipolar radial head prosthesis (CRFII, Tornier, Montbonnot, France). All type II radial head fractures were treated with open reduction and internal fixation using mini screws. Type I fractures were treated with closed reduction of the radial head dislocation under fluoroscopic control. According to the Regan and Morrey classification five type I, three type II, and two type III coronoid fractures were included. All type II and type III coronoid fractures were associated with a Bado type II fracture and stabilized using lag screws inserted through the ulnar plate or with independent lag screws after indirect reduction of the fracture. All patients were immobilized in a long-arm cast for one week. However, immobilization was discontinued for daily immediate passive motion and continuous passive motion (CPM) without any restriction of movement starting at day 2 after surgery. Full weight bearing was allowed 6 weeks postoperatively. In order to prevent heterotopic ossifications non-steroid anti-inflammatories (Indomethacin 25 mg orally 3 times daily) were administered for 3 weeks. At the clinical follow-up examination the Mayo Modified Wrist Score (MMWS) \[[@R11]\] was used to assess the wrist function, while the Mayo Elbow Performance Score (MEPS) and the functional rating index of Broberg and Morrey \[[@R12]\] were used to assess the functioning of the elbow joint. The pain, satisfaction and stress resistance of the arm were recorded using the DASH Score as a subjective parameter \[[@R13]\]. The ROM of wrists, forearms and elbow joints was measured using a standard full-circle goniometer bilateral. The elbow was tested for valgus and varus instability in maximum extension and in 30° of flexion. In order to evaluate posterolateral rotatory instability, the pivot-shift test \[[@R14]\] was performed in each patient and stability was graded as normal, mild, moderate or severely unstable. Grip strength was measured with a Jamar dynamometer (Fabrication Enterprises Inc., White Plains, New York) with the other hand serving as a control. After three consecutive bilateral measurements, the grip strength at the injured side was expressed as a percentage of the control, using a correction factor of 1.07 for the dominant hand over the non-dominant \[[@R15]\], \[[@R16]\]. Patient satisfaction was determined by a subjective satisfaction questionnaire according to Jungbluth et al. \[[@R15]\]. Pain at rest and during activity was measured with a visual analogue scale (VAS). Anteroposterior and lateral X-rays of the affected elbow joint were performed in all patients at the follow-up examination. Radiographs of the elbow were reviewed for capitellar osteopenia, degenerative changes, and heterotopic ossifications by an experienced and blinded radiologist. Capitellar osteopenia was graded as none, mild, moderate or severe according to Lamas et al. \[[@R17]\]. The degree of degenerative changes was classified according to Broberg and Morrey, as grade 0 (normal joint), grade 1 (slight joint space narrowing and minimum osteophyte formation), grade 2 (moderate joint space narrowing and moderate osteophyte formation), or grade 3 (severe degenerative changes with gross destruction of the joint) \[[@R12]\], \[[@R17]\]. Heterotopic ossification were graded as I, II, III or IV according to Brooker et al. \[[@R18]\]. In radial head prosthesis radiographic signs of loosening were assessed. Moreover, fracture union was defined as bridging bone on anteroposterior and lateral radiographs. Results ======= Clinical follow-up examinations were performed at an average of 12.3 months (range 11--13) after surgery. With an average point value of 86.5 (55--100) for the MMWS good results were obtained. An average of 89.2 points (75--100) was registered for the MEPS and 20.1 (10.8--55.8) for the DASH score, indicating good results. The average point value for the functional rating index of Broberg and Morrey was 86.5 (57--100) indicative for a good result as well. For forearm pronation mean values of 85° (70° to 90°) and for supination mean values of 75° (45° to 90°) were seen. Elbow flexion was from a mean of 9° fixed flexion (0° to 30°) to 131° (100° to 140°). Range of motion of the wrist displayed a mean palmar flexion of 75° (60° to 90°) and dorsal extension of 69° (40° to 90°). After application of the corrective factor for the non-dominant hand the grip strength was 91.8% (86% to 99%) of the contralateral side on average. No patient exhibited any wrist or elbow instability. The Pivot shift was graded as normal in all patients. According to the questionnaire for subjective patient satisfaction eight of the ten patients were satisfied with the results of the treatment and would choose the treatment regimen again. Eight patients reported no relevant restrictions of movement. All patients were able to return to their previous workplace and reached the same level of athletic activities as before the accident. The average VAS value was 1.0 (0 to 1) at rest and 1.7 (0 to 3) during activities. Eight patients exhibited neither signs of capitellar osteopenia nor degenerative changes or heterotopic ossifications in the region of the elbow on conventional radiographs. Two patients with a radial head prosthesis showed mild osteopenia and grade 1 degenerative changes. These patients exhibited grade 1 heterotopic ossifications. No patient required revision surgery. None of the patients showed signs of loosening of the radial head prosthesis. Fracture union was achieved in all patients. Discussion ========== Monteggia-like lesions are rare injuries and recommendations of treatment strategies are sparse. These injuries still represent a challenge to orthopedic surgeons. This challenge is further complicated since most studies about Monteggia injuries do not define Monteggia-like lesions as separate clinical entities \[[@R19]\]. Moreover, many studies on Monteggia injuries were carried out before the wide application of locking plates and some reports even have presented results of Monteggia injuries in children and adults together \[[@R4]\], \[[@R20]\], \[[@R21]\], \[[@R22]\]. Despite a better understanding of the biomechanical principles and advances in surgical treatment options, Monteggia injuries are still frequently associated with complications, poor functional outcomes, and high rates of revision surgery \[[@R23]\]. Furthermore, Monteggia injuries with associated radial head or neck fractures tend to have even worse outcomes \[[@R5]\], \[[@R19]\], \[[@R21]\], \[[@R24]\]. Givon et al. presented a combined series of Monteggia and equivalent lesions treated during a 10 year period \[[@R21]\]. They concluded that equivalent lesions with associated radial head fractures had worse functional outcomes than other types. Ten complications occurred in nine adult patients treated operatively. However, Givon et al. did not use validated outcome tools for assessing clinical and functional results, the equivalent lesion were not specifically evaluated, classified, and treated as separate clinical entities. Furthermore, children and adults were assessed together. Ring et al. reported on 48 patients with Monteggia injuries followed up after 6.5 years on average. The vast majority (38 patients; 79%) had Bado type II injuries of whom 68% (26 patients) had an associated radial head fracture. Ten of these patients also had a fracture of the coronoid process as a single large fragment. Most of the patients with Bado type II fractures had good or excellent results, with 112 degrees of ulnohumeral motion and 126 degrees of forearm rotation. The average point value for the functional rating index of Broberg and Morrey was 85 points indicating a good result. All five patients with a poor result suffered an associated fracture of the radial head. Four of these patients also had a fracture of the coronoid process. According to Ring et al., among the patients with a Bado type II fracture, cases with a fracture of the radial head were more likely to have an unsatisfactory result. However, a statistical difference between the patients with and without radial head fracture could not be detected. Although good clinical results were found in the majority of patients with Bado type II fractures, 50% required revision surgery within 4 months of the index procedure. Egol et al. retrospectively reviewed clinical and functional outcomes after surgical fixation of ipsilateral fractures of the proximal ulna with associated fractures of the radial head or neck, and/or radial head dislocation (Monteggia variant/Monteggia-like lesion) \[[@R19]\]. 20 patients were evaluated at a mean of 2.3 years. Egol et al. reported slightly worse results compared to the aforementioned study by Ring et al. with a mean Broberg and Morrey score of 79.1 and a mean DASH score of 64.1 (worse outcome than those of the general population). Nine of 20 (55%) patients had good or excellent scores compared to the 83% satisfactory results Ring et al. reported. In addition, all complications in their patients were associated with Bado type 2 fracture patterns. Furthermore, seven patients developed heterotopic ossification and 14 of 20 patients arthritic changes. Strauss et al. evaluated 23 patients with a Bado type II Monteggia injury associated with a fracture of the head or neck of the radius \[[@R25]\]. Six of those had an accompanied posterior ulnohumeral dislocation at the time of injury. The 17 patients without ulnohumeral dislocation were assessed at a mean follow-up of 29 months. In these patients the mean elbow flexion was 127° and the loss of extension was 5° at follow-up. The mean pronation was 60° and supination 67°. The mean standardized DASH score was 23 and the mean Broberg-Morrey functional index score was 83. This accounted for six excellent, four good, six fair and one poor outcomes. Radiographic assessment of post-traumatic arthritis showed six patients with Broberg-Morrey grade 0, nine with grade 1 and two with grade 2 changes. Mild heterotopic ossification was present anteriorly in two patients. Konrad et al. evaluated 27 Bado type II fractures as part of a population of 47 Monteggia injuries after a mean time of 8 years. In this series eleven patients suffered a radial head fracture, a fracture of a coronoid process, or a combination of both (Monteggia-like lesions) \[[@R23]\]. An intra-articular fracture of the radial head and a fracture of the coronoid process were correlated with a poor Broberg and Morrey Score. The authors concluded that Bado type II Monteggia fractures further classified as Jupiter type IIa fractures, are frequently associated with fractures of the radial head and the coronoid process. This injury pattern should be considered as negative prognostic factor for functional long-term outcome. Unfortunately, Konrad et al. did not report the score results of the evaluated Monteggia-like lesions separately. Therefore, a comparison to our results is hardly possible. However, compared to all aforementioned studies, where most of the authors did not particularly differentiate between Monteggia injuries and Monteggia-like lesion, our results are considerably better. What are the reasons for the good to excellent outcome scores and the low number of complications in our study? At first, all patients received a CT scan preoperatively to evaluate the severity of all existing concomitant injuries of the elbow joint. Hence, all radial head and coronoid fractures could be detected, classified, and addressed. Furthermore, all of our patients were treated according to the same distinct treatment strategy. This was not done in other studies persistently \[[@R21]\], \[[@R23]\], \[[@R24]\], \[[@R25]\]. According to Ring et al. the key treatment principle in Monteggia fractures is stable anatomic alignment of the ulna \[[@R24]\], \[[@R26]\]. In adults, this alignment is achieved with plate and screw fixation \[[@R24]\]. In contrast to all aforementioned studies all of our patients received osteosynthesis of the ulna using modern fixation techniques. In all cases a proximally contoured 3.5 mm LCP (locking compression plate) or precontoured 3.5 mm LCP Olecranon Plate was applied to the dorsal surface of the proximal ulna in dynamic compression mode. Since posterior tensile forces are encountered at the apex of the proximal end of the ulna with active motion, a plate applied to the lateral or medial surface of the ulna is mechanically inferior to a plate applied to the posterior surface of the ulna, which works as a tension band \[[@R24]\]. Ring et al. recommend fixation of the ulnar fracture with a stout plate, such as a 3.5-millimeter limited-contact dynamic compression plate, applied to the posterior surface of the ulna and contoured proximally to reach the tip of the olecranon \[[@R24]\]. Semitubular or one-third tubular plates as well as tension band-wire constructs seem to be not rigid or strong enough. The proximal contour allows to address the proximal fragment with more screws. The most proximal screws are oriented at 90 degrees to the more distal screws, creating a more stable construct \[[@R24]\], \[[@R26]\]. Until today, the question whether it is better to treat a severely comminuted fracture of the radial head associated with a Monteggia-like lesion with radial head excision, reconstruction or prosthetic replacement remains unanswered \[[@R19]\], \[[@R24]\]. In our case series a total of five radial head fractures were comminuted (Mason type III), while three were classified as Mason type II and two as Mason type I. All patients with a type III radial head fracture received a cemented bipolar radial head prosthesis. No patient showed signs of loosening of the radial head prosthesis at follow up. All type II radial head fractures were treated with open reduction and internal fixation using mini screws. In all cases of type I fracture the treatment of the radial head dislocation was closed reduction with verification under fluoroscopy. Egol et al. showed that patients with radial head reconstruction or resection surgery achieved similar elbow function \[[@R19]\]. However, Ring et al. found better results in patients who had resections rather than attempted internal fixation \[[@R24]\]. Reynders et al. recommended against early resection of the radial head to improve outcome \[[@R27]\]. Nevertheless, the radial head plays an important role as a secondary stabilizer of the elbow joint in the absence of the medial collateral ligament \[[@R25]\]. Preservation of the length of the radial column by fixation or replacement seems to be a mainstay in the treatment of these injuries \[[@R25]\]. Konrad et al. treated all Mason type II radial head fractures with open reduction and internal fixation, and showed good or excellent results. In the same study Mason type III fractures were treated by radial head resection or reconstruction with poor results. According to the authors, not using prosthetic replacements as a treatment option might have led to these poor results, as the preservation of radiocapitellar contact with a prosthesis or a reconstructed radial head might increase the stability of the ulnohumeral articulation \[[@R23]\]. Furthermore, radial head resection may not be advisable since this may cause a proximal migration of the radius as a result of the frequently associated lesion of the interosseous membrane \[[@R6]\]. Fractures of the coronoid process can lead to instability of the ulnohumeral joint \[[@R9]\], \[[@R23]\], \[[@R28]\]. Therefore, large fractures of the coronoid process should be reduced anatomically to restore the ulnohumeral articulation and minimize the risk of ulnohumeral arthritis \[[@R23]\], \[[@R24]\]. This can be achieved with interfragmentary compression screws inserted through the posterior surface of the ulna either through or adjacent to the plate \[[@R24]\]. According to the Morrey classification there were five type I, three type II, and two type III coronoid fractures in our case series. All coronoid fractures were associated with a Bado type II fracture and were stabilized with lag screws through the ulnar plate or with independent lag screws after indirect repositioning of the fracture. This could further account for the good results and the mild degenerative changes observed. Furthermore, patients received standardized postoperative treatment with early immediate continuous motion, which was not performed in other studies \[[@R19]\], \[[@R23]\], \[[@R24]\], \[[@R27]\]. Conclusion ========== The Monteggia-like lesions described in this series have rarely been reported before. Our findings demonstrate that good or excellent short-term results can be realized if the injury is classified correctly and a standardized surgical treatment of all components of the injury is achieved. Further studies with larger patient populations and longer follow up periods are needed to evaluate the long-term effectiveness of this treatment concept. Notes ===== Competing interests ------------------- The authors declare that they have no competing interests. ![Male patient (43 years) with a posterior Monteggia-like lesion (Bado type II, Jupiter type IIB), Mason type III fracture of the radial head, and Broberg and Morrey type III coronoid fracture. Radiographs on the day of the injury.](IPRS-04-13-g-001){#F1} ![Postoperative radiographs. Osteosynthesis of the ulna was performed using a proximally precontoured olecranon plate applied to the posterior surface of the ulna. Implantation of a cemented bipolar radial head prosthesis.](IPRS-04-13-g-002){#F2} ![Same patient 11 months after the injury: no degenerative changes and heterotopic ossifications. No signs of loosening of the radial head prosthesis. Good clinical outcome.](IPRS-04-13-g-003){#F3}	{ "pile_set_name": "PubMed Central" }
![In this work, HUVECs were mock infected or infected with HTNV for 3 days. Total RNA of the cells were analyzed by RNA-seq and obtained circRNA, miRNA, mRNA library. Differentially expressed (DE) RNAs were identified and subjected to GO analysis, KEGG analysis and ceRNA network construction. Then 8 DE circRNAs, 8 DE mRNAs, 6 DE miRNAs were verified by RT-qPCR. Besides, mRNAs (CMPK2, PARP10, GBP1, and RIG-I), circRNAs (circ_0000479), miRNAs (miR-149-5p, miR-411-3p, and miR-330-5p) in the ceRNA network were found effective to inhibit or promote virus replication. And the circ_0000479-miR-149-5p-RIG-I ceRNA axis was verified in HTNV infection.](fcimb-10-00097-g0008){#F8} Introduction {#s1} ============ Hantaviruses, which belong to the class Bunyavirales, order Hantaviridae, are widespread reemerging zoonotic pathogens. The hantaviral genome consists of three negative sense single-strand RNA segments, designated as S, M, and L, encoding nucleocapsid protein (NP), glycoprotein precursor (GPC), and viral RNA-dependent RNA polymerase (RdRp), respectively (Kariwa et al., [@B17]). Pathogenic hantaviruses infection in humans mainly causes hemorrhagic fever with renal syndrome (HFRS), hantavirus pulmonary syndrome (HPS) or hantavirus cardiopulmonary syndrome (HCPS), and a generally milder form of HFRS, which is also called nephropathia epidemica (NE) (Echterdiek et al., [@B8]). In China, HFRS is mainly caused by Hantaan virus (HTNV) and Seoul virus (SEOV) and the incidence and mortality of HFRS are highest of the world (Zheng et al., [@B55]). Research has shown that hantavirus mainly infects human vascular endothelial cells, causing destruction of capillary integrity, and altering vascular permeability (Jiang et al., [@B14]). Moreover, increased capillary leakage is the basic pathogenic feature for both HFRS and HCPS (Terajima and Ennis, [@B42]). Rapidly growing evidence is now mounting that there is a considerable overlap of clinical presentations between HFRS and HPS (Clement et al., [@B5]). The temporary capillary leak results in sometimes massive interstitial generalized fluid accumulation, leading to different degrees of multiple organ failure (MOF), mainly expressed as acute kidney injury (AKI), and/or as acute lung edema, with or without circulatory shock (Macneil et al., [@B28]; Jiang et al., [@B14]; Garanina et al., [@B9]). However, hantavirus is not a lytic virus. These findings strongly imply that hantavirus infection should regulate host cells function to coordinate viral propagation, but the association between HTNV-host cells interaction and HTNV pathogenesis has not been comprehensively studied. As much as 80% of the host genome is transcribed to non-translated RNA, which underscore the potential importance of host non-coding RNA (ncRNA) in the host-virus interaction. Accumulating evidence shows that host ncRNAs play key regulatory roles during viral replication and infection (Lamontagne et al., [@B19]; Ding et al., [@B7]; Landeras-Bueno and Ortin, [@B20]; Liu and Ding, [@B24]; Trobaugh and Klimstra, [@B43]; Tao et al., [@B41]). To date, few studies have been published about the functions of ncRNAs in HTNV infection. It was reported that the long non-coding RNA (lncRNA) NEAT1 controls HTNV infection by promoting RIG-I and DDX60 expression and interferon (IFN) responses (Ma et al., [@B27]). LncRNA NRIR facilitates HTNV infection by negatively regulating the transcription of IFITM3 (Xu-Yang et al., [@B48]). MicroRNA (miRNA) expression profiles have been studied in HTNV-infected A172 cells, but without a focus on the functions and mechanisms involved in the regulation of viral infection (Shin et al., [@B39]). Besides, our group demonstrated that miR-146a can negatively regulate the release of HTNV-induced pro-inflammatory cytokines to promote viral replication (Chen Q. Z. et al., [@B3]). Additional to lncRNAs and miRNAs, circular RNAs (circRNAs) are a new category of endogenous ncRNA (Qu et al., [@B36]). The majority of circRNAs regulate the gene expression by sponging miRNA, serving as competing endogenous RNA (ceRNA) (Hansen et al., [@B12]; Panda, [@B32]); this phenomenon has been identified in many diseases (Wang M. et al., [@B45]; Mao et al., [@B29]; Rong et al., [@B37]). More importantly, recent studies have found that circRNAs have substantial effects in host-virus interactions. For the integrative analysis of circRNAs, miRNAs and genes, it was found that the differentially expressed circRNAs were involved in cellular responses to Herpes Simplex Virus 1, human cytomegalovirus, and HIV infection through the circRNA-miRNA-gene regulatory axis. Moreover, these genes regulated by circRNAs are enriched in inflammatory response, defense response to virus, and pathways of apoptosis, which may account for viral pathogenesis and cellular immunity (Shi et al., [@B38]; Zhang et al., [@B53]; Lou et al., [@B25]). Artificial circRNAs inhibit the production of HCV viral proteins by effectively adsorbing cellular miR-122 (Jost et al., [@B16]), and the delivery of purified circRNAs can activate RIG-I-mediated immune responses and provides effective protection against viral infection (Chen Y. G. et al., [@B4]). Upon viral infection, the nuclear export of NF90/NF110 to the cytoplasm can decrease the expression of circRNA, and the dissociation of NF90/NF110 from circRNPs can inhibit viral replication (Li et al., [@B23]). However, so far, the specific characteristics and roles of circRNAs and circRNA-associated ceRNAs in HTNV infection have not yet been reported. Herein, we investigated and analyzed the profiles of circRNAs expression and the potential roles of circRNAs in HTNV-host interaction through intergrated omics data of RNA sequencing and miRNA sequencing on mock-infected and HTNV-infected HUVECs. The identified network of circRNAs and circRNA-associated ceRNAs could reveal the complex regulatory function of circRNAs in the interplay between HTNV and host. In summary, our results provide novel insight regarding the roles of circRNAs in HTNV infection and broaden our understanding of the mechanisms underlying HTNV pathogenesis. Materials and Methods {#s2} ===================== Cell Culture and Viral Infection -------------------------------- HUVECs were purchased from ScienCell (Cat No. 8000, Carlsbad, USA) and cultured in endothelial cell medium (ECM) with 5% fetal bovine serum, 1% penicillin-streptomycin, and 1% endothelial cell growth supplement (ScienCell) at 37°C and 5% CO~2~. HEK293T and Vero E6 cells were purchased from the American Type Culture Collection (ATCC) and cultured in Dulbecco\'s modified Eagle\'s medium (DMEM) supplemented with 10% FBS (Gibco) at 37°C and 5% CO~2~. The HTNV 76-118 strain was obtained from the Institute of Virology, Chinese Center for Disease Control and Prevention (CDC, Beijing, China) and propagated in Vero E6 cells. HUVECs were mock-infected or HTNV-infected at a multiplicity of infection (MOI) of 1. RNA Extraction, RNAs Libraries Construction ------------------------------------------- mRNA library construction: At 72 h post-infection, total RNA was extracted from mock-infected and HTNV-infected HUVECs with TRIzol (Invitrogen, USA), followed by the assessment of concentration, purity, and integrity using the ND-1000 Nanodrop (Thermo Fisher, USA) and Agilent 2200 TapeStation (Agilent Technologies, USA). Ribosomal RNA (rRNA) was removed from total RNA using an Epicenter Ribo-Zero rRNA Removal Kit (illumina, USA) and RNA was fragmented to \~200 bp. Single stranded cDNA was synthesized by reverse transcription, which was followed by double-stranded cDNA synthesis. Subsequently, the fragments were ligated and purified, and PCR was performed to amplify sequences. RNA was optimized and purified using the NEBNext® Ultra™ Directional RNA Library Prep Kit for Illumina (NEB, USA). The Agilent 2200 TapeStation and Qubit®2.0 (Life Technologies, USA) was employed to evaluated the purified library products. The quality-tested RNA was sequenced on a HiSeq3000 (Illumina) using the pair-end flow cell, and the mRNA library was obtained. CircRNA libraries were constructed as follows: Total RNA was isolated from HUVECs by the TRIzol Reagent (Invitrogen) according to the manufacturer\'s protocol. RNA purity was assessed by ND-1000 Nanodrop with a criteria of A260/280 ≥1.8 and A260/A230 ≥2.0. RNA integrity (RIN) was evaluated by Agilent 2200 TapeStation (Agilent Technologies, USA) requiring RIN ≥7.0. After that, \~2 μg of total RNA from each sample was subjected to the Epicenter Ribo-Zero rRNA Removal Kit (illumina, USA) to remove ribosomal RNA, followed by incubation with 40 U RNase R for 3 h at 37°C (Epicenter, USA) to remove linear RNA. Subsequently, the purified RNA were fragmented to \~200 bp and were subjected to first strand and second strand cDNA synthesis, and adaptor ligation. CircRNA libraries were prepared by using NEBNext® Ultra™ RNA Library Prep Kit for Illumina (NEB, USA) according to instructions. The Agilent 2200 TapeStation and Qubit®2.0 (Life Technologies, USA) was employed to control the purity of library products and then sequenced on HiSeq 3000 (Illumina) with 2 × 150 bp mode (Memczak et al., [@B31]). Small RNA libraries were constructed as follows: Total RNA was isolated, and the purity and integrity were evaluated as mentioned above. Each sample had a RIN^e^ (RNA integrity number) value \>7.0. In brief, RNAs successively were ligated with the 3′RNA adapter and 5′ adapter. Subsequently, the ligated RNAs were performed to RT-PCR and amplification. Then the PCR products of 18--40 nt were selected by polyacrylamide gel electrophoresis (PAGE) gel according to instructions of the NEBNext® Multiplex Small RNA Library Prep Set for Illumina® (Illumina, USA). Thereafter, the purified library products were evaluated using the Agilent 2200 TapeStation and sequenced on HiSeq2500 platform. Quantification and Differential Expression Analysis of circRNAs, miRNAs, and mRNAs ---------------------------------------------------------------------------------- ### Identification and Quantification of circRNAs As shown in [Figure S1A](#SM7){ref-type="supplementary-material"}, the raw data obtained from RNA sequencing were first filtered by Trimmomatic-0.36 program to remove adapter sequences, invalid reads and low quality reads. Afterwards, rRNA residues were removed by comparing known rRNA information in RNA central database to gain clean reads. Clean reads were then mapped to the human reference genome (hg19) from UCSC genome database (<http://genome.ucsc.edu/>) by Tophat (v2.0.13) program. The unmapped reads were performed fusion gene alignment using Tophat-Fusion (v2.0.13) to identify candidate circRNAs (Zheng et al., [@B54]). The criteria were as follows: GT/AG splicing signals, mismatch ≤ 2, back-spliced junction reads ≥1, and distance between two splice sites is no more than 100 kb as previous reported (Shi et al., [@B38]). The circRNA expression level was described as RPM (Reads Per Million mapped reads). Differential expression analysis of circRNAs were performed with edgeR package (v3.20.9), and identified by threshold values of \\|log~2~(fold-change)\\| \>1 and q \< 0.05. ### Analysis of mRNAs Expressoin As shown in [Figure S1B](#SM7){ref-type="supplementary-material"}, the reads containing adapter sequences, low-quality reads and undetermined bases were cleaned from raw data by Trimmomatic-0.36 program. Then high-quality clean reads were gained through removing rRNA residues. The clean reads were aligned to the reference human genome (hg19) using Tophat (v2.0.13). The RefSeq databases were chosen as the annotation references. The mRNAs expression were calculated as Reads Per Kilobase per Millions reads (RPKM). The differential expression of mRNAs were analyzed with edgeR package (v3.20.9). The criterion of \\|log~2~(fold-change)\\| \>1 and q \< 0.05 was used to identify differentially expressed mRNAs. ### Analysis of miRNAs Expressoin As shown in [Figure S1C](#SM7){ref-type="supplementary-material"}, the raw reads obtained from miRNA library were first filtered by Cutadapt 1.8.1. The clean reads were then mapped to hg19 reference genome using BWA v0.7.12 to identify miRNAs. The miRNAs in each group were identified basing on the known miRNAs downloaded from the miRBase database v21.0 ([www.mirbase.org](http://www.mirbase.org)) (Wu et al., [@B47]). The miRNAs expression were calculated by reads per million clean tags (RPM). Differential miRNAs were identified using thresholds: \\|log~2~(Fold-change)\\|\>1 and p \< 0.05. RNA-seq data were uploaded into the gene expression omnibus (GEO) database (<https://www.ncbi.nlm.nih.gov/geo/>) (accession nos. [GSE133634](GSE133634), [GSE133751](GSE133751), [GSE133319](GSE133319)). GO and KEGG Pathway Enrichment ------------------------------ Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed to analyze the advanced functions of DE genes, DE miRNA targeted genes, and parental genes of DE circRNAs. GO analysis labels gene with a function, such as molecular function, biological process, or cellular component. KEGG analysis provides annotation information of signal transduction and disease pathways for genes, thus providing a basis for gene function and pathway research. The analyses were based on the GO resource (<http://www.geneontology.org>), the KEGG database (<http://www.genome.jp/kegg/>), and KOBAS 2.0 software. A P \< 0.05 was considered statistically significant enrichment. Co-expression Network Analysis ------------------------------ ### Construction of circRNA-miRNA Co-expression Network Firstly, we selected 8 DE circRNAs verified by real-time quantitative PCR, and then identified miRNAs targeting on these circRNAs using miRanda (<http://www.microrna.org/microrna/home.do>), TargetScan (<http://www.targetscan.org/vert_72/>), and RNAhybrid (<https://bibiserv.cebitec.uni-bielefeld.de/rnahybrid>), which were widely used public databases. The candidate miRNA-circRNA relationships were picked out by thresholds of miRanda (Total Sore ≥ 140; Total Energy \< −17 kmol), RNAhybrid (minimum free energy ≤ −21) and Targetscan (conserved groups). ### Construction of miRNA-mRNA Co-expression Network TargetScan (<http://www.targetscan.org/vert_72/>), miRDB (<http://www.mirdb.org/>), miRTarBase (<http://miRTarBase.mbc.nctu.edu.tw/>), and miRWalk (<http://zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk2/>) were applied to predict the target DE mRNAs of miRNAs of circRNA-miRNA network. The intersection of mRNAs predicted by at least 2 databases was used to establish miRNA-mRNA network. ### ceRNA Network Analysis In the constructed miRNA-mRNA network, we selected 80 mRNAs enriched in the innate immune pathway based on the GO and KEGG analyses as candidates. These mRNAs were then confirmed to predict the miRNA-mRNA pairs. Next, circRNA-miRNA pairs sharing the same miRNAs with above miRNA-mRNA pairs were identified as candidate circRNA-miRNA-mRNA competing interaction. Pearson correlation coefficient (r) and correlation P-value (cP-value) were used to estimate the co-expression relationship between circRNAs, mRNAs, and miRNAs. The strong interactions between miRNAs, circRNAs, and mRNAs (with a threshold of r \> 0.85 or r \< −0.85 and cP-value \< 0.05) were selected to establish ceRNA network using Cytoscape (v3.4.0). Transfection and Viral Infection -------------------------------- Specific miRNA mimics, mimic control, siRNAs against circRNA and mRNA, and siRNA control were synthesized by Ribobio (Guangzhou, China). The siRNA sequences were listed in [Table 1](#T1){ref-type="table"}. HUVECs were \~50--80% confluent, followed by transfection with 50 nM of siRNAs or 10 nM of mimics using Highgene Transfection Reagent (ABclonal, China) following the manufacturer\'s protocol. At 24 h post-transfection, besides detection of uptake efficiency, HUVECs were rinsed three times by PBS and then infected with a certain volume of viral suspension gained a multiplicity of infection (MOI) of 1. After incubation for 2 h, the viral suspension was discarded and replaced with ECM medium to culture for another 72 h. ###### Target sequence of siRNAs. siRNA Target sequence ---------------- ----------------------- siRIG-I TTTGTCGCTAATCCGTGAT siCMPK2_01 CCAGGTTGTTGCCATCGAA siCMPK2_02 GGAGGAGTGTACCTCCTTT siCMPK2_03 GGCCTCCGAAATAGCTAAA siISG15_01 TGCGACGAACCTCTGAGCA siISG15_02 TCCTGGTGAGGAATAACAA siISG15_03 GGTGGACAAATGCGACGAA siPARP10_01 GGTAGAGGGATTATGACAA siPARP10_02 CTACCATGAGGACCTTCTT siPARP10_03 GCAGCATTAGCTGCCATGT siSAMD9_01 GGTTAGACCTTGACAGTGA siSAMD9_02 GAACAGGTAACCAGTTTAA siSAMD9_03 GAACCTGACAATTCTTATA siGBP1_01 AAGGCATGTACCATAAGCT siGBP1_02 GAGACGACGAAAGGCATGT siGBP1_03 GGAGTCTATGACTGATGCA siCirc_0000479 GAGCATCAGCAATACACAAGT siCirc_0006132 ATACACAAGACCGATGTTGTT siCirc_0007793 AGGAAGCAAAAGAAGGAAAAA siCirc_0046034 AGCAGGTGTTGGTCTTCTGCA Validation of circRNAs, miRNAs, and mRNAs Expression Using Real-Time Quantitative PCR (RT-qPCR) Analysis -------------------------------------------------------------------------------------------------------- To confirm the results of sequencing, firstly, the simple random sampling method was used in selecting RNAs with random numbers generated in Microsoft EXCEL. Whereafter, total RNAs of HTNV-infected and mock-infected cells were extracted using TRIzol reagent (Invitrogen, USA), and reversely transcribed by reverse transcriptase (M-MLV, Promega) according to the manufacturer\'s protocol. Then, RT-qPCR was performed with SYBR Green qPCR Master Mix (Bio-Rad, USA) on CFX96 real-time PCR system. The used primers were listed in [Table 2](#T2){ref-type="table"}. The relative quantification of miRNAs, circRNAs, and mRNAs were normalized to U6 or GAPDH with the 2^−ΔΔCT^ method. ###### Primers used for quantitative real time PCR. RNA Forward primer Reverse primer -------------- ------------------------- -------------------------- CMPK2 GCCAACAGTGTGTTTCGTCA TACCGTCTGCAGGACCTTTTC CXCL11 TGTCTTTGCATAGGCCCTGG TGATTATAAGCCTTGCTTGCTTCG ISG15 GAGAGGCAGCGAACTCATCTT CCAGCATCTTCACCGTCAGG USP18 GGCTCCTGAGGCAAATCTGT CAACCAGGCCATGAGGGTAG RIG-I ATGGGACGAAGCAGTATTTAG GCTTGGGATGTGGTCTACTC SAMD9 TGGGGGAACTACCTTGGCTA CGGTTCATTGCCCCATAGGT GBP1 TTGAGAACACTAATGGGCGACT TAGGATTTGCCTGTGCGGT PARP10 GACACAGGCCTTGAAGAGGTG CCTGGTCACCACCTGTACTGTC circ_0000479 GAAGCATTTAGAGAGCATCAGCA GCTCCGCAATTCTTTGTATCTCAT circ_0002470 CTGATGAGCTGGTCTGCGAT TCCTTACCCGCATTCTCACC circ_0006132 TGCACCCGAGTGATCATGAAA CCCTTGTCCAACTGTGTCCAA circ_0007793 ACGATCAACGGCGAATGACT GGGCACAGGGTGAAGATACC circ_0022228 AGATCTTTGCCAGCATCCCA GTTGTACAGGACCTCCCCAAT circ_0046034 TGTTCAATGACCGGCTGTGT ATCCACGTCCTGAACAGCATA circ_0066996 CTGCAATGTGGTATGCAGAGTT TCACACAGCTGACGCTCATT circ_0073948 GAACGAAACACGCCATCTGC GGGCTGAGCTGGTATGAGTC Western Blotting ---------------- As described in our previous reports (Chen Q. Z. et al., [@B3]), cells were lysed by RIPA. Protein concentration was measured with BCA protein assay kit. Equal amounts of protein were heated at 95°C for 5 min. Protein was separated by SDS-PAGE and transferred onto polyvinylidene fluoride (PVDF) membranes. Subsequently, the PVDF membrane was blocked with 5% BSA for 40 min at room temperature and then incubated with the primary antibody at 4°C overnight. Finally, the signals on the PVDF membrane were visualized. The primary monoclonal anti-HTNV 76-118 strain antibody was purchased from Abnova (MAB5482, Taiwan). GAPDH was used as internal control. The blots were scanned by ImageJ software. Luciferase Reporter Assay ------------------------- The RIG-I binding region (RIG-I-BR) was amplified by primers (forward 5′-TGC CAT TGT TCT CTT TGA CC-3′ and reverse 5′-GCA ACT ACT GTG TCA TGT AC-3′) and cloned into pmir-GLO. Circ_0000479 full length was amplified by primers (forward 5′-CAC AAG TGC ATA CAC CTT GAT AG-3′ and reverse 5′-TAT TGC TGA TGC TCT CTA AAT GC-3′) and cloned into pmir-GLO and pcDNA3.1(+). To measure relative luciferase activity, 5 nM of miRNA and 1.2 μg/ml of luciferase reporter DNA or 5 nM of miRNA, 0.6 μg/ml of luciferase reporter DNA and 0.6 μg/ml pcDNA3.1 vector DNA were co-transfected into HEK293T cells. After 48 h of transfection, cells were lysed on ice for 15 min before centrifugation. The supernatant was measured using the Promega Dual-Glo luciferase assay system. Firefly luciferase activity was normalized to that of Renilla. Statistical Analysis -------------------- Data were expressed as mean ± standard deviation (SD) from at least three independent experiments. SPSS 17.0 software was used for statistical analyses. Differences were determined to be statistically significant at values of P \< 0.05 by Student\'s t-test for two groups, one-way ANOVA analysis for multiple groups. Single, double and triple asterisks, and ns indicate statistical significance (^\^P* \< 0.05; ^\\^P \< 0.01; ^\\\^P* \< 0.001) and no significance respectively. Results {#s3} ======= Identification and Characterization of circRNAs in HTNV-Infected and Mock-Infected HUVECs ----------------------------------------------------------------------------------------- We first performed circRNA sequencing, using the rRNA-depleted samples of mock-infected (CON) and HTNV-infected (HTNV) cell cultures. We identified an average of 87 million and 84 million clean reads in the CON and HTNV groups, respectively. The clean reads were used for the identification of circRNAs, and most reads were mapped to the reference genome. The most of circRNAs (accounted for 82%) were composed of exons, while 7% were intronic, and 4% mapped to intergenic regions ([Figure 1A](#F1){ref-type="fig"}). We found that the size of the circRNAs ranged from \<200 to \>2,000 nt; 81.04% of circRNAs had a predicted spliced length of \<1,000 nt; 54.93% had a length \<500 nt and 26.11% had a length of 500--1,000 nt ([Figure 1B](#F1){ref-type="fig"}). Based on the filtering criteria (Materials and Methods), a total of 70 DE circRNAs were identified in HTNV-infected samples compared with mock-infected samples ([Table S1](#SM1){ref-type="supplementary-material"}). Of these, 65 circRNAs were up-regulated and five were down-regulated ([Figure 1C](#F1){ref-type="fig"}). DE circRNAs were widely distributed across chromosomes. Chromosome 1, chromosome 11, chromosome 17, and chromosome 20 contained many up-regulated circRNAs. The five down-regulated circRNAs were distributed on different chromosomes ([Figure 1D](#F1){ref-type="fig"}). ![circRNA expression overview. (A) circRNAs category chart. (B) circRNAs length distribution. (C) Volcano plot of DE circRNAs upon HTNV infection in HUVECs. Red dots represent up-regulated circRNAs and green dots represent down-regulated circRNAs. (D) Chromosome distribution of DE circRNAs. (E) Heatmap and clustering analysis of DE circRNAs. Each row represents one circRNA and each column represents one sample; −2, −1, 0, 1, and 2 represent fold change. Red indicates high expression and blue represents low expression. CON-1, CON-2 and CON-3 represent three mock-infected samples; HTNV-1, HTNV-2, and HTNV-3 represent three HTNV-infected samples. (F) Verification of dysregulated circRNAs. HUVECs were infected with HTNV 76-118 (MOI = 1) for 3 days. Then the total RNA was extracted and the expression levels of circRNAs were measured by RT-qPCR. Student\'s t-test, mean ± standard deviation (SD), \*P \< 0.05; \\P \< 0.01; \\\*P \< 0.001. The experiment was performed at least three times independently.](fcimb-10-00097-g0001){#F1} Further, we performed a hierarchical cluster analysis of all DE circRNAs; the results were shown in [Figure 1E](#F1){ref-type="fig"}. Most of the dysregulated circRNAs were lowly expressed in the mock-infected group, but were up-regulated in the HTNV-infected group. The hierarchical cluster analysis classified the expression patterns of mock-infected samples and HTNV-infected samples into different branches ([Figure 1E](#F1){ref-type="fig"}). To confirm the RNA-seq results, we verified the expression of eight randomly selected circRNAs (circ_0000479, circ_0002470, circ_0006132, circ_0007793, circ_0022228, circ_0046034, circ_0066996, and circ_0073948) by RT-qPCR. As shown in [Figure 1F](#F1){ref-type="fig"}, the RT-qPCR results were highly consistent with the RNA-seq results. Functional Analysis of the Parental Genes of DE circRNAs -------------------------------------------------------- To elucidate the biological function of DE circRNAs in HTNV infection, we examined the parental genes of DE circRNAs by functional analysis. Both GO and KEGG analysis were performed in this process. GO annotation analysis displayed that the parental genes of DE circRNAs were primarily enriched in diverse biological processes, including "cellular response to type I interferon," "type I interferon signaling pathway," "response to type I interferon," and "defense response to virus" ([Figure 2A](#F2){ref-type="fig"}). Moreover, the KEGG pathway enrichment analysis showed that plenty of the parental genes were principally enriched in "innate immune response," "defense response to virus," "cytokine-mediated signaling pathway," and "response to virus" ([Figure 2B](#F2){ref-type="fig"}). These data indicated that the parental genes of DE circRNAs are mainly involved in the host immune response to viral infection. ![Verification of DE circRNAs after HTNV infection and exploration of their effect on HTNV infection. (A) GO functional enrichment analysis of parental genes of DE circRNAs. The x-axis shows the P-value and gene numbers, and the y-axis shows the GO term. (B) The 30 most enriched KEGG pathways based on hosting genes of dysregulated circRNA during HTNV infection. The x-axis shows the enrichment factor, and the y-axis shows the pathway names. The point size represents the number of genes enriched in a particular pathway. (C) The silencing efficiencies of siRNAs targeting the connected sites of circRNAs. Student\'s t-test, mean ± standard deviation (SD), \*P \< 0.05; \\P \< 0.01; \\\*P \< 0.001. (D) Western blot assay of HTNV NP expression in HUVECs with knockdown of the circRNAs circ_0000479, circ_0006132, circ_0007793, and circ_0046034. One-way ANOVA with Dunnett\'s multiple comparison test. Mean ± SD, \*P \< 0.05; \\P \< 0.01; \\\*P \< 0.001; ns, no significance. The experiments were performed at least three times independently.](fcimb-10-00097-g0002){#F2} To gain insight into the effect of DE circRNAs on HTNV infection, we designed siRNAs specific for the head-tail junctions of various circRNAs, and transfected them into HUVECs. Knockdown efficiency was verified ([Figure 2C](#F2){ref-type="fig"}). Then we found that knockdown of circ_0000479 could significantly increase the expression of HTNV NP ([Figure 2D](#F2){ref-type="fig"}), suggesting circ_0000479 might be involved in the regulation of host-HTNV interplay. Nevertheless, the functional mechanisms of circRNAs remain to be further explored. Identification of DE mRNAs Between HTNV-Infected and Mock-Infected HUVECs ------------------------------------------------------------------------- We analyzed the mRNA expression profiles in HTNV-infected and mock-infected HUVECs. We detected 788 DE mRNAs, i.e., 690 up-regulated and 98 down-regulated mRNAs ([Figures 3A,B](#F3){ref-type="fig"}, [Table S2](#SM2){ref-type="supplementary-material"}). To verify the RNA-seq results, we randomly selected eight mRNAs (CMPK2, CXCL11, ISG15, USP18, RIG-I, SAMD9, GBP1, and PARP10) and measured their expression levels after HTNV infection by RT-qPCR ([Figure 3C](#F3){ref-type="fig"}). They were all markedly up-regulated, in agreement with RNA-seq data. We next examined the functions of the DE mRNAs by GO analysis and KEGG pathway analysis. The top three significantly enriched GO terms were "innate immune response," "immune response," and "response to type I interferon" i.e., biological functions ([Figure 3D](#F3){ref-type="fig"}). This implied that the cellular antiviral response was activated by HTNV infection. KEGG analysis showed that the enriched pathways were mainly involved in "TNF signaling pathway," "Influenza A," and "Herpes Simplex infection" ([Figure 3E](#F3){ref-type="fig"}). ![Identification of dysregulated mRNAs. (A) Heatmap and clustering analysis of DE mRNAs. Each row represents one mRNA and each column represents one sample; −2, −1, 0, 1, and 2 represent fold change. Red indicates high expression and blue represents low expression. CON-1, CON-2, and CON-3 represent three mock-infected samples; HTNV-1, HTNV-2, and HTNV-3 represent three HTNV-infected samples. (B) DE mRNAs count. (C) Verification of dysregulated mRNAs. HUVECs were infected with HTNV 76-118 for 3 days (MOI = 1). Then the total RNA was extracted and expression levels of the selected mRNAs were measured by RT-qPCR. Student\'s t-test, mean ± SD, \\\*P \< 0.001. The experiment was performed at least three times independently. (D) GO functional enrichment analysis of DE genes. (E) Top 10 KEGG pathways of DE genes.](fcimb-10-00097-g0003){#F3} Identification of DE miRNAs Between HTNV-Infected and Mock-Infected HUVECs -------------------------------------------------------------------------- To investigate the miRNA expression profiles of HTNV-infected and mock-infected HUVECs, six small RNA libraries were constructed and sequenced. Sequencing analysis revealed that a total of 66 miRNAs were significantly dysregulated upon HTNV infection; 25 were up-regulated and 41 were down-regulated ([Figures 4A,B](#F4){ref-type="fig"}, [Table S3](#SM3){ref-type="supplementary-material"}). We verified the up- and down-regulation of six randomly selected DE miRNAs (up-regulated miRNAs: miR-3614-5p and miR-330-5p; down-regulated miRNAs: miR-411-3p, miR-744-5p, miR-758-3p, and miR-149-5p) by RT-qPCR. The RT-qPCR data were in agreement with RNA-seq data ([Figure 4C](#F4){ref-type="fig"}). Next, we predicted the targets of the top 31 DE miRNAs; 2676 candidate target genes were obtained. GO analysis indicated that these target genes were mainly involved in "cell morphogenesis involved in differentiation," "nervous system development," and "neuron projection morphogenesis" ([Figure 4D](#F4){ref-type="fig"}). Furthermore, the KEGG analysis manifested that the target genes of the DE miRNAs were mostly involved in "neurotrophin signaling pathway," "Wnt signaling pathway," "proteoglycans in cancer," "MAPK signaling pathway," and "renal cell carcinoma" ([Figure 4E](#F4){ref-type="fig"}). ![Identification of dysregulated miRNAs. (A) Heatmap and clustering analysis of DE miRNAs. Each row represents one miRNA and each column represents one sample; −2, −1, 0, 1, and 2 represent fold change. Red indicates high expression and blue represents low expression. CON-1, CON-2, and CON-3 represent three mock-infected samples; HTNV-1, HTNV-2, and HTNV-3 represent three HTNV-infected samples. (B) DE miRNAs count. (C) Verification of DE miRNAs upon HTNV infection. HUVECs were infected with HTNV 76-118 (MOI = 1) for 3 days. Then the total RNA was extracted and the expression levels of miRNAs were measured by RT-qPCR. Student\'s t-test, mean ± SD, \\\*P \< 0.001. The experiment was performed at least three times independently. (D) GO functional enrichment analysis of target genes of DE miRNAs. The x-axis shows the P-value and gene numbers, and the y-axis shows the GO term. (E) KEGG pathway analysis revealed the top 10 enriched pathways of target genes of DE miRNAs.](fcimb-10-00097-g0004){#F4} Analysis of Regulatory Network of DE circRNAs, miRNAs, and mRNAs ---------------------------------------------------------------- It was recently identified that circRNA can competitively sponge miRNAs, by which indirectly regulate gene expression (Wang M. et al., [@B45]; Mao et al., [@B29]; Rong et al., [@B37]). To examine the involvement of DE circRNAs, an integrative analysis of interplay between circRNAs and their target miRNAs was performed to elucidate their functional connections. The analysis of miRNA binding sites in circRNAs showed that 16 miRNAs, i.e., five up-regulated and 11 down-regulated miRNAs, had potential interactions with eight selected circRNAs ([Figure 5A](#F5){ref-type="fig"}, [Table S4](#SM4){ref-type="supplementary-material"}). Then we predicted 320 target genes of these 16 DE miRNAs and constructed a miRNA-mRNA interaction network ([Figure 5B](#F5){ref-type="fig"}, [Table S5](#SM5){ref-type="supplementary-material"}). Finally, in order to investigate the possibility that circRNAs act as ceRNAs in HTNV infection, we selected 80 enriched mRNAs that were involved in the innate immune pathway based on GO and KEGG analyses and constructed a ceRNA network through integrating the above circRNA-miRNA and miRNA-mRNA interaction networks ([Figure 5C](#F5){ref-type="fig"}, [Table S6](#SM6){ref-type="supplementary-material"}). As shown in [Figure 5C](#F5){ref-type="fig"}, circRNAs (circ_0000479, circ_0046034) may function as ceRNAs and sequester miR-149-5p to relieve its binding and targeting of mRNAs (RIG-I, IL16, MX2, etc.). RIG-I, also known as DDX58, was reported to inhibit HTNV replication by facilitating HTNV-induced IFN-β production (Ma et al., [@B27]). HTNV infection enhanced the production of multiple cytokines, including interleukin (IL) 6 (Jiang et al., [@B15]; Yu et al., [@B50]; Guo et al., [@B11]). Besides, we previously demonstrated that myxovirus resistance 2 (MX2, also known as MXB) served as a potential inhibitor in HTNV infection (Li et al., [@B22]). Taken together, the circRNAs/miRNAs could play important roles in HTNV-host interaction through regulating the expression of RIG-I, IL-6, and MXB associated with HTNV infection. ![circRNA-miRNA-mRNA interaction network. (A) circRNA-miRNA regulatory network. Circles represent circRNA and triangles represent miRNA. (B) miRNA-mRNA regulatory network. Triangles represent miRNA and dots represent mRNA. (C) ceRNA co-expression network. Circles represent circRNA, triangles represent miRNA, and rhombi represent mRNA. Red and green represent up-regulated and down-regulated RNAs, respectively.](fcimb-10-00097-g0005){#F5} Function of Dysregulated mRNAs and miRNAs in ceRNA Network Upon HTNV Infection ------------------------------------------------------------------------------ In order to figure out whether mRNAs and miRNAs in the ceRNA network have effects on viral infection, we investigated the effect of selected six representative mRNAs (CMPK2, ISG15, PARP10, SAMD9, GBP1, and RIG-I) involved in the ceRNA network upon HTNV infection. Firstly, we designed and transfected three specific siRNAs into HUVECs to decrease mRNA expression ([Figure 6A](#F6){ref-type="fig"}). Then, we selected the most interference efficient siRNAs of each mRNA and transfected them into cells followed by HTNV infection to detect the expression levels of viral NP. The results showed that knockdown of CMPK2, PARP10, GBP1, and RIG-I could significantly increase the expression of HTNV NP ([Figures 6B,C](#F6){ref-type="fig"}), suggesting these mRNAs could play a role in the cellular response against HTNV infection. ![Exploration of the function of DE mRNAs and miRNAs upon HTNV infection in ceRNA network. (A) The interfering efficiencies of siRNAs targeting the different sites of mRNAs. (B) HUVECs were transfected with siCON or efficient siRNAs. 24 h after transfection, the cells were infected with HTNV at an MOI of 1 for 3 days. The expression of HTNV NP in HUVECs was measured by Western-blotting assay. (C) The expression of RIG-I and HTNV NP in HUVECs after interfering RIG-I and infected with HTNV. (D) The overexpression effect of miRNA mimics in HUVECs. (E) Western-blotting assay of HTNV NP expression in HUVECs transfected with miRNA mimics. Student\'s t-test, One-way ANOVA with Dunnett\'s multiple comparison test. Means ± SD, \*P \< 0.05; \\P \<0.01; \\\*P \< 0.001; ns, no significance. The experiments were performed at least three times independently.](fcimb-10-00097-g0006){#F6} Meanwhile, to explore the role of DE miRNAs, we first transfected specific mimics of five representative miRNAs (miR-149-5p, miR-411-3p, miR-758-3p, miR-330-5p, and miR-3614-5p) into HUVECs and measured the miRNA expression levels by RT-qPCR ([Figure 6D](#F6){ref-type="fig"}). After that, HUVECs were challenged with HTNV, and the expression level of viral NP was assessed. The data showed that miR-149-5p and miR-411-3p markedly increased the expression of HTNV NP, while miR-330-5p significantly inhibited viral replication ([Figure 6E](#F6){ref-type="fig"}). Verification of ceRNA Network ----------------------------- We selected circ_0000479, RIG-I, and miR-149-5p from the ceRNA network, which have been verified to have a significant regulatory effect on HTNV replication based on the previous results, to verify the existence of the circRNA-miRNA-mRNA regulatory axis during viral infection by dual-luciferase reporter assay. The putative binding sites of miR-149-5p to RIG-I and circ_0000479 were first identified ([Figure 7A](#F7){ref-type="fig"}). miR-149-5p significantly inhibited the luciferase activity of RIG-I-BR and circ_0000479 ([Figure 7B](#F7){ref-type="fig"}). Furthermore, overexpression of circ_0000479 abrogated the inhibition of RIG-I-BR by miR-149-5p ([Figure 7C](#F7){ref-type="fig"}). The results implicated that circ_0000479 can competitively bind miR-149-5p, thereby alleviating the inhibition of miR-149-5p on RIG-I. These results indicated that the complicated circRNA-miRNA-mRNA network is active during HTNV infection and suggested that these ceRNAs can not only regulate viral replication, but also constitute interactive networks among themselves. ![Verification of circRNA-miRNA-mRNA axis. (A) Putative binding sites of miR-149-5p to RIG-I and circ_0000479. (B) 293T cells were co-transfected with NC mimics or miR-149-5p and GLO plasmids harboring the RIG-I binding region (RIG-I-BR) and circ_0000479 for 48 h. Cells were lysed and then subjected to dual-luciferase reporter analysis. (C) 293T cells were co-transfected with miR-149-5p or negative control, GLO-RIG-I-BR, and circ_0000479 overexpressing plasmids. Cells were lysed and then subjected to dual-luciferase reporter analysis. Student\'s t-test, one-way ANOVA with Bonferroni\'s correction comparison test. Means ± SD, \*P \< 0.05; \\P \< 0.01; \\\*P \< 0.001; ns, no significance. The experiments were performed at least three times independently.](fcimb-10-00097-g0007){#F7} Discussion {#s4} ========== The function of circRNAs has primarily been illustrated in cancer and disease progression (Greene et al., [@B10]; Zhang Y. et al., [@B52]; Qu et al., [@B35]). Nevertheless, the expression profiles, functions, and mechanisms of action of circRNAs in viral infection remain largely unknown. Several studies have addressed the identification, characterization, and function of circRNAs during viral infection (Shi et al., [@B38]; Wang S. et al., [@B46]; Zhang et al., [@B53]), but the properties and potential roles of circRNAs during HTNV infection have not been explored. In the present study, we systematically analyzed circRNA, miRNA, and mRNA expression profiles in HTNV-infected HUVECs. We found that HTNV infection significantly impacts circRNA expression profiles. A total of 70 DE circRNAs were identified in HTNV-infected compared with mock-infected HUVECs. Of these, 65 circRNAs were up-regulated and five were down-regulated. The length, categories, and chromosomal distribution of DE circRNAs were similar to those observed in some other diseases (Huang et al., [@B13]; Chen J. et al., [@B2]; Zhang X. et al., [@B51]). We also identified 788 DE mRNAs and 66 DE miRNAs, i.e., 690 up-regulated mRNAs, 98 down-regulated mRNAs, 25 up-regulated miRNAs, and 41 down-regulated miRNAs, upon HTNV infection. Interestingly, most DE mRNAs (88%) and DE circRNAs (93%) were up-regulated, while most DE miRNAs (62%) were down-regulated. These results indicated that the regulatory mechanism of miRNAs in HTNV-cell interaction could be different from that of circRNAs and mRNAs. It also supports the theory that a complex regulatory ceRNA network exists. As is well-known, innate antiviral and inflammatory responses can immediately be elicited upon viral infection. These responses must be regulated delicately to prevent viral dissemination and present appropriate effective immune responses. Several researches reported that hantaviruses infection is absence of visible cytopathic effects (CPE), and concluded that HFRS might be caused by unbalanced and inappropriate immune responses elicited by hantaviruses (Pensiero et al., [@B33]; Vapalahti et al., [@B44]). Consistently, GO enrichment analysis and KEGG pathway analysis for annotating the parental genes of the DE circRNAs revealed that they were mostly involved in the host immune response, including "innate immune response," "response to virus," "defense response to other organism," "cytokine-mediated signaling pathway," "response to type I interferon," "type I interferon signaling pathway," and "cellular response to type I interferon," indicating that DE circRNAs may regulate HTNV infection through host immune signaling pathways and antiviral responses, contributing to viral pathogenic progression. GO analysis and KEGG pathway analysis were also carried out to annotate the DE mRNAs upon HTNV infection. In line with the above data, the enrichment analyses revealed that the HUVECs initiated a strong antiviral defense upon HTNV infection, as observed with other viral infections such as Influenza A virus and Herpes Simplex Virus. GO enrichment analysis and KEGG analysis of mRNAs targeted by the identified miRNAs showed that predicted target genes are involved in biological processes such as cell development, cell morphogenesis differentiation, cellular macromolecule biosynthesis, the MAPK signaling pathway, and the Wnt signaling pathway, which have previously been documented to regulate cell fate and behavior. This suggests that DE miRNAs may be the principal regulators of cellular function during HTNV infection. More and more studies have identified that circRNAs could serve as miRNA sponges, indirectly regulating gene expression. Therefore, to uncover the detailed function and mechanism of DE circRNAs in HTNV infection, we constructed a circRNA-miRNA-mRNA ceRNA network. The integrative analysis of circRNA-associated ceRNAs demonstrated the regulatory functions of circRNAs and specific interplay with other RNAs by circRNA-miRNA-mRNA regulatory axis during the cellular innate immune response against HTNV. Through ceRNA analysis, we found at least three regulatory circRNA-miRNA-mRNA axes potentially involved in the regulation of HTNV, i.e., circ_0002470-miR-3182-IFI44, circ_0006132-miR-1304-3p-OAS1, and circ_0000479-miR-337-3p-CASP1. It has been reported that IFI44 is an IFN-inducible protein, related to infection of several viruses (Balan et al., [@B1]; Power et al., [@B34]; deDiego et al., [@B6]). Other studies documented that OAS1 plays an important role in antiviral activity as antiviral factor (Knapp et al., [@B18]; Melchjorsen et al., [@B30]; Singh et al., [@B40]). CASP1 participates in the composition of the inflammasome which actives the pro-inflammatory cytokines IL-18 and IL-1β, resulting in the proinflammatory cell death known as pyroptosis (Lupfer et al., [@B26]). Recently, it has been confirmed that the NLRP3 inflammasome is responsible for hantavirus-induced expression of IL-1β (Ye et al., [@B49]). Furthermore, antiviral and proinflammatory responses to hantavirus infection play important roles in regulating host defense and disease manifestation (Shin et al., [@B39]). Therefore, we speculate that dysregulation of the identified DE circRNAs might be related to HTNV replication by antagonizing the inhibition of miRNAs on mRNAs. Moreover, we further explored the effects of six DE mRNAs, four DE circRNAs, and five DE miRNAs on viral replication. The results showed that knockdown of mRNAs (CMPK2, PARP10, GBP1, and RIG-I) and circ_0000479 could dramatically increase HTNV NP expression, suggesting that these RNAs are potential inhibitors of HTNV infection. While overexpression of miR-149-5p and miR-411-3p can facilitate viral replication; miR-330-5p can markedly inhibit viral replication. To date, there were neither characteristics nor function reports about circ_0000479. While in the ceRNA network, we predicted that circ_0000479 could sponge miR-149-5p to regulate the expression of genes, such as RIG-I, IL16, ISG15, ITGAE, MX2, MYCN, OASL, PARP14, CFH, which were functional in innate immune pathways. Consequently, these correlations preliminarily explained inhibition mechanism of circ_0000479 on HTNV replication. We further verified one of circRNA-associated regulatory axis by luciferase reporter assay. the data showed that miR-149-5p can target RIG-I and circ_0000479, and circ_0000479 could regulate the expression of RIG-I by sponging miR-149-5p. Here, for the first time, we identified and confirmed the circ_0000479-miR-149-5p-RIG-I regulatory axis in HTNV infection. As is well-known, the pattern recognition receptor RIG-I initiates the antiviral immune response and increases type I interferon expression after recognition of RNA viruses (Leung, [@B21]). In agreement with these observations, inhibition of RIG-I promotes HTNV replication (Ma et al., [@B27]). The data obtained in the present study will enrich our understanding of the interaction between HTNV and host cells. However, other circRNA-miRNA-mRNA axes in ceRNA network deserve verification and clarification. Further research on the role of ncRNAs in HTNV infection is warranted. In conclusion, this study is the first to identify a circRNA expression profile and construct a circRNA-related regulatory network involved in interaction between HTNV and host cells, in particular in the host defense to viral infection (as shown in [Graphical Abstract](#F8){ref-type="fig"}). This study therefore provides novel insightfor illuminating the pathogenesis of HTNV. Data Availability Statement {#s5} =========================== The datasets generated for this study can be found in the Gene Expression Omnibus with accession nos. [GSE133634](GSE133634), [GSE133751](GSE133751), [GSE133319](GSE133319). Author Contributions {#s6} ==================== SL, FL, and WH conceived and designed the experiments. NZ and WH contributed to the funding acquisition. SL, NZ, WG, XWa, KL, JY, CJ, SH, HXia, and XWu performed the experiments. SL, YL, HXio, LC, ZG, and FL collected and analyzed the data. SL wrote original draft. FL and WH revised the paper. All authors approved the final version of the manuscript. Conflict of Interest -------------------- The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Funding. This study was supported by grants from the National Natural Science Foundation of China (Nos. 81000734, 81271819, 81101258, 81772186), Hubei Province Natural Science Foundation of China (Nos. 2017CFB621, 2018CFB173), and Research Project of Hubei Provincial Department of Education (No. Q20172802). Supplementary Material {#s7} ====================== The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fcimb.2020.00097/full#supplementary-material> ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. ###### Click here for additional data file. [^1]: Edited by: Intawat Nookaew, University of Arkansas for Medical Sciences, United States [^2]: Reviewed by: Kai Huang, University of Texas Medical Branch at Galveston, United States; Jan Clement, KU Leuven, Belgium; Zhengwen Liu, First Affiliated Hospital of Xi\'an Jiaotong University, China [^3]: This article was submitted to Virus and Host, a section of the journal Frontiers in Cellular and Infection Microbiology [^4]: †These authors have contributed equally to this work	{ "pile_set_name": "PubMed Central" }
INTRODUCTION ============ B-cell antibody responses are initiated by the binding of antigen to surface-expressed B-cell receptors (BCRs), which induces intracellular signals that are critical for translating extracellular environmental cues into cellular behavior and activation ([@B22]). Although antigen-induced BCR signaling is a fundamental process in B-cell biology, our understanding of the mechanisms by which antigens trigger signaling is incomplete. Critical to understanding the mechanisms underlying the initiation of BCR signaling is knowledge of the spatial organization of BCRs on the surfaces of both resting and antigen-activated B-cells at the nanoscale level of individual BCRs. A variety of studies have provided evidence at the resolution of diffraction-limited light microscopy that is consistent with the existence of BCRs as predominantly monomers or small oligomers that are relatively evenly dispersed over the cell surface in resting cells ([@B10]; [@B23]). The diffusion behavior of BCRs on resting cells is also consistent with the majority of BCRs existing in a monomeric state ([@B31]), and intermolecular interactions between BCRs, as would be predicted of BCR multimers, were not detected by Förster resonance energy transfer (FRET) in resting B-cells ([@B32]). On antigen binding, BCRs form signaling-active microclusters that ultimately condense into well-ordered immune synapses ([@B2]; [@B8]). However, evidence that BCRs exist as oligomers in resting cells was provided by biochemical methods showing that BCRs isolated from detergent-solubilized B-cells were in large aggregates ([@B27]). In addition, results of quantitative bifluorescence complementation assays were also consistent with BCRs existing as oligomers on resting B-cell surfaces ([@B38]). These different results have distinct implications for the mechanisms by which BCR signaling is initiated. Indeed, two current models for the triggering of BCR signaling are predicated on either BCRs existing as monomers on the cell surface that are clustered by antigen ([@B23]; [@B33]) or BCRs existing as clusters on the resting B-cell surface that are disrupted by antigen ([@B38]). An explanation for the differences in results that support these models is lacking. The application of point localization--based superresolution fluorescence microscopy techniques ([@B3]; [@B12]; [@B25]; [@B9]) is a promising approach to gain an understanding of the nanoscale spatial organization of BCRs in the plasma membrane on the level of individual BCRs. One widely used method, direct stochastic optical reconstruction microscopy (dSTORM), is a subdiffraction resolution fluorescence imaging technique that uses single-marker switching to resolve nanoscale structures ([@B11]). This method produces superresolution images with a ∼20-nm localization precision by mathematically localizing individual fluorescent molecules stochastically activated over time. However, there are technical limitations in precisely characterizing the nanoscale organization of receptors on cell surfaces from data collected from any superresolution fluorescence imaging due to several factors. Key among these are imprecise localization and blinking of the fluorophores, which can lead to difficulties in drawing conclusions about the number and density of receptors on the cell surface, and limitations in statistical approaches used to characterize the data sets ([@B28]). Here we imaged BCRs expressed by human peripheral blood B-cells using dSTORM ([@B11]) and applied a novel combination of computational tools to overcome or at least minimize these limitations to allow for precise characterization of the spatial organization of cell surface BCRs. We used these tools to investigate an important feature of B-cell biology, namely, the nanoscale organization of immunoglobulin M (IgM) and IgG BCRs expressed by naive and memory human peripheral blood B-cells. Immune individuals acquire antibody memory characterized by rapid, high-affinity recall responses to antigen that are dominated by antibodies of the IgG isotype. Antibody memory is encoded, in part, in long-lived memory B-cells that are the differentiated product of germinal center (GC) reactions in which naive B-cells undergo somatic hypermutation and antigen selection ([@B16]). Naive B-cells that give rise to primary antibody responses upon the first encounter with antigen express IgM and IgD BCRs. In contrast, although there is considerable diversity among memory B-cells ([@B30]), many memory B-cells in human peripheral blood express IgG BCRs. BCRs are composed of a membrane-bound form of the immunoglobulin (mIg), which associates in a 1:1 molar ratio with a heterodimer of Igα and Igβ, which contain in their cytoplasmic domains key motifs that are essential to initiate signaling ([@B24]). The signaling capacities and the outcome of signaling through IgM versus IgG BCRs are not identical, and it has been proposed that differences in signaling through IgM versus IgG BCRs might account for the accelerated, high-titered antibody responses characteristic of B-cell memory ([@B37]; [@B13]; [@B36]). Because all BCRs share identical Igα and Igβ heterodimers ([@B35]), it is assumed that functional differences between the IgG and IgM BCRs are encoded, at least in part, by the mIgG and mIgM themselves or dictated by the differentiated state of the B-cell in which the BCRs are expressed. For example, several studies provided evidence that the cytoplasmic domain of the IgG BCR was essential for enhanced BCR clustering, immune synapse formation, and downstream signaling of IgG-expressing B-cells as compared with IgM-expressing B-cells ([@B6], [@B7]; [@B18], [@B19], [@B17]). A second, unexplored contribution to the signaling capacities of IgM and IgG BCRs is the spatial organization of these receptors on B-cell surfaces. It is possible that B-cell-- or BCR-intrinsic differences in the spatial organizations of IgG versus IgM BCRs on cell surfaces contribute to the accelerated and elevated antibody responses of memory B-cells compared with naive B-cells. The present results of our analysis of the nanoscale spatial organization of BCRs on B-cell surfaces provide insights into both the fundamental process of antigen-driven BCR clustering and differences in the spatial organization of IgM and IgG BCRs that may contribute to the characteristic differences in the responses of naive and memory B-cells to antigen. RESULTS ======= dSTORM images of resting and activated IgM- and IgG-expressing human B-cells ---------------------------------------------------------------------------- The B-cells from two healthy donors were purified from peripheral blood by negative selection and labeled with F(ab) of antibodies specific for human IgM or IgG conjugated to Alexa Fluor 647 (Alexa Fluor 647--Fab anti-IgM or -IgG) at saturating concentrations (300 nM). To rule out the possibility that the detecting reagents themselves activated B-cells, we measured an early event in BCR-mediated B-cell activation, namely Ca^2+^ influx, in unlabeled B-cells and in B-cells labeled with either Alexa Fluor 647--Fab anti-IgM or --Fab anti-IgG. Neither monovalent Fab reagent induced Ca^2+^ influxes in B-cells, in contrast to F(ab′)~2~ anti-Ig, which activated the B-cells (Supplemental Figure S1). Labeled B-cells were placed in chambers on supported fluid planar lipid bilayers (PLBs) that contained streptavidin-biotin--tethered F(ab′)~2~ of κ light chain--specific antibodies (anti-κ) to mimic antigen presented to the B-cells on a cell surface. Although the result of engaging the BCRs by anti-κ may not be identical to that of engaging the BCRs by antigens that bind within the antigen-combining site of the BCR, anti-Igs have proven to be a useful surrogate for antigen, particularly in studies of human B-cells, in which the frequency of B-cells specific for any given antigen is so small as to preclude studies of antigen-specific cells. Thus we refer to anti-κ as the antigen in these studies. For imaging of resting B-cells, cells were placed on PLBs that did not contain anti-κ. After 10 min of incubation at 37°C on the bilayers, the cells were fixed, and dSTORM images were acquired. To do so, the entire population of fluorophores was brought to a nonfluorescent, reversible dark ("off") state by irradiation, and then only a sparse subset of fluorophores was reactivated and imaged at a time. Repetitive photoactivation and localization of single fluorescent dye molecules in an image sequence of 20,000 frames containing only a sparsely distributed subset of labeled BCRs in each frame were performed to reconstruct a superresolved image. In each image frame of a stack, the positions of spatially well separated, activated fluorophores were precisely determined by fitting a point spread function (PSF) to the measured photon distributions to localize individual BCR peaks. Subsequently all of the fitted localizations of BCRs in all frames were summed into a superresolution image. Approximately 20--30 B-cells were analyzed for each condition, acquiring 20,000 frames per B-cell. We show both the composite total internal reflection fluorescence (TIRF) images and dSTORM images of representative resting and activated IgM- and IgG-expressing B-cells ([Figure 1A](#F1){ref-type="fig"}). Several features of the images are of interest. Although both the size of the contact area with the PLB ([Figure 1B](#F1){ref-type="fig"}) and the number of BCR single molecule (SM) localizations in the contact area ([Figure 1C](#F1){ref-type="fig"}) were similar for resting IgM- and IgG-BCR--expressing B-cells, IgG BCRs appear to be more clustered than IgM BCRs in both TIRF and dSTORM images ([Figure 1A](#F1){ref-type="fig"}). After 10 min of antigen activation, the contact area of IgM B-cells with the PLB expanded significantly ([Figure 1B](#F1){ref-type="fig"}), and the number of BCR SM-localizations in the contact area increased nearly 10-fold ([Figure 1C](#F1){ref-type="fig"}). The area of contact with the PLB also increased for IgG-BCR--expressing B-cells 10 min after antigen activation but slightly less so than for IgM B-cells ([Figure 1B](#F1){ref-type="fig"}), and the number of receptor localizations in the contact area also showed a large, sixfold increase ([Figure 1C](#F1){ref-type="fig"}). By live-cell TIRF imaging, the smaller contact areas of antigen-activated IgG B-cells compared with IgM B-cells appeared to be due to more active contraction after maximal spreading after 100 s of image acquisition ([Figure 1, D and E](#F1){ref-type="fig"}). However, it is not possible to directly compare the time frame in these live-cell images with each other or with the 10-min time frame of the fixed cell images because the exact times the B-cells contacted the PLB cannot be determined precisely. Somewhat unexpectedly, we also observed polarization of the BCRs in ∼50% of antigen-activated IgM B-cells ([Figure 1F](#F1){ref-type="fig"}) and 60% of IgG B-cells ([Figure 1G](#F1){ref-type="fig"}). Comparing the size of the contact areas in polarized and unpolarized cells, it appeared that for IgG-expressing B-cells, polarized cells had significantly larger contact areas ([Figure 1H](#F1){ref-type="fig"}), indicating that polarization may precede contraction. A similar trend was observed for IgM-expressing cells. To our knowledge, polarization of BCRs engaging antigen on a surface is a novel observation. It may be that polarization is a unique feature of human peripheral blood B-cells and could possibly reflect a requirement for peripheral blood B-cells to move toward antigen or some other factor once B-cells enter lymphoid tissues and BCR signaling is initiated. ![dSTORM imaging of resting and anti-κ activated IgM- and IgG-expressing human B-cells. Human peripheral blood B-cells were labeled with Alexa Fluor 647--Fab anti-IgM or -IgG. Labeled cells were placed on PLB with (+ anti-κ) or without (resting) streptavidin-biotin tethered anti-κ. (A) Representative pseudo-TIRFM images of unprocessed maximum projection of all acquired frames (left) and corresponding dSTORM images (middle), along with magnified regions in dashed white boxes (right). Scale bar, 1 μm. (B, C) Quantification of the contact area (B) and total number of receptor localizations in the contact area (C) of IgM- and IgG-expressing B-cells under each condition. Black line, mean. (D, E) Selected time-stamp images of live-cell TIRF imaging of (D) IgM-expressing cells and (E) IgG-expressing cells on PLB with anti-κ appearing after 440 s for IgM- and 220 s for IgG-expressing cells after acquisition start. Scale bar, 5 μm. (F, G) Representative images of IgM- (F) and IgG- (G) expressing polarized and unpolarized B-cells (left). White line depicts the boundary of the cell. Graph (middle) of intensity line profile (dashed line) through each cell and quantification of polarized and unpolarized population of IgM- and IgG-expressing cells (right). (H) Quantification of the contact area of polarized and unpolarized cells on activating bilayers. Red line, mean. Data were compared by two-tailed Mann--Whitney U test. \\\\p \< 0.0001, \\p \< 0.01, \*p \< 0.05.](511fig1){#F1} Strategy for analysis of dSTORM images to determine the nanoscale organization of individual BCRs on the surfaces of B-cells ---------------------------------------------------------------------------------------------------------------------------- Having obtained superresolution images of BCRs on the surface of IgM- and IgG-expressing human B-cells, we wanted to analyze the images to obtain information on the spatial organization of individual BCRs. Single-BCR information can be extracted from these images, but doing so requires rigorous quantitative and spatial analysis that distinguishes single labeled BCRs that fluoresce over multiple contiguous frames from actual BCR clusters. Single labeled BCRs that fluoresce across multiple contiguous frames before being irreversibly photobleached appear as clusters of BCRs in the summed frames of the reconstructed image, as depicted in [Figure 2A](#F2){ref-type="fig"}, because the localized position of the BCR in each frame varies slightly due to differences in the number of photons emitted in each frame and the uncertainty in position determination ([@B28]). To correct for this, we developed an analysis algorithm that used the spatial distribution and temporal signature of multiple peaks arising from a single molecule to calculate the contribution of such clusters to observed protein spatial organization (see Materials and Methods). This analysis showed that ∼60% of the localized peaks arose from spreading of fluorescence from single molecules across contiguous frames. Clusters of peaks arising from such spread of single-molecule fluorescence across continuous frames were grouped and collapsed into a single peak in the data set. All further analyses were performed with SM localizations derived from this grouping procedure. ![Strategy for identifying protein islands. (A, B) Human peripheral blood B-cells were labeled with Alexa Fluor 647--Fab anti-IgM or -IgG at a low concentration (10 nM) to ensure spatially and temporally well separated detection of single BCRs within the imaging area (2.4 μm × 2.4 μm) and dSTORM images were acquired. (A) Example of a single fluorophore detected in consecutive frames when the fluorescence emission from a single fluorophore is interrupted by the acquisition time. Frames in the sequence with no signal represent times when the molecule is in the dark state. (B) A second example, showing a single fluorophore detected in multiple frames by irregular intervals. Multiple appearances of a single fluorophore in both cases result in clusters of peaks originating from a single molecule. (C) Frequency of appearances of Alexa Fluor 647--Fab anti-IgM and -IgG under the foregoing conditions. (D) Representative diagram showing spatially isolated assemblies of proteins containing single and multiple receptors defined by the Hoshen--Kopelman algorithm--based cluster analysis detailed in Materials and Methods. Red arrowed line denotes 60-nm radius. (E) Radius (left) and density (right) of spatially isolated assemblies of appearances (protein island) plotted as a function of protein islands containing 3--10 appearances. Protein islands containing two appearances are excluded from these graphs because two appearances cannot define an area.](511fig2){#F2} Interpretation of images is also complicated by the fact that many fluorophores transition to reversible "off" states at unpredictable times with irregular intervals during data acquisition, as depicted in [Figure 2B](#F2){ref-type="fig"}. This phenomenon is termed blinking and results in a single labeled BCR appearing as a cluster of localized peaks in the composite superresolution image, even after the multiple appearances due to single labeled BCRs fluorescing across successive frames have been corrected by a grouping procedure. This blinking property of fluorescent dyes is not completely intrinsic to the dye but is also highly dependent on its nanoenvironment and thus can only be determined experimentally ([@B26]). The effect of the environment on the blinking behaviors of the fluorophores may be particularly complex because images are acquired at the interface of the B-cell surface and the PLB. To determine the contribution of multiple blinking to the BCR images, B-cells were sparsely labeled with either Alexa Fluor 647--Fab anti-IgG or Alexa Fluor 647--Fab anti-IgM and placed on PLB, and dSTORM images were acquired. Because the average localization precision of single molecules in our dSTORM experiments ranged between 15 and 20 nm, the two-dimensional Gaussian PSF would have a maximum 3σ value of ∼60 nm. A molecule has 99.7% probability of residing within an area with radius equivalent to 3σ of the PSF. The sparse labeling predicts that the spatial separation between any two Alexa Fluor 647--Fab anti-IgG or --Fab anti-IgM molecules would be significantly larger than the PSF of our dSTORM experiments (i.e., \>60 nm). If so, the localizations confined within a spatially isolated area with radius equivalent to the PSF would represent localizations arising from a single Alexa Fluor 647--Fab anti-Ig molecule. Such spatially isolated localizations from a single molecule would allow the counting of the number of reappearances, if any, from each isolated dye molecule. The results showed that the majority (∼80%) of labeled BCRs appeared once in the superresolution dSTORM image but that a fraction appeared more than once ([Figure 2C](#F2){ref-type="fig"}). The number of multiple occurrences was largest for two appearances, and the number of multiple occurrences decreased to nearly undetectable levels for 10 occurrences. Because the distribution of multiple appearances was nonnormal without a defined mean, it was difficult to computationally correct for the phenomenon. Thus at least a fraction of localization clusters in the final dSTORM image are likely to represent a single BCR. To quantify the distribution of BCR SM localizations in the dSTORM images of IgM- and IgG-expressing B-cells that were resting or activated with antigen, we used a modified Hoshen--Kopelman cluster analysis ([@B14]). A modified Ripley's K function analysis and a pair correlation analysis--based technique have been used for quantification of superresolution images, with the pair correlation technique being better suited to quantifying images that are subject to overcounting and prone to overestimations of clustering ([@B28]; [@B34]). Both of these techniques treat the superresolution image in its entirety and calculate average values of clustering parameters for the whole image by evaluating increased clustering of molecules in the image compared with random distributions. Consequently these methods do not provide spatial information at the single-cluster level. These methods, when applied to analyze the spatial distribution of BCR localizations, compute a single average value to describe the complex spatial clustering pattern, masking the heterogeneities in clustering properties of BCRs. Thus information concerning important biological parameters such as the size and number of BCR localizations per cluster would be obscured using either pair correlation-- or Ripley-based methods. To capture the details of BCR spatial organization, we instead used a modified Hoshen--Kopelman--based cluster analysis ([@B14]) that identifies and visualizes individual clusters of molecules and isolated individual molecules in the composite superresolution image. We refer to each spatially isolated protein assembly (i.e., a group of SM localizations separated from all other SM localizations by a distance of \>60 nm, the size of the PSF), identified by this analysis as a protein island, with the minimal island containing only one spatially isolated BCR SM localization, as depicted in [Figure 2D](#F2){ref-type="fig"} and shown for dSTORM data (Supplemental Figure S2). Identification of individual protein islands allowed us to count the exact number of BCR SM localizations present in each protein island and generate a statistical distribution of island occupancy. Note the number of BCR SM localizations detected in protein islands under saturating conditions of the Alexa Fluor 647--Fab anti-IgG or --Fab anti-IgM Alexa may not be equal to the absolute number of BCRs in the islands but will likely provide an overestimate because of blinking-induced reappearances described earlier that we could not correct for. It is also possible that under saturating conditions, more than one Alexa Fluor 647--Fab anti-IgG or --Fab anti-IgM could bind to a single BCR, which would also contribute to overcounting. However, the island occupancies (i.e., number of BCR SM localizations within the island) can be used to estimate the relative changes in the number of BCRs under different experimental conditions. We also estimated the size (radius) of the protein islands and density of BCR SM localizations within islands by calculating the convex hull or smallest convex set for the set of BCR SM localizations that formed an island. By this definition, densities could not be defined for islands with a single localization (a point) or two localizations (a line). Overcounting will not affect the evaluation of various important nanoscopic aspects of the spatial organization of BCRs such as the size and shape of the protein islands and the relative changes in density of protein islands using this clustering analysis. Overall, this strategy enables the generation of statistical distributions of characteristic cluster parameters without averaging out the underlying details of the heterogeneous distribution and thereby provides an accurate description of the spatial properties of the protein organization. Using Hoshen--Kopelman analysis to confirm presence of isolated Alexa 647--labeled IgM and IgG BCRs in sparsely labeled cells ----------------------------------------------------------------------------------------------------------------------------- The modified Hoshen--Kopelman--based cluster analysis was used to confirm that the individual SM localizations in sparsely labeled IgM- and IgG-expressing cells were in fact from individual BCRs ([Figure 2E](#F2){ref-type="fig"}). The analysis showed that the radii of the protein islands in sparsely labeled cells, irrespective of the number of SM localizations within each protein island, were ≤60 nm, indicating that the localizations of an individual protein island were always distributed within an area equivalent to the PSF of our dSTORM experiments. This supports our assumption that the Alexa Fluor 647 molecules in the sparsely labeled data sets were separated by distances larger than the PSF and that each cluster of localizations confined within an area equal to the PSF represented multiple appearances from a single dye molecule. The density of SM localizations within the islands increased linearly with occupancy (i.e., number of SM localizations) within protein islands, and islands with same number of localizations had the same density, consistent with each island having the same area (equivalent to the PSF). Alexa Fluor 647--Fab anti-IgG and --Fab anti-IgM behaved nearly identically in these analyses ([Figure 2, C and E](#F2){ref-type="fig"}), indicating that using two different Fab antibodies would not introduce biases in counting BCR SM localizations in IgG- and IgM-expressing cells. We also characterized the multiple appearances of spatially isolated IgG and IgM immobilized on glass coverslips and labeled with a saturating amount of Alexa 647--labeled Fab antibodies. Under these conditions, the labeled IgG and IgM exhibited similar blinking behavior (Supplemental Figure S3B), which indicated that IgG and IgM are labeled similarly by the Alexa 647--labeled Fab anti-IgM and -IgG, further confirming that the IgG and IgM localizations on resting and stimulated B-cell surfaces can be used to compare their spatial distribution in those situations. The spatial distribution of IgM and IgG BCRs on resting and antigen-activated human B-cells ------------------------------------------------------------------------------------------- Using the strategy described, we found that BCR islands on both IgM- and IgG-expressing cells were highly heterogeneous, varying in both size and number of receptor SM localizations per island. For resting IgM BCR--expressing B-cells, 27% of total BCR SM localizations were present as spatially isolated single localizations ([Figure 3A](#F3){ref-type="fig"}), accounting for \>60% of the protein islands on B-cell surfaces ([Figure 3B](#F3){ref-type="fig"}). The observation that at saturating conditions of Alexa Fluor 647--Fab anti-IgM labeling, 27% of the BCR localizations were present as single, isolated localization addresses the issue of multiple binding of Alexa Fluor 647--Fab anti-IgM to individual BCRs and suggests that such events would be infrequent. The remaining 73% of BCR localizations were in islands that were highly heterogeneous in the number of receptor SM localizations they contained ([Figure 3A](#F3){ref-type="fig"}). The number of BCR SM localizations in islands containing \>10 BCR SM localizations ranged from 11 to \>200 BCR SM-localizations per island ([Figure 3C](#F3){ref-type="fig"}), with a median of 16 BCR SM localizations per island. Antigen activation of IgM-expressing B-cells resulted in a large increase in the total number of BCR SM localizations in the contact area between the B-cell and the PLB ([Figure 1C](#F1){ref-type="fig"}), and although spatially isolated single BCR localizations decreased to 15% of total localizations, the percentage of protein islands that contained only a single BCR localization did not change significantly and accounted for 58% of islands. However, there was a large increase in the percentage of BCR SM localizations that were in islands that contained larger numbers of BCR SM localizations ([Figure 3, A and B](#F3){ref-type="fig"}). Strikingly, the islands that contained \>10 BCR SM localizations had significantly more BCR SM localizations in antigen-activated IgM B-cells than in resting cells, ranging from 11 to \>500 BCR SM localizations per island, with a median of 18 BCR SM localizations per island ([Figure 3C](#F3){ref-type="fig"}). The distribution of IgM BCR SM localizations in discrete islands and the effect of antigen engagement on that distribution are graphically represented in [Figure 4](#F4){ref-type="fig"}. ![Quantitative analysis of the distribution of IgM and IgG BCRs on resting and activated B-cells. (A) Percentage of total BCR peaks that reside in protein islands containing 1--10 or \>10 BCRs. (B) Frequency of spatially isolated BCR protein islands that contained 1--10 BCR localizations or \>10 BCR localizations. (C) Distribution of the number of BCR localizations per protein island for islands that contained \>10 BCR localizations. Each dot represents a single protein island. Black line, median number of receptor localizations per protein island. Data were from 21--30 cells for each condition from two or three independent experiments and compared by two-tailed Mann--Whitney U test. \\\\p \< 0.0001, \\\p \< 0.001. (D) Percentage of BCR protein islands containing monomers (filled bars) and dimers (open bars) in each individual cell expressing the total given number of BCRs.](511fig3){#F3} ![The nanoscale organization of IgM and IgG BCRs on resting and activated B-cells. Schematic representation of the distribution of IgM and IgG BCRs on the contact area of IgM- and IgG-expressing human B-cells on PLB. Monomers (black dots), dimers (green dots), trimmers (orange dots), protein islands containing 4--10 receptors (red dots), and protein islands containing \>10 receptors (blue dots) are drawn under each condition reflecting the dSTORM data. The numbers on top of each cell denote the relative size of the contact area of the cell.](511fig4){#F4} The distribution of IgG BCR SM localizations on resting B-cells differed from that of IgM BCR localizations. Of note, the percentage of spatially isolated, single IgG BCR SM localizations (15%) was nearly half of that for IgM BCR localizations ([Figure 3A](#F3){ref-type="fig"}), even though the frequency of protein islands that contained only a single BCR SM localization was similar to that of IgM-expressing resting cells, ∼60% ([Figure 3B](#F3){ref-type="fig"}). Greater than 50% of the IgG BCR SM localizations were in islands containing \>10 BCR SM localizations ([Figure 3A](#F3){ref-type="fig"}). The average number of IgG BCR SM localizations in islands containing \>10 BCR SM localizations had a wider distribution than that for IgM BCRs (11 to 4 × 10^3^ BCR SM localizations per island, with a median number of receptor localizations per island of 18; [Figure 3C](#F3){ref-type="fig"}). In antigen-activated B-cells, \<10% of the IgG BCR SM localizations were present as spatially isolated single localizations ([Figure 3A](#F3){ref-type="fig"}) even though the frequency of protein islands that contained only a single BCR localization did not change, accounting for ∼60% of all islands ([Figure 3B](#F3){ref-type="fig"}). There was a large increase in the percentage of total BCR SM localizations that were in islands that contained \>10 BCR localizations ([Figure 3, A](#F3){ref-type="fig"} and B), and the number of IgG BCR SM localizations in these islands ranged from 11 to 1 × 10^4^ BCR localizations per island, with a median value of 22 BCR localizations per island ([Figure 3C](#F3){ref-type="fig"}). The distribution of IgG BCR SM localizations on resting and antigen-activated B-cells is graphically represented in [Figure 4](#F4){ref-type="fig"}. These results indicate that IgG BCRs in resting cells are present in larger clusters than in IgM BCRs, which may facilitate the antigen-driven increase in the frequency of BCR clusters that contain larger numbers of BCR and may also be related to the rapid growth of IgG BCR clusters after antigen engagement, as previously described ([@B19]). We were interested in the relationship between small BCR islands containing either single SM localizations (which likely represent BCR monomers) or two SM localizations (a fraction of which possibly represent BCR monomers, with the rest representing BCR dimers) and larger BCR islands with multiple SM localizations on both resting and antigen-activated cells. We asked whether the frequency of the smaller islands (with one or two SM localizations) correlated with the total number of BCR SM localizations in the contact area with the PLB for each of the 21 IgM-expressing and 30 IgG-expressing B-cells analyzed. The total number of IgM BCR SM localizations per cell varied up to 48-fold, and yet the frequency of BCR islands with single or two SM localizations remained relatively constant for both resting and antigen-activated B-cells ([Figure 3D](#F3){ref-type="fig"}). A similar phenomenon was observed for IgG-expressing B-cells. These data indicate that B-cells intrinsically maintain a high frequency of small BCR islands with a single or two SM localizations (i.e., islands that likely represent BCR monomers or dimers) independently of the total number of BCR localizations in the contact area of the B-cell and the PLB, suggesting an important function for these structures. Given that the analysis of multiple blinking indicated that ∼20% of dye molecules appear more than once in the final dSTORM image ([Figure 2C](#F2){ref-type="fig"}), these data suggest that cells maintain an even higher frequency of monomeric BCRs on the surfaces of both resting and activated cells. Together these data describe the spatial organization of BCRs on the surface of resting B-cells as highly heterogeneous protein islands that contain a wide range of numbers of BCRs but with a predominance of monomers (as indicated by a large fraction of islands containing only one or two BCR SM localizations), as depicted in [Figure 4](#F4){ref-type="fig"}. The spatial organization of IgM and IgG BCRs can be distinguished by dSTORM and suggests that preclustering of IgG BCRs may be a mechanism to amplify antigen-driven clustering in IgG-expressing B-cells. For both IgM and IgG BCRs, after antigen engagement, the percentage of total BCR localizations that were present as spatially isolated single localizations decreased, but islands with single localizations still represented the predominant BCR islands, and the percentage of BCR SM localizations in islands containing \>10 BCR localizations increased. Of note, there is no evidence in these data that antigen engagement increased the frequency of islands that contained fewer of either IgM or IgG BCR localizations, at least at the 10-min time point after antigen stimulation. BCR cluster size, density, and interisland distance --------------------------------------------------- From dSTORM data, it is also possible to calculate additional important parameters of the nanoscale organization of BCRs on B-cell surfaces, including size (radius) of the BCR islands, density of the BCR SM localizations within islands, and distance between islands. The average radius of IgM-BCR islands on resting B-cells was similar to that of IgG BCR islands on resting B-cells (r* ≤ 60 nm for the majority of islands, with BCR localizations confined within an area equal to the PSF; [Figure 5A](#F5){ref-type="fig"}). Because the size of protein islands \<∼60 nm (i.e., smaller than the PSF) cannot be accurately measured, we set 60 nm as the smallest radius of protein islands for our analysis. Both IgM and IgG BCR islands with radii ≥200 nm were infrequent, and the average radii and the range of the radii of these larger clusters were similar ([Figure 5B](#F5){ref-type="fig"}). After antigen engagement, although the radii of both IgM BCR and IgG BCR islands increased, the increase for the IgG BCR was larger, and the IgG BCR islands were more heterogeneous in size, ranging in radius to \>1 μm ([Figure 5B](#F5){ref-type="fig"}). However, in all cases, namely, IgM BCRs and IgG BCRs for both antigen-stimulated and unstimulated cells, the radii of the protein islands correlated with the number of BCR SM localizations per island (correlation factors of 0.9719 for resting and 0.9224 for activated IgG-BCR islands and correlation factors of 0.8690 for resting and 0.9159 for activated IgM-BCR islands; [Figure 5C](#F5){ref-type="fig"}). ![Size of IgM and IgG BCR protein islands in B-cells on nonactivating and activating bilayers. (A) Distribution of protein island radius. Protein islands with \<60 and \>200 nm radius are categorized separately. (B) Distribution of protein island radius \>200 nm. Each dot represents a single protein island. Black line, median protein island radius with r ≥ 200 nm. Data were acquired from two or three independent experiments with a minimum of 10 cells/ experiment for each condition and compared by two-tailed Mann-Whitney U test. \\\\p \< 0.0001, \\\*p \< 0.001; ns, nonsignificant. (C) Number of receptor localizations per protein island plotted as a function of island radius for IgM and IgG BCRs under resting and activating conditions. IgM and IgG BCR protein islands in resting (left) and anti-κ activated (right) cells. ρ is Spearman's rank correlation coefficient. Protein islands containing dimers are excluded from these graphs because two receptors cannot define an area.](511fig5){#F5} Remarkably, in contrast to the positive correlation between the number of BCR SM localizations per island and the size (radius) of the islands, the density of BCR SM localizations within islands did not show a positive correlation with the number of receptor SM localizations per island ([Figure 6](#F6){ref-type="fig"}). Overall, the distribution of BCR SM-localization densities was similar for IgM and IgG islands ([Figure 6A](#F6){ref-type="fig"}). The average densities of the densest IgM and IgG BCR islands were also similar ([Figure 6B](#F6){ref-type="fig"}). However, there appeared to be no positive correlation between density and number of BCR SM localizations in any of the experimental conditions ([Figure 6C](#F6){ref-type="fig"}). In fact, the majority of islands that contained many BCR SM localizations were less dense than islands containing fewer BCR SM localizations. This suggests that the larger, less dense islands may be composed of assemblies of smaller, highly dense islands. ![Quantitative analysis showing densities of IgM and IgG BCR protein islands on resting and activated B-cells. (A) Frequency of receptor density per nanometer squared in protein islands. Protein islands with a density \<5 × 10^−4^ and \>1 × 10^−2^ receptor localizations/nm^2^ are categorized separately. (B) Distribution of the protein islands with a density \>5 × 10^−2^ receptor localizations/nm^2^. Each dot represents a single protein island. Data were compared by two-tailed Mann--Whitney U test. \\\\p \< 0.0001; ns, nonsignificant. Black line, median density of protein islands with density ≥5 × 10^−2^ receptors/nm^2^. Data were acquired from 21--30 cells from two or three independent experiments for each condition. (C) Island density per nanometer squared plotted as a function of number of receptor localizations per protein island. (D) Island radius (left) and density (right) plotted as a function of protein islands containing 3--10 BCR localizations. Protein islands containing dimers are excluded from these graphs because two receptors cannot define an area.](511fig6){#F6} Next we compared the scaling of the radius and density of protein islands with the number of IgG and IgM BCRs per protein island in data sets from stimulated and unstimulated cells. The analysis showed that protein islands with fewer than five BCR SM-localizations have a radius of 60 nm (i.e., they are confined within an area equivalent to the PSF), but protein islands with five or more BCR SM localizations have a wide range of radii (ranging from 60 to \>120 nm), with many being larger than the PSF (i.e., with radius \>60 nm). As discussed earlier ([Figure 2E](#F2){ref-type="fig"}), localizations arising from the same BCR molecule are confined within an area equal to the PSF of our experiments (i.e., 60-nm radius). This indicates that the protein islands containing five or more BCR localizations and having a radius \>60 nm ([Figure 6D](#F6){ref-type="fig"}) represent SM localizations arising from multiple Alexa Fluor 647--labeled BCR molecules. Consistent with this, the density of the protein islands with more than five localizations, unlike those in the sparsely labeled data ([Figure 2E](#F2){ref-type="fig"}), do not scale linearly with the number of localizations. Instead, the islands have a wide range of densities, further confirming that the multiple localizations in these islands do not arise from multiple appearances of a single Alexa Fluor 647 molecule. Instead, the protein islands with five or more localizations should represent assemblies of multiple IgG or IgM BCR molecules. Thus, from these results, we can conclude that the smallest oligomers contain at the most five BCRs. This would be the case if all five localizations represent unique BCR molecules. However, because some of the Alexa Fluor 647 molecules are likely to appear more than once, as discussed earlier, due to multiple appearances of ∼20% of Alexa Fluor 647 molecules ([Figure 2C](#F2){ref-type="fig"}), the smallest oligomers are likely to comprise fewer than five BCR molecules. The change in density of BCR SM localizations with the number of BCR localizations per island can be used to distinguish two possibilities concerning the molecular basis of BCR clustering. The first is that antigen-engaged BCRs cluster due to BCR protein--protein interactions, in which case the receptor assemblies should have a very high density. In this model, regardless of the number of BCRs per structure, structures of similar densities form as long as the mode of protein--protein interaction remains the same. In the second model, antigen-engaged BCRs partition into and cluster within areas on the plasma membrane defined by heterogeneity in lipid or protein composition. In this case, if the area of the protein island remains the same and the number of receptors within the protein island increases, the average density of receptors in the island will increase. Our results show that islands with few receptor SM localizations often have higher density than islands with more BCR SM localizations ([Figure 6C](#F6){ref-type="fig"}). This indicates that protein--protein interactions are more likely to mediate the formation of small BCR clusters, whereas such interactions are unlikely to be involved in the clustering of BCRs into less dense, larger protein islands. We also find that the areas of protein islands scale with occupancy, with more receptors present in islands with larger areas. This indicates that the larger clusters are not formed by partitioning of receptors into preexisting membrane domains of fixed size. However, the observed clustering pattern can still be governed, at least at some level, by membrane lipid--mediated interactions through interactions of activated receptors with membrane heterogeneities and further expansion of such heterogeneities. The average distance between the centroids of protein islands are similar in resting IgG BCR--expressing B-cells and resting IgM BCR--expressing B-cells ([Figure 7](#F7){ref-type="fig"}). As shown earlier, the total number of BCR SM-localizations at the contact area of the B-cell with the planar lipid bilayer increased with BCR stimulation ([Figure 1C](#F1){ref-type="fig"}). The increased BCR localizations can be accommodated in preexisting protein islands, leading to alteration in island occupancy (i.e., number of SM localization per island) and possible changes in size and density. In such a scenario, the average intercentroid distance would be expected to stay the same. Alternatively, new protein islands might form upon activation, which are populated by some of these BCRs. This will change the average interisland distance. After antigen activation, the interisland distance decreased by a similar extent for IgM BCR-- and IgG BCR--expressing B-cells ([Figure 7A](#F7){ref-type="fig"}). Concomitantly, the number of isolated protein islands also increased after antigen activation. Together these data indicate that antigen activation induces the formation of new protein islands, leading to greater crowding of islands on the plasma membrane. ![IgM and IgG BCR protein islands get closer after anti-κ activation. (A) Minimum interisland distance of IgM and IgG BCR protein islands on resting and activated B-cells were quantified, and means and SDs are plotted (n = 21 for IgM resting, IgM + anti-κ, IgG resting; n = 31 for IgG + anti-κ). (B) Distribution of minimum interisland distance. Black line, median. Two-tailed Mann--Whitney U test was performed comparing the minimum interisland distances of resting cells vs. anti-κ activated cells. \\\\p \< 0.0001. (C) Circularity of IgM and IgG protein islands plotted as a function of number of receptor localizations per protein island under resting and activating conditions. The circularity value ranges between 0 and 1, with 1 being a perfect circle and values closer to 1 showing smaller deviation from a circular shape. Each dot represents a single protein island.](511fig7){#F7} We also analyzed the circularity of the protein islands. The tendency of protein assemblages in membranes to form circular structures indicates a possible role of membrane lipid--driven liquid--liquid phase separation in formation of such assemblies ([@B5]). Our analysis showed that the large islands tend to have circular shapes, indicating that such lipid-mediated phase partitioning likely determines the shapes of these islands ([Figure 7C](#F7){ref-type="fig"}). DISCUSSION ========== We have only a partial understanding of the spatial organization of BCRs on B-cell surfaces and how antigens affect this organization. The recent development of superresolution fluorescence microscopy methods that allow lateral resolution of up to ∼20 nm offers a promising approach to gain an understanding of the nanoscale spatial organization of BCRs in the plasma membrane. Here we used dSTORM superresolution imaging in conjunction with a novel combination of analytical tools to characterize the spatial organization of IgM and IgG BCRs on the surfaces of resting and antigen-activated human peripheral blood B-cells. These analyses provided evidence that both IgM and IgG BCRs reside in highly heterogeneous protein islands on resting B-cells that vary in size and number of BCR localizations. IgG BCRs are more clustered than IgM BCRs on resting cells and form larger protein islands after antigen activation. Even though a portion of both IgM and IgG BCRs reside in large protein islands, the presence of a large number of spatially isolated, single BCR localizations in the dSTORM images suggest that both resting and activated B-cells intrinsically maintain a high frequency of BCR monomers. Antigen activation induces the formation of new protein islands, and protein--protein interactions likely mediate the formation of small, highly dense BCR oligomers that come together to form larger, less dense, islands, a process governed at least in part by membrane lipid interactions. A critical feature of the combination of analytical tools we used in this study is that they accounted for multiple appearances of single fluorescent receptors in contiguous frames. Overcounting is a common but an underappreciated problem in point localization--based superresolution microscopy methods, and studies have shown that distinguishing a single receptor with multiple appearances from clusters of receptors is challenging without particular attention to this phenomenon ([@B28]; [@B34]). To address overcounting in our data, we developed an analysis algorithm to correct for multiple appearances of a single receptor in contiguous frames. This analysis determined that ∼60% of localized peaks arise from the spreading of fluorescence from a single molecule over several contiguous frames. We could not completely eliminate multiple localizations originating from a single receptor because multiple random blinking of Alexa Fluor 647 that occurred in discontiguous frames showed nonnormal distribution without a defined mean in our system. Our analysis of sparsely labeled receptors shows that ∼20% of receptors appear more than once in the final processed dSTORM image under our imaging conditions, with nearly all of these appearing only twice. Thus it is likely that a portion of dimers identified by our cluster analysis represent single BCRs. Even accounting for this possibility, our data provided novel insight into the fundamental process of antigen-driven BCR clustering, as well as into differences in the spatial organization of IgM and IgG BCRs that might contribute to the characteristic differences in the response of naive and memory B-cells to antigens. The nanoscale data presented here fill a critical gap in our knowledge of the process of BCR clustering and synapse formation by providing evidence that the basic BCR signaling unit is an oligomer that contains five or fewer BCRs and that these oligomers do not come into molecular proximity with one another as they form larger BCR islands. The key finding that supports this conclusion is that the density of BCR islands does not increase with the number of BCR localizations per island. In fact, the densest islands contained few BCR localizations, whereas there was a significant decrease in density in most of the larger islands. The drop in density for islands containing a larger number of BCR localizations indicates that BCR oligomers do not come into close molecular proximity in larger structures but maintain the structural characteristics of smaller oligomers. We also determined that even on resting B-cell surfaces, even though most islands contained only a single BCR localization, accounting for 15--30% of all BCR localizations, BCRs existed in a wide array of islands that differed in size and the number of BCR localizations per island. This observation raises a question about the relationships between these different structures on resting and antigen-activated B-cells. In earlier studies, we used FRET between fluorescent proteins expressed in the cytoplasmic domains of Igα and Igβ to measure interactions between BCRs and observed little or no FRET in resting B-cells, leading us to conclude that most BCRs were monomers ([@B32]). FRET was detected upon BCR antigen engagement, indicating clustering; however, the FRET between the cytoplasmic domains of antigen-engaged BCRs was highly transient. Because the loss of FRET was dependent on the phosphorylation of the Igα and Igβ cytoplasmic domains by the Src family kinase Lyn, we interpreted the loss of FRET as the "opening" of the BCR cytoplasmic domains. We concluded that the initiation of BCR signaling involved oligomerization of BCRs bringing the BCR cytoplasmic domains into close molecular proximity, followed by the recruitment of Lyn and the opening of the BCR's cytoplasmic domains to allow the further assembly of a signaling complex. In related studies, we also observed that as BCR clusters grew in size, as measured by fluorescence intensity of labeled BCRs, the FRET between BCRs did not continue to increase. This observation suggested that the cytoplasmic domains of antigen-clustered BCRs remained in an open (no-FRET) conformation as they formed larger structures and, moreover, that the BCRs forming larger structures did not come into the same molecular proximity as did the BCRs in structures that first formed upon antigen engagement ([@B29]). Relating these findings to the present dSTORM data suggests that the observed monomers may form signaling-active oligomers that contain five or fewer BCRs that remain in an open conformation as they come together to form larger structures. If this is the case, at any given time in resting B-cells, the FRET signal would be low or undetectable not because BCRs were not in close molecular proximity in oligomers, but because those oligomers were in an open conformation. What, then, is the function of BCR oligomers on resting and activated cells and their relationship to the larger structures observed? We speculate that oligomers may form spontaneously on resting BCR surfaces and that, once formed, the oligomers come together to form larger structures. The function of the oligomers could be for tonic signaling. Our earlier studies provided evidence for a mechanism of spontaneous BCR oligomerization describing a change in the membrane-proximal domains of IgG (γ3) and IgM (µ4) that occurred with the binding of BCRs to monovalent antigens on an opposing membrane ([@B31]). The γ3 and µ4 domains were necessary for BCR clustering and, when expressed on the surface of B-cells alone, spontaneously clustered and signaled. We hypothesized that BCR spontaneously undergo this "conformational" change at some low rate and that, upon random bumping with other BCRs that had undergone a conformational change, form signaling-active oligomers. In such a model, the process of oligomerization and the coming together of oligomers into larger structures would be similar in resting and activated B-cells, with the process greatly accelerated by antigen binding. Description of the dynamics of BCR movement in the B-cell membrane using single-molecule tracking in TIRF microscopy (TIRFM) may also lend insights into the interpretation of these dSTORM data ([@B31]). These studies showed that most receptors on the surface of resting B-cells are freely diffusing. These data suggest that the BCRs on resting cells observed in the present studies to be in large structures may be freely diffusing within a confined area. The single-molecule tracking results also showed that diffusion of BCRs after antigen engagement dropped markedly, indicating that these were immobilized. Taken together with the results presented here, these data suggest that the BCRs in large structures on resting B-cells are freely diffusing and BCRs in larger structures on antigen-activated cells are immobilized. If this is indeed the case, then large BCR islands on resting versus activated B-cells are qualitatively different. Concerning differences between the nanoscale organization of IgM and IgG BCRs, we determined that the organization of IgG BCRs was distinct from that of IgM BCRs. On resting B-cells, a smaller fraction of IgG BCR localizations appear as spatially isolated monomers than do IgM BCRs. In addition, there were more islands with \>10 IgG BCR localizations on the surfaces than in IgM BCR cells. Thus there appears to be an inherently distinctive nanoorganization of BCRs on B-cell surfaces according to receptor isotypes or possibly the differentiated state of the B-cell expressing the receptors. For both IgM and IgG BCRs, activation on membrane-bound antigen resulted in the formation of larger islands containing tens to thousands of BCR localizations. Compared with IgM BCRs, the frequency of larger islands was higher for IgG BCRs, with the IgG BCR islands having a larger average number of BCR localizations per island, as well as a wider range in the number of BCR localizations. These observations from dSTORM images are consistent with results from TIRFM images, in which we observed that IgG BCRs clustered more rapidly and formed larger structures than did IgM BCRs ([@B19]). The nanoscale organization of the IgG BCRs, which shows more clusters even in resting cells, may facilitate rapid growth of IgG BCR clusters into larger islands. Three recent studies also used superresolution microscopy to characterize BCRs on the plasma membranes of B-cells. [@B21]) pioneered the use of dSTORM to determine the nanoscale organization of IgM and IgD BCRs on the surfaces of naive resting mouse splenic B-cells as part of a study to determine the contribution of actin and tetraspanin networks to the cell surface organization of BCRs. They showed that both IgM and IgD BCRs were nonrandomly distributed and organized into small nanoclusters on resting primary mouse B-cells. They observed that ∼38% of IgM BCRs were in nanoclusters of 80 nm radius that contained ∼20--50 molecules per cluster. However, they did not report on the frequency of BCRs in monomers. For IgD BCRs, 70% were in clusters of 80 nm radius in resting cells that contained ∼30--120 BCRs. They also found that IgD nanoclusters were more densely packed than IgM clusters. Thus IgM and IgD had different nanoscale spatial organizations on resting splenic B-cells. Because both receptor types were expressed in the same cellular context, the differences in organizations were presumably BCR intrinsic. Both IgM and IgD BCRs showed marked increases toward larger clusters upon activation with soluble antigen. Although it is not possible to directly compare details of the results from their study and ours, primarily because of differences in data analysis, the overall picture is similar, namely that BCR of different isotypes have distinct nanoscale organizations on B-cell surfaces and that antigen drives BCRs toward larger clusters. It is of interest that the organization of IgM and IgD BCRs did not change upon treatment with latrunculin A, suggesting that these organizations are independent of actin cytoskeleton. Similar results to those presented here were obtained by [@B1]), who used dSTORM to analyze the spatial organization of IgG1 BCRs on ovalbumin (OVA)-specific B-cells from mice generated by somatic cell nuclear transfer using the nucleus of an OVA-specific B-cell as a donor. These mice provided a unique opportunity to investigate the behavior of isotype-switched BCRs in the context of naive B-cells. They demonstrated that \>90% of IgG1 BCRs were monomers or dimers on resting cells. They also observed a significant increase of the percentage of clusters with more than three BCRs after monovalent antigen engagement. The predominant organization of BCRs into small clusters, monomers and dimers, and the antigen-induced formation of clusters containing larger numbers of BCRs are similar to the observations reported here for both IgM and IgG BCRs on the surface of human B-cells. This is of particular interest because the IgG1 BCRs were expressed on naive B-cells rather than on differentiated memory B-cells as in our studies, suggesting the possibility that the nanoscale organization of IgG BCRs may be BCR intrinsic and not influenced by the differentiated state of the B-cell in which they are expressed. [@B20]) used dSTORM to study triply deficient (RAG, λ5, and SLP65) mouse B-cell lines expressing single-chain IgM BCRs and IgD BCRs. They normalized overcounting in some of the analyses by using a pair autocorrelation function. However, they calculated the blinking properties of their fluorophore by measuring blinking on poly-[l]{.smallcaps}-lysine surfaces and showed better and more predictable behavior of the fluorophores than that observed here. They showed a heterogeneous, wide range of BCRs per protein island, with a median of 30 IgM BCRs per island and 48 IgD BCRs per island. The median radius for the IgM and IgD BCR islands was 218 and 290 nm, respectively. However, no information was provided on the percentage of BCRs that occupied these islands. Nonetheless, these data are consistent with the conclusion that BCRs of different isotypes have distinctive nanoscale organization on cell surfaces. Strikingly, both IgM and IgD BCRs were found in dissociated monomers upon stimulation. In contrast, our data provided no evidence for an increase in BCR monomers or dimers after antigen activation, at least not at 10 min after antigen exposure. It may be a limitation to the interpretation of results of [@B20]) that only ∼10% of the total number of surface BCRs was imaged by dSTORM, reconstructed from 6000 frames. The data in our study were reconstructed from 20,000 frames to map most of the receptors present on the contact surfaces because 10,000 frames were not enough to fully capture the stochastic activation and consequent localization of BCRs in our cells. These authors also report that latrunculin A treatment reduced the number of BCRs per island, in contrast to the findings of [@B21]) showing that the nanoscale organization of BCRs was not dependent on the actin network. This finding raises the possibility that the transfected cell lines may differ from naive B-cells in their molecular requirement for nanoscale organization. Such differences could account for the differential effect of antigen activation on BCR organization. Clearly, much more remains to be learned about the nanoscale organization of immunoreceptors and the effect of antigen engagement on this organization. The application of new combinations of analysis of dSTORM data that allowed for the resolution and quantitation of individual BCRs presented here provide a new means to gain insight into the mechanisms underlying the initiation of BCR signaling in the future. MATERIALS AND METHODS ===================== Sample preparation ------------------ Primary human B-cells were isolated from the peripheral blood of healthy donors by negative selection with a magnetic bead--based human B-cell isolation kit (Stemcell Technologies) according to the manufacturer's instructions. Cells were labeled with Alexa Fluor 647--conjugated F(ab) of antibodies specific for IgM or IgG (Jackson ImmunoReseach) at 300 nM for 10 min on ice and washed with phosphate-buffered saline (PBS). The fluorophore-to-F(ab) ratio was 1.57 and 1.67 for IgM and IgG, respectively. The purity of the F(ab) preparations was \>99% (tested by immunoelectrophoresis and double immunodiffusion assay vs. anti--Fc-specific antibodies). Cells were allowed to settle on fluid planar lipid bilayer with or without anti-human κ light chain at 37°C for 10 min. The samples were then fixed in 4% paraformaldehyde at 37°C for 10 min and imaged in PBS with 100 mM 2-mercaptoethylamine (MEA) and an oxygen scavenger system consisting of 0.5 mg/ml glucose oxidase (Sigma-Aldrich), 40 μg/ml catalase (Sigma-Aldrich), and 5% (wt/vol) [d]{.smallcaps}-glucose. For live-cell, time-lapse TIRF imaging, cells were labeled with Alexa Fluor 647--conjugated F(ab) of antibodies specific for IgM or IgG at 300 nM concentration each, washed twice, and added to the fluid planar lipid bilayer containing anti--human κ light chain at 37°C. Preparation of fluid planar lipid bilayer ----------------------------------------- Fluid planar lipid bilayers were prepared as described previously ([@B4]). Briefly, the manufacturer's coverslip was removed from LabTek I eight-well chambers (Thermo Fisher Scientific) and new \#1.5 glass coverslips were cleaned with NanoStrip (Cyantek), rinsed, and glued to the chamber bottom. Biotin-containing lipid bilayers were then prepared by fusing unilamellar vesicles of 99% 1,2-dioleoyl-sn-glycero-3-phosphocoline and 1% 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-cap-biotin (Avanti Polar Lipids) to the coverslips and then incubated with 50 nM streptavidin (Invitrogen) for 10 min. For stimulating conditions, bilayers were further incubated with 10 nM biotin-F(ab′)~2~ anti--human κ light chain for 20 min. Microscopy ---------- For dSTORM, cells were imaged with an Olympus IX-81 inverted microscope equipped with a TIRF port, an oil-immersion 100× objective (TIRF 1.1, numerical aperture 1.45), an autofocus system (CRISP; Applied Scientific Imaging), and an electron-multiplying charge-coupled device camera (Evolve Delta; Photometrics), typically following a published protocol for dSTORM imaging ([@B15]). Excitation of Alexa Fluor 647 was accomplished using a 647-nm diode laser (OBIS). For each cell, 20,000 frames were acquired with an exposure time of 33 ms using MetaMorph software (Molecular Devices). For live-cell time-lapse imaging, objective lens and heated stage were maintained at 37°C with 5% CO~2~, and TIRF images were acquired at 2-s intervals with an exposure time of 33 ms after addition of cells to chambers containing bilayers. dSTORM data analysis -------------------- Single-molecule localizations in dSTORM images were accomplished with the superresolution module in MetaMorph software (Molecular Devices) after correction of the drift using fiduciary markers (TetraSpeck microspheres, 100 nm; Life Technologies). The contact area of the B-cells and gray value intensity for polarized and unpolarized cells were measured using ImageJ software (National Institutes of Health, Bethesda, MD). The fluorescence signal of single Alexa Fluor 647 molecules often lasts longer than the exposure time of a single image in a time series and thus spreads across consecutive images until it either blinks off or is irreversibly photobleached. When such peaks are fitted and displayed, an artifactual cluster of peaks results. Typically, these peaks spread within an area equivalent to the PSF of the imaging system, which is a two-dimensional Gaussian with a σ of ∼20 nm. To correct for such artifactual clustering arising from interruption of the single-molecule signal by frame transition, all peaks appearing in consecutive frames within a radius of 3σ (equivalent to an area representing 99.7% of probability density) were grouped together and substituted with a single peak whose position coordinates were calculated as a weighted average of the position coordinates and localization precision of the individual peaks comprising the group. Next a composite superresolution image was generated by combining all the grouped peaks from the entire image series. The single-molecule localizations in this composite image were organized in a heterogeneous pattern composed of single, isolated localizations and clusters with varying size and number of localizations. The commonly used pair correlation-- and Ripley's function--based cluster analysis methods ([@B28]; [@B34]) compute average estimates of cluster parameters (e.g., size, density, and occupancy number) over the entire image and thus are not able to provide an appropriate quantitative description of the heterogeneous spatial patterning of the BCRs. To quantify the heterogeneous spatial distribution of the BCRs in the composite dSTORM images, we instead used a modified Hoshen--Kopelman ([@B14])--based cluster analysis that uses an iterative spatial clustering analysis to identify and visualize individual clusters of molecules and spatially isolated single molecules within the superresolution image. The localization precision of the single molecules was used as a metric for assigning the molecules to individual protein islands, which are defined as spatially isolated assemblies of proteins. Briefly, an iterative grouping process was used to identify all neighboring single molecules within a distance of 60 nm (∼3σ of the PSF of the imaging system) and assigned them to the same protein island. By performing this operation over the composite superresolution image, we could evaluate the spatial location of individual protein islands and the distribution of proteins within the islands. This enabled the computation of cluster parameters such as size, shape, density, and protein numbers of each individual protein island. Because at least three points are required to define a two-dimensional space, the size and shape analyses of protein islands were performed on islands containing at least three protein peaks. The space occupied by each protein island was demarcated by calculating the convex hull (the smallest convex set) for the set of molecules comprising the protein island. The lines joining the vertices of the convex hull served to circumscribe the protein islands. The area of the convex hull was used as an estimate of the area of the cluster. The radius of a circle of same area as the convex hull gave an estimate of the cluster radius. The cluster density was calculated as the density of molecules within the convex hull. The shape of the protein islands was assessed by evaluating the circularity of the islands. The circularity of each protein island was calculated from the ratio of its area to its perimeter (circularity = 4π × area/perimeter^2^). The circularity value range was \[1, 0\], with 1 representing a perfect circle and lower values representing greater deviation from a circle. Combining the cluster parameter values of all the protein islands generated statistical distributions of the different cluster parameters, which provided an accurate quantitative description of the heterogeneous BCR organization in the dSTORM composite images. Thus the Hoshen--Kopelman algorithm--based cluster analysis of the superresolution data set provided an objective measure of the physical properties of the heterogeneous ensemble of receptor clusters on the surface of B-cells. Supplementary Material ====================== ###### Supplemental Materials This article was published online ahead of print in MBoC in Press (<http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E16-06-0452>) on December 14, 2016. This work was supported by the Intramural Research Program of the National Institutes of Health, National Institute of Allergy and Infectious Diseases, Eunice Kennedy Shriver National Institute of Child Health and Human Development, and Howard Hughes Medical Institute. dSTORM : direct stochastic optical reconstruction microscopy FRET : Förster resonance energy transfer PLB : planar lipid bilayer TIRF : total internal reflection fluorescence. [^1]: ^†^These authors contributed equally to this study.	{ "pile_set_name": "PubMed Central" }
1. Introduction =============== The relatively non-specific clinical manifestations of gastrointestinal disease can make it difficult for clinicians to distinguish between functional and organic intestinal disease, especially in patients presenting without rectal bleeding or systemic upset.[@CIT0005] The gold standard for identifying bowel inflammation, colonoscopy and histology, is an expensive and invasive procedure. Although attitudes to clinical targets have changed, endoscopic services are limited in many countries and a non-invasive tool to select individuals for early referral and investigations would enable the most cost effective use of resources. Faecal calprotectin (FC), a 36.5 kDa calcium-binding cytosolic protein found in neutrophils, is increasingly being used in clinical practice as a surrogate marker for intestinal inflammation. FC correlates with faecal excretion of white cells and a number of studies have demonstrated that FC is significantly elevated in the stool of patients with active inflammatory bowel disease (IBD) compared to control groups.[@CIT0015] There is a large amount of existing literature relating to FC and its use in differentiating IBD and irritable bowel syndrome (IBS). However, the majority of these studies use data obtained from patients with a pre-existing diagnosis of IBD and IBS. Few studies assess the use of FC in undiagnosed populations; those that do analyze small sample sizes.[@CIT0015] FC is described by the British Society for Gastroenterology IBD guidelines as accurate in detecting colonic inflammation, and a NICE review was completed in October 2013.[@CIT0050] The systematic review that has been produced as part of this assessment reported that 'calprotectin testing will lead to considerable savings to the NHS, as well as the avoidance of an unpleasant invasive procedure in people whose symptoms are due to IBS.'[@CIT0060] The current recommended upper limit of FC in the faeces of healthy individuals is 50 μg/g. A meta-analysis of adult patients has previously given sensitivity of 95% and a specificity of 91% when using a 50 μg/g cut-off threshold for differentiating IBD from healthy controls.[@CIT0065] A more recent meta-analysis of prospective studies using patients with suspected IBD found the pooled sensitivity and specificity of FC to be 93% and 96% respectively, although this analysis used studies with variable cut off values ranging from 24 to 150 μg/g.[@CIT0070] Importantly, FC is a poor test for colorectal cancer with a sensitivity and specificity of 36% and 71% respectively.[@CIT0065] However FC could potentially be used in clinical practice to identify young adult patients who require further invasive investigation to exclude intestinal inflammation. When used in the correct clinical scenario, with no alarm symptoms present, a negative FC result could be highly suggestive of an absence of organic gastrointestinal disease, thus usually avoiding the need for invasive investigation. Patients over the age of 50 years presenting with lower GI symptoms will require colonoscopy to exclude colorectal cancer. Since 2005 a reliable FC assay has been available in the biochemistry department at the Western General Hospital, Edinburgh. More than 8000 assays had been performed by 2008. Our clinical practice has evolved to utilise FC values in two main areas. First, FC has been used to monitor disease activity in patients with established IBD; second, to exclude IBD in patients presenting to the out-patient clinic with lower GI upset. As confidence in the utility of FC has grown, we have tended in recent years to avoid invasive endoscopic and radiological investigation in young adult patients with a negative FC (\< 50 μg/g) and no alarm symptoms. This study aimed to determine the most effective use of FC in the diagnosis of GI disease in patients with no prior known GI disease, at the first presentation to GI services. We assessed how FC can be used as a non-invasive tool to aid referral to GI services, and how this improves cost effectiveness of resource allocation through reduction of unnecessary colonoscopies. Comparison was made against other serum markers to determine the optimal initial diagnostic workup of patients aged 16--50 years. 2. Methods ========== 2.1 Patient population ---------------------- Patient data were analysed from two large teaching hospitals within the same healthcare board (NHS Lothian): Western General Hospital (WGH), Edinburgh, and the Royal Infirmary Edinburgh (RIE). These were identified using the Edinburgh Faecal Calprotectin Register (EFCR), a record kept by the Biochemistry department at WGH. The EFCR contains the name, patient I.D., date of birth, referring hospital/department, and FC concentration for all of the samples analysed. 2.2 Derivation of cohort ------------------------ The EFCR contains the data of 22,204 FC samples from 16,267 patients ([Fig. 1](#F1){ref-type="fig"}). Patients were identified who had had their first FC between January 2005 and April 2009 to allow sufficient follow-up time to pick up cases of latent IBD. 1544 patients were aged 50 or under and had at least one sample originating from the WGH or RIE from the index period. Where multiple FC samples were listed for the same patient, the initial FC value from the patient\'s first presentation was included in all analyses. Subjects were excluded from the study if they had a confirmed GI diagnosis at time of sample (n = 247) or if they had already started treatment for presumed IBD (n = 14). ![Derivation of the cohort.](eccojc_jju802_f0001){#F1} For the primary analysis, patients suffering from severe intercurrent illnesses (n = 19) were excluded as were patients receiving NSAID or aspirin therapy (n = 52), aminosalicylates and/or corticosteroids (n = 2), leaving 895 patients in the final cohort. 2.3 FC assay technique ---------------------- FC was measured in a faecal extract using a standard enzyme-linked immunosorbent assay (ELISA) technique as previously described (Calpro AS, Norway).[@CIT0075] Faecal extract was added to a microtitre plate pre-coated with polyclonal antibodies to FC. Bound FC was detected using an alkaline phosphatase labelled human antibody to FC and quantified by comparison with a known standard preparation (numerical values given between 20 and 2500 μg/g). This assay was performed in the Department of Clinical Biochemistry at the Western General Hospital, Edinburgh. The reported assay precision for the calprotectin ELISA is a coefficient of variation (CV) of less than 6%. When including the faecal extraction step, the CV for the entire assay has been estimated to be less than 10% (unpublished data; 2014 email from Susan Walker, Department of Clinical Biochemistry, Western General Hospital, Edinburgh). 2.4 Data collection ------------------- Data was collected retrospectively by review of electronic patient records and recorded on a standardised data collection form. The electronic patient record system (Trak, Intersystems, Cambridge MA, USA) logs all patient contacts with secondary care (throughout NHS Lothian), including all endoscopic and radiological investigations, clinic appointments and hospital admissions. This data was then cross-referenced with other hospital electronic databases that store clinic letters and laboratory results in order to ensure the maximum retrieval and accuracy of data. Patients were followed up until at least three years after first presentation using Trak, ensuring all re-presentations and subsequent diagnoses were noted. Parameters recorded were: age, gender, FC level and date of sample, presenting complaints (bloody diarrhoea, watery diarrhoea, rectal bleeding, constipation, abdominal pain, weight loss, flatulence/bloating, vomiting, dyspepsia, fatigue, possible extraintestinal manifestations, other), past medical history, family history (ulcerative colitis (UC), Crohn\'s disease (CD), IBD unclassified (IBDU), coeliac disease, colon cancer), smoking history (current at time of FC, ex- or never), drug history (including NSAIDs, antibiotics, laxatives, opioids, immunosuppresants, loperamide, aminosalicylates, acetaminophen, aspirin, corticosteroids), investigations performed (stool culture, colonoscopy, flexible sigmoidoscopy, upper GI endoscopy, abdominal ultrasound scan, abdominal X-ray, small bowel MRI, abdominal/pelvic CT, barium enema, barium follow-through, capsule endoscopy and radio-labelled white cell scan) and blood results (full blood count, liver function tests, erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), albumin, hematinics (ferritin, vitamin B12 and serum folate), thyroid function tests, glucose, 7 alphahydroxycholestenone and anti-tissue transglutaminase IgA titre). Rectal bleeding, bloody diarrhoea, nocturnal symptoms, weight loss and anaemia were grouped as "alarm symptoms". Where a laboratory test was reported as greater or less than a threshold, for statistical purposes it was assigned to one more or one less than the threshold respectively. Any investigations performed were recorded as normal, abnormal or incomplete. "Abnormal" endoscopy findings included mucosal abnormalities, such as histologically proven malignancies and inflammation. The normal group includes those where no abnormalities were found as well as non-adenomatous polyps and haemorrhoids. 2.5 Diagnosis ------------- Diagnosis was recorded as had been stated in the clinical notes. The Lennard-Jones criteria were used to diagnose IBD and the Montreal criteria to classify clinical phenotypes.[@CIT0080] The ROME III criteria were used to classify patients diagnosed with IBS.[@CIT0090] In cases where a diagnosis had not been recorded in the clinical notes, anonymised patient\'s notes were reviewed independently by two gastroenterologists blinded to the FC level (CWL and IDRA). Organic GI diagnoses were grouped as IBD, 'abnormal gastrointestinal (GI) tract' where a diagnosis would be expected to demonstrate a macroscopically abnormal GI tract and other GI where bidirectional endoscopy and capsule endoscopy would be expected to be normal. Details can be seen in Table S1. Patients with a definitive organic diagnosis or who had undergone full colonoscopy (n = 467) were censored at the time of initial case note review. Those cases where an organic diagnosis was not made at the time of the FC or where no colonoscopy had been performed (n = 428) were reviewed in the last quarter of 2012 to ensure that no further cases of IBD or other significant GI pathology had been missed. Patients whose symptoms resolved spontaneously, who did not require further investigation and who did not re-present to hospital with GI symptoms were classified as 'symptoms resolved.' Those who were lost to follow-up without a definitive diagnosis were classed as 'lost to follow-up'. The main comparisons have been made in those with functional disease vs. those with IBD or another condition associated with an abnormal tract, since these are the patients in whom endoscopy would be a potentially useful test. 2.6 Cost analysis ----------------- Potential cost savings were calculated using 2012 tariff prices quoted by the Department of Health.[@CIT0095] One colonoscopy with biopsies in a patient aged 19 years or older was stated to cost £563, while a flexible sigmoidoscopy plus biopsy cost £360. The in-house processing cost of a single FC assay at WGH in 2008 was £24.47. 2.7 Statistical analysis ------------------------ Statistical analyses of functional vs. organic groups and functional vs. IBD groups were performed. Medians and inter-quartile range are provided. Mann--Whitney U, Kruskal--Wallis, chi-squared and Fisher\'s exact tests were used to determine statistical significance. Receiver operating characteristic (ROC) curves were used to determine the best cut off point for FC when predicting organic disease and IBD. Comparison of area under the curve (AUC) was performed using the Delong and bootstrap methods. Positive predictive values (PPV) and negative predictive values (NPV) were calculated. Pre-test probabilities were calculated using all individuals regardless of FC concentration. Post-test probabilities were calculated with respect to different thresholds of FC. A two-tailed p-value of less than 0.05 was considered significant. Confidence intervals for sensitivity and specificity were calculated using the method described by Newcombe with continuity correction.[@CIT0100] Confidence intervals for likelihood ratios were calculated using the method described by Simel et al.[@CIT0105] Statistical analyses were performed using R 3.0.1 (R Foundation for Statistical Computing, Vienna, Austria). 3. Results ========== 3.1 Demographics ---------------- 64.9% of patients were female, and the median age (interquartile range) at the time of FC was 33.1 years (25.6--40.7) ([Table 1](#T1){ref-type="table"}). 566/895 (63.2%) of patients were diagnosed with a functional disorder ([Table 2](#T2){ref-type="table"}). 91/895 (10.2%) were diagnosed with IBD, while a further 58 (7.3%) had conditions associated with an abnormal gastrointestinal tract. 63 patients (7.0%) had other miscellaneous gastrointestinal disorders. 78 patients (8.7%) did not have a final diagnosis, of whom 48 had complete symptomatic resolution and have not re-presented in ≥ 3 years, one has had further presentations with abdominal pain without a diagnosis while the remainder were lost to follow-up. ###### Demographics of study population. ------------------------------------------------------------------------------------------------------------------ Variable All\ Primary analysis cohort\ n (%) or median (IQR) n (%) or median (IQR) --------------------------------------------- ----------------- ----------------------- -------------------------- Sex Female 627/968 (64.8%) 581/895 (64.9%) Age at calprotectin/years 33.3 (25.7--41.0) 33.1 (25.6--40.7) Smoking status at calprotectin Current 204/641 (31.8%) 183/594 (30.8%) Ex 72/641 (11.2%) 68/594 (11.4%) Never 365/641 (56.9%) 343/594 (57.7%) Unknown 327/968 (33.8%) 301/895 (33.6%) Drugs at calprotectin NSAIDS 22/769 (2.9%) 0/701 (0.0%) Antibiotics 50/769 (6.5%) 0/701 (0.0%) Laxatives 16/769 (2.1%) 12/701 (1.7%) Opiates 39/769 (5.1%) 38/701 (5.4%) Immunosuppressants 82/769 (10.7%) 61/701 (8.7%) Loperamide 6/769 (0.8%) 2/701 (0.3%) Aminosalicylates 47/769 (6.1%) 42/701 (6.0%) Acetaminophen 1/769 (0.1%) 0/701 (0.0%) Aspirin 73/769 (9.5%) 52/701 (7.4%) Corticosteroids 2/769 (0.3%) 0/701 (0.0%) Unknown 199/968 (20.6%) 194/895 (21.7%) Family history None^a^ 862/968 (89.0%) 794/895 (88.7%) UC 22/968 (2.3%) 21/895 (2.3%) CD 27/968 (2.8%) 26/895 (2.9%) IBDU 63/968 (6.5%) 60/895 (6.7%) Coeliac disease 11/968 (1.1%) 11/895 (1.2%) Colon cancer 14/968 (1.4%) 13/895 (1.5%) Previous medical history None^a^ 920/968 (95.0%) 868/895 (97.0%) Inflammatory disease (non-gastrointestinal) 30/968 (3.1%) 24/895 (2.7%) Ankylosing spondylitis 30/968 (3.1%) 24/895 (2.7%) HIV 3/968 (0.3%) 2/895 (0.2%) Alcoholic liver disease 9/968 (0.9%) 0/895 (0.0%) Severe intercurrent illness 6/968 (0.6%) 1/895 (0.1%) ------------------------------------------------------------------------------------------------------------------ NSAIDS: non-steroidal anti-inflammatory drugs; UC: ulcerative colitis; CD: Crohn\'s disease; IBDU: inflammatory bowel disease unclassified.a It has been assumed for this table that in the absence of any recorded previous medical history or family history in the patient records that there is none. ###### Faecal calprotectin, age and time from calprotectin to diagnosis by diagnostic category. ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Diagnosis category n (%) \% female Median age/years\ Median faecal calprotectin/μg/g\ Median time from calprotectin to diagnosis/days\ (IQR) (IQR) (IQR) --------------------------------------- ----------------- ----------- ------------------- ---------------------------------- -------------------------------------------------- ----------------- ----- ----------- Functional 566/895 (63.2%) 68.40% 32.7 (26.0--40.3) 20 (\< 20--50.0) 95 (40--190) IBD 91/895 (10.2%) 51.60% 29.8 (24.2--39.7) 1251 (532.5--2325.0) 7 (0--64) Abnormal GI tract 65/895 (7.3%) 53.80% 37.7 (26.1--44.4) 50 (20.0--145.0) 92 (41--206) Other GI 63/895 (7.0%) 65.10% 35 (27.0--42.8) 20 (\< 20--70.0) 92 (35--153) Other organic 32/895 (3.6%) 68.80% 31 (25.3--41.4) 22.5 (\< 20--86.2) 106 (34--192) Lost to Fup 29/895 (3.2%) 62.10% 35.8 (26.5--43.2) 135 (35.0--325.0) None 1/895 (0.1%) 100.00% 20.8 1825 Symptoms resolved --- no GI pathology 48/895 (5.4%) 62.50% 34.3 (25.3--42.7) 35 (\< 20--76.2) ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 3.2 FC and demographic variables -------------------------------- FC was not significantly associated with age (p = 0.21), sex (p = 0.18) or current smoking (p = 0.80). 3.3 FC and other clinical parameters assessed by final diagnosis ---------------------------------------------------------------- FC was significantly higher in patients diagnosed with IBD (median FC 1251 μg/g, IQR 532--2325 μg/g) than those with other conditions associated with an abnormal gastrointestinal tract (median FC 50 μg/g, IQR 20--145 μg/g) or with a functional diagnosis (median FC 20 μg/g, IQR \< 20--50 μg/g) (p ≤ 0.0001 in each case, see [Fig. 2](#F2){ref-type="fig"}). ![Box plot showing difference in faecal calprotectin between patients with functional diagnoses and those with IBD and other conditions associated with an abnormal GI tract.](eccojc_jju802_f0002){#F2} 3.4 FC in patients taking non-steroidal anti-inflammatories (NSAIDs) -------------------------------------------------------------------- Patients taking NSAIDs or aspirin were excluded from the primary analysis. In patients with a functional diagnosis, the FC was significantly higher in those taking NSAIDs or aspirin when compared with those on neither drug (median FC 52 μg/g \[IQR \< 20--181 μg/g\] vs. 20 μg/g \[IQR \< 20--50 μg/g\], p = 0.001). 3.5 FC in patients with IBD --------------------------- Of the 91 patients ultimately diagnosed with IBD, 40 (44%) had CD, 41 (45%) had UC and 10 (11%) had IBDU. There was no significant difference in FC between the three subtypes of IBD (p = 0.56). Within the group with CD, there were 10 (25%) with L1 (ileal) disease, including one patient with L1 + 4, 18 (45%) with L2 (colonic) disease and 12 (30%) with L3 (ileocolonic) disease. FC was significantly higher in those with L2 or L3 disease, with a median (IQR) of 1280 (714--2295) μg/g than in those with L1 disease where median (IQR) FC was 495 (288--822) μg/g (p = 0.009) (Supplementary Fig. 1). Within the group with UC, there were 3 (7%) with E1 disease (proctitis), 12 (29%) with E2 disease (left-sided colitis) and 21 (51%) with E3 disease (extensive colitis). In the remaining 5 patients, the disease extended beyond the point of insertion of the sigmoidoscope and complete staging of extent was not achieved during the initial diagnostic period. There was no significant difference in FC by disease extent when those without complete staging were excluded (p = 0.25 by Kruskal--Wallis test; Supplementary Fig. 2). Across all patients diagnosed with IBD, there was no significant association between time to diagnosis and faecal calprotectin. 3.6 Diagnostic accuracy of FC compared to other clinical parameters ------------------------------------------------------------------- Receiver operating characteristic (ROC) analysis revealed an area under the curve (AUC) for FC of 0.85 for prediction of conditions with an abnormal GI tract (including IBD) vs. functional disease, and 0.97 for prediction of IBD vs. functional disease ([Fig. 3](#F3){ref-type="fig"}). This was significantly higher than that seen for CRP, albumin, ESR or white cell count in both cases (p \< 0.001 for all comparisons). The sensitivities, specificities, and positive and negative predictive values for faecal calprotectin can be seen in [Table 3](#T3){ref-type="table"} at different thresholds. Summaries of the number of available tests, medians and interquartile ranges for each parameter can be seen in Supplementary Table S2. ![Receiver operating characteristic curves for calprotectin, CRP, albumin, ESR and white cell count as predictors of inflammatory bowel disease (IBD) or abnormal gastrointestinal tract versus functional disease (A) and IBD versus functional disease (B).](eccojc_jju802_f0003){#F3} ###### Diagnostic accuracy of fecal calprotectin at different thresholds PPV: positive predictive value; NPV: negative predictive value; PLR: positive likelihood ratio; CI: confidence interval. Threshold fecal calprotectin (μg/g) Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI) NPV (95% CI) PLR (95% CI) ----------------------------------------------------------------------------------- ---------------------- ---------------------- -------------- -------------- -------------- -------------- ------ -------------- ------ -------------- A: Inflammatory bowel disease (IBD) or abnormal GI tract vs. functional disease 20 0.89 (0.83--0.93) 0.49 (0.44--0.53) 0.32 (0.28--0.37) 0.94 (0.91--0.96) 1.73 (1.57--1.91) 50 0.79 (0.71--0.85) 0.74 (0.70--0.77) 0.45 (0.39--0.52) 0.93 (0.90--0.95) 3.02 (2.57--3.54) 70 0.73 (0.65--0.80) 0.80 (0.76--0.83) 0.50 (0.44--0.57) 0.92 (0.89--0.94) 3.66 (3.03--4.43) 100 0.70 (0.62--0.77) 0.87 (0.84--0.90) 0.60 (0.52--0.67) 0.91 (0.89--0.93) 5.42 (4.27--6.87) B: IBD vs. functional disease 20 0.99 (0.93--1.00) 0.49 (0.44--0.53) 0.24 (0.20--0.28) 1.00 (0.98--1.00) 1.92 (1.77--2.09) 50 0.97 (0.90--0.99) 0.74 (0.70--0.77) 0.37 (0.31--0.44) 0.99 (0.98--1.00) 3.70 (3.20--4.27) 70 0.97 (0.90--0.99) 0.80 (0.76--0.83) 0.44 (0.37--0.51) 0.99 (0.98--1.00) 4.84 (4.09--5.74) 100 0.96 (0.89--0.99) 0.87 (0.84--0.90) 0.54 (0.46--0.62) 0.99 (0.98--1.00) 7.41 (5.96--9.22) 3.7 Synergistic effect of FC sampling and alarm symptoms -------------------------------------------------------- Alarm symptoms were present in 25% (140/566) of those ultimately diagnosed with functional disease, 86% (78/91) of those diagnosed with IBD and 54% (35/65) of those diagnosed with another condition associated with an abnormal GI tract (p \< 0.0001). The positive predictive value of alarm symptoms for IBD or an abnormal GI tract vs. functional disease was 0.45 (95% CI 0.38--0.51), and the negative predictive value was 0.91 (0.88--0.93) with a sensitivity of 0.72 (0.65--0.79) and specificity of 0.75 (0.71--0.79). For prediction of IBD vs. functional disease, the PPV was 0.36 (0.29--0.43) and NPV was 0.97 (0.95--0.98), with a sensitivity of 0.86 (0.76--0.92) and specificity of 0.75 (0.71--0.79). As can be seen in [Table 4](#T4){ref-type="table"}, FC is helpful in improving the prediction of an abnormal GI tract or IBD compared with alarm symptoms alone. Within the cohort with functional disease or an abnormal GI tract, none of the 329 patients with no alarm symptoms and a FC of \< 50 μg/g was found to have IBD, while 11/36 (31%) of patients with no alarm symptoms and a FC of ≥ 200 were found to have IBD. ###### Pre- and post-test probabilities when combining alarm symptoms and fecal calprotectin. Pre-test probability Post-test probability for different values of fecal calprotectin (μg/g) --------------------------------------------------------------------------------- ---------------------- ------------------------------------------------------------------------- ------ ------ ------ ------ A: inflammatory bowel disease (IBD) or abnormal GI tract vs. functional disease Alarm symptoms 0.45 0.15 0.18 0.24 0.50 0.91 No alarm symptoms 0.09 0.03 0.06 0.12 0.20 0.41 B: IBD vs. functional disease Alarm symptoms 0.36 0.02 0.05 0.05 0.41 0.89 No alarm symptoms 0.04 0.00 0.00 0.00 0.06 0.33 Thirteen patients had no alarm symptoms and a FC of \< 50 μg/g, but were found to have a disease associated with an abnormal GI tract. These were 1 case of appendicitis, 1 coeliac disease, 3 with colonic adenomatous polyps, 1 with diverticulosis, 3 with GI infections (1 Fasciola hepatica, 1 giardiasis, 1 presumed infection with response to metronidazole), 2 with gastro-oesophageal reflux disease and 2 non-specific bowel inflammation. One of these patients with non-specific bowel inflammation was initially thought to have CD but had non-specific changes on her index colonoscopic biopsies and subsequently normal colonoscopy and biopsies. 3.8 Multivariable analysis -------------------------- Multiple logistic regression analysis of predictors of IBD vs. functional disease showed that elevated FC, elevated CRP, male sex, alarm symptoms and albumin were independently significant. Age at FC and white cell count were not ([Table 5](#T5){ref-type="table"}). ###### Multiple logistic regression of predictors of inflammatory bowel disease vs. functional disease. CRP: C-reactive protein. Variable Odds ratio (95% CI) P ------------------------------- --------------------- ------------ Fecal calprotectin≥50μg/g 65.3 (12.1--351.5) 1.1×10^−6^ Alarm symptoms 19.5 (7.9--127.5) 3.0×10^−6^ Albumin\<40g/L 18.7 (4.1--85.4) 3.0×10^−5^ Male sex 14.1 (3.8--52.2) 7.0×10^−5^ CRP≥5g/L 6.9 (2.0--23.7) 0.002 Age at calprotectin \>0.05 White cell count \>11×10^9^/L \>0.05 Comparing different strategies of investigation ([Table 6](#T6){ref-type="table"}) demonstrated that FC alone provided the optimum specificity for both IBD vs. functional disease and IBD or abnormal GI tract vs. functional disease. The optimal combination of sensitivity and specificity was attained using the approach of alarm symptoms or FC ≥ 50 μg/g. Sensitivity and specificity for IBD vs. functional disease were 1.00 and 0.54 with this strategy, while for IBD or abnormal GI tract vs. functional disease they were 0.96 and 0.55. Adding CRP to this combination had minimal effect on sensitivity, while reducing specificity. ###### Comparison of different strategies for identifying IBD or abnormal gastrointestinal (GI) tract vs. functional disease. IBD: inflammatory bowel disease; CRP: C-reactive protein. Strategy IBD vs. functional IBD or abnormal GI tract vs. functional -------------------------------------------- -------------------- ----------------------------------------- ------ ------ Alarm symptoms only 0.85 0.73 0.76 0.74 CRP≥5g/L only 0.85 0.70 0.71 0.70 Faecal calprotectin (FC)≥50μg/g only 0.97 0.74 0.86 0.75 Alarm symptoms or CRP≥5g/L 0.99 0.50 0.89 0.51 Alarm symptoms or FC≥50μg/g 1.00 0.54 0.96 0.55 Alarm symptoms or CRP≥5g/L or FC≥50μg/g 1.00 0.39 0.97 0.39 Alarm symptoms or (CRP≥5g/L and FC≥50μg/g) 0.99 0.65 0.88 0.67 3.9 Low FC in patients diagnosed with inflammatory bowel disease ---------------------------------------------------------------- Three patients had a low FC (\< 50 μg/g) and were diagnosed with inflammatory bowel disease. All three had alarm symptoms (two had blood in their stool and one had weight loss). Two of these patients were diagnosed with ulcerative proctitis which has not extended further in \> 4 years of follow-up. One had mild terminal ileal CD with no subsequent progression. 3.10 Cost effectiveness of FC: reducing the number of invasive investigations ----------------------------------------------------------------------------- Between 2005 and 2008, our practice evolved with increasing use of FC and reduction in the percentage of these patients subsequently undergoing invasive investigation. In the 2005, 63 patients underwent stool analysis for FC with 84.1% of them undergoing either sigmoidoscopy or colonoscopy. In 2008, 409 patients had stool sent for FC with 56.7% subsequently undergoing invasive investigation (Table S3). Over the study period, 581/895 (64.9%) patients presented without alarm symptoms. 395 of these (68.0%) had a FC of \< 50 μg/g. 150 of these patients (38%) had a subsequent colonoscopy and 50 (13%) a flexible sigmoidoscopy, identifying incidental adenomatous polyps in 3 patients and no other significant pathology. If the low FC had been used to triage these patients to a non-invasive approach, this would have saved £88,233 over that time period. 4. Discussion ============= This study uses the largest, 'real-world' population of undiagnosed patients to determine the best way of using FC at first presentation to the GI clinic to differentiate non-invasively between organic and functional disease. This allows identification of those in need of efficient and effective further investigation. Incorporating FC into the standard work-up of patients presenting with lower GI symptoms may potentially relieve pressure on hospital services by identifying patients who can be managed solely in primary care. Our findings corroborate existing data showing that FC reliably distinguishes between patients with functional disease and IBD. Von Roon et al.\'s meta-analysis of adult patients demonstrated a sensitivity of 95% and specificity of 91% when using a 50 μg/g cut-off point for differentiating IBD patients from healthy controls.[@CIT0065] At the same cut-off, our study found 95% sensitivity but only 75% specificity. This agreement in sensitivity reinforces the diagnostic ability of FC in identifying patients with IBD in a large cohort of patients. The lower specificity seen in our study may be due to the patient population used, all of which have presented to services with GI symptoms, unlike the healthy control population used by Von Roon et al. Van Rheenen et al\'s more recent meta-analysis of six adult studies found a pooled sensitivity and specificity of 93% and 96% respectively.[@CIT0110] However, inconsistent FC thresholds were used in these six studies, with 47.7% of included patients analysed using a cut-off greater than 100 μg/g, and this may have influenced the specificity. Both ESR and CRP are markers that are commonly used to identify systemic inflammation in patients with IBD-like symptoms. In accordance with previous research, we show CRP and ESR are raised in patients with organic disease and IBD.[@CIT0020] ROC analysis demonstrates however, that FC is superior to CRP and ESR in the diagnosis of IBD --- a finding that agrees with the recent economic report produced by the NHS Centre for Evidence Based Purchasing.[@CIT0120] Furthermore, we demonstrate that the NPV of FC in patients presenting with no alarm symptoms is superior to the NPV of CRP for both organic GI disease and IBD. Cost savings could be made by solely checking FC in patients presenting with lower GI symptoms, rather than checking CRP and ESR in these patients One of the most clinically relevant findings from our data is the NPV for IBD of 99.0% when a FC threshold of 50 μg/g is used. When FC less than 50 μg/g is combined with the absence of alarm symptoms, NPV is 100.0% for IBD. This allows the exclusion of IBD from the differential diagnosis of these patients. Furthermore, in patients meeting these criteria, NPV for any GI tract abnormality is 96.1%. Of the 13 patients with no alarm symptoms and FC less than 50 μg/g who had a diagnosis of abnormal GI tract, colonoscopy was helpful in only four patients and these (diverticular disease and colonic polyps) were likely incidental findings. Clinicians can therefore be reassured that referral for colonoscopy will not identify severe organic disease in patients in whom no abnormalities are found in initial investigations. This finding could potentially be applied to a primary care scenario and aid selection of patients for colonoscopy. With the Department of Health pricing a single colonoscopy in adults at £563 there is great potential for FC to aid more cost-effective decision making with regard to further investigation.[@CIT0095] Von Rheenen et al.\'s meta-analysis demonstrated that screening with FC could reduce unnecessary colonoscopies by 67% in those suspected of having IBD.[@CIT0110] Similar results were documented by Mindemark and colleagues, with a reduction of colonoscopies by 50% using the FC cut off of \< 50 μg/g and 67% using a FC cut off of \< 100 μg/g.[@CIT0125] During the study period of the present study, if patients with a FC \< 50 μg/g and no alarm symptoms had not undergone lower GI endoscopy there could have been 150 fewer colonoscopies and 50 fewer flexible sigmoidoscopies. Our data reflects real world practice, with proportionally fewer patients being investigated by colonoscopy as our knowledge and experience of FC increased. A reducing trend in the numbers of those patients investigated with colonoscopy can clearly be seen as the number of FC assays received by the labs increase over the three years. The number of potential colonoscopies saved quoted above may even be more than this had our unit not been internally evaluating FC\'s use in clinical practice. Furthermore, the numbers we have analyzed only include patients who attended the GI clinic and had a FC sample sent. These findings could be applied to all patients who attend the GI clinic with lower GI symptoms, potentially reducing further the number of colonoscopies and resulting in even greater cost savings. It is important to take into consideration that this study uses patients referred to hospital GI services, and by virtue of this the spectrum of symptoms seen in this population is more severe when compared to all the patients presenting to GPs with GI symptoms. In primary care, FC could identify the small numbers of patients with IBD, whilst excluding its presence in a large number of patients presenting with GI symptoms. Not only could this streamline the referral of appropriate patients to hospital, but it will also reduce the number of unnecessary referrals and invasive investigations. This does, however, require detailed pilot testing before any formal recommendations about the roll-out of FC into primary care can be made. Moreover, it is important that FC is used in the context of a defined protocol to ensure that it does not delay referral of patients with alarm symptoms and that consideration is given to possible false positive tests from aspirin and non-steroidal inflammatory drugs. One of the strengths of this study is that all individuals without a definitive diagnosis or in whom a functional diagnosis had been made without colonoscopy were re-reviewed three years later to identify any possible latent cases of IBD or other GI disease. This study clarifies important, clinically relevant information about FC. Awareness of the high negative predictive value of FC allows clinicians to effectively exclude IBD as a cause for gastrointestinal symptoms in patients with FC levels under 50 μg/g. FC can thus be used as an adjunct to other presenting complaints and investigations, allowing the risk stratification of patients presenting with gastrointestinal symptoms in a cost-effective manner. Statement of interests ====================== NAK is funded by a research training fellowship from the Wellcome Trust\[grant 097943\] and has had financial support to attend meetings from Warner Chillcott, Shire, MSD and Norgine. He has had speaker fees from MSD and Warner Chilcott. JS has had research funding from Abbvie, speaker fees from Ferring and travel support from Shire. IDRA has been on advisory boards for MSD, Hospira and P&G. CWL has been on advisory boards for and had lecture fees from Abbvie, Hospira, MSD, Vifor, Pharmacosmos and P&G. Guarantor Dr Charlie Lees is the guarantor of this article. Authorship CWL had the initial concept and managed the study. AC, AW, NAK, JCWC, FFS, MM collected the data. WGB and KK provided the calprotectin data and biochemical expertise. IDRA and CWL reviewed cases where the diagnosis was uncertain. NAK conducted the statistical analyses. NAK, AC and AW wrote the initial draft of the manuscript. All the authors contributed to revision of the manuscript and approved the text. AUC : area under the curve CD : Crohn\'s disease CRP : C-reactive protein ESR : erythrocyte sedimentation rate FC : faecal calprotectin IBD : inflammatory bowel disease IBDU : inflammatory bowel disease unclassified NHS : National Health Service NICE : National Institute of Clinical Excellence NPV : negative predictive value PPV : positive predictive value ROC : receiver operating curve UC : ulcerative colitis Supplementary data to this article can be found online at <http://dx.doi.org/10.1016/j.crohns.2014.07.005>. [^1]: These authors contributed equally to the manuscript. [^2]: All authors represent the Gastroenterology Department of the Western General Hospital Edinburgh.	{ "pile_set_name": "PubMed Central" }
Core tip: Non-alcoholic steatohepatitis (NASH) is anticipated to account for a greater proportion of hepatocellular carcinoma (HCC) incidence due to the growing epidemic of obesity and diabetes. Currently NASH-related HCC is the fastest growing indication for liver transplant in HCC candidates. Increased efforts to implement effective screening and preventative strategies particularly in non-cirrhotic NASH patients possibly based on genetic susceptibility are needed to reduce the future impact imposed by NASH-related HCC. INTRODUCTION ============ Hepatocellular carcinoma (HCC) accounts for 90% of primary liver cancers Worldwide, HCC being the sixth most common cancer, and is the second leading cause of cancer-related death\[[@B1]\]. HCC largely occurs in the background of chronic liver disease and cirrhosis of the liver\[[@B2]\]. The leading liver disease etiologies for cirrhosis in patients with HCC include but are not limited to chronic hepatitis B, chronic hepatitis C virus (HCV), and alcoholic liver disease. With advent of curative treatments for HCV, the risk of progression to cirrhosis and development of HCC secondary to HCV is anticipated to decline. However, in recent years, non-alcoholic fatty liver disease (NAFLD) has quickly risen as one of leading etiologies for liver disease. NAFLD is a spectrum of chronic liver disease ranging from simple hepatic steatosis to liver cell injury and inflammation known as non-alcoholic steatohepatitis (NASH). The rising incidence of NAFLD/NASH has subsequently led to a dramatic rise in NASH-related HCC incidence\[[@B3]\]. Numerous studies have demonstrated that NASH can lead to advanced fibrosis and cirrhosis, thereby increasing the risk of developing HCC\[[@B4]-[@B6]\]. Among patients with NAFLD or NASH, liver disease is the third leading cause of death\[[@B4]\], while HCC represents the main cause of death in this group\[[@B7]\]. The cumulative annual incidence rate for developing HCC in patients with NASH-related cirrhosis is approximately 2.4%-12.8%\[[@B8]\]. In the absence of NASH or cirrhosis, NAFLD can present with HCC. These patients usually present with less aggressive tumors and are less likely to diagnosed by surveillance compared to HCC that develops in the setting of viral hepatitis\[[@B9]-[@B11]\]. A similar rising trend has been reported in NASH progressing to HCC in the absence of cirrhosis\[[@B12]-[@B14]\]. In NASH, several risk factors for HCC development have been identified including metabolic syndrome and insulin resistance causing changes in serum cytokines, persistent inflammation, and altered gut microflora and bile composition\[[@B15]\]. EPIDEMIOLOGY ============ Currently, NAFLD affects more than 80 million Americans, making it the most common etiology for liver disease in the United States. With the incidence of obesity, diabetes and metabolic syndrome continuing to increase in the United States and Europe, NAFLD/NASH may become the most common cause of HCC in developed countries in the near future\[[@B16]\]. In 2012, primary liver cancer was recognized overall as the second most common cause of cancer-related death in the world. In the United States, HCC is the most rapidly rising cause of cancer and cancer-related deaths with an incidence that has tripled over the last decade. This high likelihood for mortality reflects a poor prognosis without therapeutic intervention\[[@B17]\]. HCC is the most prevalent histological subtype accounting for 70%-85% of primary liver malignancies\[[@B18]\]. Compared to HCC in alcoholic liver disease and viral hepatitis, there is a lack of strong epidemiological data regarding the incidence and prevalence of HCC in NAFLD\[[@B19]\]. While the prevalence of NAFLD is thought to be highest among Hispanics and Caucasians, the ethnic distribution among NAFLD/NASH-related HCC patients has yet to be defined\[[@B20]\]. NASH-related HCC patients are predominantly male; however, gender has not been proven to be a statistical risk factor NASH progression to HCC\[[@B21]\]. Studies analyzing demographic and clinical characteristics of NASH-related HCC patients are outlined in Table [1](#T1){ref-type="table"}. Reports indicate that NASH can be verified by histological evaluation in up to 47% of all NAFLD cases among obese individuals\[[@B22],[@B23]\]. Amongst a growing population of diabetes which has surpassed 26 million in the United States, the prevalence of biopsy-proven NAFLD and NASH has been reported to be as high as 74% and 11%, respectively\[[@B24],[@B25]\]. ###### Reported studies of hepatocellular carcinoma in patients with cirrhotic and non-cirrhotic non-alcoholic fatty liver disease/non-alcoholic steatohepatitis, and their clinical characteristics Ref. *All (n)* *NASH/NAFLD (n)* Study type Clinical characteristics Cirrhotic NASH with HCC Non-cirrhotic NASH with HCC ---------------------------- --------------- ------------------------------------ ------------------------------- --------------------------------------------------------------------- ------------------------------------- ------------------------------------ ------------------------------------- ------------------------------------ Cotrim et al\[[@B97]\] 110 110 Cohort Age, 67 ± 11 yr; male, 72 (65.5%); non-Hispanic white, N/A 32 (29.1%) 58 (52.7%) 20 (18.2%) 0 Van Meer et al\[[@B98]\] 933 91[1](#T1FN1){ref-type="table-fn"} Cohort Age, 64 yr; male, 60 (66%); non-Hispanic white, N/A N/A N/A 91 (100%) N/A Shrager et al\[[@B99]\] 9 9 Case series Age, 58 yr; male, 8 (88.9%); non-Hispanic white, N/A 5 (55.5%) N/A 4 (44.4%) N/A Kikuchi et al\[[@B93]\] 42 38 Case series Age, 66.5 yr; male 26 (62%); non-Hispanic white, N/A 34 N/A 4 N/A Chagas et al\[[@B100]\] 394 7 Prospectiv Age, 63 ± 13 yr; Male 4 (57%); non-Hispanic white, N/A 6 N/A 1 N/A Ertle et al\[[@B101]\] 150 36 Cohort Age, 68.6 ± 8.4 yr; male 32 (88.9%); non-Hispanic white, N/A 5 14[2](#T1FN2){ref-type="table-fn"} 10 7[2](#T1FN2){ref-type="table-fn"} Tokushige et al\[[@B102]\] 2299 292 Cohort Age, 72 ± 8.4 yr; male, 181 (62%) 181[3](#T1FN3){ref-type="table-fn"} N/A 111[3](#T1FN3){ref-type="table-fn"} N/A Hashizume et al\[[@B103]\] 1310 10 Case series Age 71.5 yr; male 6 (66.7%) 5 N/A 4 N/A Kawada et al\[[@B13]\] 807 8 Cohort Age 73 yr; male 3/6 (50%); non-Hispanic white, N/A 2 N/A 6 N/A Malik et al\[[@B104]\] 143 143 Case control Age 59 ± 7.6 yr; male 44 (44.9%); 16 non-Hispanic White, 1 Asian 17 N/A 0 N/A Takuma et al\[[@B105]\] 11 11 Case series/Literature review Age 73.8 ± 4.9 yr; male 5 (45%) 4 N/A 7 N/A Perumpail et al\[[@B106]\] 44 6 Cohort Age 72 ± 8 yr; male 5 (83.3%) NA NA 6 N/A Ascha et al\[[@B107]\] 510 195 Cohort Age 56.5 yr; male 86 (44.1%) NA NA N/A 25[4](#T1FN4){ref-type="table-fn"} Mohamad et al\[[@B108]\] 83 83 Cohort retrospective Age 64.8 ± 10.4 yr; male 54 (65.1%); non-Hispanic White, 77 (92.8%) 47 N/A 36 N/A Histological data available in 86 patients only; AASLD Radiological criteria used for diagnosis; Results based on both liver biopsy and abdominal imaging. Differentiating data not available in the study; Histologic confirmation obtained in 59% of the patients diagnosed with HCC. HCC: Hepatocellular carcinoma; EMT: Epithelial to mesenchymal transition; NAFLD: Non-alcoholic fatty liver disease; NASH: Non-alcoholic steatohepatitis; N/A: Not available. This rise in the incidence of NASH-related HCC has impacted trends in liver transplantation as well. A retrospective cohort study amongst adult liver transplant recipients from 2002-2012 indicated that there was 4-fold increase in patients undergoing liver transplant for NASH-related HCC compared to 2-fold increase in number of patients undergoing transplantation for HCV-related HCC\[[@B26]\]. During this 10-year span, NASH also became the second leading cause of HCC-related liver transplantation in America, steadily increasing from 8.3% in 2002 to 10.3% in 2007 and to 13.5% in 2012\[[@B16]\], and most likely will surpass 15% by 2017. PROGRESSION OF NASH/NAFLD TO HCC ================================ NAFLD is the hepatic manifestation of metabolic syndrome, with insulin resistance driving the alteration in physiology. As mentioned earlier, it ranges from isolated hepatic steatosis, to NASH with or without cirrhosis, and progression to HCC. The diagnosis of NASH is based on histological evidence of hepatic steatosis or magnetic resonance spectroscopic evidence \> 5% fat accumulation of liver weight without the presence of secondary causes such as alcohol abuse, endocrine disorders, chronic HCV infection or familial hypobetalipoproteinemia\[[@B27]\]. Recent evidence has demonstrated an association between NASH and HCC to be exclusive to patients who had progressed to cirrhosis, suggesting causality\[[@B8]\]. Compared to benign course of simple steatosis, patients with NASH are more likely to develop progressive advanced liver disease. Matteoni et al\[[@B28]\] demonstrated increased rates of cirrhosis in patients with NASH compared to those with fatty liver without NASH (25% vs 3%, respectively), and increased risk of liver disease-related death (11% vs 2%, respectively). In a much larger study across the entire spectrum of NAFLD which included 420 patients, Rafiq et al\[[@B29]\] demonstrated a higher mortality in those with NASH/NAFLD when compared to the general population; liver-related deaths occurred in 13% vs \< 1% in general population, and 3% of those with NAFLD developed cirrhosis. Another study further confirmed increased rate of liver-related deaths among patients with NASH when compared with those without NASH (17.5% vs 3%, respectively). In patients with compensated cirrhosis, NASH-related cirrhosis patients had better survival outcomes compared to HCV-related cirrhosis patients. However, in decompensated cirrhosis both cohorts had comparable poor outcomes\[[@B30],[@B31]\]. Currently, both the American Association for the Study of Liver Diseases, and the European Association for the Study of Liver Disease recommend screening for HCC in patients with NASH related cirrhosis every 6-12 mo\[[@B32]\]. HCC IN NON-CIRRHOTIC NAFLD/NASH =============================== Emerging evidence suggests that a significant proportion of patients with NAFLD-associated HCC, do not have histologic evidence of cirrhosis. In a study conducted by Kawada et al\[[@B13]\], of 1168 patients who underwent hepatic resection for HCC, 6 of 8 patients with NASH-related HCC did not demonstrate cirrhosis. This study suggested that the presence of cirrhosis in NASH-related HCC was lower compared to HCV-related HCC. These data suggest that compared to patients HCV, HCC may develop at an earlier stage those with NASH. Paradis et al\[[@B33]\] analyzed 128 HCC patients who were recruited over 12 years, and reported significant number of patients with NASH developed HCC in the absence of fibrosis when compared to HCC in the setting of other underlying chronic liver disease (65% with F0-F2 in NASH group vs 26% in chronic liver disease)\[[@B33],[@B34]\]. To explain this phenomenon in non-cirrhotic NAFLD patients, one proposed hypothesis is the malignant transformation of hepatic adenoma. Few published reports have suggested that in the presence of metabolic syndrome, hepatocellular adenoma may incur a malignant transformation\[[@B19],[@B35]\]. HCC IN CIRRHOTIC NAFLD/NASH =========================== During the last two decades, various studies have tried to determine the relationship between NAFLD/NASH, cryptogenic cirrhosis and HCC. A recent meta-analyses by White et al\[[@B8]\] showed that approximately 60% HCC cases attributed to NAFLD/NASH had cirrhosis either before or at the time of diagnosis. This meta analyses also included review of cohort and longitudinal studies which showed that NASH-associated cirrhosis consistently carried an increased HCC risk ranging between 2.4% and 12.8%\[[@B8]\]. Additionally, this study reported the risk of developing HCC is lower in patients with cirrhosis due to NAFLD/NASH when compared to those with chronic HCV (NAFLD/NASH, 26.9% vs HCV, 19.7%). The true prevalence of NASH and NASH-related HCC is likely underestimated. In up to 6.9%-29% of HCC, the underlying etiology of liver disease is unknown and is considered secondary to cryptogenic cirrhosis\[[@B19]\]. Features suggestive of NASH are more frequently observed in HCC arising in patients with cryptogenic cirrhosis than in age- and sex-matched HCC patients of well-defined viral or alcoholic etiology\[[@B36]\]. Although the prevalence of NAFLD/NASH-related HCC is not well defined, the increasing incidence of obesity and diabetes, suggests the impact of NAFLD/NASH-related HCC will continue to grow. MORTALITY IN NAFLD/NASH ======================= Long term outcomes in NAFLD and NASH has been evaluated in several studies and distinctive differences between NASH and non-NASH subtypes of NAFLD have been shown\[[@B28],[@B29],[@B37]-[@B42]\]. Type 2 diabetes mellitus has been shown to increase the risk of both liver- related mortality and overall mortality in NAFLD patients\[[@B43],[@B44]\]. In light of these findings NAFLD patients with type II diabetes should be prioritized in future treatment protocols\[[@B44]\]. A population-based study published in 1996 followed 153852 subjects and found that diabetic patients had a standardized incidence ratio of 4.1 for HCC\[[@B45]\]. However, another retrospective analysis from United States Veteran Registry noted increased the risk of primary liver cancer in patients with diabetes only in the presence of other risk factors such as hepatitis C or B or alcoholic cirrhosis\[[@B46]\]. These observations were not supported by further analysis that found an incremented HCC risk in diabetic patients independently from alcoholic liver disease and viral hepatitis\[[@B47],[@B48]\]. In a recent meta-analyses, Younossi et al\[[@B49]\] reported that in NAFLD patients, annual incidence of HCC was 0.44 per 1000 person-years (95%CI: 0.29-0.66), whereas for those with NASH, the annual HCC incident rate was 5.29 per 1000 person-years (95%CI: 0.75-37.56). Among NAFLD cohort, the pooled liver-specific and overall mortality incidence rates were 0.77 per 1000 person-years (95%CI: 0.33-1.77 events) and 15.44 per 1000 person-years (95%CI: 11.72-20.34 events), respectively. Among the NASH cohort, the pooled liver-specific and overall mortality incidence rates were 11.77 per 1000 person-years (95%CI: 7.10-19.53 events) and 25.56 per 1000 person-years (95%CI: 6.29-103.8 events), respectively. Although cardio-vascular (CV) events remain the major cause of death in patients with NAFLD and NASH, the CV mortality rate amongst the NASH and non-NASH subtypes of NAFLD is similar\[[@B42],[@B50]-[@B52]\]. Since patients with NASH have significantly higher liver-related mortality than those with non-NASH NAFLD, treatment strategies should be designed to ameliorate the risks for cardiovascular mortality\[[@B28],[@B29],[@B38],[@B40]-[@B42],[@B49],[@B50]\]. Further, patients with NASH and type 2 diabetes mellitus, will need increased attention and linkage of care to reduce liver disease-related compliactions and to reduce their risk of HCC\[[@B53]-[@B55]\]. RISK FACTORS AND PROPOSED MECHANISMS FOR NASH-RELATED HCC ========================================================= Development of HCC in the setting of chronic liver disease is a complex but gradual process that requires transition through a dysplasia-carcinoma sequence. Several putative oncogenic mechanisms has been incriminated that lead to genomic instability, including telomere erosion, chromosome segregation defects and alterations in the DNA-damage-response pathways\[[@B56],[@B57]\]. Obesity and diabetes are involved in the mechanisms involved in the development of HCC in NAFLD. The development of HCC in NAFLD is likely multifactorial; involving low grade chronic systemic inflammatory response, increased lipid storage and lipotoxicity, gut disbiosis with elevated levels of lipopolysaccharide (LPS) and hyperinsulinemia with insulin resistance and increased IGF levels\[[@B19]\]. In addition patients with HCC from NAFLD in general has a distinctive phenotype with presentation in older age, being less aggressive and less likely to be diagnosed by surveillance compared with HCC caused by viral hepatitis\[[@B9]-[@B11]\]. Other factors such as genetic polymorphism and, increased iron absorption may also lead to development of HCC in NASH\[[@B14]\]. Proposed mechanisms for NASH-related HCC are depicted in Figure [1](#F1){ref-type="fig"}. ![Risk factors and proposed mechanisms for non-alcoholic fatty liver disease and non-alcoholic steatohepatitis-related hepatocellular carcinoma. The development of NAFLD and NASH-related HCC is multifactorial. Proposed pathogenic mechanisms include obesity, peripheral and hepatic insulin resistance from type 2 diabetes, increased hepatic lipid storage and lipotoxicity, EMT, genetic mutations and intestinal mibrobiota dysregulation. HCC: Hepatocellular carcinoma; EMT: Epithelial to mesenchymal transition; NAFLD: Non-alcoholic fatty liver disease; NASH: Non-alcoholic steatohepatitis; FFA: Free fatty acid; IGF: Insulin-like growth factor; LPS: Lipopolysaccharide; PNPLA3: Patatin-like phospholipase domain-containing 3; TM6SF2: Transmembrane 6 superfamily member 2.](WJH-9-533-g001){#F1} Cytokines carry out the intercellular communication signals, cellular interactions along with growth and differentiation. Disease states cause imbalances in cytokine levels promoting aberrant signaling and modulating inflammatory responses seen in epithelial to mesenchymal transition pathologic process\[[@B15]\]. Imbalances in the levels of cytokines such as tumor necrosis factor (TNF)-alpha, leptin, adiponectin and interleukin-6 (IL-6) play a pivotal role in NASH\[[@B58]-[@B60]\]. Obesity ------- Obesity is a significant risk for the development of HCC particularly in patients with NASH, who have a higher predisposition for obesity. Obese (body mass index \> 30 kg/m^2^) patients have a reported 1.93-fold higher risk of developing primary liver cancer. Obesity and excessive visceral adipose tissue has been associated with a chronic inflammatory state due to increased levels of leptin. Leptin, a profibrotic and proangiogenic cytokine, activates the Janus kinase (JAK) pathway, thereby initiating an intracellular signaling cascade of pro-inflammatory cytokines\[[@B61],[@B62]\]. Obesity has also been associated reduced level of adiponectin, an anti-inflammatory cytokine. Additionally, obesity has been associated with other risk factors including insulin resistance, increased hepatic lipid storage and alteration of intestinal microflora. Insulin resistance ------------------ Diabetes has shown to be an independent risk factor for the development of HCC in NASH\[[@B61],[@B63]\]. Excessive fat accumulation and obesity lead to hepatic and peripheral insulin resistance causing compensatory hyperinsulinemia. Evidence supports that insulin and insulin-like growth factor (IGF) may promote the development of primary liver cancer by activating various oncogenic pathways\[[@B61]\]. Both IGF-1 and insulin receptor substrate stimulates growth by activating the mitogen-activate protein kinase (MAPK) pathway and increases the transcription of c-fos and c-jun, known proto-oncogenes. Activation of MAPK pathway subsequently activates the Wnt/β-catenin signaling cascade leading to fibrosis and hepatocarcinogenesis\[[@B61],[@B62]\]. Lipotoxicity ------------ Increased lipid accumulation in the liver arises from lipolysis within peripheral adipose tissue, dietary sources and de novo hepatic lipogenesis\[[@B19],[@B64]\]. This increased lipid accumulation causes hepatic lipotoxicity resulting in the excessive production of saturated and monounsaturated free fatty acids (FFAs)\[[@B65]\]. These FFAs undergo β-oxidation leading to formation of reactive oxygen species. Reactive oxygen species induce endothelial reticulum stress, mitochondrial damage and gene transcription promoting inflammatory cell signaling pathways. Intestinal microflora dysregulation ----------------------------------- Other novel pathogenic pathway between the gut and liver has been demonstrated, which is driven by dietary changes leading to gut dysbiosis that has the potential to generate hepatic inflammation can ultimately influence HCC. In NASH patients, small intestinal bacterial overgrowth\[[@B66],[@B67]\] and increased TNF-α levels, elevated expression of Toll-like receptor (TLR) 4 and increased levels of serum IL-8\[[@B67]\] has been demonstrated. LPS, a major component of outer membrane of gram-negative bacteria, is an endotoxin that causes inflammation upon entering the systemic circulation. The involvement of LPS in the development of HCC is suspected by the observation that LPS removal by gut sterilization results in diminished tumor growth in patients with chronic liver injury\[[@B68],[@B69]\]. In two recent studies, the investigators observed in NASH patients, increased levels of TNF-alpha, interleukin-8 and elevated expression of TLR 4 and small intestinal bacterial overgrowth\[[@B66],[@B67]\]. NASH patients also have less gut gram-negative Bacteroidetes and an increase in alcohol producing bacteria when compared to patients with simple steatosis, which raises a question as to whether these strains are involved in the pathogenesis of NASH\[[@B70],[@B71]\]. Several recent studies have identified potential link between gut dysbiosis and NAFLD in both in animal models and human\[[@B66]-[@B72]\]. There is incremental evidence for gut microbiome in the pathogenesis of NASH based on these findings, suggesting potential therapeutic role of correcting of gut dysbiosis to a more healthy phenotype in limiting progression of NASH. Evidence linking gut microbiota, NASH, and HCC development is reported from Dapito et al\[[@B69]\]. They treated mice with diethylnitrosamine (DEN) followed by carbon tetrachloride (CCL4) to promote fibrosis-driven HCC\[[@B69]\]. They found that TLR4-deficient mice had limited HCC growth; DEN/CCL4-treated wild-type mice that received antibiotics also had reduced tumor growth, suggesting that the microbiota played a role in HCC progression possibly via LPS-TLR4 axis. Gut microbiota can catalyse generation of secondary bile acids such a sDCA, which is known to induce DNA damage\[[@B72]\]. Yoshimoto et al\[[@B68]\] found that DCA can promote the activation of a senescence-associated secretory phenotype in HSCs, reflected by the secretion of IL-1β. Further they observed limited obesity-induced HCC development in the absence of IL-1β, and alleviation of HCC development with antibiotic treatment. In addition, lowering of DCA or feeding of DCA, limited or enhanced HCC growth respectively. Although the role for bile acids in NASH HCC progression need further exploration, these studies certainly lay the foundation for future exploratory studies in both animal models and human. Genetic polymorphisms --------------------- Genetic polymorphism is also one of the factors that may account for development of HCC in NAFLD. Genetic predisposition plays an important role in susceptibility to the metabolic syndrome and NASH. Recent genome-wide association studies have identified a single nucleotide polymorphism in the patatin-like phospholipase domain-containing 3 (PNPLA3) gene. Specifically, a C-to-G genotype in the rs738409 gene, encoding the I148M protein variant, determines differences in hepatic fat accumulation\[[@B73]\]. Although the physiological and biological functions of PNPLA3 within the liver, which effect fat accumulation and NASH, remain unclear, the association of rs738409 polymorphisms with HCC is evident\[[@B74]\]. It has also been suggested a polymorphism in the transmembrane 6 superfamily member 2 gene (TM6SF2) may increase the risk of NASH progression to HCC\[[@B75]\]. TM6SF2 mutation encodes for a loss of function substitution of lysine to glutamic acid. This TM6SF2 variant was associated with liver injury in NAFLD and NASH patients. While there is an increased prevalence of the TM6SF2 variant in NAFLD and NASH patients, conflicting preliminary data exists regarding its role in the progression to HCC. Other risk factors ------------------ Increased intrahepatic iron accumulation has been associated with NASH progression to HCC. Although clinical data is limited, Sorrentino et al\[[@B76]\] demonstrated higher hepatic iron storage levels among NASH-related HCC patients compared to NASH patients. The underlying mechanism of increased iron absorption in NASH patients may be related to oxidative DNA damage but further studies are required to understand the role of iron accumulation in NAFLD and HCC\[[@B19],[@B76]\]. Other significant risk factors for NASH progression to HCC include advanced age and concomitant chronic alcohol consumption\[[@B77]\]. Alcohol consumption among NASH patients has an associated 3.6-fold increased risk for development of HCC. Also emerging evidence has suggested a possible correlation between obstructive sleep apnea and NAFLD and NASH but its association to development of HCC has not been investigated\[[@B78]\]. SURVEILLANCE ============ With the increasing prevalence of NAFLD/NASH and associated HCC, chemopreventive and perhaps reconsideration of current surveillance guidelines are needed\[[@B16]\]. The current AASLD guidelines recommend screening for HCC every 6 mo in patients with cirrhosis. However, the current guidelines lack recommendations for surveillance of NASH patients without cirrhosis who are at risk for developing HCC. This is further supported by a study performed by Mittal et al\[[@B79]\] in which the data collected on about 1500 HCC patients where HCC related to NASH received less surveillance and treatment compared with HCC arising in underlying etiologies related to HCV and alcohol. The lack of longitudinal data in the non-cirrhotic NASH population makes it difficult to develop good evidence - based screening guideline. There is a need for studies addressing the screening guidelines for surveillance of HCC in NASH particularly for non-cirrhotic individuals. We suspect that earlier screening may be needed in patients with NASH who have multiple risk factors for HCC\[[@B19]\]. CURRENT THERAPEUTIC OPTIONS =========================== The biological heterogeneity of HCC makes it difficult to clarify the key mechanisms of cancer development and thus to develop and implement effective therapies\[[@B80]\]. A few chemopreventive agents have shown promise in the prevention and treatment of steatohepatitis and fibrosis; however these are small individual studies and thus there is a lack of a general consensus due to paucity of data. There is currently no effective chemoprevention to decrease the incidence of HCC. Exceptions include nucleoside analogues used to reduce viral replication in those with hepatitis B, and DAAs for HCV which have very high cure rates\[[@B81]\]. Medical therapy --------------- Regular exercise and controlled caloric intake is the mainstay of therapy for NAFLD, however the extent to which these are effective to prevent the development of HCC is unclear. Physical activity has been reported to have a preventive effect on development of HCC. A large prospective cohort study, which included over 400000 participants suggested that increased physical activity might have a role in HCC prevention that is independent of weight reduction\[[@B82]\]. Preliminary data suggests that statins, metformin and S-Adenosylmethonine are potential chemopreventive agents\[[@B16]\]. Patients with NASH have been found to be deficient in vitamin E and D; vitamin D deficiency is thought to play a role in hepatic carcinogenesis\[[@B83],[@B84]\]. Other dietary antioxidants such as vitamin C, selenium, coenzyme Q12 and certain phytochemicals have also been touted have chemopreventive potential\[[@B85]\]. NASH patients have been shown to have low levels of serum lycopene\[[@B83]\]. There is a strong inverse relationship between serum lycopene levels and the risk of GI cancers\[[@B86]\]. Metformin has an antitumor effect in HCC via suppression of mTOR pathway\[[@B87]\]. Although it may not have a role in the treatment of NASH, metformin may have a role in decreasing the incidence of HCC in NASH\[[@B16]\]. A review of two recent meta-analyses included 22650 cases of HCC in approximately 334000 patients with type 2 diabetes revealed that metformin reduced incidence of HCC by 50% whereas sulfonylurea and insulin increased incidence of HCC by 62% and 161% respectively\[[@B88]\]. The use of metformin has also been shown to increase survival of HCC patients who have cirrhosis\[[@B89]\]. Statins have shown a protective effect in individuals who are at risk for development of steatohepatiits and F2-F4 fibrosis\[[@B90]\]. The protective effect of statins in diabetics is thought to be due to anti-inflammatory properties of statins mediated through the inhibition of JAK\[[@B91]\]. A recent Swedish case control study which evaluated almost 4000 HCC patients treated with statin that were matched with 19970 controls showed that the odds ratio for HCC amongst statin users was 0.88, suggesting a modest but beneficial effect of statins in reducing the risk of HCC\[[@B92]\]. The heterogeneity of HCC makes it difficult to clarify the mechanism of cancer development and to develop effective therapeutics. However, an integrative functional genomics approach will contribute to the discovery of potential molecular features critical for HCC development. These studies will provide us with better treatment strategies that may be effective to treat all HCC patients including those with NASH. Surgical therapy ---------------- Curative treatment options including liver resection and liver transplantation in select early-stage HCC candidates. The Barcelona Clinic Liver Cancer staging system and therapeutic algorithm has been applied to HCC candidates including those with NASH-related HCC\[[@B93]\]. Non-cirrhotic NASH-related HCC patients who underwent curative surgical resection have shown to have superior survival than those with HCV and alcohol-related HCC\[[@B11]\]. Since the implementation of the Model for End-Stage Liver Disease (MELD) system for liver allocation in 2002 the number of HCC liver transplantations has dramatically increased. In 2012, they accounted for 23.2% of all liver transplantations in the United States\[[@B26]\]. HCC liver candidates are eligible to receive a MELD exception which upgrades their priority and thus, increases their likelihood of receiving liver transplant and survival. Subsequently, a higher number of HCC candidates have sought listing for liver transplant. A recent study using United Network for Organ Sharing data from 2004-2013\[[@B94]\] demonstrated that NASH-related HCC candidates have lower rates for receiving MELD exception and have longer time to transplant compared to HCV-related HCC. Despite this, NASH-related HCC was the fastest growing indication for liver transplantation from 2002-2012\[[@B26]\]. NASH-related HCC liver transplant recipients have better outcomes compared HCV-related HCC with a 5-year post-transplant survival approaching 68%\[[@B95]\]. NASH-related HCC liver transplant recipients with morbid obesity and CV risk factors tend to have poorer outcomes\[[@B96]\]. Further research is needed to evaluate NASH-related HCC post liver transplant survival risk factors and exploring why this growing cohort is less likely to receive a MELD exception. CONCLUSION ========== With the prevalence of HCV expected to decline, NASH is anticipated to account for a greater proportion of HCC incidence in the near future due to the growing epidemic of obesity and diabetes. The annual incidence rate of developing HCC in patients with NASH-related cirrhosis is not clearly understood with rates ranging from 2.6%-12.8%. Recent evidence has shown a significant proportion of patients with NAFLD and NASH progress to HCC in the absence of cirrhosis. While liver resection and transplantation represent curative therapeutic options in select NASH-related HCC candidates, they also have placed a significant burden to our healthcare resources and utilization. Currently NASH-related HCC is the fastest growing indication for liver transplant in HCC candidates. Increased efforts to implement effective screening and preventative strategies, particularly in non-cirrhotic NASH cohort, are needed to reduce the future impact imposed by NASH and NASH-related HCC. Manuscript source: Invited manuscript Specialty type: Gastroenterology and hepatology Country of origin: United States Peer-review report classification Grade A (Excellent): 0 Grade B (Very good): B, B Grade C (Good): C, C, C Grade D (Fair): 0 Grade E (Poor): E Conflict-of-interest statement: All authors have no conflicts of interest to disclose related to the research or data presented in this manuscript. There was no funding for this study. This manuscript is not being considered for publication elsewhere. Peer-review started: July 31, 2016 First decision: September 8, 2016 Article in press: March 14, 2017 P- Reviewer: Chetty R, Chiu KW, Chuang WL, Hernanda PY, Kamiyama T, Zhang X S- Editor: Kong JX L- Editor: A E- Editor: Li D [^1]: Author contributions: Cholankeril G was responsible for drafting the article, critical revision of the article and approval of the article; Patel R was responsible for drafting the article, critical revision of the article and approval of the article; Khurana S was responsible for critical revision of the article, intellectual input, and approval of the article; Satapathy SK was responsible for the concept/design, drafting the article, critical revision of the article and approval of the article. Correspondence to: Sanjaya K Satapathy, MBBS, MD, DM, FACG, FASGE, Associate Professor of Surgery Transplant Hepatologist, Division of Gastroenterology and Hepatology, Methodist University Hospital Transplant Institute, University of Tennessee Health Sciences Center, 1211 Union Avenue, Suite 340, Memphis, TN 38104, United States. <ssatapat@uthsc.edu> Telephone: +1-901-5169179 Fax: +1-901-5168993	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== The World Health Organization estimates that nearly 530,000 women worldwide are diagnosed with cervical cancer every year and that 275,000 die from the disease. Cervical cancer is renowned as the third most common cause of cancer in women globally, of which almost 70% occurs in developing countries \[[@CR1], [@CR2]\]. In Ethiopia, the age-standardized incidence and mortality rates are estimated as 26.4 and 18.4 per 100,000, respectively, four- and ninefold higher that the incidence and mortality rates in Western Europe \[[@CR1]\]. Cervical cancer has been recognized as an unusual outcome of a sexually transmitted infection, and the etiology is limited to a few human papillomavirus (HPV) genotypes. The association between HPV and cervical cancer is a universal fact, and variability among the different types is geographically limited. With optimal testing systems, HPV DNA can be identified in almost all specimens of invasive cervical cancer, and infection of the cervix with HPV is the main cause of cervical cancer \[[@CR2]\]. One of the major reasons identified for the progression and development of cervical neoplasia among women who are repeatedly infected is ineffective cell-mediated immunity \[[@CR3]\]. Of all HPV genotypes, more than 40 have been identified from anogenital mucosa samples and most are transmitted sexually. HPV genotypes 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68 are classified as the high-risk (HR) group, which predicts cervical cancer \[[@CR4]\]. The major phases in cervical oncogenesis include infection of the metaplastic epithelium of the cervical transformation zone with high-risk HPV infection, viral persistence and clonal progression of the persistently infected epithelium to cervical pre-cancer, and invasion \[[@CR5]\]. In sub-Saharan Africa, HPV-associated cervical cancer is one of the major causes of morbidity and mortality. A lack of strong initiatives as well as sustainable cervical cancer prevention programs and services have been identified as potential causes of the high incidence rate in most countries \[[@CR6]\]. In Eastern Africa, approximately 35.8% of women are estimated to harbor cervical HPV infection at any given time, and 76.5% of invasive cervical cancers are associated with HPV 16 or 18 \[[@CR7]\]. Moreover, only 0.6% of the total female population aged 18--69 years in Ethiopia is screened every 3 years, representing 1.6% urban women and 0.4% rural women, which demonstrates that screening practice is underdeveloped and that the overall coverage is very limited \[[@CR8], [@CR9]\]. This study produced substantial information with relevant data regarding the burden of HR HPV infection and cervical cytology abnormalities in the intended setting. Methods {#Sec2} ======= An institutional-based cross-sectional study design was employed at three selected obstetrics and gynecology clinics of Addis Ababa, Ethiopia, from June to October 2015. The study was conducted among women who visited the Family Guidance Association of Ethiopia Addis Ababa Area Reproductive Health Clinic, Hemen Maternal and Children Health Specialty Center, and SinamokshEthio Women's Health Special Clinic. The study population consisted of women who visited the clinics for any gynecological purposes, including cervical cancer screening, and fulfilled the inclusion criteria. A nonprobability convenience sampling technique was used to select the study sites, considering the scope and volume of services provided. As these health facilities provide cervical cancer screening services and have a significant volume of client visits, they were potential sites for this study and among the very few sites providing this service consistently in the city. All women who visited each clinic during the study period and who were eligible for this study were consecutively added until the number of clients reached the calculated minimum sample size. A total of 366 women were enrolled in the study. Sociodemographic characteristics, sexual behaviors and other risk-factor variable responses were gathered using a structured questionnaire (Additional file [1](#MOESM1){ref-type="media"}). HR HPV DNA and Pap screenings were performed following the standard operating procedure (Additional file [2](#MOESM2){ref-type="media"} and [3](#MOESM3){ref-type="media"}). The cytological examination was performed by two pathologists whose degree of expertise was Medical Doctor with Diploma in Pathology and Cytology. Agreement between HR HPV and Pap smear results was assessed by Cohen's Kappa coefficient by recoding the findings into two categories (Negative and Positive). The results were entered into EpiData software Version 13.0, and the data were analyzed using STATA Software Version 11.0. Descriptive statistics, proportions and the actual number of cases were used to describe frequency outputs for categorical variables and arithmetic means for the average age of the participants. Cross-tabulations were performed to explore and display relationships between two categorical variables. Chi-square statistics were employed to assess differences between two categorical variables. Multivariate logistic regression analysis (adjusted odds ratio) was applied to evaluate the strength of the association of the various potential risk factors with the presence of HR HPV infection and cervical cytology abnormalities. Positive and negative percentage agreement and overall percentage agreement were assessed for HR HPV DNA PCR and Pap smear screening methods. A P-value of less than 0.05 was considered statistically significant. Results {#Sec3} ======= Study subjects and sociodemographic characteristics {#Sec4} --------------------------------------------------- A total of 366 participants between 18 and 68 years of age were enrolled in this study. The mean age was 42.7 ± 10.7 SD. Most study subjects, 296/366 (80.9%), were within the range of 31--60 years. In terms of residence, 352 (96.2%) participants visited the study clinics from the Addis Ababa area. Of the total number of participants, 287 (78.4%) were married; among these, 71 (24.7%) was married for the first time before 18 years of age. Regarding parity, 29 (7.9%) of the participants had \> 5 complete pregnancies and deliveries; 281 (76.8%) women were parity 1 to 5. Participant employment status was also assessed, and 248 (67.8%) of the study participants were self-employed. Regarding educational status, the highest proportion comprised those with Diploma or Degree and above qualification (158/366; 43.4%), and only 39 (10.7%) were unable to read and write (Table [1](#Tab1){ref-type="table"}).Table 1Sociodemographic characteristics of the study participants, Addis Ababa, Ethiopia, June to October 2015VariableNumber%Age 18--305314.48 31--6029680.87 \> 60174.64Residence Addis Ababa35296.2 Outside Addis Ababa143.8Marital Status Single328.7 Married28778.4 Widowed287.7 Divorced195.2Age at first marriage \< 153910.66 15--17328.74 \> =1829580.6Parity 05615.3 1 to 528176.8 \> 5297.9Employment status Employed (Government/Private/NGO)10829.5 Self-employed24867.8 Unemployed102.7Education Unable to read and write3910.7 Elementary6417.5 High school10528.7 Diploma/Degree and above15843.2 Burden of High-risk Human papillomavirus and its genotypes {#Sec5} ------------------------------------------------------------ The overall burden of HR HPV infection in this study was 50/366 (13.7%). Among the HR HPV-positive cases, 8 (16%) were identified as having HR HPV 16 genotype, 38 (76%) had "other HR HPV" (HR HPV genotypes 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, or 68), 2 (4%) had genotype 16 together with "other HR HPV" genotypes, 1 (2%) had genotype 18 together with "other HR HPV" genotypes, and 1 (2%) had genotype 18. The HR HPV genotype distribution showed that "other HR HPV" types dominated over genotypes 16 and 18 (Fig. [1](#Fig1){ref-type="fig"}).Fig. 1HR HPV genotype proportional distribution among all HR HPV-positive participants, Addis Ababa, Ethiopia, June to October 2015 Despite the low proportion, multiple infections were identified for HR HPV 16 and "other HR HPV" as well as for HR HPV 18 and "other HR HPV", with proportions of 4 and 2%, respectively. The proportion of HR HPV-positive cases was 26, 72, and 2% in age ranges 18--30, 31--60, and \> 60, respectively. Age range 31--60 was found to have the highest proportion of positivity (76, 95% CI (71.3--80.1%)), which was statistically significant. The association between HR HPV infection with sociodemographic and reproductive health, sexual behavior, and other risk factors was analyzed through bivariate analysis using the chi-square test. Age (P = 0.000), parity (P = 0.017), age at first marriage (P = 0.027), education (P = 0.003), condom use during sexual intercourse (0.011), cigarette smoking (0.000), and family history of cervical cancer (0.003) were significantly associated with HR HPV infection. Ever use of any type of contraceptive, age at first sexual intercourse, more than one lifetime sexual partnership, history of STD, alcohol consumption, and HIV serostatus, with P-values of 0.106, 0.266, 0.334, 0.824, 0.227, and 0.688, were not significantly associated. In multivariate analyses using logistic regression, only "other HR HPV" type was significantly associated with residence, employment status, and HIV serostatus, with P-values of 0.037, 0.01, and 0.041, respectively (Tables [2](#Tab2){ref-type="table"} and [3](#Tab3){ref-type="table"}). Individuals who visited the clinics from outside Addis Ababa were 8.12 times more likely to have "other HR HPV" type infection than those who were from Addis Ababa. Furthermore, the likelihood of having "other HR HPV" infection among unemployed individuals was 9.2 times higher than for employed individuals. Compared to diploma or degree holders, women who were not able to read and write were less likely to be infected with "other HR HPV" types (Table [2](#Tab2){ref-type="table"}).Table 2Association of "other HR HPV" genotypes with sociodemographic factors, Addis Ababa, Ethiopia, July to October 2015SociodemographicResponse Category"other HR HPV" positivity\COR(95% CI)P-value\\AOR(95% CI)P-valueAge18--3010 (26.32)Ref^a^31--6027 (71.05)0.43 (0.19,0.96))0.0380.6 (0.01,1.77.9)0.36\> 601 (2.63)0.27 (0.03,2.27)0.2280.33 (0.02,4.9)0.42ResidenceAddis Ababa35 (70.0)Ref^a^Out of Addis Ababa3 (6.0)2.35 (0.62,8.81)0.2078.12 (1.14, 57.9)0.037^b^Marital statusMarried28 (56.0)Ref^a^Unmarried7 (14.0)2.795 (1.096,7.128)0.0313.22 (0.73, 14.29)0.123Widowed3 (6.0)1.05 (0.299,3.716)0.9341.80 (0.40,8.24)0.445Divorced00Age at first marriage\< 152 (4.0)Ref^a^15--174 (8.0)0.25 (0.025,2.59)0.2482.15 (0.17,0.48)0.55\> = 1832 (64.0)0.14 (0.029,0.646)0.0124.98 (0.47,52.2)0.18Parity07 (14.0)Ref^a^1 to 528 (56.0)0.72 (0.3,1.8)0.4714.15 (0.90,19.2))0.067\> 53 (6.0)0.79 (0.2,3.3)0.7436.78 (0.66,72.4)0.101Employment statusEmployed (Government/Private/NGO)12 (24.0)Ref^a^Self employed22 (44.0)0.79 (0.37,1.66)0.5311.22 (0.548, 3.07)0.67Unemployed4 (8.0)5.1 (1.26,20.73)0.0239.17 (1.6, 52.22)0.01^b^EducationUnable to read and write1 (2.0)Ref^a^Elementary6 (12.0)3.75 (0.43,32.48)0.234.86 (0.4,59.4)0.215High school11 (22.0)4.33 (0.54,34.77)0.16812.39 (01.01151.12.5))0.049Diploma/Degree and above20 (40.0)5.34 (0.69,41.21)0.10814.06 (1.12,176.6)0.041^b\^COR-Crude Odds Ratio, ^\\^AOR-Adjusted Odds Ratio, ^a^Reference, ^b^There is a statistically significant associationTable 3Association of "other HR HPV" genotypes with sexual behavior and other risk factor variables, Addis Ababa, Ethiopia, June to October 2015Sexual behavior and other risk factor variablesResponse Category"other HR HPV" positivity\COR(95% CI)P-value\\AOR(95% CI)P-valueEver use of contraceptiveYes24 (48.0)Ref^a^No14 (28.0)0.01 (0.3,1.2)0.1640.62 (0.28,1.40)0.251Age at first sexual intercourse\< 153 (6.0)Ref^a^15--1710 (20.0)1.39 (0.35,5.53)0.6370.30 (0.03,2.62)0.275\> = 1825 (50.0)0.82 (0.23,2.9)0.7530.13 (0.01,1.04)0.054More than one lifetime partnershipYes25 (50.0)Ref^a^No13 (26.0)0.65 (0.32,1.31)0.2290.79 (0.34,1.86)0.591Condom use during sexual intercourseYes14 (28.0)Ref^a^No24 (48.0)0.53 (0.3,0.9)0.0760.66 (0.27,1.66)0.38History of STDYes6 (12.0)Ref^a^No32 (64.0)1.24 (0.5,3.1)0.6412.62 (0.7,9.8)0.151Cigarette smokingYes2 (4.0)Ref^a^No36 (72.0)0.53 (0.1,2.6)0.4350.86 (0.13,5.6)0.87Family history of cervical cancerYes4 (8.0)Ref^a^No34 (68.0)0.01 (0.2,1.9)0.4021.24 (0.3,5.)0.76Alcohol consumptionUsually3 (6.0)Ref^a^Occasionally16 (32.0)1.32 (0.4,4.9)0.681.33 (0.27,6.2)0.72Never19 (38.0)0.75 (0.2,2.7)0.670.8 (0.16,4.0)0.79HIV serostatusNegative25 (50.0)Ref^a^Positive3 (6.0)2.67 (0.7,10.3)0.1565.73 (1.06,30.9)0.042^b\^COR-Crude Odds Ratio, ^\\^AOR-Adjusted Odds Ratio, ^a^Reference, ^b^There is a statistically significant association Abnormal cervical cytology {#Sec6} -------------------------- Overall, Pap smear abnormalities were observed in 13.1% (48/366) of the study subjects. Among the abnormalities, 3 (6.3%), 39 (81.3%), and 6 (12.5%) were ASCUS, LSIL, and HSIL, respectively (Fig. [2](#Fig2){ref-type="fig"}). Among the abnormal cytology categories, LSIL abnormality showed the highest frequency. Low-grade squamous intraepithelial lesion (LSIL) and high-grade intraepithelial lesion (HSIL) rates in the 31--60 age category were 33 (84.62%) and 6 (66.67%), respectively, higher compared to the other age categories (Table [4](#Tab4){ref-type="table"}). The association between age category and abnormal cytology was assessed by Fisher's exact test and found not to be statistically significant (P-value = 0.180).Fig. 2Proportion of Pap smear cytology results among the study population, Addis Ababa, Ethiopia, June to October 2015Table 4Frequency of abnormal Pap smear cytology by age category, Addis Ababa, Ethiopia, June to October 2015Abnormal Pap smear cytologyAge Categories (year)P-valueTotal18--3031--60\> 60No (%)No (%)No (%)ASCUS0 (0.00)3 (7.5)0 (0.00)3 (6.25)LSIL4 (10.26)33 (84.62)2 (5.13)0.18039 (81.3)HSIL0 (0.00)4 (66.67)2 (33.33)6 (12.5)Total4 (8.33)40 (83.33)4 (8.33)48 (100.0)Table 5Association of HSIL abnormal cytology with "age, age at first marriage and educational status", Chi-square analysis, Addis Ababa, Ethiopia, June to October 2015VariableFrequency\ No (%)Chi-squareP-ValueAge18-300 (0)11.840.00331-604 (66.67)\>602 (33.33)Age at first marriage\<151 (16.67)10.990.00415-173 (50)\>=182 (33.33)Educational statusIlliterate3 (50)10.660.014Elementary0 (0)High School1 (16.67)Diploma/Degree and above2 (33.33) In addition, a significant association with any of the risk factor variables was not observed for LSIL abnormalities according to the Chi-square test. In contrast, HSIL abnormal cytology was significantly associated with age, age at first marriage and educational status, with P-values of 0.003, 0.004 and 0.014, respectively (Table [5](#Tab5){ref-type="table"}). Ever use of any type of contraceptive, oral contraceptive use, age at first sexual intercourse, more than one lifetime sexual partnership, frequency of condom use, frequency of cigarette smoking, history of STD, and alcohol consumption were not significantly associated with abnormal Pap cytology (P-value \> 0.05). HR HPV genotypes were compared with cytological abnormalities, and the results are summarized in Table [6](#Tab6){ref-type="table"}. The overall HR HPV genotype frequency among the total normal cytology results was 40/318 (12.6%). As shown in Tables [6](#Tab6){ref-type="table"}, 40 (80%) of HR HPV-positive individuals had normal cytology (NILM); 5 (10%), 4 (8%), and 1 (2%) had LSIL, HSIL, and ASCUS, respectively. Among the total number of HR HPV-positive individuals, HR HPV 16 was found in two of the cases of HSIL abnormality, and the remaining two were caused by "other HR HPV" genotypes. HR HPV 18 was only found in NILM, but HR HPV 16 was identified both in NILM and HSIL. Nonetheless, "other HR HPV" genotypes were found across all stages (Table [6](#Tab6){ref-type="table"}).Table 6HR HPV genotypes compared to Pap smear cytology findings, Addis Ababa, Ethiopia, June to October 2015HR HPV GenotypesNILMASCUSLSILHSILHR HPV 166 (12%)0 (0%)0 (0%)2 (4%)HR HPV 181 (2%)0 (0%)0 (0%)0 (0%)Other HR HPV30 (60%)1 (2%)5 (10%)2 (4%)HR HPV 16 and other HR HPV2 (4%)0 (0%)0 (0%)0 (0%)HR HPV 18 and other HR HPV1 (2%)0 (0%)0 (0%)0 (0%)Total40 (100%)1 (100%)5 (100%)4 (100%) Percent agreement between HR HPV DNA PCR and conventional pap smear cytology {#Sec7} ---------------------------------------------------------------------------- Agreement between HR HPV DNA PCR and conventional Pap smear cervical cancer screening methods was analyzed using positive, negative, and overall percent agreement and the Kappa statistic. The positive and negative percent agreement was found to be 87.7 and 22.4%, respectively. However, the overall percent agreement was 79.0%, and the Kappa value was 0.12 (95% CI (0.00--0.24), P-value =0.01). The overall percent agreement findings reveal significant agreement between HR HPV DNA PCR and conventional Pap smear cytology screening methods (P \< 0.05). Discussion {#Sec8} ========== This study mainly aimed to assess the burden of HR HPV and cervical cytology abnormalities, along with potentially associated sociodemographic, sexual behavior, and reproductive health variables, in three Obstetrics and Gynecology and reproductive health clinics in Addis Ababa, Ethiopia. In this study, the overall HR HPV burden was 13.7%. "Other HR HPV" genotypes (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, or 68 types) were the most frequent (76%) genotypes identified in this study, followed by HR HPV 16 (16%). The overall prevalence of abnormal cytology was also 13.1%. Approximately three-fourth (72%) of the HR HPV-infected women were in the age range of 31 to 60 years, and this was significantly associated with abnormal cytology. HR HPV was found in 12.6% of normal cytology reports. Moreover, residence, occupation, and HIV serostatus were significantly associated with HR HPV infection. In this study, the overall HR HPV burden was 13.7%, a finding that was consistent with previous studies reported from different parts of Ethiopia \[[@CR14], [@CR15]\], at 13.2 and 15.8%, respectively. In contrast, our finding was much lower than those in two other studies from Ethiopia \[[@CR16], [@CR17]\]. This difference might be because the participants in the first study \[[@CR16]\] were women with cervical complaints and all samples were cases of cervical dysplasia, which may result in higher values. In our study, however, cytological samples were obtained from women who did not necessarily have cervical dysplasia or cervical complaints. Similarly, the difference from the other report \[[@CR17]\] may be due to the study site chosen, as that study was conducted in the only specialized cancer center in Ethiopia, which would increase the probability of observing a large number of positive HR HPV cases. In addition, approximately 34% of those study participants were HIV positive, which might also have contributed to the higher rate of HR HPV \[[@CR12]\]. Our finding was lower than the estimated prevalence reported for all HPV genotypes (high-risk and low-risk types) from sub-Saharan African countries (21.8%) \[[@CR11]\] and Nigeria (21.6%) \[[@CR12]\]. To date, studies conducted in Ethiopia \[[@CR14], [@CR16]\] have reported that HR HPV 16 is the predominant type. In contrast, the most frequent genotypes identified in the present study were "other HR HPV" genotypes (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, or 68 types), contributing 76%, followed by HR HPV 16 (16%). Our finding is comparable with that in a worldwide meta-analysis review \[[@CR10]\], which reported that the predominant genotype in Eastern Africa was HR HPV 52, followed by HPV 16. Another study \[[@CR18]\] found that HPV-positive women in sub-Saharan Africa were less likely to be infected by HPV 16 than were women in Europe. Similarly, another study \[[@CR19]\] examining paraffin-embedded cervical tissues reported that HPV 52 (25.5%) and 58 (22%) were the most frequent genotypes. This difference in genotype frequency in various studies might be due to geographic variation and host immunogenetic factors. Regardless, HPV 16 appears to be less influenced by immune status than other HPV genotypes. This fact, coupled with impairment in cellular immunity, may contribute to the presence of HPV genotypes other than HPV 16 in some populations \[[@CR19]\]. Multiple HR HPV type infections were found in 7.9% of HR HPV-positive individuals in a study by Mohammed et al. in Northeastern Nigeria \[[@CR20]\], which is comparable to the findings of the present study (6%). In contrast, the 17.5% of multiple infections in a study conducted on Ethiopian and Sudanese women \[[@CR19]\] was relatively higher than that in the present study. This may be due to the nature of the samples processed in that study \[[@CR19]\], which included tissue blocks with cervical intraepithelial neoplasia or carcinoma, and the possibility of infection by more than one type of HR HPV may increase in such cases \[[@CR16]\]. An age-specific HPV infection study in South Africa \[[@CR13]\] reported that the highest frequency (74.6%) of infections was found in women older than 25 years. Similarly, another study from Addis Ababa, Ethiopia, reported that 50.6% of HR HPV-infected women were in the age range of 30--50 years \[[@CR17]\]. These studies are consistent with our finding that 72% of the HR HPV-infected women were 31--60 years of age. However, the significant association between age group in the bivariate analysis (P \< 0.05) was not significant in the multivariate analysis. This is similar to the results of the study conducted in Gurage Zone, Ethiopia \[[@CR14]\]. In contrast, a study by Andall B in Trinidad (33) showed that the highest (63%) prevalence of HPV infection was observed among women aged \< 30 years (P \< 0.0001), with a peak in the age range of 21 to 25 years. This might be due to the detection of low-risk HPV in addition to HR HPV. In this study, residence, occupation, and HIV serostatus were significantly associated with HR HPV infection in multivariate analysis. This finding was comparable with a study \[[@CR24]\] reporting that occupation and residence are significantly associated with HPV infection. Nonetheless, the study by Muluken et al. in Tikur Anbessa Specialized Hospital, Addis Ababa, Ethiopia \[[@CR17]\], reported that HIV and residence were not significantly associated with HR HPV prevalence. This might be due to differences in sampling, type of participants, and data collection methods. In our study, ever use of any type of contraceptive, age at first sexual intercourse, and more than one lifetime sexual partnership were not associated with HR HPV infection. This outcome is comparable to the findings of Mega AC et al. in rural Nigeria \[[@CR21]\]. The overall abnormal cytology burden in the present study was 13.1%, which was lower than that in a similar study from Ethiopia \[[@CR17], [@CR22]\] and another from South Africa \[[@CR13]\]. This difference might be due to the presence of a large number of HIV-infected individuals, who are not easily able to resolve infection and experience progression to the development of precancerous to cancerous lesions \[[@CR23]\]. In our bivariate analysis, age at first marriage and educational level were significantly associated with HSIL Pap smear abnormality (p-value 0.004 and 0.014), consistent with a study reported by Abel et al. \[[@CR22]\]. Furthermore, our study presents high-risk HPV genotypes with cervical cytology findings. HR HPV 16 was found in 50% of HSIL reports, and "other HR HPV genotypes" were the most frequent finding for LSIL. Similarly, for women who had normal cervical cytology results, the most frequent genotypes were "other HR HPV" genotypes. According to the meta-analysis by Gary C. et al. \[[@CR18]\], the most common HR HPV type in HSIL among women with and without cervical neoplastic diseases was HR HPV 16, which was consistent with our findings. In contrast to the same study \[[@CR18]\], which reported HR HPV 16 as the predominant genotype in LSIL and NILM, all the LSIL and NILM results in our study were attributed to "other HR HPV" genotypes. Moreover, "other HR HPV" genotypes were observed across all grade levels of cytological findings. As reported in various studies, HPV-positive women in sub-Saharan Africa are less likely to be infected with HR HPV 16 than are their counterparts in Europe (\[[@CR18]--[@CR20]\], and). Interestingly, the present study also revealed that 12.6% of women with NILM were positive or any type of HR-HPV infection. This is comparable to a study \[[@CR25]\] from West Africa reporting that 13% of women with normal cytology results were positive for HR HPV. In such situations, the women may continue to have an increased risk of HSIL during the interval between the first and next screening \[[@CR26]\]. Conclusions {#Sec9} =========== The burdens of HR HPV infection and cervical cytology abnormalities presented in this study are consistent with the few previous local studies and reviews in Ethiopia but somehow lower than the estimated prevalence for sub-Saharan Africa. Unlike previous studies, "other high-risk HPV" genotypes contributed considerably to the overall HR HPV burden. Multiple-type infections were found in sexually active women. The highest frequency of HR HPV positivity was in women without cervical cytology abnormalities. Hence, the interval between the primary and secondary HPV screening for HR HPV positives and negatives needs to be defined separately. The performance of the Abbott Real-Time HR HPV DNA PCR and Pap smear cytology screening methods may need to be further evaluated against histologically confirmed results. In addition, the screening program for early-age sexually active women should be further promoted in various health settings. The Ministry of Health should also further consider the possibility of introducing vaccines targeting other oncogenic HPV types in addition to genotypes 16 and 18. A large-scale community-based cohort study shall also be designed and implemented to determine the national burden and the molecular epidemiology of persistent HR HPV types and cervical cytology abnormalities which will help to recommend the ideal screening algorithm considering the local context. This will significantly contribute to the national preventive public health strategies against cervical cancer. Additional files ================ {#Sec10} Additional file 1:Questionnaire. (DOCX 20 kb) Additional file 2:HR HPV detection procedure using the Abbott Real-Time PCR method. (DOCX 17 kb) Additional file 3:Abnormal cytology diagnosis procedure using the conventional Pap smear method. (DOCX 17 kb) ASCUS : Atypical squamous cells of undetermined significance HIV : Human immunodeficiency virus HR HPV : High-risk human papillomavirus HSIL : High-grade squamous intraepithelial lesion LSIL : Low-grade squamous intraepithelial lesion NILM : Negative for intra-epithelial lesions and malignancy PCR : Polymerase chain reaction STD : Sexually transmitted disease Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. We are indebted to Abbott Molecular Diagnostics, especially Mr. Daan J. Potgieter, for the donation of the HR HPV reagents and consumables, and International Clinical Laboratories (ICL), in particular, Mr. Tamrat Bekele, for the support of all Pap smear collection and examination kits and for authorizing the laboratory workbench and diagnostic platform. I would like to extend acknowledgement to Mr. Wudneh Yitayew (ICL) and Eyuel Lema for the entire support and facilitation of the shipment of reagents and supplies. Mr. Melaku Tesfaye, Mr. Tadele Getachew, Mrs. Azeb Adamu, Mr. Giorgis Okabegzi, Mr. Yonas Abay, and Ms. Ruth Getachew (ICL) are also thanked for their generous technical support for the entire laboratory work and technical support. American Journal Expert (AJE) is acknowledged for the language editorial service. To all study participants, we thank them for their consent and cooperative participation. All selected health facility management and data collectors and coordinators (Dr. Selamawit Ashagre, Dr. Adanech Belay, Sr. Hamelmal Kelemeork from Hemen MCH, and Sr. Husniya Shash, Sr. Shibre Beshah, and Sr. Etetu Kassa from FGA) are also acknowledged for their willingness and commitment throughout this project. Finally, we thank the Department of Medical Laboratory Science, College of Health Sciences, Addis Ababa University, for administrative facilitation of the work from beginning to end. Endnotes {#FPar1} ======== ^a^References, ^b^There is a statistically significant association Conceived and designed the experiments: KE, IA, KD, MN, DS, ZL, and TH. Performed the experiments and investigations: KE, MN, and DS. Analyzed and reviewed the data: KE, THK, and KD. Wrote the paper: KE, KD, and RT. Reviewed and edited the manuscript draft: IA, MN, DS, ZL, TH, THK, EL, and RT. All authors read and approved the final manuscript. Not applicable The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. The study proposal was reviewed and approved by the Departmental Research and Ethics Review Committee (DRERC) of the Medical Laboratory Sciences, School of Allied Health Sciences, College of Health Sciences; Addis Ababa University (Letter Ref Number: MLS/388/15) on 08/04/2015. Formal individual written consent was collected from each participant. Informed consent was obtained by interpretation in the participant's local dialect. The privacy and confidentiality of each individual participant was ensured. An appropriate coding system was used rather than any personal identifiers. All client results were reported to their clinicians in the standard result format with a specified turnaround time and utilized for clinician judgment in addition to being used for these research purposes. Not applicable The authors declare that they have no competing interests.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Esophageal cancer (EC) has a high incidence worldwide ([@b1-mco-0-0-1909]) and carries a poor prognosis. EC develops mainly in individuals aged \>50 years, and the number of older patients with EC in Japan is increasing concomitantly with the aging of the population. Older patients frequently have comorbidities, cognitive decline, polypharmacy and social issues ([@b2-mco-0-0-1909]). The currently available treatment modalities for EC include endoscopic submucosal dissection (ESD), esophagectomy, radiotherapy, chemotherapy, chemoradiotherapy (CRT) and best supportive care (BSC) ([@b3-mco-0-0-1909]). ESD was developed for superficial EC restricted to the mucosal layer (T1a) ([@b3-mco-0-0-1909]). Despite the increasing number of older patients with EC, the majority of clinical trials have involved only, or mostly, younger patients ([@b4-mco-0-0-1909]). Although some studies have focused on older patients ([@b5-mco-0-0-1909]), these involved a relatively limited number of subjects and a single arm ([@b6-mco-0-0-1909],[@b7-mco-0-0-1909]). In addition, older patients have high rates of morbidity and mortality ([@b8-mco-0-0-1909],[@b9-mco-0-0-1909]). Therefore, an optimal treatment modality for older patients with EC has yet to be established. Older patients have been defined as those aged \>70, 75 or 80 years, depending on the study in question ([@b6-mco-0-0-1909]--[@b11-mco-0-0-1909]). A multicenter questionnaire survey indicated that most institutions consider EC patients aged \>75 or 80 years as older patients ([@b12-mco-0-0-1909]). In the present study, older patients were defined as those who were at least 75 years old. Patients and methods ==================== ### Patients Data on EC patients diagnosed at the Niigata University Medical and Dental Hospital (Niigata, Japan) between January 2007 and December 2017 were retrospectively collected using a hospital-based cancer registry and electronic medical records. The study protocol was approved by the Ethics Review Committee of the School of Medicine of Niigata University (Niigata, Japan; approval no. 2485). All procedures performed in studies involving human participants were in accordance with the ethical standards of the Institutional and National Research Committee and the 1964 Declaration of Helsinki and its later amendments, or comparable ethical standards. Patient informed consent was not applicable, as this study was retrospective and data were collected only from medical records. All patients were informed of the opportunity to opt out of this study through the internet home page of Niigata University School of Medicine. The collected data included age at diagnosis, sex, tumor histology, cancer stage (TNM), treatment modality and prognosis. The Union for International Cancer Control guidelines, version 7 ([@b13-mco-0-0-1909]), were used for TNM staging. The patient population was divided into younger and older patients, with 75 years at the time of EC diagnosis as the cut-off. ### Statistical analysis The Chi-squared or Fisher\'s exact tests were used to evaluate the significance of the differences between the two groups. Disease-specific survival duration was defined as the period from the date of diagnosis to that of death due to EC. Patients who died from causes other than EC were censored at the date of death. Surviving patients were censored on the date of their last visit to the hospital. Survival curves were drawn using the Kaplan-Meier method and were compared by log-rank tests. P\<0.05 was considered to indicate a statistically significant difference. Statistical analyses were performed using IBM version 22.0 (IBM Corp.). Results ======= ### Patient characteristics A total of 990 patients with EC who visited Niigata University Medical and Dental Hospital between January 2007 and December 2017 were identified. The patient characteristics are summarized in [Table I](#tI-mco-0-0-1909){ref-type="table"}. Of these patients, 359 (36.3%) were aged 75 years or older, and 631 patients (63.7%) were aged \<75 years. The median age was 71 years in the entire population, 66 years in the younger group and 79 years in the older group. The majority of the EC patients in both groups had squamous cell carcinoma (90.1%) and early-stage EC (stage 0 or I). The frequency of adenocarcinoma was significantly higher (P\<0.01), and the rates of different stages were significantly different (P\<0.01) in the older compared with the younger group. ### Treatment modality Treatment modalities were classified according to the main treatment, as precise classification of treatment modalities would be overly complicated. The majority of the patients underwent ESD (n=629, 63.5%) followed by surgery (n=144, 14.5%) and CRT (n=108, 10.9%) ([Table II](#tII-mco-0-0-1909){ref-type="table"}). Older patients were more likely to undergo radiotherapy alone (n=24, 6.7%) or BSC (n=48, 22%) compared with younger patients (P\<0.01). The complete multidisciplinary treatment modalities in stage I and II/III cases are presented in [Table III](#tIII-mco-0-0-1909){ref-type="table"}. Among younger patients with stage I EC treated by ESD (n=207), 38 (18.4%) also received chemotherapy/radiotherapy, as advanced-stage EC was detected by histological analysis of ESD specimens. By contrast, only 8 (6.5%) of the older patients with stage I EC (n=124) underwent chemotherapy/radiotherapy after ESD. A total of 38 (14.1%) younger patients and 14 (8.2%) older patients underwent surgery with/without additional treatment for stage I disease. Of the younger (n=65) and older (n=21) patients with stage II/III EC who underwent surgery, 46 (71.0%) and 10 (47.6%), respectively, received chemotherapy or/and radiotherapy before or after surgery. Among 10 older patients treated with chemotherapy after surgery, 2 succumbed to treatment-related adverse effects of severe bone marrow suppression. Although patients with stage IV EC in both groups received various treatment modalities, 40% of the older patients (n=25) underwent BSC (P\<0.01). Patients with EC of unknown stage were only found in the older group, and 24 (65%) of those opted for BSC. ### Disease-specific survival Disease-specific survival rather than overall survival was analyzed, due to the shorter life expectancy of the older patients. The median follow-up time was 39.6 months (range, 1--120 months). The 5-year disease-specific survival rate was 59.1% in the older group and 75.2% in the younger group at all stages ([Fig. 1](#f1-mco-0-0-1909){ref-type="fig"}). The disease-specific survival duration of the older patients was significantly shorter compared with that of the younger patients (P\<0.001); moreover, the disease-specific survival rate of the older patients decreased markedly after 3.5 years. The disease-specific survival duration of the older patients with stage 0, I or II/III EC was significantly shorter compared with that of the younger patients with stage 0, I or II/III EC ([Fig. 1](#f1-mco-0-0-1909){ref-type="fig"}). The disease-specific survival rate of the older patients with stage I EC decreased considerably after 3.5 years, similar to the older group as a whole. The survival curves differed markedly between older and younger patients with stage I and II/III EC compared with those with stage 0 EC. No difference in survival was observed between older and younger patients with stage IV disease. In older patients with stage II/III EC, the disease-specific survival rate did not differ significantly between patients who underwent surgery and those who received CRT; by contrast, in the younger group, patients who received surgery exhibited a significantly higher survival rate and duration compared with those who were treated with CRT ([Fig. 2](#f2-mco-0-0-1909){ref-type="fig"}). Older patients who underwent surgery had a poorer prognosis compared with younger patients. A total of 3 older patients died from adverse events due to chemotherapy for stage II/IIII (2 cases) and IV (1 case) disease; no treatment-related deaths were reported among younger patients. Discussion ========== In the present study, the clinicopathological characteristics and treatment modalities and outcomes of 990 patients diagnosed with EC in our institution were reviewed. Compared with younger patients with stage I EC, older patients with stage I EC less frequently received additional treatment following ESD (6.5% of the older vs. 18.4% of the younger patients). Compared with younger patients with stage II/III EC, older patients with stage II/II EC less frequently received perioperative chemotherapy/radiotherapy (47.6 vs. 71.0%, respectively) and definitive CRT (21.7 vs. 34.2%, respectively). Among the older patients, 13% selected BCS compared with 2% of the younger patients. Older patients had a significantly shorter disease-specific survival duration compared with younger patients, specifically for stage I and II/III disease. Two large studies involving EC patients in the United States and Taiwan ([@b14-mco-0-0-1909],[@b15-mco-0-0-1909]) reported 5-year survival rates of \<20%. In addition, older patients (≥70 years of age) were less likely to undergo surgery or/and radiotherapy, and had a lower survival rate. In the US study, 33.1 and 24.0% of the older and younger patients, respectively, opted for BSC ([@b14-mco-0-0-1909]). These survival rates are lower compared with those reported in the present study, likely because the majority of our patients had early-stage EC, were treated with ESD, and only 13% of the older patients selected BSC. ESD can completely remove superficial EC and EC confined to the lamina propria mucosae, and is only indicated for Tis and T1a ([@b3-mco-0-0-1909],[@b16-mco-0-0-1909]). Patients diagnosed with T1a (m3) or T1b, i.e., tumor invasion of the muscularis mucosae or submucosal layer, respectively, require adjuvant therapy after ESD. In a retrospective analysis in an adjuvant treatment setting, both the 3-year relapse-free survival and overall survival rates of patients with T1a (m3) or T1b EC after ESD were significantly improved by adjuvant treatments ([@b17-mco-0-0-1909]). ESD followed by CRT for stage I \[m3(T1a) + T1b\] is reportedly effective and safe, and improves the prognosis compared with definitive CRT ([@b18-mco-0-0-1909]). In the present study, 6.5 and 18.4% of the older and younger patients, respectively, with stage I EC received adjuvant therapy. This low rate of adjuvant treatment may partially explain the marked decrease in the survival rate after 3.5 years among older patients ([Fig. 1](#f1-mco-0-0-1909){ref-type="fig"}). Of the 8 older patients who received adjuvant therapy after ESD, 2 (25%) experienced grade 3 or 4 adverse effects of neutropenia, but recovered. Therefore, adjuvant treatments after ESD should be considered for older patients with T1a (m3) and T1b EC. The low rate of surgical treatment (8.2 vs. 14.1%) may also explain the worse prognosis of the older patients. The prognosis of the older patients with stage II/III EC did not differ significantly between those who underwent surgery and those treated with CRT ([Fig. 2](#f2-mco-0-0-1909){ref-type="fig"}). Older patients with EC who received surgery had a poorer prognosis compared with younger patients ([Fig. 2](#f2-mco-0-0-1909){ref-type="fig"}). Esophagectomy is a viable alternative treatment option for patients aged \>80 years, if the surgical indication is strictly determined ([@b19-mco-0-0-1909]). In a prior study, the poor prognosis of older patients (aged ≥75 years) who underwent surgery was suggested to be due to the low rate of neoadjuvant chemotherapy ([@b6-mco-0-0-1909]). Indeed, in the present study, half of the older patients did not receive perioperative therapy, and of the 21 older patients who received adjuvant therapy after surgery, 2 succumbed to treatment-related adverse events. However, the lack of precise data regarding treatment-related complications and cause of death in older patients is a limitation of the present study. CRT is reportedly effective without major toxicity in older patients with locally advanced EC ([@b20-mco-0-0-1909]). Therefore, CRT may be considered as the optimal treatment strategy for older patients with locally advanced EC. Although 54 and 20% of the younger and older patients, respectively, with stage IV EC were treated with CRT, the median survival duration was 10 months in both groups. CRT should only be considered for patients with tumor-induced esophageal stenosis. Novel and effective treatment modalities for treating metastatic EC, such as immune-checkpoint drugs ([@b21-mco-0-0-1909]), are required. In this context, it may be hypothesized that the main explanation for the poorer prognosis in older patients is the more conservative approach to treatment. In addition, the higher frequency of multiple comorbidities among older patients may also partially explain the poorer prognosis in older compared with that in younger patients with EC. However, there was a lack of information regarding comorbidities among EC patients in the present study. Lack of disease-free survival data was also a limitation of this study. A more precise analysis is required to fully elucidate the exact causes of poorer prognosis among older patients. In conclusion, the poor prognosis of older patients (aged ≥75 years) with stage I EC may be improved with multidisciplinary treatment after ESD. Although the optimal treatment for older EC patients with stage II/III disease may be CRT at present, more efficient and safer treatment modalities, such as immune checkpoint drugs, are urgently needed. Not applicable. Funding ======= No funding was received. Availability of data and materials ================================== All data generated or analyzed during the present study are included in this published article. Authors\' contributions ======================= YM contributed to the drafting of this manuscript. YM, KK, TS, QZ, KS and MM contributed to the collection and analysis of the data. MM and YS contributed to conception, design and editing of the manuscript. All authors have read and approved the final version of this manuscript for publication. Ethics approval and consent to participate ========================================== This study was approved by the Ethics Review Committee of the School of Medicine of Niigata University (no. 2485). Patient consent for publication =============================== This was a retrospective observational study, carried out by the opt-out method of School of Medicine of Niigata University website. Competing interests =================== The authors declare that they have no competing interests. EC : esophageal cancer ESD : endoscopic submucosal dissection CRT : chemoradiotherapy BSC : best supportive care ![Disease-specific survival according to EC stage. Disease-specific survival of older (n=80) and younger (n=205) patients with stage 0 EC, older (n=171) and younger (n=270) patients with stage I EC, older (n=46) and younger (n=117) patients with stage II/III EC, older (n=25) and younger (n=39) patients with stage IV EC, and older (n=359) and younger (n=631) patients with all-stage EC. EC, esophageal cancer.](mco-11-04-0383-g00){#f1-mco-0-0-1909} ![Disease-specific survival of patients with stage II/III EC after surgery and CRT. Disease-specific survival after surgery and CRT of the younger (A) and older (B) patients with stage II/III EC. EC, esophageal cancer; CRT, chemoradiotherapy.](mco-11-04-0383-g01){#f2-mco-0-0-1909} ###### Characteristics of patients with esophageal cancer. Characteristics Total (n=990) Age ≤74 years (n=631) Age ≥75 years (n=359) P-value --------------------------- --------------- ----------------------- ----------------------- --------- Median age 71 (33--91) 66 (33--74) 79 (75--91) \<0.01 Sex 0.8 Male 853 545 308 Female 137 86 51 Histology \<0.01 Squamous cell carcinoma 897 588 309 Adenocarcinoma 69 32 37 Others 24 11 13 Stage \<0.01 0 285 205 80 I 441 270 171 II 50 33 17 III 113 84 29 IV 64 39 25 Unknown 37 0 37 ###### Main treatment modality for each stage of esophageal cancer. Patients --------- ------------------------------------------------------- ---------- ------ ----- ------ ----- ------ -------- All N \% N \% N \% \<0.01 ESD 629 63.5 411 65.1 218 60.7 Surgery^[a](#tfn1-mco-0-0-1909){ref-type="table-fn"}^ 144 14.5 103 16.3 41 11.4 CRT 108 10.9 81 12.8 27 7.5 RT 36 3.6 12 1.9 24 6.7 Chemotherapy 13 1.3 12 1.9 1 0.3 BSC 60 6.1 12 1.9 48 22 0 \<0.01 ESD 274 96.5 202 98.5 72 90 Surgery^[a](#tfn1-mco-0-0-1909){ref-type="table-fn"}^ 1 0.4 0 0 1 1.3 CRT 2 0.7 1 0.5 1 1.3 BSC 8 2.8 2 1 6 7.5 I \<0.01 ESD 331 75.1 207 76.7 124 72.5 Surgery^[a](#tfn1-mco-0-0-1909){ref-type="table-fn"}^ 52 11.8 38 14.1 14 8.2 CRT 31 7 20 7.4 11 6.4 RT 13 2.9 2 0.7 11 6.4 Chemotherapy 2 0.5 2 0.7 0 0 BSC 12 2.7 1 0.4 11 6.4 II/III \<0.05 Surgery^[a](#tfn1-mco-0-0-1909){ref-type="table-fn"}^ 86 52.8 65 55.5 21 45.7 CRT 50 30.7 40 34.2 10 21.7 RT 11 6.7 3 2.6 8 17.4 Chemotherapy 5 3.1 4 3.4 1 2.2 BSC 11 6.7 5 4.3 6 13 IV \<0.01 Surgery^[a](#tfn1-mco-0-0-1909){ref-type="table-fn"}^ 8 12.5 3 0.1 5 20 CRT 26 40.6 21 53.8 5 20 RT 11 17.2 6 1.5 5 20 Chemotherapy 5 7.8 5 1.3 0 0 BSC 14 21.9 4 0.1 10 40 Unknown ESD 13 35.1 0 0 13 35.1 \<0.01 BSC 24 64.9 0 0 24 64.9 Esophagectomy. ESD, endoscopic submucosal dissection; surgery; CRT, chemoradiotherapy; RT, radiotherapy; BSC, best supportive care. ###### Multidisciplinary treatment in stage I and II/III. Patients -------- ---------------------------------------------------------------------- ---------- ----- ----- -------- I \<0.05 ESD alone 281 169 116 ESD + CRT 41 33 6 ESD + RT 7 5 2 Surgery^[a](#tfn2-mco-0-0-1909){ref-type="table-fn"}^ alone 34 24 10 Surgery^[a](#tfn2-mco-0-0-1909){ref-type="table-fn"}^ + chemotherapy 2 2 0 Surgery^[a](#tfn2-mco-0-0-1909){ref-type="table-fn"}^ + CRT 15 12 3 Surgery^[a](#tfn2-mco-0-0-1909){ref-type="table-fn"}^ + RT 1 0 1 CRT 31 20 11 RT 13 2 11 Chemotherapy 2 2 0 BSC 12 1 11 II/III \<0.01 Surgery^[a](#tfn2-mco-0-0-1909){ref-type="table-fn"}^ alone 29 19 11 Surgery^[a](#tfn2-mco-0-0-1909){ref-type="table-fn"}^ + chemotherapy 49 38 10 Surgery^[a](#tfn2-mco-0-0-1909){ref-type="table-fn"}^ + CRT 8 8 0 CRT 50 40 10 RT 11 3 8 Chemotherapy 5 4 1 BSC 11 5 6 Esophagectomy. ESD, endoscopic submucosal dissection; CRT, chemoradiotherapy; RT, radiotherapy; BSC, best supportive care.	{ "pile_set_name": "PubMed Central" }
![](envhper00367-0096.tif "scanned-page"){.95} ![](envhper00367-0097.tif "scanned-page"){.96} ![](envhper00367-0098.tif "scanned-page"){.97}	{ "pile_set_name": "PubMed Central" }
Introduction {#sec1} ============ It is essential to understand the behavior of actinide complexes for the development of separation technologies and future nuclear fuel recycling processes.^[@ref1]^ Organic ligands, which coordinate to lanthanides/actinides, have been studied extensively^[@ref2]^ and will be a key element in the management and long-term storage of nuclear waste. In this context, the tridentate N-donor ligand 2,6-bis(2-benzimidazyl)pyridine (H~2~BBP), shown schematically in [Figure [1](#fig1){ref-type="fig"}](#fig1){ref-type="fig"}, has been studied in the past as a low-valent actinide sequestering agent and as a sensitizer for lanthanide luminescence.^[@ref3]−[@ref9]^ We have previously reported the experimental synthesis and spectroscopic characterization of a series of uranyl ({UO~2~}^2+^) complexes featuring the BBP ligand and its derivatives.^[@ref10],[@ref11]^ Interestingly, crystallographic characterization of the complexes revealed nonplanarity in the equatorial ligand coordination structure about the uranyl moiety and a modulation of the nonplanarity with changing ligand character, all without significant structural changes in the uranyl moiety. Structural distortions of uranyl complexes have been examined extensively, and much of the focus has been upon the uranyl moiety itself.^[@ref12]−[@ref15]^ Altering both the steric and electronic characteristics of the equatorial ligands has been shown to play a role in changing the U=O bond lengths of the ({UO~2~}^2+^) structural unit as well as opening up new avenues for reactivity.^[@ref16],[@ref17]^ There has also been interest in distorting the dioxo O--U--O bond angle from its typical linear arrangement.^[@ref18]−[@ref22]^ Recently, Hayton and co-workers reported a series of uranyl complexes with macrocyclic ligands, whose steric constraints push the O--U--O bond angle slightly away from its desired linear arrangement.^[@ref22]^ Distortion of the equatorial ligands has also been seen previously in other uranyl complexes.^[@ref22]−[@ref24]^ For example, Berthet and co-workers reported a variety of uranyl complexes featuring different N-donor ligands, including \[(UO~2~)(terpy)(OTf)~2~\] (terpy = 2,6-bis(2-pyridyl)pyridine) and \[U(phen)~3~\]\[OTf\]~2~ (phen = 1,10-phenanthroline), which both exhibit a significant deviation from planarity.^[@ref24]^ ![Schematics of the organic ligands used in this study. \[Left\] 2,6-Bis(2-benzimidazyl)pyridine (H~2~BBP); \[right\] smaller ligand obtained from H~2~BBP by truncating the benzene rings of the benzimidazole groups.](ao-2016-00459p_0002){#fig1} The aim of this study is to investigate in detail the coordination characteristics of BBP--uranyl complexes through first-principles theoretical simulations and to determine the origin of the nonplanarity of the equatorial ligands, either steric or electronic. To this end, a series of computational experiments are carried out, which clarify the relationship between ligand character and nonplanarity. The three complexes, \[(UO~2~)(H~2~BBP)Cl~2~\] (1), \[(UO)~2~(HBBP)(Py)Cl\] (2), and \[(UO~2~) (BBP)(Py)~2~\] (3), that were experimentally synthesized and form the primary subjects of this investigation are shown in [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}. The complexes have similar BBP ligands, which provide three coordinating bonds, but differ in the character of the two additional ligands that complete the pentagonal equatorial coordination structure around the uranyl moiety. Compound 1 has two Cl^--^ groups, compound 2 has one Cl^--^ and one pyridine, whereas compound 3 has two pyridines as the additional ligands. There are also concomitant differences in the charge state of the BBP ligand. As shown in [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}, N4 and N5 on the BBP ligand are protonated in compound 1 and are both deprotonated in 3. In 2, N5, which is on the side of the pyridine attachment, is protonated, whereas N4 is not. In terms of coordination, 1 and 2 exhibit distinct noncoplanarity with the equatorial plane compared to 3, which is coplanar about the equatorial plane and symmetric under reflection. One of the main objectives of the current study is to determine whether electronic or steric effects drive the observed coordination structure differences in these complexes. A density functional theory (DFT)-based analysis of the crystal structures of the three molecular complexes, 1--3, has been reported previously.^[@ref10]^ However, as will be shown later, crystal packing has very little effect on the interesting structural differences related to nonplanarity about the uranyl. Additionally, some of the complexes we simulate to investigate trends in bonding and nonplanarity have not been synthesized experimentally, and because the molecular crystals of these systems often depend on incorporated solvent units, trying to predict their crystalline structures from first-principles simulations would distract from the main focus of this work. Therefore, in this study, we investigate the intrinsic structural characteristics of the isolated molecular units of BBP--uranyl complexes free from crystal packing effects. ![Labeling and structure of three BBP--uranyl complexes. Side views are shown on the right side. Compounds 1 and 2 show the distinct asymmetric uranyl structure with respect to pyridine ligand planes. Color code: U, light blue; N, dark blue; O, red; Cl, green; C, gray; H, white.](ao-2016-00459p_0006){#fig2} Results and Discussion {#sec2} ====================== Relevant bond lengths and bond angles extracted from optimized geometries of the isolated complexes are compared to experimental molecular crystal values, as summarized in [Table [1](#tbl1){ref-type="other"}](#tbl1){ref-type="other"}. We have shown previously^[@ref10]^ that generalized gradient approximation (GGA)-optimized structural parameters of molecular crystals agree reasonably well with the experiment with errors within the 3% range. In the case of isolated complexes, we find slightly larger differences, with bond lengths typically longer by 3--5% compared to those of experimental crystal structures, and we attribute this to the absence of the crystalline environment in the calculations. However, the utility of the isolated molecular calculations is that they enable us to identify trends in the structural properties that are intrinsic to the complexes rather than driven by crystal packing. For instance, the experimentally observed trend that the BBP N--U bond lengths get progressively shorter going from 1 to 3 is reproduced for the isolated complexes. Accordingly, as can be inferred from [Table [1](#tbl1){ref-type="other"}](#tbl1){ref-type="other"}, the average calculated BBP N--U bond length decreases from 2.722 to 2.505 Å going from 1 to 3. Meanwhile, the crystalline O=U=O bond angles deviate from 180° by an increasing degree going from 1 to 3, but this trend is not observed in the simulated structures of the isolated complexes. This is because the O=U=O bond angle is significantly affected by crystal packing and DFT simulations reproduce the observed trend in crystals.^[@ref10]^ Meanwhile, the U=O bond lengths do not change significantly across the three complexes in both the experimental and theoretical structures. ###### GGA-Optimized Structural Parameters for Isolated Molecules of Compounds 1--3 Are Compared to Experimental Molecular Crystal Data[a](#t1fn1){ref-type="table-fn"} compound 1 compound 2 compound 3 ---------------- ---------------- ---------------- -------------- -------- ------- -------------- -------- ------- U--O1 1.766 1.825 U--O1 1.762 1.821 U--O1 1.768 1.827 U--O2 1.772 1.827 U--O2 1.773 1.822 U--O2 1.768 1.827 U--N1 2.556 2.707 U--N1 2.593 2.646 U--N1 2.501 2.513 U--N2 2.606 2.754 U--N2 2.573 2.614 U--N2 2.519 2.489 U--N3 2.581 2.705 U--N3 2.533 2.522 U--N3 2.501 2.512 U--Cl1 2.661 2.603 U--N6 2.554 2.586 U--N6 2.538 2.642 U--Cl2 2.692 2.605 U--Cl1 2.685 2.637 U--N7 2.538 2.642 ∠O1 = U = O2 178.1 169.7 ∠O1 = U = O2 176.53 177.2 ∠O1 = U = O2 176.2 169.5 ∠N1--N2--N3 113.0 115.3 ∠N1--N2--N3 114.43 113.8 ∠N1--N2--N3 114.94 113.8 ∠Eq./N1N2N3 30.90 28.4 ∠Eq./N1N2N3 13.20 24.3 ∠Eq./N1N2N3 3.5 2.1 Equatorial planes are generated by sets of three points, namely, Cl1--U--Cl2 in compound 1, Cl1--U--N6 in compound 2, and N6--U--N7 in compound 3 (see [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}). Units for distances and angles are angstroms and degrees, respectively. To better understand the origin of the observed nonplanarity in 1 and 2, we investigated several potential driving factors, such as crystal packing, ligand character, and steric hindrance. The degree of nonplanarity about the uranyl is quantified by first defining an equatorial plane in each complex built by sets of three points: (U1, Cl1, Cl2) in 1, (U1, Cl1, N6) in 2, and (U1, N6, N7) in 3. The angle between the equatorial plane and a plane formed by the BBP N sites (N1, N2, N3) shows how much the uranyl is elevated with respect to the plane of the BBP, yielding the degree of nonplanarity. As shown in [Table [1](#tbl1){ref-type="other"}](#tbl1){ref-type="other"}, compound 1 shows the highest angle, whereas compound 3 is virtually flat and symmetric, in both the experiments and computations. Complex 2 shows an intermediate nonplanar angle between compounds 1 and 3. The realization of a nonplanar configuration in isolated units of complexes 1 and 2 suggests that it is an intrinsic feature of the complexes themselves and not a result of crystal packing effects. However, given the structural complexity of these compounds, we still need to verify the energetic stability of the nonplanar structures with respect to planar motifs. To this end, we consider a model unit of complex 1 in a supercell geometry, with the structure of the model initially configured so that all of the equatorial ligands form a plane as shown in [Figure [3](#fig3){ref-type="fig"}](#fig3){ref-type="fig"}. We then carry out a constrained structural optimization whereby all of the atoms of the BBP ligand are constrained to lie in the Cl1--U--Cl2 plane during the optimization. Once all of the forces in this planar structure fall below 0.01 eV/Å, we release the previously imposed constraint and allow all of the atomic positions to further relax until all of the forces are smaller than 0.001 eV/Å. Thus, we find that a planar configuration with almost no forces acting on the atoms can be realized in the calculation even though its energy is 0.124 eV higher than that of the optimized nonplanar configuration, suggesting that the planar configuration is at best a local minimum. A vibrational analysis of the planar structure confirms the instability through an imaginary low-frequency (26.4 cm^--1^) vibrational mode with an out-of-plane displacement of the entire {UO~2~}^2+^ unit relative to the BBP ligand. This establishes that the nonplanar geometry is indeed the intrinsically stable minimum-energy configuration of 1 and not a result of crystal packing. Interestingly, the geometric parameters of the optimized planar and nonplanar configurations of the isolated complex, 1, provide some insight into the role of sterics (see [Table [2](#tbl2){ref-type="other"}](#tbl2){ref-type="other"}). We find that in the out-of-plane configuration, BBP N-uranyl bond lengths are shorter on average by ∼0.034 Å, the U--N1 and U--N2 bonds being significantly shorter, and ∠N1--N2--N3 of the BBP ligand at 115.3° is also smaller by 2.9° compared to that in the in-plane configuration. In contrast, ∠Cl1--U--Cl2 is smaller by 5.6° and the Cl1--H1 and Cl2--H2 distances are 9.5% shorter in the planar geometry. Thus, accommodating {UO~2~}^2+^ in the plane of the BBP ligand requires a more open ∠N1--N2--N3, longer BBP N-uranyl bonds, and a narrower ∠Cl1--U--Cl2, all of which suggest steric interaction in the equatorial plane between the Cl^--^ ions and the nearby H groups on the benzimidazole rings of BBP. Conversely, being out-of-plane and thereby increasing the Cl1--H1 and Cl2--H2 distances not only allows the uranyl to strengthen its bonding to the BBP ligand but also allows the latter to achieve an unstrained ∠N1--N2--N3 that is similar to that found in complexes 2 and 3. These findings point to the steric interaction between the Cl^--^ ions and the H sites on the outer benzimidazole groups as potentially driving the nonplanarity. ![Side view of model complex 1 with all of the BBP ligand atoms constrained to lie in the equatorial plane formed by the Cl1--U--Cl2 group of atoms.](ao-2016-00459p_0001){#fig3} ###### GGA-Optimized Structural Parameters for an Isolated Molecule of Complex 1 Are Compared to Those of a Constrained Model of Complex 1 Wherein the Structure Is Optimized While Forcing All the BBP Ligand Atoms To Lie in the Same Equatorial Plane Defined by the Three Points, Cl1--U--Cl2 (See [Figure [3](#fig3){ref-type="fig"}](#fig3){ref-type="fig"})[a](#t2fn1){ref-type="table-fn"} bond/angle complex 1 planar complex 1 -------------- --------------- ---------------------- U--O1 1.825 1.824 U--O2 1.827 1.824 U--N1 2.707 2.751 U--N2 2.754 2.765 U--N3 2.705 2.752 U--Cl1 2.603 2.607 U--Cl2 2.605 2.606 Cl1···H1 2.592 2.322 Cl2···H2 2.542 2.324 ∠N1--N2--N3 115.3 118.2 ∠N1--U--N3 118.9 122.5 ∠Cl1--U--Cl2 84.2 78.6 ∠Eq./N1N2N3 28.4 0.0 Units for distances and angles are angstroms and degrees, respectively. To further investigate this, we consider the effect of substituting the Cl^--^ ligands in complex 1 with other halide ligands characterized by different ionic radii and electronegativities. We note that in contrast to when the Cl^--^ ligands are replaced by pyridine ligands the substitution of Cl^--^ by other halides leaves the formal charge on the BBP ligand unchanged, enabling us to vary the properties of one set of ligands, that is, the halides whereas leaving the properties of the BBP ligand nominally unchanged. In [Table [3](#tbl3){ref-type="other"}](#tbl3){ref-type="other"}, we present GGA-optimized geometric parameters for a series of molecules derived from complex 1 by substituting Cl^--^ with F^--^ through I^--^ (see [Figure [4](#fig4){ref-type="fig"}](#fig4){ref-type="fig"}). We find that the out-of-plane angle increases monotonically with increasing ionic radius (r~ion~) of the halide ligands (indicated by labels α1 and α2), going from an almost planar structure for F^--^ to a nonplanar angle of 39.9° in the case of I^--^. Additionally, ∠α1--U−α2 is also seen to decrease monotonically from F^--^ to I^--^ with an overall change of 5%. In particular, as shown in [Table [3](#tbl3){ref-type="other"}](#tbl3){ref-type="other"}, (r~ion~ + 0.9 Å) is seen to be a good approximation to the distance between each halide ion and the nearest H site on the BBP benzimidazole groups across all of the halide ligand complexes. This suggests that in these complexes the α1--uranyl−α2 moiety moves out of the BBP plane and simultaneously ∠α1--U−α2 reduces until the halide to benzimidazole H distance reaches approximately (r~ion~ + 0.9 Å). An effective radius of 0.9 Å is therefore indicated for H groups in their interaction with halide ions. Meanwhile, the U--N1 and U--N3 bond lengths show neither significant variation nor a systematic trend across the halide-substituted series, whereas the U--N2 bond length actually decreases slightly going from F^--^ to I^--^. This is in contrast to complexes 2 and 3 where the reduced nonplanar angle is also accompanied by reduced U--N1, U--N2, and U--N3 distances relative to 1. Thus, we can infer that once the halide--benzimidazole H steric interaction is accommodated through nonplanarity, the BBP N-uranyl bond lengths are modulated primarily by the effective charge state of the BBP ligand. A similar computational experiment as described before in the case of complex 1, with structures generated by initially enforcing perfect planarity of all atoms in the BBP ligand and the α1--U−α2 group, indicates that the optimized planar structures are higher in energy than the corresponding fully unconstrained equilibrium structures by 0.025, 0.124, 0.197, and 0.316 eV in the F^--^-, Cl^--^-, Br^--^-, and I^--^-substituted complexes, respectively. A vibrational frequency analysis further indicates that such structures are unstable with respect to a motion of the α1--U−α2 moiety out of the plane of the BBP ligand. Soft imaginary frequency modes corresponding to this out-of-plane motion are obtained at 18, 26, 29, and 36 cm^--1^ for the F^--^-, Cl^--^-, Br^--^-, and I^--^-substituted complexes, respectively. We note in this instance that both the energy difference between the perfectly planar and unconstrained structures and the magnitude of the unstable mode energies increase monotonically with the halide ligand size. Interestingly, for F^--^, although a weakly unstable mode for out-of-plane motion of F^--^--U--F^--^ is detected in the perfectly planar structure, in practice, the F^--^ complex assumes its lower energy equilibrium geometry through a slight buckling of the BBP ligand itself rather than such an out-of-plane motion. ![Structural models of isolated molecular units derived from complex 1 by incorporating different halide ligand substitutions; (a) F^--^, (b) Cl^--^, (c) Br^--^, and (d) I^--^; are shown. Side views of the complexes are shown on the right. For those atoms not explicitly labeled, the same labeling convention as used in [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"} is followed.](ao-2016-00459p_0003){#fig4} ###### GGA-Optimized Structural Parameters for Isolated Molecular Units Derived from Complex 1 by Different Halide Ligand Substitutions Are Compared[a](#t3fn1){ref-type="table-fn"} bond/angle F \[r~ion~ = 1.19 Å\] Cl \[r~ion~ = 1.67 Å\] Br \[r~ion~ = 1.82 Å\] I \[r~ion~ = 2.06 Å\] ------------- --------------------------- -------------------------- -------------------------- ------------------------- U--O1 1.838 1.825 1.822 1.821 U--O2 1.840 1.827 1.827 1.826 U--N1 2.668 2.707 2.683 2.684 U--N2 2.780 2.754 2.748 2.735 U--N3 2.671 2.705 2.681 2.683 U−α1 2.137 2.603 2.776 3.022 U−α2 2.137 2.605 2.777 3.020 α1···H1 2.071 (2.09) 2.592 (2.57) 2.795 (2.72) 3.001 (2.96) α2···H2 2.068 (2.09) 2.542 (2.57) 2.796 (2.72) 3.013 (2.96) ∠N1--N2--N3 115.2 115.3 113.2 113.1 ∠N1--U--N3 121.4 118.9 116.4 116.1 ∠α1--U−α2 87.7 84.2 85.1 83.3 ∠Eq./N1N2N3 1.7 28.4 36.5 39.9 In column 1, α1 and α2 represent the placeholders for the two halide ligand sites (see [Figure [4](#fig4){ref-type="fig"}](#fig4){ref-type="fig"}). The equatorial planes are generated by sets of three points, namely, α1, U, and α2. Units for distances and angles are angstroms and degrees, respectively. The numbers in parentheses indicate (r~ion~ + 0.9 Å). Having established that for a given N-donor ligand, that is, BBP, reducing the ionic radius of the equatorial halide ligands leads to a reduced out-of-plane angle, we next investigated if for a given halide ligand, reducing the effective size of the N-donor ligand leads to a similar trend. We reason that if steric hindrance is the primary driver of nonplanarity then the effect should be similar whether it is the size of the halide ligands or the N-donor ligand that is reduced. Accordingly, we truncated the outer benzene rings of the benzimidazole groups on the BBP ligand to obtain a smaller pi-conjugated ligand (see [Figure [1](#fig1){ref-type="fig"}](#fig1){ref-type="fig"}) such that for a given nonplanar angle, the H atoms on the imidazole rings of this ligand are farther away from the halide ions. This truncated BBP (tr-BBP) ligand has the same nominal charge state as that of the unmodified BBP ligand. We constructed four modified complexes featuring the tr-BBP and F^--^, Cl^--^, Br^--^, and I^--^ ligands as shown in [Figure [5](#fig5){ref-type="fig"}](#fig5){ref-type="fig"} and optimized their geometries. We find that for each type of halide ion the degree of nonplanarity with the tr-BBP ligand is significantly smaller than that with the original BBP ligand as is evident from the nonplanar angles reported in [Table [4](#tbl4){ref-type="other"}](#tbl4){ref-type="other"}. Furthermore, we see that in the case of the larger halide ions where perfect planarity is not attained, the final distance between the halide ion and the nearest H atom on tr-BBP is similar to the corresponding value observed in the case of the unmodified BBP ligand (see [Table [3](#tbl3){ref-type="other"}](#tbl3){ref-type="other"}). For Cl^--^, Br^--^, and I^--^, (r~ion~ + 0.9 Å) is once again a good predictor of the optimized halide to H group distance, whereas for the smaller F^--^ ion where planarity is reached, this distance is larger than (r~ion~ + 0.9 Å). Finally, as before, we find a monotonic decrease of ∠α1--U−α2 and the U--N2 bond length with increasing r~ion~, whereas the U--N1 and U--N3 bond lengths show a weak but opposite trend. ∠N1--N2--N3 also show a weak monotonically increasing trend across the series but with values that are comparable to those obtained with the full BBP ligand. Weaker trends are also apparent in perfectly planar analogues of the tr-BBP complexes where, as before, planarity of all atoms in the tr-BBP ligand and the α1--U−α2 moiety is initially enforced. We find that the optimized planar structures are higher in energy relative to the equilibrium structures by 0.058, 0.085, 0.061, and 0.064 eV, respectively, for the F^--^, Cl^--^, Br^--^, and I^--^ ligand complexes. Interestingly, a vibrational analysis of the planar structures does not yield unstable modes corresponding to a clear out-of-plane motion of the α1--U−α2 moiety except in the case of the F^--^ complex. This suggests that the planar Cl^--^-, Br^--^-, and I^--^-substituted tr-BBP--uranyl complexes represent higher energy local minima, whereas the planar F^--^ complex, with more compact U--N1 and U--N3 bonds and a smaller ∠N1--N2--N3, is weakly unstable. Despite subtle variations in the behavior of the planar structures, given that the same choice of the halide ligand yields equilibrium structures with very different nonplanar angles based on the size of the N-donor ligand, we are once again able to establish the primacy of steric effects and rule out halogen electronegativity differences, as being a significant factor in the structural trends observed across the series of halide ligand complexes. Our computational study shows that the structural properties of N-donor uranyl complexes are sensitive to both the effective charge states of the ligands and steric interactions and can be systematically tuned by incorporating ligands of different sizes. ![Structural models of isolated molecular units derived from complex 1 by truncating the benzene rings on the benzimidazole groups of the BBP ligand and substituting various halide ligands (a) F^--^, (b) Cl^--^, (c) Br^--^, and (d) I^--^ in the equatorial plane are shown. Side views of the complexes are shown on the right. For those atoms not explicitly labeled, the same labeling convention as used in [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"} is followed.](ao-2016-00459p_0004){#fig5} ###### GGA-Optimized Structural Parameters for Isolated Molecular Units Derived from Complex 1 by Truncating the Benzene Rings of the Benzimidazole Groups and Incorporating Different Halide Ligands Are Compared (See [Figure [5](#fig5){ref-type="fig"}](#fig5){ref-type="fig"})[a](#t4fn1){ref-type="table-fn"} bond/angle F \[r~ion~ = 1.19 Å\] Cl \[r~ion~ = 1.67 Å\] Br \[r~ion~ = 1.82 Å\] I \[r~ion~ = 2.06 Å\] ------------- ------------------------- -------------------------- -------------------------- ------------------------- U--O1 1.840 1.826 1.824 1.821 U--O2 1.840 1.826 1.824 1.822 U--N1 2.621 2.636 2.643 2.651 U--N2 2.806 2.776 2.765 2.750 U--N3 2.622 2.637 2.643 2.651 U−α1 2.142 2.621 2.791 3.035 U−α2 2.141 2.621 2.793 3.037 α1···H1 2.466 (2.09) 2.629 (2.57) 2.714 (2.72) 2.874 (2.96) α2···H2 2.472 (2.09) 2.634 (2.57) 2.708 (2.72) 2.870 (2.96) ∠N1--N2--N3 113.1 114.1 114.5 114.7 ∠N1--U--N3 120.1 121.5 122.0 122.1 ∠α1--U−α2 95.6 86.7 84.4 81.8 ∠Eq./N1N2N3 1.7 3.8 4.9 11.8 In column 1, α1 and α2 represent the placeholders for the two halide ligand sites. The equatorial planes are generated by sets of three points, namely, α1, U, and α2. Units for distances and angles are angstroms and degrees, respectively. The numbers in parentheses indicate (r~ion~ + 0.9 Å). Conclusions {#sec3} =========== The structural properties of a series of uranyl complexes; \[(UO~2~)(H~2~BBP)Cl~2~\] (1), \[(UO)~2~(HBBP)(Py)Cl\] (2), and \[(UO~2~)(BBP)(Py)~2~\] (3); based on the N-donor ligand 2,6-bis(2-benzimidazyl)pyridine (H~2~BBP) are investigated theoretically with a particular emphasis on the nonplanarity observed in these complexes about the equatorial plane of uranyl. Through a series of computational experiments, the role of different factors such as crystal packing, ligand electronegativity, and steric effects in driving the nonplanarity is explored. We find that the steric interaction between bulky halide ligands characterized by large ionic radii and nearby H groups on the benzimidazole groups of the BBP ligand is the primary reason for the nonplanarity and the role of other factors, such as axial uranyl distortions, is insignificant in this regard. Furthermore, these steric effects can be exploited to systematically tune the structural properties of these complexes by incorporating ligands of different sizes, and we believe this will be useful for the design of actinide complexes. Computational Details {#sec4} ===================== Electronic structure calculations were carried out using the DFT^[@ref25],[@ref26]^ platform provided by the Vienna ab initio simulation package,^[@ref27],[@ref28]^ which implements a planewave basis set framework in conjunction with projector augmented wave^[@ref29]^ pseudopotentials. The DFT exchange-correlation energy is modeled within the Perdew--Burke--Ernzerhof^[@ref30]^ GGA.^[@ref31]^ A planewave cutoff of 400 eV is used for the wave functions, and the Brillouin zone is sampled at the Γ point. As the uranium ion in {UO~2~}^2+^ features a nominally empty 5f shell in its U^6+^ oxidation state with an electronic configuration of 1s^2^2s^2^2p^6^3s^2^3p^3^d^10^4s^2^4p^6^4d^10^4f^14^5s^2^5p^6^5d^10^6s^2^6p^6^, f-electrons do not play a significant role in the ground state electronic structure. Therefore, on-site Coulomb repulsion corrections are not employed in our structural analysis. Relativistic effects are taken into account within the core regions of the atoms in constructing the pseudopotentials employed but are neglected in the valence self-consistent field calculations. Supercells of the isolated complexes are constructed within a periodic boundary condition approach so that atoms in neighboring image cells are separated by at least 8 Å of vacuum, and the Brillouin zone is sampled at the zone center. Geometry optimizations are carried out until all forces are smaller than 0.01 eV/Å. Structural models shown were created using the VESTA-3^[@ref32]^ program. The authors declare no competing financial interest. The authors wish to acknowledge ByoungSeon Jeon for early contributions to the theoretical simulations presented in this work. This work was supported by the Director, Office of Science, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences, and Biosciences of the U.S. Department of Energy at Lawrence Berkeley National Laboratory under Contract No. DE-AC02-05CH11231. Calculations were performed at the National Energy Research Scientific Computing Center (NERSC-LBNL) and the Molecular Foundry computing resources managed by the High Performance Computing Services Group, LBNL. Both facilities are supported by the Office of Science of the U.S. Department of Energy.	{ "pile_set_name": "PubMed Central" }
Introduction {#Sec1} ============ Melanoma, a highly malignant neoplasm of the melanocytes, is the most aggressive form of skin cancer^[@CR1]^. It accounts for less than 5% of all skin cancer cases, but the vast majority (80%) of skin cancer related-deaths^[@CR2]^. The incidence of malignant melanoma has been increasing at a steady rate in fair-skinned populations around the world for decades^[@CR3],\ [@CR4]^. However, currently available chemotherapeutics against malignant melanoma are often expensive, with toxic side effects, low response rates, and/or high tendency to develop tolerance^[@CR5]--[@CR8]^. These disappointing but harsh realities highlight the urgency of exploring novel, safe and effective alternative approaches for melanoma management. Because of their biological activity and low toxicity, natural products (i.e., food, herbs) have been demonstrated to be promising candidates for melanoma prevention and treatment^[@CR9],\ [@CR10]^. Signal transducer and activator of the transcription 3 (STAT3), which is constitutively activated at 50 to 90% frequencies in diverse human cancers including melanoma, has been considered as a potential target for melanoma treatment^[@CR11]^. Previous experimental findings have demonstrated that targeting STAT3 in melanoma tumor models induces tumor cell death/tumor regression^[@CR12],\ [@CR13]^ and inhibits metastasis^[@CR14]^. Sophorae Flos (SF), the flower and flower-bud of Sophora japonica L., is commonly consumed as a vegetable and used to make jam and snacks in China. It has skin-care benefits^[@CR15]^. The dye extract from SF has been shown to possess ultraviolet protective properties^[@CR16]^. Lonicerae Japonicae Flos (LJF), the flower bud of Lonicera japonica Thunb., is widely consumed as tea, and has long been used for treating skin carbuncles and pyocutaneous diseases in Asian countries^[@CR17],\ [@CR18]^. Both SF and LJF are commonly used in treating melanoma in traditional Chinese medicine (TCM) practice^[@CR19]--[@CR21]^. In Yi Xue Qi Meng (a Chinese medicine classic issued 600 years ago), a formula (SL) consisting of SF and LJF simmered in rice wine is documented as a remedy for subcutaneous ulcer, skin carbuncle and abscess, which have TCM symptoms resembling those of melanoma. In addition, constituents in SF and LJF, such as rutin, quercetin and luteolin, have been shown to possess anti-melanoma properties^[@CR22]--[@CR24]^. Some of these constituents have also been demonstrated to inhibit STAT3 signaling in different types of tumor cells^[@CR23],\ [@CR25]^. However, there is no report about the pharmacological effect of SL on melanoma so far. In the present study, we evaluated the in vivo and in vitro anti-melanoma effects of an ethanolic extract of SL (SLE). Human A375 and murine B16F10 melanoma cells, together with the B16F10 melanoma allograft model in C57/BL6 mice were employed. The involvement of STAT3 signaling in the anti-melanoma effects of SLE was also explored. Results {#Sec2} ======= SLE restrained tumor growth and STAT3 activation in a B16F10 allograft model {#Sec3} ---------------------------------------------------------------------------- The in vivo anti-melanoma effect of SLE was evaluated using a B16F10 melanoma allograft C57/BL6 mouse model. At the end of the experimental period, each mouse only had one tumor. As shown in Fig. [1A](#Fig1){ref-type="fig"}, daily intragastric administration of 1.2 g/kg SLE for 15 days significantly inhibited tumor growth in mice. In comparison with the control group, the average tumor size and tumor weight in SLE-treated group were remarkably reduced by 54.1% and 55.3% after SLE intervention, respectively (Fig. [1B and C](#Fig1){ref-type="fig"}). No animal death occured during the experimental period. No abnormalities were found in all mice at necropsy on day 15. No significant differences were observed in the food and water consumption (Data not shown), and body weight (Fig. [1D](#Fig1){ref-type="fig"}) between the two groups. It is well recognized that constitutive activation of STAT3 plays a critical role in melanoma development^[@CR10]^. To determine whether SLE affects STAT3 activation, we examined the expression of phosphorylated STAT3 in tumor tissues by immunoblotting. SLE potently decreased the protein levels of phospho-STAT3 (Tyr705). SLE also inhibited the expression of phospho-Src (Tyr416) in tumor tissues. The non-receptor tyrosine kinase Src is an upstream kinase of STAT3. SLE did not affect the expressions of total STAT3 and total Src in B16F10 allograft tumors (Fig. [1E](#Fig1){ref-type="fig"}).Figure 1Anti-melanoma effects of SLE in mice. (A) The photo of B16F10 tumors dissected from mice. (B) Mean weights of the dissected tumors. (C) Time-dependent effect of SLE on melanoma growth in a B16F10 allograft mouse model. (D) Body weights at different time points. In (B), (C) and (D), data were means ± SD of 7 mice. (E) Protein expression levels of phospho-STAT3 (Tyr705), STAT3, phospho-Src (Tyr416), Src and GAPDH in B16F10 tumors collected from 3 individual mice were examined by immunoblotting (left panel) and the relative band intensity was analyzed by Image J software (right panel). The images shown were the cropped blots for the corresponding proteins. The original blots from which images were cropped were shown in Supplementary Dataset. \*P \< 0.05, \\P \< 0.01 versus vehicle control group. SLE exhibited higher cytotoxicity in melanoma cells than in normal skin cells {#Sec4} ----------------------------------------------------------------------------- SLE treatment (50, 100, 200, 300, 400, 600, 800 and 1000 μg/mL, 24 and 48 h) decreased the viabilities of A375 and B16F10 cells in time- and dose- dependent manners (Fig. [2A and B](#Fig2){ref-type="fig"}). The IC~50~ values of SLE were 321.60 ± 3.45 μg/mL against A375 cells and 281.29 ± 5.15 μg/mL against B16F10 cells at 48 h. The crystal violet staining assay visually validated the inhibitory effects of SLE on B16F10 cell proliferation (Fig. [2C](#Fig2){ref-type="fig"}). Moreover, the inhibition rates of 100, 200, 300 and 400 μg/mL of SLE on human HDFa and HaCaT normal skin cells were much lower than that on A375 and B16F10 melanoma cells (Fig. [2D](#Fig2){ref-type="fig"}).Figure 2Cytotoxic effects of SLE on melanoma cells and normal skin cells. Effects of SLE on the viabilities of A375 cells (A) and B16F10 cells (B) were determined by the MTT assay and trypan blue exclusion assay, respectively. (C) Proliferation inhibitory effects of SLE on B16F10 cells measured by the crystal violet staining assay. Cells were treated with the indicated concentrations of SLE for 24 or 48 h. (D) Comparison of viability inhibitory effects of SLE on melanoma cells and normal skin cells. Cells were treated with the indicated concentrations of SLE for 48 h. Data were presented as means ± SD of three independent experiments. SLE induced apoptosis of A375 human melanoma cells {#Sec5} -------------------------------------------------- Annexin V/7-AAD double staining assay was performed to quantify the apoptotic effects of SLE on A375 cells. SLE treatment for 48 h dose-dependently induced apoptosis in A375 cells. SLE at the doses of 100, 200, 300 and 400 μg/mL significantly increased the rate of Annexin V positive cells to 14.2 ± 4.7%, 16.9 ± 4.0%, 20.4 ± 4.3% and 31.2 ± 4.0%, respectively, as compared with the untreated cells (5.7 ± 1.3%) (Fig. [3A and B](#Fig3){ref-type="fig"}). To further validate the pro-apoptotic effects of SLE, we determined the impact of SLE on the cleavage of poly (ADP-ribose) polymerase (PARP), which is considered as a hallmark of apoptosis. SLE (100, 200 and 300 μg/mL) dose-dependently increased the protein levels of cleaved-PARP in A375 cells (Fig. [3C](#Fig3){ref-type="fig"}).Figure 3Pro-apoptotic effects of SLE on A375 melanoma cells. Cells were treated with the indicated concentrations of SLE for 48 h. (A) Representative flow cytometry plots of cell apoptosis. Apoptosis was analyzed using the Annexin V/7-AAD double staining assay. (B) The percentage of apoptotic cells after SLE treatment. Data were shown as means ± SD of three independent experiments. \*P \< 0.05, \\P \< 0.01 versus control. (C) Effects of SLE on PARP cleavage in A375 cells. Protein levels were examined by immunoblotting and the cropped blots were shown. GAPDH was included as a loading control. Original uncropped western blot images were shown in Supplementary Dataset. SLE suppressed the migratory and invasive abilities of melanoma cells {#Sec6} --------------------------------------------------------------------- Wound healing assay and matrigel invasion assay were performed to determine the effects of SLE on melanoma cell migration and invasion, respectively. As shown in Fig. [4A and B](#Fig4){ref-type="fig"}, SLE (50 μg/mL) treatment for 24 h significantly inhibited the migration of A375 and B16F10 cells. For the cell invasion assay, A375 and B16F10 cells with or without SLE treatment were allowed to invade for 24 h. 50 μg/mL of SLE markedly reduced the cell invasiveness by 32.9% for A375 cells and 55.4% for B16F10 cells (Fig. [4C and D](#Fig4){ref-type="fig"}). As shown in cell viability assays, SLE at 50 μg/mL did not significantly affect the viabilities of A375 and B16F10 cells (Fig. [2A and B](#Fig2){ref-type="fig"}).Figure 4Inhibitory effects of SLE on melanoma cell migration and invasion. Representative photographs of A375 (A) and B16F10 (B) cell migration after SLE (50 μg/mL) treatment. Cells seeded in 6-well plates were grown to 80--90% confluence. A single scratch was made in the cell monolayer, and followed by vehicle control or SLE treatment for 24 h. Pictures were taken at 0 and 24 h after treatment. Effects of SLE on A375 (C) and B16F10 (D) cell invasion were examined using the matrigel invasion assay. Representative photographs of invaded cells (left) and quantification of invasiveness (right) were shown. Cells (1.5 × 10^5^) in DMEM-0.1% BSA with SLE (50 μg/mL) or vehicle control were allowed to pass through matrigel coated membrane for 24 h. Cells on the lower surface of the membrane were considered as invaded cells. Number of invading cells was determined as the average cell number in 5 random microscope areas for each condition for each separate replicate experiment. Data were shown as mean ± SD from three independent experiments, \\P \< 0.01. SLE inhibited STAT3 activation and STAT3 nuclear localization in melanoma cells {#Sec7} ------------------------------------------------------------------------------- Immunoblotting assay was employed to further determine the effects of SLE on STAT3 activation in melanoma cells. As expected, SLE (100, 200 and 300 μg/mL) treatment for 48 h apparently suppressed the phosphorylation of STAT3 (Tyr705) and Src (Tyr 416), but did not affect total STAT3 and total Src expressions, in both A375 and B16F10 cells (Fig. [5A and B](#Fig5){ref-type="fig"}). After phosphorylation, STAT3 homodimerizes and translocates into the nucleus where it binds to promoter elements of responsive target genes to regulate their transcriptions^[@CR26]^. Hence, we examined whether SLE could affect STAT3 nuclear localization. The immunofluorescence images and immunoblotting results showed that SLE at 100, 200 and 300 μg/mL caused evident decreases in nuclear STAT3 levels in A375 cells (Fig. [5C and D](#Fig5){ref-type="fig"}).Figure 5SLE suppressed STAT3 and Src phosphorylation, and STAT3 nuclear localization in melanoma cells. (A) The expression levels of phospho-STAT3 (Tyr705), STAT3, phospho-Src (Tyr416) and Src in A375 (A) and B16F10 (B) cells. Cells were treated with the indicated concentrations of SLE for 48 h. Protein levels were examined by immunoblotting. GAPDH was used as a loading control. (C) Immunofluorescence staining of STAT3 in A375 cells treated with SLE for 48 h. Cells were fixed and labeled with STAT3 antibody and counterstained with DAPI. (D) Immunoblot analysis of STAT3 levels in cytoplasmic and nuclear fractions of A375 cells treated with SLE for 48 h (left panel). GAPDH represents a loading control for the cytoplasmic fraction, and PCNA represents a loading control for the nuclear fraction. Relative protein levels were analyzed by Image J software (right panel). Results were shown as mean ± SD from three independent experiments, \\P \< 0.01. The corresponding western blot images represented were cropped sections and the original images were shown in Supplementary Dataset. SLE inhibited the expression of STAT3 target genes in melanoma cells {#Sec8} -------------------------------------------------------------------- Mcl-1 and Bcl-xL are STAT3-target genes involved in cell survival, and matrix metalloproteinase (MMP)-2 and MMP-9 are STAT3-target genes involved in cell migration and invasion^[@CR11]^. As shown in Fig. [6A and C](#Fig6){ref-type="fig"}, SLE effectively decreased the mRNA and protein levels of Mcl-1, Bcl-xL, MMP-2 and MMP-9 in A375 cells. SLE also dose-dependently inhibited mRNA and protein expressions of Mcl-1, Bcl-xL, MMP-2 and MMP-9 in B16F10 cells (Fig. [6B and C](#Fig6){ref-type="fig"}). These data further suggest that STAT3 signaling may be involved in the anti-melanoma action of SLE.Figure 6SLE downregulated the expression levels of STAT3 target genes in melanoma cells. Protein expressions of Mcl-1, Bcl-xL, MMP-2 and MMP-9 in A375 (A) and B16F10 (B) cells. Cells were treated with indicated concentrations of SLE for 48 h. Protein levels were examined by immunoblotting and cropped blots were displayed. The original blot images were shown in Supplementary Dataset. (C) The mRNA expressions of Mcl-1, Bcl-xL, MMP-2 and MMP-9 in A375 and B16F10 cells. Cells were treated with indicated concentrations of SLE for 24 h. Total RNA was extracted with Trizol. Gene expression levels were detected using RT-PCR analysis. Data were shown as fold change ± SD of three independent experiments, \*P \< 0.05, \\P \< 0.01 versus vehicle control. Overexpression of STAT3C in A375 cells diminished the effects of SLE on cell viability, apoptosis and invasion {#Sec9} -------------------------------------------------------------------------------------------------------------- To confirm the involvement of STAT3 signaling in SLE's anti-melanoma effects, A375 cells were stably transfected with a plasmid containing STAT3C, a constitutively active variant of STAT3. Immunoblotting results showed that transfection of STAT3C caused remarkable elevation in both total STAT3 and phospho-STAT3 (Tyr705) levels, compared to transfection with an empty vector (Fig. [7A](#Fig7){ref-type="fig"}). The effects of SLE on the viability, apoptosis and invasion of A375 control cells and cells stably expressing STAT3C were compared. When treated with 100, 200 and 300 μg/mL of SLE for 48 h, the viability inhibition rate was significantly decreased from 10.78 ± 3.64%, 30.88 ± 2.86%, 41.54 ± 2.43% in A375 cells transfected with empty vector to 1.38 ± 2.76%, 7.28 ± 2.04%, 15.24 ± 2.75% in cells transfected with STAT3C, respectively (Fig. [7B](#Fig7){ref-type="fig"}). Overexpression of STAT3C evidently diminished the apoptotic effects of SLE on A375 cells (Fig. [7C](#Fig7){ref-type="fig"}). STAT3C overexpression in A375 cells also remarkably attenuated 50 μg/mL of SLE-induced cell invasion inhibition rate from 36.8% to 16.7% (Fig. [7D](#Fig7){ref-type="fig"}).Figure 7Overexpression of STAT3C in A375 cells diminished the effects of SLE on cell viability, apoptosis and invasion. (A) Expression levels of phospho-STAT3 (Tyr705), STAT3 and Flag in A375 cells stably expressing Flag-tagged STAT3C or empty vector (pcDNA3). GAPDH was included as a loading control. Proteins from cultured cells were examined by immunoblotting and cropped blot images were shown. The original uncropped images were shown in Supplementary Dataset. Results shown were the representatives of three independent experiments. (B) Effects of STAT3C overexpression on SLE-mediated cytotoxic activity. (C) Effects of STAT3C overexpression on SLE-mediated apoptotic effects. In (B) and (C), A375 cells stably expressing STAT3C or empty vector were treated with indicated concentrations of SLE for 48 h. Cell viability was measured by the MTT assay. Apoptosis was determined by flow cytometric analysis of cells double stained with Annexin V/7-AAD. Data were presented as the mean ± SD of three independent experiments, \*P \< 0.05, \\P \< 0.01. (D) Effects of STAT3C overexpression on SLE-mediated inhibitory effects on cell invasion. A375 cells stably transfected with STAT3C or vector were treated with SLE (50 μg/mL) or vehicle control for 24 h. Cell invasion was determined by matrigel invasion assay. Statistical analyses were performed by ANOVA followed by LSD Test. \\P \< 0.01, SLE treatment versus vehicle control treatment. ^\#\#^ P \< 0.01, SLE-mediated cell invasion inhibition in A375 cells stably transfected with STAT3C versus that in A375 cells stably transfected with empty vector. Discussion {#Sec10} ========== Although mutant BRAF targeted-therapy and immunotherapy are showing exciting clinical results, the 5-year survival rate for patients with distant metastatic melanoma is merely 17%^[@CR4]^. Alternative management options for melanoma are urgently needed. TCM-based herbs, especially edible herbs, have been recognized as promising resources for alternative/complementary approaches for cancer management^[@CR9],\ [@CR10]^. In a TCM classic Yi Xue Qi Meng written 600 years ago, a herbal formula (SL) consisting of SF and LJF was documented to be used by Chinese medicine practitioners to treat skin diseases with TCM symptoms that resemble melanoma. However, there is no modern clinical or experimental evidence about the anti-melanoma action of this formula. In this study, we demonstrated that SLE, an ethanolic extract of SL, exerted potent prophylactic effects on melanoma growth in B16F10 melanoma-bearing mice. It is generally believed that the impacts of food or food products on diseases are modest. However, we observed that the treatment with 1.2 g/kg of SLE for 15 days exerted potent anti-melanoma effects in animals. Specifically, oral administration of SLE caused 54.1% and 55.3% reduction in the average tumor size and tumor weight in mice, respectively. SLE, utilized in this study, was prepared with 30% ethanol. This was based on the original documentation in a TCM classic that the formula SL (SF and LJF in a 5/1 ratio) is decocted with rice wine (about 30% alcohol) for treating skin disorders. The dose of 1.2 g/kg of SLE used for the in vivo experiment is the human equivalent dose. Whether SLE treatment with higher dose and longer duration exhibits more potent anti-melanoma effects is a question to be addressed. It is noteworthy that 1.2 g/kg of SLE did not cause observable reduction in animals' body weight (Fig. [1D](#Fig1){ref-type="fig"}) and any other abnormalities in clinical signs and gross pathology, suggesting that SLE has very low toxicity. Furthermore, in vitro MTT assays showed that SLE exhibited much lower cytotoxicity on human normal skin cells than on tumorous A375 cells (Fig. [2D](#Fig2){ref-type="fig"}). Indeed, SF and LJF are consumed as vegetable and tea, respectively, in Asian countries, indicating their edibility and safety^[@CR27]^. These suggest that 1.2 g/kg of SLE is effective and safe for melanoma prevention. Further studies are warranted to identify the bioactive constituents responsible for the anti-melanoma activity of SLE. We hope to discover compounds with potent anti-melanoma efficacy from SLE, just like the discovery of artemisinin from a traditional anti-malaria Chinese medicinal herb Herba Artemisiar Annuae, which was inspired from a record in a TCM classic. In our previous studies, it was found that quercetin, one of the bioflavonoids occurring in both SF and LJF, exhibited anti-melanoma action partially by inhibiting STAT3 signaling^[@CR23]^. In this study, SLE treatment also effectively inhibited the phosphorylation of STAT3 and one of its upstream kinases Src in B16F10 tumors. STAT3, persistently activated in melanoma, has been shown to regulate the transcription of a panel of genes involved in several oncogenic features, such as cell proliferation, apoptosis and metastasis^[@CR28]^. Therefore, we evaluated the potential effects of SLE on the viability, apoptosis, migration and invasion of human A375 and murine B16F10 cells. Our results showed that SLE exhibited potent in vitro anti-melanoma effects, as evidenced by the significant inhibition of cell viability, migration and invasion, as well as the induction of apoptosis in melanoma cells treated with SLE. In melanoma cells, we also found that SLE obviously suppressed the phosphorylation of STAT3 (Tyr 705) and Src (Tyr 416), which is consistent with the results of our animal experiments. STAT3 is a cytoplasmic transcription factor that transmits signals from plasma membrane to nucleus^[@CR26]^. It contains Src homology 2 (SH2) domains and can be transcriptionally activated by Src family kinases through phosphorylating its tyrosine 705 residue^[@CR29],\ [@CR30]^. Upon phosphorylation, STAT3 forms homodimers, which translocate into the nucleus to mediate the transcription of STAT3 target genes^[@CR26]^. Thus, we speculated that SLE-triggered inhibition of STAT3 Tyr 705 phosphorylation should cause reduction of STAT3 nuclear localization and suppression of STAT3 target genes expression. STAT3 contributes to melanoma growth by targeting the antiapoptotic proteins Bcl-xL and Mcl-1, which are upregulated during melanoma progression^[@CR11],\ [@CR31]^. STAT3 activation also promotes melanoma metastasis via increasing the expressions of vascular endothelial growth factor (VEGF), MMP-2 and MMP-9^[@CR14],\ [@CR32]--[@CR34]^. We found that SLE inhibited nuclear accumulation of STAT3 in A375 cells, decreased the protein and mRNA levels of Mcl-1, Bcl-xL MMP-2 and MMP-9 in both A375 and B16F10 cells; whereas, SLE did not reduce the expression of VEGF (Data not shown). These results demonstrate that the suppressive capabilities of SLE on melanoma cell survival, migration and invasion may be attributed to the inhibition of STAT3-targeted Mcl-1, Bcl-xL, MMP-2 and MMP-9. To further validate the contribution of STAT3 signaling in the anti-melanoma effects of SLE, we established A375 stable cells persistently expressing STAT3C, a constitutively active variant of STAT3. We found that STAT3C overexpression in A375 cells significantly diminished the inhibitory effects of SLE on cell viability and invasion, and also the pro-apoptotic effects of SLE. Furthermore, MTT assays showed that the cytotoxic effect of SLE could be observed in various melanoma cell lines with different genetic backgrounds, and the effect on p-STAT3-positive cells was stronger than on p-STAT3-negative cells (see Supplementary Fig. [1](#MOESM1){ref-type="media"}). These results demonstrate that SLE exerts anti-melanoma effects, at least in part, by inhibiting STAT3 signaling. One of the major challenges in developing targeted cancer therapies is reducing or eliminating the severe adverse events resulting from on-target and off-target effects^[@CR35],\ [@CR36]^. On-target side effects refer to the pharmacodynamic effect on normal tissues, and off-target side effects are unexpected toxicities derived from the inhibition of unintended or unknown functions^[@CR36]^. For specific STAT3 inhibitors (WP1066, AZD9150, STAT3 DECOY, OPB-31121 and OPB-51602), several clinical trials have been approved to use them to treat melanoma and other tumors (<https://clinicaltrials.gov/>). However, some trials have been discontinued due to the severe adverse events^[@CR37]--[@CR39]^. The herbal formula SL has long been used in melanoma management by TCM practitioners in ancient China without unfavorable effects reported, and our experimental findings indicate that SLE is effective and safe for fighting melanoma. These are probably the comprehensive effects of SLE with its multicomponent and multitarget natures, which are exactly the advantages of TCM formula. Therefore, we deem that SLE may have greater clinical significance compared to specific STAT3 inhibitors. Considering the multitarget characteristic of SLE, we believe that STAT3 is one of the targets of SLE. To find more molecular targets of SLE, we analyzed the differentially expressed genes in B16F10 tumors collected from mice with or without SLE treatment using transcriptome re-sequencing analysis. We found that SLE altered the expressions of some genes related to tumor immune evasion (e.g. IL-6, IL-17, STAT1, TNF-α, CD45, MHCII) (Data not shown). In the future, therefore, we will investigate the impact of SLE on melanoma immune microenvironment. Our results showed that a clinically relevant dose of SLE suppressed melanoma growth in a mouse model. SLE also inhibited cell viability, induced cell apoptosis, restrained cell migration and invasion, suppressed STAT3 activation and nuclear localization, and downregulated STAT3 target genes in melanoma cells. Overexpression of constitutively active STAT3C diminished these effects. Thus, we conclude that SLE, an ethanolic extract of a herbal formula comprising SF and LJF, exerts anti-melanoma effects, and these effects are partially due to the inhibition of STAT3 signaling. This study provides pharmacological justification for the traditional use of this formula in melanoma treatment, and suggests that SLE and SLE-derived compounds have the potential to be developed as modern alternative and/or complimentary agents for melanoma management. Based on the ancient record and our experimental data, we recommend clinical trials of SLE for melanoma prevention and treatment. Methods {#Sec11} ======= Preparation of SLE {#Sec12} ------------------ SF and LJF were purchased from the Mr. & Mrs. Chan Hon Yin Chinese Medicine Specialty Clinic and Good Clinical Practice Centre in Hong Kong Baptist University. Both of them were authenticated by the corresponding author. Voucher specimens have been deposited at the Centre for Cancer and Inflammation Research, School of Chinese Medicine, Hong Kong Baptist University. It is documented in Yi Xue Qi Meng that SF and LJF (at a ratio of 5:1) are mixed and decocted with rice wine (about 30% alcohol) for treating skin disorders, suggesting that some active constituents are lipophilic. Therefore, we extracted the mixture of SF and LJF (at a 5/1 ratio) with 30% (v/v) ethanol aqueous by heat refluxing. The obtained ethanolic extract of SL (SLE) was then lyophilized with a Virtis Freeze Dryer (The Virtis Company, New York, USA). The yield of powdered SLE was 8.49%. The contents of rutin and chlorogenic acid (see Supplementary Fig. [2](#MOESM1){ref-type="media"}) in SLE was quantified by HPLC analyses. Animal experimentation {#Sec13} ---------------------- Male C57/BL6 mice (6 weeks old) were obtained from The Laboratory Animal Services Centre, The Chinese University of Hong Kong. All care and handling of animals were performed with the approval of the Department of Health, Hong Kong. Experimental procedures were approved by the Committee on the Use of Human & Animal Subjects of the Hong Kong Baptist University. B16F10 cells (1 × 10^6^) were resuspended in 0.1 mL of phosphate-buffered saline (PBS) and inoculated subcutaneously into the backs of C57/BL6 mice^[@CR40]^. Immediately after cell injection, the mice were randomly divided into two groups of seven each. Mice were then intragastrically (i.g.) administered with 0.5% Carboxymethyl cellulose-Na (CMC-Na) (vehicle control) or 1.2 g/kg (human equivalent dose) of SLE once per day for consecutive 15 days. To monitor the toxicity of SLE, general clinical observations were made once a day. Clinical signs, which included, but were not limited to, changes in skin, fur, eyes, mucous membranes, occurrence of secretions and excretions and autonomic activity (e.g., lacrimation, piloerection, pupil size, unusual respiratory pattern) were recorded. Changes in gait, posture and response to handling as well as the presence of colonic or tonic movements, stereotypes (e.g., excessive grooming, repetitive circling) or bizarre behavior (e.g., self-mutilation, walking backwards) were also recorded. Food and water consumption, and body weight of mice were measured once every 3 days. Tumor volumes were determined using a vernier caliper at days 7, 10 and 15 after cell injection. At the end of the experimental period, the tumor of each mouse was dissected and weighed. The organs including liver, kidneys, adrenals, testes, epididymides, spleen, brain and heart of each mouse were trimmed of any adherent tissue and weighed. Gross necropsy was performed for all the dissected organs and tissues. Proteins from tumor tissues were extracted with RIPA lysis buffer and examined by immunoblotting. Cell culture {#Sec14} ------------ Human A375 and murine B16F10 melanoma cells were obtained from American Type Culture Collection (ATCC, USA). Human keratinocyte cell line HaCaT and human adult dermal fibroblasts HDFa were obtained from Invitrogen (Casecade Biologics, Invitrogen cell culture, CA). A375, B16F10 and HaCaT cells were cultured in DMEM (GIBCO, USA) supplemented with 10% fetal bovine serum (FBS, GIBCO, USA) and 1% penicillin/streptomycin (P/S, GIBCO, USA). HDFa cells were cultured in medium 106 (GIBCO, USA) supplemented with LSSG kit (GIBCO, USA). All cells were maintained at 37 °C and in a humidified atmosphere of 5% CO~2~. Cell viability assay {#Sec15} -------------------- MTT (3-\[4,5-dimethylthiazol-2-yl\]-2,5-diphenylterazolium bromide) assay^[@CR41]^ was performed to determine the cytotoxicity of SLE on A375, HaCaT and HDFa cells. Cells (5,000 cell/well) seeded in 96-well plates were treated with SLE at various concentrations (0--1000 μg/mL) for 24 or 48 h. 20 μL of MTT solution (5 mg/mL) was added to each well and incubated for additional 4 h. The medium was then aspirated, and 100 μL of DMSO was added to dissolve the formed formazan crystals. The absorbance was measured at 570 nm by a microplate spectrophotometer (BD Biosciences, USA). The cytotoxic effect of SLE on B16F10 cells was determined by trypan blue exclusion assay. Cells (1 × 10^5^ cell/well) were seeded overnight in 6-well plates, and subsequently treated with various concentrations of SLE (0--1000 μg/mL) for 24 or 48 h. After treatment, cells were collected, stained with 0.4% trypan blue, and counted using a Countess Automated Cell Counter (Invitrogen Japan K.K.). Crystal violet staining assay^[@CR41]^ was also employed to visualize the effects of SLE on the proliferation of B16F10 cells. Treated cells were fixed with 10% formalin for 5 min, followed by staining with 0.05% crystal violet solution in distilled water for 30 min. Cells were then washed and photographed. Apoptosis assay {#Sec16} --------------- Apoptotic effects of SLE on A375 cells were quantified by Annexin V/7-AAD double staining assay^[@CR42]^ with the Apoptosis Detection Kit (BD Biosciences, San Jose, CA, USA). After treated with SLE for 48 h, both detached and adherent cells were harvested and then incubated in 100 μL labeling solution (5 μL of AnnexinV-PE, 5 μL of 7-AAD, 10 μL of 10X binding buffer and 80 μL of H~2~O) in darkness for 15 min. 400 μL of 1X binding buffer was added to stop the staining reaction. Flow cytometric analysis was performed on a FACSCalibur^TM^ (BD, San Jose, CA). In vitro cell migration assay {#Sec17} ------------------------------- The effect of SLE on cell migration was evaluated by wound healing assay^[@CR43]^. Cells were allowed to grow to 80--90% confluence in 6-well plates. Cell monolayers were scratched with a sterile 10 μL pipette tip across the center of the well to generate a clean, straight wound area. Cells were then washed with PBS to remove the detached cells and incubated with vehicle (control) or SLE in serum-free DMEM medium. Cell migration was photographed at the 0 and 24 h time points with a digital camera installed on the microscope (Leica, Germany). Five images were taken of each well. In vitro cell invasion assay {#Sec18} ------------------------------ Cell invasion was determined using BD BioCoat^TM^Matrigel^TM^ invasion chamber (24 well plate, 8 mm pore size) according to the manufacturer's instructions^[@CR43],\ [@CR44]^. Invasion chambers were pre-hydrated for 1 h. 0.75 mL of DMEM with 10% FBS was added into the lower chamber. 1.5 × 10^5^ cells in 500 μL of DMEM-0.1% BSA with SLE or vehicle (control) were placed in the upper chamber. After incubation at 37 °C for 24 h, chambers were washed with PBS. The matrigel and cells remaining in the upper chamber were discarded by scrubbing with a cotton swab. Afterward, cells on the lower surface of the membrane were fixed with 4% paraformaldehyde and stained with crystal violet. Each chamber was washed 3 times with distilled water. Invaded cells (cells on the lower surface of the membrane) in five microscope areas (200 × magnification) were counted and imaged by a microscope (Leica, Germany). Western blot analysis {#Sec19} --------------------- Whole cell lysates were prepared from cultured cells and melanoma tissues. Standard western blotting assay was performed as previously described^[@CR41]^. Immunoreactive bands were visualized using the Enhanced chemiluminescence (ECL) detection system (Invitrogen, Carlsbad, CA, USA). Subcellular fractionation {#Sec20} ------------------------- Cells were washed once with cold PBS and scraped into 400 μL of hypotonic lysis buffer^[@CR23]^. After incubation on ice for 15 min, 12 μL of Nonidet P-40 (NP-40) (10%, v/v) was added and cells were kept on ice for another 10 min. The supernatants were collected as the cytoplasmic extracts after centrifugation at 14,000 rpm at 4 °C for 1 min. The nuclei pellets were then washed once with hypotonic buffer and incubated on ice in high salt buffer for 30 min^[@CR23]^. After centrifuging at 14,000 rpm at 4 °C for 10 min, the lysates were taken as nuclear fractions. Immunofluorescence assay {#Sec21} ------------------------ A375 cells were seeded on glass coverslips and incubated without or with SLE for 48 h. After treatment, cells were fixed with 4% formaldehyde in PBS for 15 min, permeabilized with ice-cold 100% methanol for 10 min and blocked in blocking buffer. Then the coverslips were incubated overnight with specific primary antibody against STAT3 (1:1600) at 4 °C. After washing with PBS, coverslips were incubated with fluorescein isothiocyanate (FITC)-conjugated secondary antibody (1:500) at room temperature in darkness for 1 h. Coverslips were then mounted on glass slides using Fluoroshield^™^ with 4′,6-Diamidino-2-Phenylindole (DAPI) (Sigma-Aldrich, USA). Images were viewed and photographed using a fluorescence microscopy (DMI 3000B, Leica, Wetzlar, Germany)^[@CR41]^. Quantitative real-time polymerase chain reaction analysis (RT-PCR) {#Sec22} ------------------------------------------------------------------ Total RNA was extracted with Trizol reagent (Invitrogen, USA) and reverse-transcribed into cDNA using PrimeScriptTM RT reagent Kit (Takara, Japan). Quantitative RT-PCR was performed in triplicate using iTaq™ Universal SYBR Green Supermix (Bio-Rad, USA) with a ViiA 7 Real Time PCR System (Applied Biosystems, USA). Quantification analysis was performed by the comparative C~T~ method^[@CR41],\ [@CR45]^. Relative gene expression was normalized to the endogenous control GAPDH. Stable transfection {#Sec23} ------------------- A375 cells were transfected with pcDNA3-STAT3C plasmid or empty vector using Lipofectamine 2000 (Life Technologies, Inc., Rockville, MD) following the manufacturer's instructions^[@CR40]^. 72 h after transfection, cells were selected with 1 mg/mL of G418 in DMEM with 10% FBS for 14 days. G418-resistant colonies were pooled and regarded as STAT3C-expressing stable cells. Statistical Analysis {#Sec24} -------------------- The data were expressed as mean ± SD and analyzed by one-way analysis of variance (ANOVA) followed by Fisher's Least Significant Difference (LSD) Test using IBM SPSS Statistics Version 20.0 (IBM Corp., Armonk, New York). The difference was considered significant if P \< 0.05. Electronic supplementary material ================================= {#Sec25} Supplementary Dataset Electronic supplementary material Supplementary information accompanies this paper at doi:10.1038/s41598-017-03351-2 Publisher\'s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This work is supported by National Natural Science Foundation of China (81673649), Science, Technology and Innovation Commission of Shenzhen (JCYJ20140807091945050, JCYJ20150630164505508 and JCYJ20160229210327924); Research Grants Council of Hong Kong (12125116); Natural Science Foundation of Guangdong Province (2016 A030313007); and Hong Kong Baptist University (FRG1/15-16/050, FRG2/15-16/020 and FRG2/16-17/033). T.L. performed the research, analyzed and interpreted the data, and drafted the manuscript. X.Q.F. helped in the animal study and revised the manuscript. A.K.T. helped in data analysis and reviewed the manuscript. H.G. helped in the chemical analysis. L.K.W. helped in flow cytometric analysis. B.L., T.S. and X.Y.W. helped in the experimental work. Z.L.Y. designed the study, reviewed the original data and finalized the manuscript. All authors reviewed and approved the final version of manuscript. Competing Interests {#FPar1} =================== The authors declare that they have no competing interests.	{ "pile_set_name": "PubMed Central" }
1.. Introduction {#s1} ================ Making informed decisions about the appropriate focus of conservation investment has become a central theme of both academic research and conservation action. The discipline has been driven by both academia and conservation by the often-cited 'agony of choice'---in essence, how to make the best decision with limited resources \[[@RSTB20140013C1]\]. A variety of methods have been proposed for prioritizing those species most deserving of conservation attention; one candidate that is gaining traction prioritizes those species that maximize phylogenetic diversity (PD), or the sum of the edge lengths of the phylogenetic tree linking species \[[@RSTB20140013C2]\]. Because time and divergence are likely to be correlated, species representing a greater proportion of independent evolutionary time on the tree represent more total evolution and so should be prioritized. Therefore, using metrics to assess and prioritize PD attempts to account for the fact that species differ substantially in the amount of complementary genetic information that they embody \[[@RSTB20140013C3]--[@RSTB20140013C5]\]. Such techniques are a practical way of accounting for the unequal contribution of species of conservation concern to biodiversity conservation. Building on this recognition, species-level measures that integrate evolutionary history and likelihood of loss of a given species have been developed over the past two decades \[[@RSTB20140013C2],[@RSTB20140013C5]--[@RSTB20140013C7]\]. The field of research has moved rapidly, with the consequent translation of theory \[[@RSTB20140013C1],[@RSTB20140013C4],[@RSTB20140013C8]--[@RSTB20140013C10]\] into conservation practice \[[@RSTB20140013C6],[@RSTB20140013C11],[@RSTB20140013C12]\]. Despite this welcome development of global-scale metrics for species prioritization, the majority of conservation action is taken not under global level priority setting schemes, but at a local scale, typically at the level of ecological communities and populations. Understanding change in population and community-level metrics of diversity over time is critical to understanding change in biodiversity. The loss of populations from communities is a prelude to species extinction \[[@RSTB20140013C13]\], and local reductions in taxonomic, genetic and functional diversity impact many different aspects of biodiversity. Whether and how global-scale metrics based on evolutionary history map to a local scale has received far less scrutiny to date than the development and implementation of global prioritization methods. The scale of distinction could be important. At a local scale, conservation-relevant assemblages of species are likely to be made up of relatively few species, possibly spread across a large section of the phylogenetic tree. This phylogenetic spread means that one or a few species may contribute a great deal to local PD, making the choice of priority setting algorithm important. A local perspective on conservation phylogenetics intersects with a related field as well: as phylogenetic relationships among species has rapidly become easier and more reliable, community ecologists have become increasing interested in the evolutionary relationships among coexisting species for evaluating clustering, community assembly, functional differences and for mechanistic insights into community structure \[[@RSTB20140013C14]\]. Importantly, community ecologists have also presented provocative evidence that the total PD in an assemblage is a good predictor of ecosystem function \[[@RSTB20140013C15]\]. In this article, we seek to take a step towards integrating community-level phylogenetics with global conservation approaches for prioritizing evolutionary history. We ask to what extent global species-centred metrics of evolutionary isolation \[[@RSTB20140013C6]\] are useful for conservation priority setting at the community level, specifically asking whether species that score highly for global metrics also score high when the metrics are measured on the tree spanning the species in a community assemblage and whether these species also contribute substantially to the total PD and trait diversity of the local community. We use a rich dataset from communities of Nearctic and Neotropical breeding birds to test the following: 1.---how do global and local metrics of three measures of evolutionary isolation (evolutionary distinctiveness (ED); unique PD contribution (PE) and average pairwise distance (APD)) relate to one another?2.---does prioritizing globally high ED scoring species within a community select species with unusual sets of ecological traits at the local level? and3.---is more local PD and trait variation lost when removing species from a community that score high in global evolutionary isolation compared with random species removal? 2.. Material and methods {#s2} ======================== (a). Community data {#s2a} ------------------- We organized published lists of species in bird communities across the Nearctic and Neotropics. We drew species identity from two vetted long-term datasets: the breeding bird survey (United States) and Christmas bird counts (North and South America). Species lists from all years for the same survey route were collapsed so that all species seen during all years\' surveys were taken to represent the local 'community' along that route. To account for under-sampling, we removed those sites where surveys had recorded fewer than 25 species. We analysed a final dataset of 4628 communities of Nearctic and Neotropical birds containing 2662 species in total (median community size = 94 species, range 26--534). Phylogenetic and trait data were derived from previously published sources. We downloaded 10 000 full species-level phylogenies from \[[@RSTB20140013C12],[@RSTB20140013C16]\] and compiled trait data from \[[@RSTB20140013C17]\] for all surveyed species. We chose ecological traits based on their suitability for explaining ecological separation, following \[[@RSTB20140013C18]\]. Traits included body mass, primary habitat, habitat breadth, vertical distribution (ground to aerial), vertical breadth (range between seven levels from ground to aerial), diet (for each species, relative amount of food types consumed add up to 10), diet breadth, guild, social structure, nest type (14 types), nest substrate (11 types), mean clutch size and activity pattern. All quantitative data were standardized by their respective mean. We recognize that these are relatively few ecological variables from a large possible set; however, we do not know how much extra (orthogonal) information would be added by including more than this ecologically significant subset. We matched species among the community, phylogenetic and trait datasets using Avibase as a reference \[[@RSTB20140013C19]\]. Those species that were found in the community data but not in the phylogeny were dropped; in the case of sub-species, we simply substituted the parent species in the phylogeny. (b). Analyses {#s2b} ------------- We calculated the median of each of three measures of evolutionary isolation across the 10 000 phylogenies, each at three scales: the length of the terminal branch \[[@RSTB20140013C20],[@RSTB20140013C21]\] linking a species to the tree (unique PD contribution, also called phylogenetic endemism or pendant edge (PE) cophenetic, R package ape \[[@RSTB20140013C22]\]); the APD, i.e. the mean APD to all other species in the tree (cophenetic, R package ape \[[@RSTB20140013C22]\]) and the fair proportion measure of ED (ed.calc, R package caper \[[@RSTB20140013C23]\]). These metrics include the two ends of an axis of evolutionary isolation measures that weights information nearer the root (APD) or nearer the tip (PE \[[@RSTB20140013C24]\]), alongside the only currently used measure in active conservation prioritization (ED: \[[@RSTB20140013C6],[@RSTB20140013C11]\]). For each species in each community, we calculated two 'global' scores, one based on the entire tree of the birds and another set of scores based on a continental level tree consisting of all the species in our dataset. Differences between the two sets of ED and PE scores were minimal (Spearman\'s rank correlation ρ = 0.975--0.999) and, given that we had only trait data for all species in our North American communities, hereon we used the scores calculated using the continental tree; for brevity, we refer to these as 'global' scores. Then, for comparison, we calculated the same three measures of isolation for each community using just the sub-tree linking the species in that community ('local' scores). We calculated trait-based metrics based on distances between all species in the surveyed communities using the Gower approach (daisy R package cluster \[[@RSTB20140013C25]\]). Any species pairs that had no data to compare were awarded the median distance for all other species. Minimum distance to any other species in the global tree (trait uniqueness, TU) and mean APD to all other species in the global tree (trait APD (TAPD)) were then calculated directly from the resulting distance matrix. The same approach as above was then applied to each local community separately, resulting in local values for these two measures. We then produced datasets with and without species known to be alien invasives \[[@RSTB20140013C26]\] and recalculated all local scores. To investigate the role of abundance, we also produced a dataset after removing all species that accounted for less than 1% of all individuals seen during the course of all surveys at the location, again recalculating local scores. Using all three datasets (full, no invasives, rare species removed), we calculated a set of rank correlation (Spearman\'s ρ; cor, R package base \[[@RSTB20140013C27]\]) for each community, between global scores (APD, ED and PE) and all versions of local scores (i.e. 'local' APD, ED, PE, TU and TAPD). To visualize the variation in the correlation between local and global scores across communities, we mapped the Spearman\'s ρ values onto the survey point locations in our study area. The observed pattern was spatially non-random ([figure 2](#RSTB20140013F2){ref-type="fig"}). To further explore this variation among communities, we chose one of the scores, ED, and modelled the correlation between local ED and global scores using linear mixed-effects models that account for the spatial autocorrelation between the survey points and the variables: years surveyed, latitude, species richness (total species seen), PD/species richness (total community PD divided by total species seen, as a measure of dispersal across the tree), distance to coast (measured as kilometres from Global Self-consistent Hierarchical High-resolution Geography coastlines <http://www.ngdc.noaa.gov/mgg/shorelines/gshhs.html> using R spDist function) and habitat class (data taken from Globcover 2009 calculated using R function over). We chose ED as it is the only measure of evolutionary isolation so far (to our knowledge) used actively in global conservation \[[@RSTB20140013C6],[@RSTB20140013C11]\]. Diagnostic analyses (not reported) confirmed a spatial influence and by trial and error a model using spherical distribution had the best Akaike information criteria (AIC) score. AIC scores for the full model were calculated and terms removed manually, starting with the least correlated variable until AIC did not improve. Finally, we ran a simulation to test the effect of using global conservation priority setting on capturing biodiversity at the community level. We first ranked all species by their three respective global evolutionary isolation scores. We selected the top scoring 500 species for each global metric and determined how many of these top-ranking species were in each community. We then removed these species from each community they were found in and measured biodiversity change in terms of mean reduction in local PD and change to mean average trait pairwise difference (measured using TAPD). The latter value measures how closely related species in a group are on average, with a reduction in this value occurring if species with more unusual traits being removed, i.e. the remainder is then more similar. For a comparison, we then removed the same number of species randomly from each community (replicated 500 times), to create an expectation of PD lost and TAPD change given random loss of species from communities. 3.. Results {#s3} =========== (a). Phylogenetic diversity {#s3a} --------------------------- The correlations between global ED and local ED scores among the 4628 communities of birds were moderate (mean ρ across communities = 0.52, s.d. = 0.09). Correlations were marginally higher for global ED versus local PE (mean ρ = 0.59; s.d. = 0.09). There was a weaker relationship between global PE and local PE (mean ρ = 0.31, s.d. = 0.13). Surprisingly, the global APD was very closely related to local APD ([figure 1](#RSTB20140013F1){ref-type="fig"}; mean ρ = 0.96, s.d. = 0.04, 98% of communities significant at the 0.05 level) and global APD also covaried strongly with local ED (mean ρ = 0.79, s.d. = 0.15) and to a lesser extent with local PE (mean ρ = 0.50, s.d. = 0.17), though these latter results had relatively higher variance. Figure 1.Mean correlation between three global evolutionary isolation scores (ED, PE and APD---definitions in text) and local scores measured at the community level for 4628 bird communities across North and tropical South America. On the whole, the removal of alien species from the communities had a negligible effect on the correlations between global ED and local ED. Removing rarely seen (vagrant) species also had a limited but slightly positive effect on the majority of correlations. The strength of the correlation between global ED and local ED was spatially non-random ([figure 2](#RSTB20140013F2){ref-type="fig"}), appearing to be strongest in the western Nearctic, and in central areas and towards northern limits of where bird communities were surveyed. There were especially strong relationships for communities in the central prairie regions of the United States as well as the northern boreal forest communities, and communities from the tropical wet forests in parts of South America, particularly Brazil (though for the latter, numbers of communities were low). Conversely, global APD showed weaker associations with local ED and local PE in the central prairie regions, and stronger correlations in eastern forest communities ([figure 2](#RSTB20140013F2){ref-type="fig"}). Figure 2.Map of correlations between local evolutionary isolation scores (l) and global evolutionary isolation scores (g) for 4628 bird communities for three different measures of evolutionary isolation---descriptions in text. Blue colours are higher correlation (light blue is a coefficient of approximately 0.7; dark blue represents a value tending to 1), orange-red colours show weaker correlation (red represents a correlation coefficient of 0 to around 0.4; orange around 0.55). (Online version in colour.) These biogeographic patterns were apparent even when accounting for other factors, such as species richness and survey effort ([table 1](#RSTB20140013TB1){ref-type="table"}). Areas near the coast, and in forested habitats and grasslands had higher correlations between local and global ED scores, though the lack of differentiation in the land cover data into forest types, e.g. eastern (mainly deciduous) forests and boreal (mainly coniferous) forest, prevented more detailed comparisons ([table 1](#RSTB20140013TB1){ref-type="table"}). Table 1.Regression coefficients from a linear mixed-effects model of the correlation coefficient between global ED scores and local ED scores based on six significant explanatory variables. (n = 4628 Nearctic and Neotropical bird communities.)global versus local ED corr. predicted byβ estimateplatitude0.0010.0001species richness0.120.0001PD/species richness−0.010.26years surveyed−0.00070.003distance to coast−0.000010.0001coastal (habitat)0.0150.02cropland (habitat)0.0010.76flooded (habitat)0.0050.72forested (habitat)0.0160.0001urban/bare (habitat)−0.010.35 Importantly, in spite of the moderate relationships between global and local measures of evolutionary isolation, nearly twice as much local PD was lost when removing the top 500 high scoring most evolutionary isolated species from each community compared with the removal of species at random (mean p \< 0.03 for ED, PE and APD, n = 4628, multiple testing accounted for using false discovery rate control R \[[@RSTB20140013C28]\]; [table 2](#RSTB20140013TB2){ref-type="table"}). Table 2.Mean loss of biodiversity value when high global scoring species (in global top 500) are lost from each of 4628 communities. (PD value represents the (total community PD minus PD of community when all of the top 500 species are removed). Random represents the loss when an equal number of species are removed randomly. Mean pairwise distance (trait) value represents change in mean pairwise trait distance from the unaltered community to the community with all of the top 500 species removed. Positive values represent more closely related species in depleted community, and zero represents no change in the average relatedness of communities.)metricbiodiversity measuretop 500randompAPDPD1080.39446.67\>0.001EDPD556.74264.37\>0.001PEPD489.99294.870.03APDmean pairwise trait distance (TAPD)0.0320\>0.001EDmean pairwise trait distance (TAPD)0.01200.08PEmean pairwise distance (TAPD)0.00800.14 (b). Trait scores {#s3b} ----------------- We found a weak positive correlation (ρ = 0.21) between ED score and TU when calculating both sets of scores globally. For the other two global measures of evolutionary isolation, APD was more strongly correlated to global TU (ρ = 0.43), while global PE was only weakly related to globally calculated TU (ρ = 0.09). Similar patterns were seen with TAPD (results not shown). Alternatively, if we compare the correlation of the global phylogeny-based scores (ED, PE and APD) to TU and trait pairwise difference calculated at the community level, a different relationship is revealed. At this smaller spatial scale, only one of the globally calculated evolutionary isolation metrics, APD, remains positively correlated to our community based trait measures (TAPD ρ = 0.42, s.d. = 0.09; TU ρ = 0.38 s.d. = 0.13; [figure 3](#RSTB20140013F3){ref-type="fig"}). Global ED scores and global PE scores were unrelated or negatively related to local TU ([figure 3](#RSTB20140013F3){ref-type="fig"}). Removing alien species and vagrant species from the analyses had minimal impact. Figure 3.Mean correlation between three global evolutionary isolation scores (ED, PE, APD---definitions in text) and local trait minimum distance to other species in the community (local TU) and average trait pairwise distance to other members of the community (local TAPD) for 4628 bird communities across North and tropical South America. Removing the top 500 high scoring global APD species from all communities resulted in reduction in an average mean trait pairwise distance of 8%, meaning that the remaining species tended to be more similar in terms of their functional traits compared with the unperturbed community. This difference was much higher than expected from random species removal (0; p \< 0.001; [table 2](#RSTB20140013TB2){ref-type="table"}). Removing the globally high scoring ED and PE species had the same effect in reducing the mean trait relatedness of the community but not significantly more than random removal ([table 2](#RSTB20140013TB2){ref-type="table"}). 4.. Discussion {#s4} ============== There appears to be potential for close agreement between different metrics used to create conservation priorities at different spatial scales for Nearctic and Neotropical birds, but this relationship varies widely among communities and depends on the precise metric used. While metrics to measure different facets of evolutionary isolation have proliferated over the past two decades, the principal metric employed in active global-scale conservation thus far has been ED \[[@RSTB20140013C6]\]. However, we find that the global version of this metric had varying effects. Prioritizing evolutionary isolated species using ED safeguarded local PD on average (retaining almost twice the local tree compared with losing species on average), but the effect on prioritizing overall local trait variation showed no strong trend. The application of this global-scale metric of evolutionary isolation to community-scale conservation may therefore depend on what one wants to conserve. To date, there has been little evaluation of the relative benefits of using ED versus other metrics that are available, and also little evaluation of the properties that might matter in selecting an evolutionary isolation score for conservation \[[@RSTB20140013C24]\]. Our results, however, suggest that an as-yet-unused and unproven metric of evolutionary isolation, global APD, is strongly related not only to its local community-level version but also to community-level TU. Therefore, even when setting conservation priorities for an entire clade at broad spatial scale, this metric will preferentially choose species that, within their community, have unusual sets of traits. Furthermore, APD is also the most effective metric at capturing community PD (despite performing worse than ED in global-based analyses \[[@RSTB20140013C29]\]). If these community-level properties of APD are considered desirable, our results suggest APD may be a useful metric for setting conservation priorities across spatial scales, although more work is needed to confirm this. A significant downfall of the APD metric approach is that little is known about how it relates to evolutionary processes and this represents a significant barrier to it being recommended as a conservation tool in the future. Interestingly, the only process-based evolutionary isolation metric known to the authors, character rarity \[[@RSTB20140013C30]\], which attempts to model genetic information on phylogenies under different models of evolution, produces values that are almost identical to APD with its default parameter setting. By altering these settings and comparing outputs, character rarity might prove to be a useful tool to offer insights into what APD is measuring. Finally, as there is double counting of branches when calculating scores, APD may also prove to have analytically undesirable properties. Two of the metrics examined here (ED and PE) principally or wholly are measuring the length of species\' terminal branch \[[@RSTB20140013C24]\]. One benefit of taking such an approach is that it is conceptually simple: this terminal branch represents all of the features that have evolved since a species split from its nearest extant relative. However and importantly, in large and incomplete phylogenies, the terminal branches are often likely to be incorrect: for instance, in the phylogenies used thus far to set conservation priorities for the Evolutionarily Distinct Globally Endangered (EDGE) programme \[[@RSTB20140013C11],[@RSTB20140013C12],[@RSTB20140013C31]\], species from data-poor species groups have either been awarded all the same terminal branch or a simple evolutionary model has been used to roughly estimate branch lengths. The third metric evaluated in this study (APD) takes into account all branches in the tree by averaging the distance along the tree from a target species to all the other species in the phylogeny. Any occasional incorrect branch lengths (particularly the terminal branch) will, therefore, provide only a limited amount of incorrect information to a final metric value. Using a metric that is not so strongly reliant on the terminal branch may also have other benefits associated with it. Most changes to taxonomic identity and phylogenetic relationships are likely to occur in localized areas of the tree, for example, through taxonomic revisions in a particular genus or family. Once a reasonable phylogeny is established for a given group \[[@RSTB20140013C16],[@RSTB20140013C32]\], very large changes in the hypothesized relationships among species are less likely \[[@RSTB20140013C8]\]. It then follows that metrics of evolutionary isolation that are less affected by tip-level phylogenetic data should be less susceptible to changes in conservation rank (e.g. species being elevated to higher categories of risk, or being down-listed as a result of concerted conservation intervention \[[@RSTB20140013C11],[@RSTB20140013C33]\]). Interestingly, APD is commonly used to measure the overall relatedness of communities \[[@RSTB20140013C14]\]. How this approach links to the distribution of species-level measures could be an interesting avenue to explore in a conservation context. Previous work has shown support for the hypothesis that high ED scoring mammal species are moderately ecologically unusual \[[@RSTB20140013C11],[@RSTB20140013C34]\], supporting the use of ED as a conservation metric. The moderate correlations we report here are consistent with these earlier studies. Notably, these previous studies were carried out at a clade level (i.e. comparing species within families and orders), which makes intuitive sense for a global-scale scheme of conservation priority setting for an entire taxon. However, species tend broadly to interact at the community level, where species from a number of clades may also interact. Here, we consider the situation where conservation priorities are set at the global level and show that there is no link between global ED and unusual local trait distinctiveness or complementarity. By contrast, global APD again is a fairly good surrogate for both. Furthermore, species-level APD mirrors previous work examining the functional 'originality' \[[@RSTB20140013C35]\] of species at the community level, suggesting a potential to develop analytical links between phylogenetic and ecological conservation priorities. Certainly, if such a link can be made, then the APD metric may have some appealing properties: being able to capture aspects of ecosystem function while setting priorities at the global level using data-cheap approaches would be a valuable goal. However, this remains untested. We observed strong spatial patterns of correlation between metrics among the communities of Nearctic and Neotropical bird species tested. The strongest correlations between global and local measures of all three metrics were observed in central continental areas, the tropics and towards northern limits of where bird communities were surveyed, and particularly for communities in the central prairie regions of the United States as well as the northern boreal forest. While we did not start out with a testable prediction of geographical patterns of the correlations of global and local values for the three metrics that we considered, we offer some observations. In the communities with strongest correlations (typically those in north and central United States), birds tend to have larger global ranges on average than those species in tropical and sub-tropical communities \[[@RSTB20140013C36]\]. This may mean that, at each location, tropical communities are more likely to just sample just one or two species from the many large tropical bird clades, e.g. antbirds (Thamnophilidae). In addition, these communities with strong correlations tend to have more species from distinct but widespread guilds such as raptors, waterfowl and waders. By contrast, most tropical bird communities in our analysis, perhaps especially those on Christmas Bird Count routes (the design of which is not systematic), are tropical forest and woodland communities, which are almost entirely dominated by passerines and near passerines. This general difference suggests that the presence of a few species with a high isolation score (generally non-passerines) could weakly affect rank correlations. Further, previous work has also suggested areas in central North America have higher numbers of species with top 10% highest ED scores than surrounding areas \[[@RSTB20140013C12]\]. An alternative driver here may be owing to some type of ecological filtering, such that some communities with a high correlation coefficient between global and local metrics of evolutionary isolation are comprised of few relatively complete clades (e.g. waders), such that tip lengths (PE scores) are similar at the two geographical scales. Our analysis of Neotropical and Nearctic birds provides one of the first examples of how relationships between global and local measures of evolutionary isolation might be evaluated. In light of the patterns we report, we need more work, including studies uncovering the relationship between phylogenetic distance and meaningful trait distances at both global and, importantly, local levels. As the use of approaches that help prioritize at risk species that comprise disproportionately large amounts of evolutionary history continues to grow, it is important that local community-scale conservation projects can make use of the many available techniques for evaluating those species which need focused conservation attention. It is feasible that priority setting at a local scale will start to make greater use of these approaches in order to manage and integrate cost savings to projects that have very limited funding resources. However, the next step for conservation must be to determine how such metrics should be applied. C.H.S. is grateful to Sherron Bullens, Debbie Fisher, David Hayes, Beth Karpas, Kathleen McMullen and especially Jason Socci for their dedicated help with the world bird ecology database. We thank Dan Faith and an anonymous referee for comments that improved an earlier version of this article. Funding statement {#s5} ================= C.H.S. thanks the University of Utah and the Christensen Fund for their support. A.O.M. acknowledges NSERC Canada\'s Discovery Grant Programme. A.O.M. and D.W.R. are also grateful for the support offered by sDiv, the Synthesis Centre for Biodiversity Sciences (DFG FZT 118). [^1]: One contribution of 17 to a discussion meeting issue '[Phylogeny, extinction and conservation](http://rstb.royalsocietypublishing.org/content/370/1662.toc)'.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== Anesthesia in obese patients has always been considered a significant problem. The decreased volume of cerebrospinal fluid (CSF) in obese patients can cause spinal anesthesia to expand in the cephalic direction \[[@B1], [@B2]\]. Some positions of surgery and the surgical techniques themselves can cause changes in hemodynamic parameters \[[@B3], [@B4]\]. The tissue concentration of local anesthetic in spinal anesthesia is controlled by the tissue blood flow in addition to the CSF \[[@B5]\]. Transurethral prostate resection (TUR-P), a surgical treatment used in benign prostatic hyperplasia, is carried out in the lithotomy position \[[@B6]\]. This position meaningfully increases the systolic arterial pressure from the subextremities with the effect of autotransfusion \[[@B7]\]. At the same time, during TUR-P, an entrance of liquid occurs into the system that is proportional to the liquid used, the hydrostatic pressure of the solution, the number of opened venous sinuses, the time of the irrigation, the absorption speed, and amount of irrigation liquid \[[@B8]\]. All these factors cause changes in cardiovascular system. The purpose of our study is to determine the effects of hemodynamic changes on the level and time of spinal block that may occur during TUR-P in obese versus nonobese individuals. 2. Patients and Methods {#sec2} ======================= This study was approved by the Institutional Review Board of Dicle University Medical School and all patients provided informed consent. A total of 60 male patients who had undergone TUR-P operations under spinal anesthesia with an American Society of Anesthesiologists (ASA) score of I or II in the Urology Department were included in this prospective study. Patients with BMIs lower than 25 kg/m^2^ constituted the nonobese group (Group N, n = 30) and those with BMIs equal to or greater than 30 kg/m^2^ constituted the obese group (Group O, n = 30). The body mass indices (BMIs) of the patients were calculated by dividing patient weight by height squared in meters (kg/m^2^). Neurological disease, deformities of the spinal column, sensitivity to bupivacaine or other contraindications for spinal anesthesia, and skin infection at the site of injection were defined as exclusion criteria All patients were premedicated with 0.03 mg/kg IV midazolam 30 min before the anesthesia. Patients in both groups were monitored with continuous electrocardiography (ECG), cyclic noninvasive blood pressure (NIBP), and peripheral oxygen saturation (SpO2) in the operating room. All patients were administered 10 mL/kg of lactated Ringer\'s solution before the spinal anesthesia. Dural puncture was performed at the L3-4 interspace using a 25-gauge Quicke spinal needle in the sitting position. 3 mL of 0.5% hyperbaric bupivacaine was injected over 20 s. Patients were immediately placed in a supine position after the spinal anesthesia and then all patients were immediately placed in the lithotomy position. In this study, the standard lithotomy position was used; both thighs were lifted 90° toward the trunk and the lower legs were hung on poles with ties. The operating table was in the horizontal position. Oxygen was given at 2L/min via nasal cannula during surgery. Analgesia was defined as the inability tosense pinprick. Success of spinal anesthesia was defined as a bilateral T10, sensory block to pinprick within 15 min of intrathecal drug administration. Motor block in the lower limbs was classified using the Bromage Scale \[[@B9]\]: 0 = ability to lift an extended knee at the hip; 1 = ability to flex the knee but not to lift an extended leg; 2 = ability to flex toes only; and 3 = inability to move hip, knees, or toes. Sensory and motor block assessments weremeasured and then recorded at 2, 4, 6, 10, 20, and 30 min afterthe lithotomy position and after intrathecal drug administration at 120 min by an assistant. Sensory and motor block at 6 and 120 min (Marcaine Spinal Heavy, time to onset of effect in minutes: 5--8 min, duration of effect in hours: 2-3 h for urological surgery \[[@B10]\]) and the levels of peak sensory and motor block were compared between the groups. The patient\'s systolic blood pressure (SBP), diastolic blood pressure (DBP), mean arterial blood pressure (MAP), heart rate (HR), and peripheral oxygen saturation (SpO2) were measured and then recorded at five-minute intervals from there after the lithotomy position for 30 min and then at ten-minute intervals until the end of surgery. Systolic blood pressure (SBP), heart rate (HR), and peripheral oxygen saturation (SpO2) were compared between the groups. Hypotension was defined as SBP less than 70% of the baseline value or less than 90 mmHg. When hypotension occurred, repeated intravenous ephedrine bolus doses of 5 mg were administered. Bradycardia was defined as HR less than 60 beats/min. If the heart rate was \<60 beats/min, atropine (0.5 mg) was administered. Nausea and vomiting events were recorded. Intraoperative nausea and vomiting were treated with intravenous metoclopramide. The data were analyzed using the SPSS 16.0 for Windows (SPSS Inc., Chicago, IL, USA) packet program. Descriptive statistics of the demographic data and constant variables were indicated as mean ± standard deviation. Data were analyzed using Student\'s t-test depending on the normality of data. Chi-square test was used to analyze incidence data. Parametric repeated data were evaluated by a repeated measures ANOVA test. The value of P \< 0.05 was accepted as statistically significant. 3. Results {#sec3} ========== No significant differences were noted between the two groups with respect to age, height, duration of surgery, or male/female ratio. The two groups differed with regard to weight and BMI, as expected ([Table 1](#tab1){ref-type="table"}). Basal hemodynamic parameters were similar between the groups. Cardiovascular responses are shown in [Table 2](#tab2){ref-type="table"}. In both groups, SBP, 5 min after spinal block, decreased from the baseline value. Systolic blood pressure (SBP) values were measured at 10, 15, and 20 min after the lithotomy position were significant increased for Group N compared to Group O (P \< 0.001; P \< 0.001; P \< 0.05, resp.) ([Figure 1](#fig1){ref-type="fig"}). HR values were similar between the groups. Sensory and motor block levels are shown in [Table 3](#tab3){ref-type="table"}. Peak sensory and motor block levels and 6th minute sensory and motor block levels were similar between the groups. The 120th minute sensory and motor block levels were higher for Group O than for Group N (P = 0.017; P = 0.008, resp.). No significant intergroup differences were observed with respect to adverse effects ([Table 4](#tab4){ref-type="table"}). Comparisons made between both groups revealed no significant differences for SpO2 values. No complications were encountered for any of the patients. 4. Discussion {#sec4} ============= Many investigators have stated that obesity affects the time and level of spinal anesthesia \[[@B1], [@B12]\]. Pressure on the inferior vena cava, which is dependent on increased intra-abdominal pressure, causes distension in the lumbar plexus \[[@B13]\]. This distension can cause a decrease in the volume of CSF, which in turn can affect the level and time of spinal block. In addition, according to different viewpoints, the increased epidural adipose tissue presses on the dural sac in obese patients and causes a decrease in the volume of CSF. Previous studies on obese patients have been conducted with the patients in the supine and lateral decubitus positions \[[@B14]--[@B17]\]. However, some surgical positions, such as the head-down tilt and lithotomy positions, may affect arterial tension in the patients. The resulting changes in blood flow can affect the pharmacokinetics of local anesthetics. In our study, which differed from previous studies, we compared the effects of hemodynamic changes occurring in patients who underwent TUR-P in the lithotomy position on the block levels and times in obese and nonobese patients. In the lithotomy position, a volume of blood of approximately 500--1000 mL passes from the lower extremity to the central circulation as an effect of autotransfusion \[[@B7]\]. Similarly, during transurethral prostate resection (TUR-P), the absorption of irrigation solution in small amounts from the veins in the resection area can reach dimensions that may threaten life depending on the speed of the absorption \[[@B8]\]. In our opinion, blood and irrigation solutions that pass to the central circulation cause fullness in the epidural veins, so that decreases in the volume of CSF and spinal anesthesia time and level may occur in obese patients as well as in nonobese patients. A study conducted by Miyabe and colleagues \[[@B7]\] showed that SBP increased meaningfully in patients in the lithotomy position compared to the supine position. In our study, the SBP value was meaningfully higher in the nonobese group than in the obese group. Although the difference between the peak block levels was not statistically significant, an unexpected decrease was seen in the obese group. The reason for this may be attributed to the presence of increased intra-abdominal pressure to the inferior vena cava in obese patients and the prevention of venous return. At the same time, the movement of intra-abdominal organs in an upward direction due to the lithotomy position may have increased the pressure on the vena cava and more strongly prevented venous return. The venous circulation of the prostate is realized by vesicle veins and the internal iliac vein. These vessels drain into the vena cava inferior. In obese patients, abdominal pressure increases due to the weight of abdominal content. Intra-abdominal pressure increases the pressure on inferior vena cava in parallel and decreases the venous return to the heart, thereby resulting in reduced cardiac output \[[@B18]\]. In addition to the lithotomy position, the occlusion force on the inferior vena cava, which is located retroperitoneally, is increased with the effect of gravity due to the weight of the abdominal contents. This increase in the hydrostatic pressure in the inferior vena cava reduces the amount of irrigation solution that joins systemic circulation in obese patients compared to nonobese patients and TUR-P may be less effective in obese patients on hemodynamics. No statistically significant difference was noted during the 6th minute after the lithotomy position in the sensory and motor block evaluation following spinal anesthesia or in the intergroup comparison of the block peak level. However, previous studies \[[@B2], [@B13]\] have shown that the volume of CSF decreases due to the increased intra-abdominal pressure in obese patients and it may cause deeper blocks because of the decrease in the dilution of local anesthetic. A study conducted by Carpenter and colleagues \[[@B2]\], conducted using a magnetic resonance monitoring technique, showed that lumbosacral CSF volume is important in the distribution of spinal anesthesia and in the peak action time. The probable reason for our patients\' failure to show any difference in block level may be due to the increased arterial pressure leading to fullness in epidural veins in nonobese patients and a decrease in the volume of CSF. The sensory and motor block levels at 120 min following the spinal anesthesia were statistically higher in the obese patients than in the nonobese patients. In spinal anesthesia, the local anesthetic, which is given at the subarachnoid interval, is absorbed by the veins in this interval and eliminated and diffused from the arachnoid or dural membranes. The elimination of local anesthetics is affected by tissue blood flow, where high blood flow increases the elimination of local anesthetics. Because the flow of blood is much higher at the anterior aspect of the spinal cord, the elimination of local anesthetics is faster \[[@B5]\]. In our study, the low level of sensory and motor block levels at the 120th minute in nonobese patients, when compared to obese patients, may have been caused by an increase in blood flow and a subsequent increase in arterial pressure in the spinal cord and the elimination of local anesthetic. Previous studies conducted with the use of isobaric bupivacaine in the supine position showed a meaningful expansion in the cephalic direction in obese patients when compared to nonobese patients \[[@B1], [@B12], [@B14]\]. However, other studies conducted with hyperbaric bupivacaine in the supine position detected no statistically significant differences in peak sensory and motor block levels and times \[[@B15], [@B16]\], contrary to our hypothesis. However, studies conducted with methods of monitoring showed that radio opaque materials expand more in the cephalic direction in the supine position and in obese patients when compared to nonobese patients \[[@B2], [@B13]\], in agreement with the findings of our study. The volume of CSF may show individual differences and CSF volume is difficult to predict when relying on length, weight, and BMI \[[@B13], [@B19]\]. In our study, our patients had greater average age and height than those in the other studies \[[@B15], [@B16]\]. These factors may have caused the difference in the outcomes. In the same study \[[@B15], [@B16]\], the calculated sensory block time was longer in obese patients. In our study, the sensory and motor block levels were significantly higher in obese patients at the 120th minute. However, another topic which must be taken into consideration is the decrease in the elimination of local anesthetic by decreased blood flow, in addition to the individual differences in CSF volume. In obese patients, increased intra-abdominal pressure also prevents the venous return to the heart in addition to being the reason for the fullness in the lumbar venous plexus due to pressure on the vena cava. It may be the reason for the elongation in block time in obese patients when compared to the nonobese patients. Consequently, lithotomy position and TUR-P meaningfully increased the systolic blood pressure in nonobese patients when compared to obese patients. The increase in hemodynamic parameters increases the blood flow in the spinal cord and may form sensory and motor block levels in nonobese patients that are similar to those in obese patients. However, the increased blood flow enhances the elimination of local anesthetic and may lead to earlier retreatment time for spinal anesthesia. Ethical Approval ================ This study was approved by the Institutional Review Board of Dicle University Medical School (Trial Registry no. 79, Date: January 22, 2014). Conflict of Interests ===================== The authors declare that there is no conflict of interests regarding the publication of this paper. ![After lithotomy position systolic blood pressure (SBP) (between Group O and Group N); ^∗^ P : 0.001 10 min and 15 min, ^‡^ P: 0.05 20 min.](BMRI2015-453939.001){#fig1} ###### Demographic characteristics of obese and normal weight patients and operation duration. ----------------------------------------------------------------------------------------- Obese patients \ Normal weight patients \ P (n = 30) (n = 30) ----------------------------- ------------------ -------------------------- ------------- Age, years 62.4 ± 13.0 56.5 ± 16.3 0.290 Body weight, kg 92.9 ± 8.2 68.2 ± 5.6 \<0.001^\^ Height, cm 169.3 ± 8.8 174.9 ± 4.6 0.452 Body mass index, kg/m^2^ 32.4 ± 3.7 22.3 ± 1.7 \<0.001^\^ Operation duration, minutes 41.0 ± 21.4 50.9 ± 23.9 0.151 ----------------------------------------------------------------------------------------- ^∗^Extremely significant at the level of P \< 0.001. ###### Hemodynamic data. Group Minutes after the lithotomy position --------- -------------------------------------- --------------- ------------------- ------------------- ------------------ --------------- -------------- Group N SBP 130.83 ± 16.9 115.17 ± 20.1 123.61 ± 19.3^\^ 118.67 ± 12.7^\^ 106.49 ± 16.8^‡^ 102.71 ± 15.1 96.31 ± 17.8 HR 90.07 ± 14.3 87.43 ± 12.4 80.23 ± 8.1 75.50 ± 11.8 74.48 ± 10.1 72.36 ± 14.6 70.91. ± 8.4 Group O SBP 124.49 ± 18.7 109.49 ± 20.6 104.78 ± 12.3 102.33 ± 15.2 98.87 ± 13.8 97.77 ± 15.8 91.39 ± 14.5 HR 86.45 ± 16.5 83.11 ± 3.7 76.63 ± 12.2 72.27 ± 14.1 70.87 ± 9.3 70.19 ± 10.7 67.66 ± 14.3 Changes in systolic (SBP) and heart rate (HR), in the obese and nonobese groups. ^\^Extremely significant at the level of P* \< 0.001, ^‡^Significant at the level of P \< 0.05. ###### The levels of sensory and motor block. ----------------------------------------------------------------------------------------------------------- Obese \ Normal weight \ P patients \ patients \ (n = 30), n (%) (n = 30), n (%) ----------------------------------------------- --------------------- --------------------- --------------- Sensory block 120th minute level T4 0 (0.0) 0 (0.0) 0.017 ^‡^ T6 7 (23.3) 0 (0.0) T8 10 (33.3) 8 (26.7) T10 10 (33.3) 14 (46.6) T12 3 (10.0) 8 (26.7) 6th minute level T4 0 (0.0) 0 (0.0) 0.410 T6 1 (3.3) 0 (0.0) T8 3 (10.0) 5 (16.7) T10 9 (30.0) 5 (16.7) T12 17 (56.7) 20 (66.6) Motor block 120th minute Bromage 0 0 (0.0) 0 (0.0) 0.008 ^‡^ 1 9 (30.0) 21 (70.0) 2 15 (50.0) 6 (20.0) 3 6 (20.0) 3 (10.0) 6th minute Bromage 0 9 (30.0) 12 (40.0) 0.714 1 16 (53.3) 14 (46.7) 2 5 (16.7) 4 (13.3) 3 0 (0.0) 0 (0.0) The maximum levels of sensory and motor block Maximum Bromage Scale 0-1-2-3, (n) 0-3-17-10 0-4-19-7 0.676 Peak level of sensory block T7 (T4--T10) T7 (T4--T10) 0.753 ----------------------------------------------------------------------------------------------------------- ^‡^ P \< 0.05 (between Group O and Group N) 120th minute the levels of sensory and motor block. ###### Incidence of adverse effects. ---------------------------------------------------- Adverse effects Group O\ Group N\ P n (%) n (%) ----------------- ---------- ---------- ------------ Hypotension 5 (16.6) 2 (6.6) \>0.05 Bradycardia 2 (6.6) 1 (3.3) Vomiting 5 (16.6) 4 (13.3) Nausea 7 (23.3) 5 (16.6) ---------------------------------------------------- [^1]: Academic Editor: Patricia Khashayar	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-biomolecules-07-00080} =============== The human body is exposed to harsh environmental conditions and various infectious diseases. Infectious diseases are the global cause of mortality and morbidity. The host's immune system plays a vital role in protection. There are various levels of immunity such as innate factor, adaptive immunity, and anatomical and physiological barriers. Innate immunity can be further categorized into humoral immunity and cellular mechanisms. In addition, a part of innate factor is build up from a broad category of antimicrobial peptides (AMPs). In the past decades, various innovations and research methods have used antimicrobial peptides as a tool to combat against intruding pathogens, and, hence, are renowned as natural antibiotics \[[@B1-biomolecules-07-00080]\]. The in vitro experiments using AMPs displayed a wide range of antimicrobial activity including antibacterial, antifungal, and antiviral activity against the offending organism \[[@B2-biomolecules-07-00080]\]. These peptides provide defence against the offending organism at the initial stages. Recently, AMPs have been shown to have wound healing potential, alteration potential of the host gene expression, and the ability to induce cytokines production, all of which fall under the category of immunomodulatory function. This has been described as their indirect role in the defence \[[@B2-biomolecules-07-00080],[@B3-biomolecules-07-00080]\]. The immunomodulatory properties of human AMP are to reduce the level of inflammatory cytokines, help in wound healing, leukocytes activation, and macrophage differentiation \[[@B4-biomolecules-07-00080]\]. AMPs are found in both prokaryotes and eukaryotes. In mammals, two primary groups of AMPs have been recognised: defensins and cathelicidins. Among various types of antimicrobial peptides, this review is focusing on cathelicidins antimicrobial peptides. Cathelicidins (LL-37) is an antimicrobial peptide (Mw \~ 18 kDa) that belong to the cationic amphipathic family found in both mammals (such as rabbits, cattle, horses, pigs, rats, rodents, and ungulates) and non-mammals (such as hagfish, chickens, and salmon) \[[@B5-biomolecules-07-00080],[@B6-biomolecules-07-00080],[@B7-biomolecules-07-00080],[@B8-biomolecules-07-00080]\]. In mammals, LL-37 are produced by various cells, including skin epithelial cells, leukocytes \[[@B9-biomolecules-07-00080]\], B-cells, keratinocytes, melanocytes \[[@B10-biomolecules-07-00080]\], neutrophils, bone marrow cells \[[@B11-biomolecules-07-00080]\], breast milk \[[@B12-biomolecules-07-00080]\], mast cells, seminal plasma \[[@B13-biomolecules-07-00080]\], salivary glands \[[@B14-biomolecules-07-00080]\], inflamed gingival tissues, and respiratory epithelium \[[@B15-biomolecules-07-00080]\]. Hence, cathelicidins are among the first defence peptides that come in contact with foreign pathogens and aids in first line defence. Various functions of human LL-37 peptides in the maintenance of human health reference are illustrated in [Figure 1](#biomolecules-07-00080-f001){ref-type="fig"}. Historically, following the discovery of Bac5 and seven antimicrobial peptides from bovine neutrophils (in the 1980s), cathelicidins were isolated from pig's intestine and were classified as mammalian cecropin \[[@B10-biomolecules-07-00080]\]. In 1995, the human cathelicidins (CAP-18) were isolated from neutrophils \[[@B16-biomolecules-07-00080]\]. Later on, it was isolated from a variety of human cells, body fluids, and tissues \[[@B17-biomolecules-07-00080]\]. We have highlighted the biochemistry, mode of action, and the importance of cathelicidins. The antibacterial property of cathelicidins peptides is due to its ability to destabilize the bacterial cell membrane through interactions with proteins and non-polar components of bacterial membranes. Therefore, cathelicidins are an active part of humoral immunity and play a significant role in the direct defence and activation of the inflammatory cells. The role of these peptides is not limited to the antibacterial activity, but its antiviral and anti-fungal actions makes it a biomolecule of greater interest. 2. Types and Biochemistry of LL-37 {#sec2-biomolecules-07-00080} ================================== Cathelicidins and their precursor molecules are synthesized after proteolytic cleavage \[[@B17-biomolecules-07-00080]\]. Based on their structures and molecular weight diversities, these peptides are characterized and are found in a variety of species. The particular gene that is reported to be responsible for the synthesis of cathelicidins in mammals is organized as 4 exons and 3 introns. Four other different genes (CATH1, CATH2, CATH3 and CATH-B1) have been reported in birds that have structure similar to mammalian peptides. Three genes (CATH1, CATH2, CATH3) encodes for the major part of cathelin-like domain, signal peptides and 5′ untranslated terminal. While the fourth exon (CATH-B1) encodes for mature peptides and 3′ untranslated terminals \[[@B18-biomolecules-07-00080]\]. Cathelicidins are generated as inactive precursor molecule comprises of three parts \[[@B8-biomolecules-07-00080]\]; (1)N-terminal that is composed of 29--30 amino acid molecules and is assumed to guide the liberation of biologically active peptides;(2)cathelin-domain comprising of 98--114 amino acid molecules with its function not yet examined; and(3)C-terminal that comprises of 12--100 amino acid molecules as an active peptide with wide range of antimicrobial property against bacteria, viruses, and fungi. The reported members of cathelicidins family include; (1)LL-37 (leucine-leucine 37) that is found in humans \[[@B19-biomolecules-07-00080]\],(2)CRAMP (cathelicidins related antimicrobial peptide) found in rats and mice \[[@B16-biomolecules-07-00080]\],(3)Flowlicidin 1,2,3 and cathelicidins β-1 found in chickens \[[@B20-biomolecules-07-00080]\],(4)CATH-1 and CATH-2 both are found in the Atlantic salmon \[[@B21-biomolecules-07-00080]\],(5)p15s found in rodents, and CAP18 in rabbits \[[@B22-biomolecules-07-00080]\], and(6)CAP11 is found in guinea pigs and LL-37 in rhesus monkeys \[[@B23-biomolecules-07-00080]\]. In humans, the gene encoding for the LL-37 is found on chromosome number 3, it is generated by the cleavage of hCAP18 (human cationic antimicrobial peptide 18 precursor protein). The cleavage of the C-terminus end of this precursor protein leads to the formation of LL-37 which is cationic in nature. Both the precursor and the active product are found in a number of human cells and body fluids (human wound fluid, saliva, gingival crevicular fluid (GCF), seminal plasma, vernix, and tracheal aspirates) \[[@B14-biomolecules-07-00080],[@B24-biomolecules-07-00080],[@B25-biomolecules-07-00080]\]. The production of cathelicidin can be regulated by variable number of cytokines, growth factors and activated vitamin D, which constitute the endogenous products \[[@B26-biomolecules-07-00080]\]. This is important, as vitamin D is known to induce LL-37 in oral epithelial. Moreover, LL-37 also plays an important role in the maintenance of oral health. Presence of LL-37 in saliva and GCF has the potential role of antimicrobial activity against gram negative and gram positive bacteria in the oral cavity and an immunomodulatory function. LL-37 expression has a role in the protection of tooth structures and oral mucosa \[[@B27-biomolecules-07-00080]\]. 3. Mechanism of Action against Microbes {#sec3-biomolecules-07-00080} ======================================= Antimicrobial peptides are components of host defence proteins that act against the microbial invasion by various mechanisms, such as: (a) barrel-stave model; (b) carpet model; and (c) toroidal model. A comprehensive review on oral antimicrobial peptides, their types and role in the oral cavity, including how these peptides are secreted and inhibit bacterial activities, has been reported elsewhere \[[@B28-biomolecules-07-00080]\]. Cathelicidins and other antimicrobial peptides exhibited the potential of eliminating foreign pathogens through various processing pathways such as membrane disrupting activity, antiseptic activity, apoptosis, angiogenesis, wound healing, chemotaxis and immune modulation (both humoral and cellular components) \[[@B29-biomolecules-07-00080]\]. Furthermore, these peptides target only pathogens, not human cells, due to the diversities within the biological membrane, including structure and composition. The broad spectrum antimicrobial property of cathelicidins is because of their ability to disrupt the bacterial cell membrane and ultimate death of bacteria. The three mechanisms proposed on how these peptides act on cell membranes include carpet model, barrel stave, and toroidal pore models \[[@B30-biomolecules-07-00080],[@B31-biomolecules-07-00080]\]. The most lethal mode of action of an antimicrobial peptide is considered to be its interaction with the cytoplasmic membrane ([Figure 2](#biomolecules-07-00080-f002){ref-type="fig"}). In the carpet model theory, peptides assemble themselves parallel to the surface of bacterial cell membrane, interacting electrostatically with phospholipid head groups on cell membrane surfaces. Hence, disintegration occurs due to the pressure created by the molecules of peptides and cell membrane in a detergent-like manner \[[@B24-biomolecules-07-00080]\]. The barrel stave theory suggests that the formation of a pore in a manner that a single peptide is associated with the surface of the membrane at its hydrophobic region. The hydrophobic region is inserted further into the membrane. A number of peptide monomers self-aggregate and insert further into layer aligning each other perpendicular to the membrane to form a water filled pore. In the toroidal pore mechanism, the interactions of peptides with the phospholipid head groups, causing a fold in lipid bilayer rather than insertion and hydrophobic interactions \[[@B32-biomolecules-07-00080],[@B33-biomolecules-07-00080]\]. In last decade, cathelicidins based drugs that are patented by different research groups. Zaiou et al. claimed that native LL-37 from human sweat have antimicrobial activity \[[@B34-biomolecules-07-00080]\]. Shorter peptides with low hemolytic properties toward human blood cells were identified. In addition, enhanced antimicrobial activity, and the ability to synergize with the parent peptides were observed. Ståhle-bäckdahl and coworkers claimed an additional therapeutic potential of LL-37 and its derivatives for wound healing applications. These peptides facilitated the regenerative potential of the traumatized skin \[[@B35-biomolecules-07-00080]\]. The Octoplus N.V. biopharmaceutical company (Amsterdam, Netherlands) reported two synthetic derivatives of LL-37 namely; peptide P10 (LAREYKKIVEKLKRWLRQVLRTLR-OH) for the treatment of infections related to atopic dermatitis and P60.4Ac (IGKEFKRIVERIKRFLRELVRPLR-OH) that can be used as a bioactive peptide layer for the prevention of bacterial growth and biofilm formation on the surfaces of biomaterials and metal implants \[[@B35-biomolecules-07-00080],[@B36-biomolecules-07-00080]\]. Kosikowska et al. \[[@B37-biomolecules-07-00080]\] reported antimicrobial peptides-based drugs and highlighted the importance of AMPs as a novel class of antibiotics ([Table 1](#biomolecules-07-00080-t001){ref-type="table"}). 4. Importance of LL-37 in Oral Cavity {#sec4-biomolecules-07-00080} ===================================== The environment contains an infinite number of microorganisms. For instance, bacteria are covering our skin, throat, gut, nasal cavity, ear, eyes, and oral cavity. Nature has protected the human body from the pathogenic bacteria by providing three lines of defense, including physiological, anatomical barrier i.e., skin, and certain immune cells to kill such pathogens. LL-37 has an immunomodulatory effects comprised of cellular and humoral components. The cellular component stimulates cells that play a crucial role in immunity and killing of foreign pathogens, for instance: neutrophils, macrophages, mast cells, dendritic cells, monocytes, and eosinophils. The humoral component (includes proteins, complement system and cytokines, cathelicidins) plays a different role. When considering the effects of cathelicidins on cells, for example, neutrophils, which play an important role during injury and inflammation, release in large number and are first to reach on such sites or infections. Neutrophils produce and express mediators, which include certain chemokines, cytokines, and fibrinogen and angiogenic factors. In addition, neutrophils stimulate the release of certain enzymes in the cytoplasm. Furthermore, they act as chemotactic agents, the source of prostaglandin, and leukotrienes, and are capable of generating reactive oxygen species. It has been reported that cathelicidins aid neutrophil in a number of ways; it increases the life span of neutrophils by expressing Bcl-XL proteins, and thus inhibiting early apoptosis of neutrophils \[[@B38-biomolecules-07-00080]\]. LL-37 also increases chemotaxis activity and migration of neutrophils by inhibiting expression of surface receptors CXCR2. One of the principal mechanisms of innate immunity in which neutrophil extracellular traps (NETs) is formed (NETosis), cathelicidins aids to the formation of these NETs \[[@B39-biomolecules-07-00080]\]. LL-37 affects other inflammatory cells activity as well, which includes monocyte or macrophages. In these cells it is also reported that they enhance the receptor expression on the site of injury, stimulate mediator release, aids in decreasing the endotoxin of Neisseria meningitis. In short, LL-37 stimulates these inflammatory cells and helps in the abolition of infections and inflammation, hence, playing a useful role in wound healing \[[@B3-biomolecules-07-00080]\]. As discussed earlier, LL-37 also act as a broad-spectrum antibiotic in the human body. It was reported that this peptide provides essential role in innate response against Mycobacterium tuberculosis, as they stimulate alveolar macrophages as first line of defense against tuberculosis \[[@B40-biomolecules-07-00080]\]. Another important role of cathelicidins is that it inhibits certain gastrointestinal (GI) disorders, such as ulcers, inflammation, and cancer, as these conditions commonly invade GI mucosa. Bacteria associated with gastritis and peptic ulcers are killed, which helps in repairing and angiogenesis of damaged tissues \[[@B41-biomolecules-07-00080],[@B42-biomolecules-07-00080]\]. Furthermore, LL-37 is seen to be expressed in tongue and buccal mucosa, and also can be detected in GCF and saliva. While, inflamed gingival tissues has shown to have upregulated expression of LL-37, signifying its diagnostic activity in inflammatory periodontal disorders \[[@B43-biomolecules-07-00080]\]. The role of LL-37 in saliva suggests its antimicrobial activity in the protection of tooth structure, which, in turn, can be correlated to resistance to caries. LL-37 presence in GCF contributes to the oral health status as it is seen to be associated with the severity of periodontal disease. For this reason, LL-37 antimicrobial quality makes them an excellent candidate for broad spectrum antimicrobial therapeutics, and plays a distinctive role in maintaining oral health \[[@B44-biomolecules-07-00080]\]. In addition, LL-37 has an antiviral effect against certain viruses that includes influenza \[[@B45-biomolecules-07-00080]\], Herpes simplex-1 \[[@B46-biomolecules-07-00080]\], adenovirus, and human immuno-virus-1 (HIV-1) \[[@B47-biomolecules-07-00080]\]. It has been reported that these peptides suppress the viruses by acting on their membrane envelopes and their protein capsules. On HIV, it works by disrupting the HIV-1 reverse transcriptase pathway through blocking its binding. In addition to antiviral effects of LL-37, there is antifungal role against important fungi such as Candida albicans that is present in our normal oral flora. Research using various types of cathelicidins, it is hypothesized that cathelicidins interact with the cell wall and generate reactive oxide species within fungi, leading to their fungicidal action \[[@B26-biomolecules-07-00080]\]. 5. Diagnostic Biomarker in Oral Health and Research {#sec5-biomolecules-07-00080} =================================================== There are different approaches that are available to diagnose any clinical situation, according to the need, such as biopsy and bio-fluids analysis (saliva, blood, semen, sperm, cervico-vaginal secretions). Currently, human saliva plays an important role in the diagnostic sciences due to its beneficial features, such as non-invasive collection, cheap, easy to collect, and no need of any special clotting agents in contrast to blood testing. Saliva is rich in proteins and peptides, and has the ability to diagnose many human diseases \[[@B48-biomolecules-07-00080],[@B49-biomolecules-07-00080],[@B50-biomolecules-07-00080],[@B51-biomolecules-07-00080]\]. Pakistan human salivary research group recently highlighted the importance of saliva as a diagnostic fluid for the detection of oral squamous cell carcinoma (OSCC) and the capability of salivary interleukins (IL-6 and IL-8) and tumor necrosis factor-α as a diagnostic biomarker for the detection of OSCC \[[@B52-biomolecules-07-00080],[@B53-biomolecules-07-00080]\]. LL-37 is found in epithelial cells lining the oral cavity, tongue, buccal mucosa, inflamed gingival tissues, saliva, and GCF \[[@B54-biomolecules-07-00080]\]. It's concentration in saliva is approximately 0.14--3 μg/mL, and maintains protective activity against gingival lesions in addition to its significant role in wound healing \[[@B55-biomolecules-07-00080]\]. Cathelicidins are well known for innate defensive barrier against various microbial pathogens, including gram negative and gram positive bacteria. Murakami et al. investigated the expression of messenger RNA of cathelicidins in the sialadenitis through reverse transcriptase-polymerase chain reaction and immunohistochemically staining. Cathelicidins protein expressions are upregulated in chronic sialedinitis compared to normal salivary glands and providing defense mechanisms in the salivary glands \[[@B14-biomolecules-07-00080]\]. Moreover, cathelicidins has been broadly studied in relation to their immunomodulatory and antibacterial properties, whereas, salivary LL-37 is also being released by neutrophils in gingival crevicular fluid, salivary glands, and expressions of LL-37 indicates its role in the protection of tooth structure, oral mucosa, and enhances the production of immunoglobulins (IgA and IgG). Such measures control the bacterial checks for the prevention of biofilm and hence behaving as a natural antibiotic against dental caries \[[@B27-biomolecules-07-00080]\]. Guo et al. tested children saliva in biofilm formation assay to evaluate the inhibitory effects of LL-37 on the biofilm that is produced by Streptococcus mutans and LL-37 interaction with epigallocatechin gallate (EGCG), which also has anti-infective property towards biofilm production. LL-37 enhanced the effects of EGCG on Streptococcus mutans, which, in turn, makes LL-37 as an anti-biofilm compound that can be potentially used for dental treatments \[[@B56-biomolecules-07-00080]\]. LL-37 activates metalloproteinase via transactivation of the epidermal growth factor receptor (EGFR) receptor to elicit growth stimulatory properties, which helps in wound closure \[[@B57-biomolecules-07-00080]\]. Kajiya et al. determined the effect of LL-37 on migration of human pulp cells by wound healing assay and immunoblotting. LL-37 plays a role in the enhancement of regeneration of pulp-dentin complex by activation of EGFR and c-Jun N-terminal kinase by the induction of heparin binding cell migration \[[@B58-biomolecules-07-00080]\]. Furthermore, Tsai et al. performed competition assay and concluded that LL-37 has the property of reducing and inhibiting the infectivity of Candida albicans in the oral cavity by having effects on cell wall carbohydrates and can be used as a screening tool to check the involvement of fungal infections \[[@B59-biomolecules-07-00080]\]. LL-37 seems to have role in the development of Papillon-Lefèvre syndrome. Eick et al. compared the levels of LL-37 in GCF, saliva, and neutrophil-derived enzymes, which revealed the fact that dysfunctional cathepsin C in patients of Papillon-Lefevre syndrome caused the deficit in immunomodulatory and antimicrobial functions of LL-37 in GCF, leading to severe periodontal disease. This occurs because of the lack of activation of cathepsin C in Papillon-Lefèvre syndrome, which ultimately results in deficit immunomodulatory and antimicrobial functions of LL-37 in gingiva that will allow bacteria such as Aggregatibacter actinomycecomitans to infect gingiva and periodontium to develop periodontal diseases. Hence, proving the ability of LL-37 as a diagnostic tool for inflammatory periodontal diseases \[[@B60-biomolecules-07-00080]\]. In the same way, Montreekachon et al. investigated through methylthiazolyldiphenyl-tetrazolium bromide assay and real time-polymerase chain reaction in the gingival epithelial cells and reported the effects of LL-37 on Th1/Th2 cytokine expression. The results indicated the involvement of LL-37 in induction of IL-8 secretion and the recruitment of neutrophil at inflammatory sites in the periodontal tissues \[[@B61-biomolecules-07-00080]\]. Several experimental studies have been conducted on evaluating the effects and outcomes of cathelicidins against bacteria (gram positive and gram negative), enveloped viruses, and fungi \[[@B17-biomolecules-07-00080]\]. According to a study where unstimulated saliva was collected from children in correlation with their caries activity revealed that peptide based oral care products are protective in nature and provide defense against dental caries \[[@B62-biomolecules-07-00080]\]. In patients that are diagnosed and recorded with periodontitis, the levels of cathelicidins were noted to be higher when compared to the healthy individuals. Long term exposure to smoking leads to a reduction of cathelicidins, hence, the protective and antibacterial effects of the peptides are lost enhancing the probability of acquiring periodontitis \[[@B63-biomolecules-07-00080]\]. The protective activity of cathelicidins was clearly demonstrated when a comparison of the levels was carried out using quantitative analysis (ELISA) between patients that were diagnosed with oral lichen planus and healthy individuals. Based on the study, higher levels were noted in severe forms of oral lichen planus (erosive) compared to reticular form, which, in turn, had higher levels when compared to their healthy counter-parts. These patients were periodically evaluated and as the lesion subsided clinically, the levels of the protective peptides also subsided \[[@B64-biomolecules-07-00080]\]. This clearly states its anti-inflammatory and protective nature. Saliva collected from volunteers with healthy periodontium showed that this peptide has antimicrobial properties that are enabled in saliva. Thus, saliva acts as a mediator of its antimicrobial activity when tested against E. coli, but only in the presence of the organism Porphyromonas gingivalis proteases \[[@B65-biomolecules-07-00080]\]. In vitro experiments have led to the conclusion that this could be a promising antimicrobial peptide supporting the host against livestock-associated methicillin-resistant Staphylococcus aureus (LA-MRSA), primarily because it is not influenced by common resistance genes \[[@B66-biomolecules-07-00080]\]. Yi-jie Guo et al. experimented with EGCG, a constituent of tea catechins, which has potential of inhibiting not only microbial growth but on biofilm as well. Cathelicidins enhanced the activity of EGCG and opened a new horizon for potential anti-biofilm compounds \[[@B56-biomolecules-07-00080]\]. Different sources of cathelicidins in the oral cavity were also evaluated, the quantitative study on the levels of cathelicidins using ELISA indicated lower levels in edentulous patients when compared to dentate patients. This has led to the theory that cathelicidins are released from the gingival tissues \[[@B25-biomolecules-07-00080]\]. Keeping in view the above mentioned experimental studies, cathelicidins reinforce the protective and defensive antimicrobial properties, and play a vital role in the immunity. 6. Conclusions {#sec6-biomolecules-07-00080} ============== Cathelicidins is a group of antimicrobial peptides that are secreted in the oral fluids, such as saliva, gingival crevicular fluid and can be used for diagnostic significance of oral health. The enhanced level of oral cathelicidins is associated with inflammatory conditions, such as gingivitis and immune disorders, such as oral lichen planus. In addition, the localized availability of cathelicidins may influence the immune response to oral microbial conditions, such as caries, periodontitis, and oral malignancies. Besides the diagnostic role, such an increase in the level of cathelicidins prevents infection (antimicrobial) and promotes wound healing of the effected tissues. Although the mechanism of action is not fully understood, the most likely mode of action of these antimicrobial peptides is through microbial membrane disruption. We are thankful to the Pakistan Human Saliva Research Group (PakSRG) for their guidance and for providing a platform that enabled us to develop the manuscript; we also thank the PakSRG for their funding during the preparation of this manuscript. All authors equally contributed in the preparation of the manuscript. The authors declare no conflict of interest. ![Different functions of human cathelicidins (LL-37) peptides in the human body \[[@B4-biomolecules-07-00080]\]. hCAP18: human cationic antimicrobial peptide 18 precursor protein.](biomolecules-07-00080-g001){#biomolecules-07-00080-f001} ![Illustration representing the cathelicidins mechanism of action against microbes specifically for LL-37 and oral cavity.](biomolecules-07-00080-g002){#biomolecules-07-00080-f002} biomolecules-07-00080-t001_Table 1 ###### Reported microbial inhibitory concentration (MIC) of different antimicrobial peptides sequences from LL-37 (adapted from Gallo and Murakami \[[@B37-biomolecules-07-00080]\]). Peptide Sequence Staphylococcus aureus (MIC (µM)) Escherichia coli (MIC (µM)) Candida albicans (MIC (µM)) --------- ------------------------------------------ ------------------------------------ ------------------------------- ------------------------------- LL-37 LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES-OH \>64 64 20 RK31 RKSKEKIGKEFKRIVQRIKDFLRNLVPRTES-OH 16 8 4 KS30 KSKEKIGKEFKRIVQRIKDFLRNLVPRTES-OH 16 8 2	{ "pile_set_name": "PubMed Central" }
Introduction ============ β-Enaminones and β-enamino esters represent important synthetic building blocks for the development of versatile carbon--carbon bond-formation reactions and heterocyclic constructions \[[@R1]--[@R5]\]. In recent years, the in situ generated β-enamino esters, which could be easily obtained from the addition of aliphatic or aromatic primary amines to the activated alkynes, have been widely recognized as practical synthons for the synthesis of a wide variety of heterocycles and pharmaceutical compounds \[[@R6]--[@R11]\]. Many domino reactions have been developed by trapping this kind of β-enamino ester with sequential adding of nucleophilic or electrophilic reagents to give versatile nitrogen-containing compounds and N,O-heterocycles \[[@R12]--[@R23]\]. Recently, Perumal and we have both developed an efficient synthetic procedure for functionalized spiro\[indoline-3,4'-pyridines\] by domino reactions of in situ generated β-enamino esters, isatin and malononitrile with triethylamine as the base catalyst \[[@R24]--[@R25]\]. A literature survey indicated that there has been an explosion of activity around the synthesis of spirooxindoles in the past years \[[@R26]--[@R31]\]. Encouraged by these results, and hunting for new synthetic methods for functionalized spirooxindoles, we investigated the domino reactions of arylamines, methyl propiolate, aromatic aldehydes and malononitrile (ethyl cyanoacetate) and successfully developed a facile synthetic procedure for functionalized spiro\[indoline-3,4'-pyridines\] and spiro\[indoline-3,4'-pyridinones\]. Results and Discussion ====================== The efficient formation of functionalized spiro\[indole-3,4'-pyridines\] via the four-component reaction prompted us to study the reaction scope further \[[@R25]\]. Another widely used electron-deficient alkyne reagent, methyl propiolate, was utilized to replace dimethyl acetylenedicarboxylate as one component. The addition reaction of aniline to methyl propiolate to give the active adducts, 3-arylaminoacrylates, usually needs more than twelve hours. Thus, we decided firstly to let arylamine and methyl propiolate react in ethanol at room temperature for 24 hours. TLC analysis indicated that the addition reaction had finished, and TLC analysis indicated that almost exclusively the desired β-arylaminoacrylates existed in the solution. Then isatin and malononitrile as well as triethylamine were added to the system and the mixture was heated under reflux for an additional 24 hours. The expected spiro\[indoline-3,4'-pyridines\] 1a--1p were obtained in good yields by using this one-pot domino reaction procedure ([Table 1](#T1){ref-type="table"}). It can be seen that aniline bearing methoxy, methyl or chloro groups all reacted smoothly to give the expected spiro\[indoline-3,4'-pyridines\] with marginal effect. Benzylamine also afforded good yields of the spiro product. The structures of the prepared spiro\[indoline-3,4'-pyridines\] 1a--1p were fully characterized by spectroscopic methods and were further confirmed by the determination of the single crystal structures of the spiro compounds 1c ([Fig. 1](#F1){ref-type="fig"}) and 1h ([Fig. 2](#F2){ref-type="fig"}). ###### Synthesis of spiro\[indoline-3,4'-pyridine\] 1a--1p by domino reaction. ------------------------------------------- -------- -------------------- ---- --------------- ----------- ![](Beilstein_J_Org_Chem-09-846-i001.jpg) Entry Compd Ar R R' Yield (%) 1 1a C~6~H~5~ H H 79 2 1b p-CH~3~C~6~H~4~ H H 76 3 1c p-CH~3~OC~6~H~4~ H H 74 4 1d p-ClC~6~H~4~ H H 81 5 1e p-BrC~6~H~4~ H H 80 6 1f p-ClC~6~H~4~ Me H 75 7 1g p-CH~3~C~6~H~4~ Me H 73 8 1h p-ClC~6~H~4~ Cl H 84 9 1i p-CH~3~C~6~H~4~ Cl H 72 10 1j p-ClC~6~H~4~ H CH~2~C~6~H~5~ 64 11 1k p-CH~3~C~6~H~4~ H CH~2~C~6~H~5~ 75 12 1l p-CH~3~OC~6~H~4~ H CH~2~C~6~H~5~ 72 13 1m C~6~H~5~ H CH~2~C~6~H~5~ 66 14 1n m-ClC~6~H~4~ H CH~2~C~6~H~5~ 61 15 1o m-CH~3~C~6~H~4~ H CH~2~C~6~H~5~ 62 16 1p C~6~H~5~CH~2~ H CH~2~C~6~H~5~ 50 ------------------------------------------- -------- -------------------- ---- --------------- ----------- ![Molecular structure of the spiro compound 1c.](Beilstein_J_Org_Chem-09-846-g002){#F1} ![Molecular structure of the spiro compound 1h.](Beilstein_J_Org_Chem-09-846-g003){#F2} When ethyl cyanoacetate was utilized in the domino reaction under similar conditions, the reaction usually resulted in a complicated mixture of spiro\[indoline-3,4'-pyridines\] 2 and spiro\[indoline-3,4'-pyridinones\] 3 depending on whether the cyano group or the ester group was taking part in the cyclization process ([Scheme 1](#C1){ref-type="fig"}). In order to simplify the separation process, only the main product was separated from the reaction mixtures by column chromatography. The other product was not separated or structurally characterized. The results are listed in [Table 2](#T2){ref-type="table"}. It is interesting to find that anilines bearing p-chloro, m-chloro or p-bromo groups afforded spiro\[indoline-3,4'-pyridines\] 2a--2h as the main products ([Table 2](#T2){ref-type="table"}, entries 1--8). Due to fact that m-nitroaniline and p-nitroaniline could not react with methyl propiolate to give the desired intermediate β-enamino ester, they were not utilized in this reaction. On the other hand p-methoxyaniline, p-methylaniline and aniline itself gave spiro\[indoline-3,4'-pyridinones\] 3a--3j as the main products ([Table 2](#T2){ref-type="table"}, entries 9--18). Thus, it seems that anilines with electron-withdrawing groups preferably give spiro\[indoline-3,4'-pyridines\] 2, while anilines with electron-donating groups preferably produce spiro\[indoline-3,4'-pyridinones\] 3. ###### Synthesis of spiro\[indoline-3,4'-pyridines\] 2a--2h and spiro\[indoline-3,4'-pyridinone\] 3a--3n via domino reaction. ------------------------------------------- -------- -------------------- ---- --------------- ----------- ![](Beilstein_J_Org_Chem-09-846-i002.jpg) Entry Compd Ar R R^'^ Yield (%) 1 2a p-ClC~6~H~4~ H H 49 2 2b p-ClC~6~H~4~ H CH~2~C~6~H~5~ 52 3 2c p-ClC~6~H~4~ Me C~4~H~9~ 47 4 2d p-ClC~6~H~4~ Me CH~2~C~6~H~5~ 35 5 2e m-ClC~6~H~4~ Me CH~2~C~6~H~5~ 36 6 2f p-BrC~6~H~4~ H H 52 7 2g p-BrC~6~H~4~ H C~4~H~9~ 36 8 2h p-BrC~6~H~4~ Me CH~2~C~6~H~5~ 56 9 3a p-CH~3~OC~6~H~4~ H H 39 10 3b p-CH~3~OC~6~H~4~ H CH~2~C~6~H~5~ 53 11 3c p-CH~3~OC~6~H~4~ H C~4~H~9~ 54 12 3d p-CH~3~OC~6~H~4~ Me CH~2~C~6~H~5~ 51 13 3e p-CH~3~C~6~H~4~ H CH~2~C~6~H~5~ 56 14 3f p-CH~3~C~6~H~4~ H C~4~H~9~ 47 15 3g p-CH~3~C~6~H~4~ Me CH~2~C~6~H~5~ 49 16 3h C~6~H~5~ Me C~4~H~9~ 56 17 3i C~6~H~5~ H CH~2~C~6~H~5~ 47 18 3j C~6~H~5~ Me CH~2~C~6~H~5~ 52 19 3k C~6~H~5~CH~2~ H H 68 20 3l C~6~H~5~CH~2~ Me CH~2~C~6~H~5~ 45 21 3m C~6~H~5~CH~2~ H C~4~H~9~ 63 22 3n C~6~H~5~CH~2~CH~2~ H CH~2~C~6~H~5~ 48 ------------------------------------------- -------- -------------------- ---- --------------- ----------- Benzylamine and 2-phenylethylamine could also be used in the domino reactions to give the spiro\[indoline-3,4'-pyridinones\] 3k--3n as the main products ([Table 2](#T2){ref-type="table"}, entries 19--22). The structures of the prepared spiro compounds 2a--2h and 3a--3n were fully established by spectroscopic methods. The single-crystal structures of spiro compounds 2b ([Fig. 3](#F3){ref-type="fig"}) and 3b ([Fig. 4](#F4){ref-type="fig"}) were successfully determined by X-ray diffraction methods. It should be pointed out that the ^1^H NMR spectra of compounds 2a--2h showed some distinguishing features. The characteristic resonance of the NH~2~ group usually displays one broad peak at about 7.40 ppm, which is overlapped with the peaks of aromatic protons. Comparing with the one singlet at 5.80 ppm of the NH~2~ group in compounds 1a--1p, the peak of the NH~2~ group in compounds 2a--2h is shifted to a much lower field. Secondly, the characteristic peaks of the CH~2~ unit of the ethoxy group in most spiro\[indoline-3,4'-pyridines\] 2a--2h split into two mixed peaks at about 3.75--3.72 (m, 1H), 3.37--3.34 (m, 1H), which indicated that these two protons existed in different circumstance. The CH~3~ unit of the ethoxy goup showed a normal peak. ![Molecular structure of spiro compound 2b.](Beilstein_J_Org_Chem-09-846-g004){#F3} ![Molecular structure of spiro compound 3b.](Beilstein_J_Org_Chem-09-846-g005){#F4} From the molecular structure of the spiro compound 3b shown in [Fig. 4](#F4){ref-type="fig"}, we could clearly see that the cyano group and the phenyl group of oxindole moiety in the newly formed dihydropyridinone ring exist in cis-configuration. ^1^H NMR spectra of 3a--3n all display one singlet at about 5.40 ppm for the one proton in the dihydropyridinone ring, which indicated that only one isomer exists in each product. Based on the single crystal structure and ^1^H NMR data we could tentatively conclude that the spiro\[indoline-3,4'-pyridinones\] 3a--3n have cis-configuration of the cyano group and the phenyl group of the oxindole moiety. To explain the mechanism of this domino reaction, a plausible reaction course was proposed to account for the different products based on the published reactions containing methyl propiolate \[[@R20]--[@R22]\], which are illustrated in [Scheme 1](#C1){ref-type="fig"}. The first reaction is the formation of the key intermediate β-enamino ester A from the addition of arylamine to methyl propiolate. The second reaction is a Knoevenagel condensation of isatin with malononitrile or ethyl cyanoacetate under the catalysis of triethylamine to give the isatylidene deriatives B. The third reaction is a Michael addition of β-enamino ester intermediate A with isatinylidene derivative B to yield intermediate C. In the case of the reaction containing malononitrile, the nucleophilic addition of the amino group to the C--N triple bond in intermediate C resulted in spiro compound 1 with the tautomerization of the imino group to an amino group. In the case of the reaction containing ethyl cyanoacetate, the amino group could react with both the cyano group and the ester group in the adduct C. The nucleophilic addition of the amino group to the C--N triple bond finally afforded spiro\[indoline-3,4'-pyridines\] 2. On the other hand the amino group attacked the ester group to give spiro\[indoline-3,4'-pyridinones\] 3 with the elimination of ethanol. In this reaction process, the reasons that anilines with electron-withdrawing groups preferably attack the cyano group and anilines with electron-donating groups preferably attack the ester group are not very clear. ![The proposed mechanism for the domino reaction.](Beilstein_J_Org_Chem-09-846-g006){#C1} Conclusion ========== In summary, we have successfully developed a one-pot four-component reaction of arylamines, methyl propiolate, isatin and malononitrile or ethyl cyanoacetate with triethylamine as base catalyst. This reaction can proceed smoothly under mild conditions to afford the functionalized spiro\[indoline-3,4'-pyridines\] and spiro\[indoline-3,4'-pyridinones\] in moderate to good yields. The advantages of this reaction included readily available starting materials, mild reaction conditions, operational simplicity, a widely variety of substrates, and molecular diversity of the products. The potential uses of this reaction in synthetic and medicinal chemistry may be quite significant. Experimental ============ Reagents and apparatus: Aromatic aldehydes, arylamines, methyl propiolate and other reagents are commercial reagents and were used as received. Solvents were purified by standard techniques. All reactions were monitored by TLC. Melting points were taken on a hot-plate microscope apparatus. IR spectra were obtained on a Bruker Tensor 27 spectrometer (KBr disc). NMR spectra were recorded with a Bruker AV-600 spectrometer. HPLC--MS were measured with a Fennigan LCQ Deca XP MAX instrument. High-resolution mass spectra (ESI) were obtained with a Bruker UHR--TOF maXis spectrometer. X-ray data were collected on a Bruker Smart APEX-2 CCD diffractometer. Single-crystal data for compounds 1c (CCDC 843674), 1h (CCDC 843676), 2b (CCDC 904924) and 3b (CCDC 904923) have been deposited in the Cambridge Crystallographic Data Centre. General procedure for the synthesis of spiro\[indoline-3,4'-pyridine\] derivatives 1a--1p: In an analogous manner to our procedure published in \[[@R25]\], a solution of arylamine (2.0 mmol), methyl propiolate (2.0 mmol) in 5 mL ethanol was stirred at room temperature overnight. Then isatin (2.0 mmol), malononitrile (2.0 mmol) and triethylamine (0.4 mmol) were added. The mixture was heated under reflux for about 24 hours. Then the solution was concentrated to approximately half the volume. The resulting precipitates were collected and washed with ethanol to give the pure product for analysis. General procedure for the synthesis of spiro\[indoline-3,4'-pyridines\] 2a--2h and spiro\[indoline-3,4'-pyridinone\] derivatives 3a--3n: A similar procedure to that above was used. A solution of arylamine (2.0 mmol), methyl propiolate (2.0 mmol, 0.168 g) in 5 mL ethanol was stirred at room temperature overnight. Then isatin (2.0 mmol), ethyl cyanoacetate (2.0 mmol, 0.226 g) and triethylamine (0.4 mmol) were added to it, and the whole mixture was heated under reflux for about 24 hours. Then the solution was concentrated to approximately half the volume, which was subjected to column chromatography with ethyl acetate and light petroleum (v/v = 1:3) as eluent to give the pure product for analysis. Supporting Information ====================== ###### Experimental details and detailed spectroscopic data. This work was financially supported by the National Natural Science Foundation of China (Grant No. 21172189) and the Priority Academic Program Development of Jiangsu Higher Education Institutions.	{ "pile_set_name": "PubMed Central" }
Background ========== The modulated arc (mARC) technique \[[@B1],[@B2]\] has recently become available for Siemens Artiste linear accelerators as an analogue to RapidArc and volume modulated arc therapy (VMAT) available for Varian and Elekta machines (\[[@B3]\]; for a recent review see \[[@B4]\]). Both techniques offer highly conformal treatment, since inversion is performed similarly to intensity-modulated radiotherapy (IMRT) for a large number of beam directions, which may create a complete or partial gantry rotation; a notable decrease in treatment time is accomplished by the continuous gantry rotation and multi-leaf collimator (MLC) movement. However, mARC differs from VMAT in underlying philosophy and practical implementation. Whereas for VMAT, the beam is kept on during the complete arc, while the MLC, collimator and gantry move to reach predetermined optimized configurations at the control points, the mARC only switches the beam on inside short "arclets" around the optimization points. Between the "control points" that define the start and end of an arclet, the MLC and collimator configuration are kept fixed; the width of the arclet is chosen as small as possible from the point of view of monitor units (MU) and gantry rotation velocity, while always remaining within user-defined upper limits. This method offers an additional degree of freedom in treatment planning in the sense that the treatment planning system (TPS) is not confined in its choice of MLC pattern and collimator by the gantry rotation velocity; rather, the gantry can be stopped between arclets until the desired configuration is reached. This expands the solution space of the optimization; at the same time, it may be argued that dosimetric agreement between the treatment plan and the delivered dose is improved, since intermediate MLC/collimator configurations are excluded. At the time of writing, only the Prowess Panther TPS supports mARC plannning; the certification for RayStation is pending. The present study presents a novel implementation of mARC planning based on any kind of TPS capable of IMRT planning and RTP export. For clinics relying on treatment planning systems other than Prowess/RayStation, our conversion algorithm offers the possibility of implementing mARC without switching to Prowess/RayStation, within the familiar environment of their own TPS. We hope that this will make the excellent functionality of mARC available to a much larger range of clinical users and researchers. In the clinical context, this brings Siemens users to the point of arc treatments which have hitherto been missing; in research, the different approach to arc planning merits further research, since a comparison with VMAT and RapidArc may be interesting for future developments in this field. As an additional advantage, our conversion method provides an equivalent step-and-shoot plan for each mARC plan, which can be used as a back-up concept in institutions where only one linac is equipped with mARC. Furthermore, the implementation presented offers improved functionality for the planning of hybrid arcs (compare \[[@B5]\]), in which the arc is complemented by one or a number of static fields with several segments, similar to IMRT beams. In contrast to hybrid planning in existing TPS, our approach offers the possibility of planning in hybrid mode without a priori choosing the static gantry angles; instead, the optimum hybrid beam directions are determined by the inversion algorithm. This work is meant as a feasibility study to outline a practical implementation of mARC planning using Philips Pinnacle as an example, with an outlook on the possibilities for further research that can be based on this modality. As a basis to the demonstration of feasibility, we show and discuss some examples of how plans may be created and converted, but without the intention of providing a complete planning study. The same conversion code was also tested for conversion of Prowess Panther IMRT plans to ensure that the algorithm works independently of the original planning system. To our knowledge, this is the first proof-of-principle of converting an IMRT treatment plan into an equivalent single/hybrid arc. Method ====== Technical implementation ------------------------ The Department of Radiotherapy of the Saarland University Medical Centre is equipped with three linacs (two Siemens Artiste, one Siemens Oncor) with 160 MLC, out of which one Artiste with 6 MV and flattening-filter-free (FFF) 7 MV photon energies received the mARC upgrade. The 6 MV energy is matched with the other two linacs, which do not yet support mARC treatment. CT-based (Philips BigBore) treatment planning is routinely performed using the Philips Pinnacle TPS (V9.2 and V9.4); for the mARC, Prowess Panther was additionally commissioned. Treatment plans are exported from Pinnacle via DICOM and loaded into Mosaiq (Elekta), then transferred to Syngo (Siemens) and sent to the linac for treatment. Treatment planning for mARC is generally performed along the same lines as "standard" IMRT treatment planning, the main difference being that the optimization points along an arc are placed equidistantly (generally at 5°-10° intervals), with one "segment" (MLC/collimator configuration) intended for each gantry direction. The user defines the maximum arclet length over which the dose for each optimization point may be spread; the Artiste firmware will restrict this angle as much as technically feasible, but never exceed it. As the TPS only calculates the dose at the optimization points, the delivered dose will deviate from the calculated dose slightly, which is why long arclets (5° or more) should be excluded. In mARC planning, the user may decide beforehand to favour particular beam directions for which several segments are allowed, resulting in a hybrid arc. For Pinnacle IMRT plans, the creation of multiple-segment beams can be permitted by setting the maximum number of segments larger than the number of beams. In contrast to mARC, this will let the TPS decide for which gantry direction these segments will be applied. The optimization will therefore automatically choose the directions of hybrid beams (if any), hence providing an additional degree of freedom in comparison with hybrid mARC planning. Our approach to mARC planning in Pinnacle relies on performing "normal" IMRT planning using direct machine parameter optimization (DMPO) for 18 to 36 beams, and then reformatting the output RTP-file \[[@B6]\] to conform to mARC standards (Figure [1](#F1){ref-type="fig"}). In principle, this is the same method as is applied in SmartArc planning, where a first IMRT-like optimization is performed using beams spaced between 12° and 24° apart, the segments of which are then reorganized to create an arc with control points every 2° to 6° \[[@B7]\]. Therefore, each RapidArc plan can be interpreted as an equal-quality IMRT plan with one beam and one segment placed at each control point (resulting in a larger number of beam directions than routinely used in IMRT treatments). Vice versa, we move from the IMRT planning with equidistantly spaced beams, possibly including more than one segment, to an arc treatment to be irradiated as mARC. This method combines the advantage of arc-based treatment, which is both the speed and the many gantry angles, with the advantage of IMRT, i.e. free choice of MLC configuration and optimal distribution of hybrid segments. ![Schematic workflow of treatment plan creation, conversion into mARC until transfer to the linear accelerator.](1748-717X-8-193-1){#F1} Formally, the RTP-file output of a hybrid arc plan is similar to the output of an IMRT plan, with the main difference that the Pinnacle IMRT plan interprets each gantry direction as a separate beam with an arbitrary number of segments, whereas a hybrid arc is interpreted as a single beam. Consequently, the coding of the field and control points differs somewhat between the two output formats (mARC vs. step-and-shoot). In particular, the Pinnacle export RTP-files of single-segment beams in IMRT plans have only one control point rather than two (at the beginning and end of the segment) for mARC, and the dose is normalized for each beam rather than for a complete arc. Several minor pieces of information in the field and control point definitions also require adjustments (gantry rotation direction from one arclet to the next, field size, dose rate etc.). The task of our conversion algorithm is to perform all these input and reformatting adjustments in a way to create a new RTP-file which will look like an mARC plan and can be imported for treatment. Since the RTP-file can be opened with any text editor, reformatting can be performed automatically: we used tcsh-scripting with awk, but any kind of ASCII file manipulation code will serve the purpose. The main steps of the algorithm are summarized in Figure [2](#F2){ref-type="fig"}. ![Principle of mARC conversion algorithm.](1748-717X-8-193-2){#F2} After importing the modified RTP-file into Mosaiq, the remaining part of the workflow remains as before. If it is so desired, the original plan can be imported as an alternative back-up concept, which would dosimetrically produce the same plan, but be interpreted as a step-and-shoot IMRT with correspondingly longer treatment time. Planning examples ----------------- We present planning examples for two cases routinely treated with IMRT at our institutions: a prostate patient and a head-and-neck patient, which are both planned using three different methods for the two available energies: flat 6 MV and FFF 7 MV, respectively. Since this is meant only as a proof of principle and not as a planning study, the main focus of our paper is on the feasibility of converting the plans into mARC; the decision exactly how the original step-and-shoot plans are created is left to the user depending on the treatment planning system used and on the clinical requirements. However, we believe it is of interest to show different ways in which planning can proceed using the Pinnacle TPS, and the resulting quality of the plans, compared with "normal" IMRT plans irradiated at our institution (Table [1](#T1){ref-type="table"}). We limit ourselves to PTV plans, since there is more variability in the boost depending on the tumour location; the boost planning and conversion can be performed along the same lines as shown for the PTV. ###### Overview of planning and conversion scenarios Original step-and-shoot plan Conversion Converted mARC plan --------------- --------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------ Prostate Version a 19 beams Convert directly as is 8 arclets 35 segments 11 hybrid fields Version b 19 beams 35 segments Same plan as version a Split beams with more than 1 segment into two beams spaced 4° apart (one beam −2°, one +2° relative to original), this gives 26 beams, then convert 22 arclets 4 hybrid fields Version c 30 beams 35 segments Convert directly 25 arclets 5 hybrid fields Head and neck Version a 18 beams 50 segments Split each beam with More than 1 segment into two beams spaced 6° apart (one beam −3°, one +3° relative to original) 26 arclets 5 hybrid fields Version b 36 beams Convert directly 27 arclets 50 segments 9 hybrid fields Version c 36 beams 72 segments Do not change beams, but order segments in a way that two rotations are made, so the first segment of each beam is irradiated on the cw arc; the rest of the segments are included in the ccw arc Clockwise: pure mARC including only 36 arclets (not hybrid) Counter-clockwise: 23 arclets, 6 hybrid fields For the prostate tumor, the first plan (version a) is based on an original 19 beam IMRT plan with 35 segments (1--3 segments per beam). This is directly converted into mARC, corresponding to an "extreme hybrid" case in which over 10 beams are hybrids. This plan is still much more coarsely spaced than a normal arc plan; compared with IMRT, the converted plan already has the advantage of static jaws and arclets. Starting from this original plan version a, manual field separation is performed to create more closely spaced beams, by separating those beams with two segments into two beams spaced 4° apart; beams with one segment or three segments are kept unchanged. After conversion of this second (version b) plan into mARC, beams with three segments appear as hybrid fields, whereas beams with one segment and the divided beams (originally with two segments) become normal arclets. This process results in a hybrid plan with a much smaller number of hybrid fields and a more "arc-like" beam configuration. As a third alternative, a "real" hybrid arc plan is optimized using 30 equidistant beams spaced 12° apart, again allowing 35 segments (version c), and converting the resulting plan to mARC. This plan now has only five segments which can create a maximum of five hybrid fields, whereas the rest of the beams become normal arclets. This technique gives a similar "arc-like" beam configuration to version b, but with much less manual manipulation and splitting of beams. For the head-and-neck case, planning is more complicated since a larger number of segments is usually needed to achieve good quality IMRT plans. This is mirrored in mARC planning -- using Prowess, we also find it difficult to obtain good quality head-and-neck plans using just one single arc. We therefore create a first version IMRT plan with 18 beams and 50 segments, which has more than one segment in each beam. If the conversion to mARC shall make sense, at least one single-segment beam must exist to be converted into an arclet. We therefore split all those beams with two segments into two beams spaced 6° apart -- this version (a) is converted into mARC. A second version (b) is created using 36 beams at 10° increments and 50 segments, so that conversion into mARC results in at least 22 arclets. Head-and-neck target volumes generally require more segments for good plan quality than prostate plans, and arc treatments for these target volumes often involve two or three arcs to achieve a good dose distribution with regard to PTV coverage and sparing of organs at risk. Therefore, we performed a third inversion (version c) using 36 beam directions with 72 segments. Most beam directions now have two segments, a few have one segment or more than two. The final plan can be manually split into two rotations, one clockwise and the other counter-clockwise, which are converted separately. For this, beams with two segments can be distributed evenly to the two arcs, beams with one segment only appear in one arc, and beams with more than two segments are used to create a single-segment beam in the first arc and a hybrid field with the remainder of the segments in the second arc. The manual conversion of the two plans with different gantry rotation directions is somewhat cumbersome; however, this is intended as a proof of principle, a direct splitting and conversion to two or more arcs are envisioned for future implementation in the conversion algorithm. For all plans, direct machine parameter optimization over 40 iterations is performed in Pinnacle using the same objectives and constraints we usually apply for IMRT. CT-based dose calculation is on a 0.2 × 0.2 × 0.2 cm^3^ dose grid with the collapsed cone algorithm. Reporting of IMRT dose distributions is based on the guidelines of the International Commission on Radiation Units and Standards \[[@B8]\]. The conversion algorithm works on the RTP-file-level, which is in principle independent of the planning system used to create the original IMRT plans. Nonetheless, we demonstrate the feasibility for a second choice of planning system using a Prowess Panther prostate IMRT plan with 30 beams of one segment each. The IMRT plan was optimized in Prowess Panther using our standard criteria and the dose was calculated on a 0.3 × 0.3 × 0.3 cm^3^ grid using the collapsed cone algorithm. After exporting the treatment plan, conversion and irradiation were performed in the same way as for the Pinnacle plans. Dosimetric verification ----------------------- Each plan is converted into an mARC plan according to the conversion algorithm presented above, with a maximum arclet angle of 4°. Dosimetric verification involves comparison of the converted mARC plan with the original IMRT step-and-shoot plan and with the calculated dose from the Pinnacle TPS. Both absolute dose and fluence distribution are assessed. For the absolute dose, point dose measurements are performed in an acrylic phantom (BrainLab) with a semiflex ionization chamber (PTW type 31010) in the same way routinely performed for IMRT verification. The planar dose distributions of the step-and-shoot plan and mARC plan are compared using GafChromic film measurements inside the same acrylic phantom and PTW Octavius with 729 2D-Array, analyzed in the PTW VeriSoft software. For a comparison of the measured 3D dose distribution (both absolute dose and fluence distribution) with the calculated dose from the treatment planning system, the PTW Octavius with 729 2D-Array is used. Results ======= Planning and plan quality ------------------------- The creation of the plans in Pinnacle followed the same procedure as routinely applied for IMRT planning and was hence straightforward. Splitting the beams (version b for the prostate patient and version a for the head-and-neck example) involved some time-consuming manual changes in the plans, but without difficulties. The same applies to the head-and-neck version c, which rotates first clockwise and then counter-clockwise: again, splitting the original plan into two plans involved some manipulation of the plans, but was straightforward in principle. Two plans were then exported and separately converted into arcs, they could be irradiated in sequence without problems. The main focus of this study is the technical feasibility of the conversion of step-and-shoot plans into mARC, disregarding how the original plans were created. We have mentioned several ways just to give an example of how one may proceed from IMRT to mARC using Pinnacle, but the possibilities vary with the treatment planning system, and our examples are neither meant to be exclusive nor offer optimum advice on how best to create good quality plans with a maximum of beam angles and a minimum of segments per beam. The creation of good step-and-shoot plans for mARC conversion is not trivial particularly for complicated target volumes such as head-and-neck tumours -- in fact the up-front creation of mARC plans for these indications is complicated even when using a dedicated planning system such as Prowess. We therefore believe it is of interest to demonstrate the quality of the plans we created for our example cases, so the reader gets an impression of the outcome from the different methods. We stress again that this is not intended as a planning study and offered only as an aside, which is why in-detail analysis of quality measures (Table [2](#T2){ref-type="table"}) are omitted. ###### Quality measures of the plans presented as examples Prostate patient --------------------------- ------------------------ --------------------------------- ------- ------- PTV D~2~ (Gy) 51.73 53.98 53.59 D~50~ (Gy) 50.60 51.62 51.45 D~98~ (Gy) 48.13 45.59 44.97 HI 0.07 0.16 0.17 Rectum D~2~ (Gy) 50.60 51.05 50.55 D~30~ (Gy) 35.18 36.25 35.63 D~50~ (Gy) 24.94 21.61 19.93 Bladder D~2~ (Gy) 51.28 52.69 52.63 D~30~ (Gy) 48.97 43.06 43.12 D~50~ (Gy) 34.50 14.69 13.57 Femoral head (right) D~2~ (Gy) 32.99 37.43 34.50 D~30~ (Gy) 24.09 25.39 24.37 D~50~ (Gy) 20.49 18.46 18.13 Head-and-neck patient Measures of quality Standard IMRT plan 2 arcs FFF 7 MV (version c) PTV D~2~ (Gy) 54.47 53.32 D~50~ (Gy) 45.55 44.86 D~98~ (Gy) 50.55 50.67 HI 0.18 0.17 Right parotid D~2~ (Gy) 7.30 5.41 D~50~ (Gy) 48.37 51.82 D~mean~ (Gy) 15.36 13.62 Spinal cord D~2~ (Gy) 21.28 21.28 D~50~ (Gy) 25.36 25.25 D~mean~ (Gy) 19.77 19.30 The values presented are based on the ICRU Report 83 \[[@B8]\]. The homogeneity index is defined as HI = (D~2~ -- D~98~) / D~50~. We only give dose values for the right femoral head and right parotid, since they receive higher doses than their left counterparts. Creation of plans with very few segments per beam is generally complicated in Pinnacle, even more so when flattening-filter-free beams are used. For the prostate patient, decent quality plans could be achieved with little difficulty using IMRT with either 19 beams or 30 beams and 35 segments; in both cases the flat 6 MV plan gave better dose distribution, so we present these two plans as compared with the clinically radiated plan for the same patient (Figure [3](#F3){ref-type="fig"}). The splitting of the beams did not introduce great changes in the plans, in particular the very symmetrical and conformal prostate plans. Prostate plans version a and b are virtually identical with respect to DVH and quality measures, only few slices show a slightly different dose distribution in several voxels, which is negligible. We therefore present prostate plan versions a/b combined without distinguishing between the two. All prostate 6 MV plans are acceptable according to our in-house standard in that the PTV is covered by the 95% isodose, although coverage is slightly reduced and the maximum slightly higher (i.e. homogeneity is effectively worse) than for the treated plan. The dose to the organs at risk is similar to the original plan, with a decrease in dose to the bladder for the arc plans. Depending on whether PTV homogeneity or dose to the bladder is weighted higher, the decision would be either for the arc plans or for the originally treated plan. The FFF 7 MV plans were discarded because although satisfying all the criteria (PTV coverage and dose to organs at risk), they had higher dose gradients close to the rectum, which would endanger the rectum in the case of positioning uncertainties or intra-fraction motion. ![Example dose distributions of prostate treatment plan. A standard prostate plan (irradiated for the patient) is shown on top, the dose distributions of plan versions a/b and c (6 MV) in the middle panel, with dose volume histogram (lower panel). In the DVH, thick lines correspond to the standard IMRT plan, thin lines to prostate plan version a/b (no noteable difference between the two plans), and dashed lines to prostate plan version c.](1748-717X-8-193-3){#F3} For the head-and-neck example, little difference was observed between the plan qualities of 6 MV and FFF 7 MV plans. Possibly, this may be indebted to a better choice of inversion objectives or constraints, which is equally adequate for flat and FFF beams. In fact, the inversion objectives for the head-and-neck case were taken from a planning study comparing IMRT performance of the flat and FFF beam (Dzierma et al., in prep.), which found equally good performance of both as long as the inversion objectives were modified to be suitable for both. Since no experience exists at our institution for FFF prostate IMRT planning (FFF beams are being introduced for prostate IMRT, but no standard template of objectives has yet evolved), the poorer plan quality of the FFF beam versions in the prostate example may just be caused by a less-than-optimal choice of inversion objectives. The prostate plans a posteriori created for mARC conversion have somewhat worse quality than the original treated plan. They were included here to demonstrate several possible planning procedures; however, the optimization might still be improved to achieve better PTV coverage. We show an example of a good quality plan created for clinical treatment in Figure [4](#F4){ref-type="fig"}. ![Example dose distribution and DVH of a clincally treated prostate plan (PTV IMRT planned with 30 beams and 30 segments and converted to mARC for treatment). Prescription was for the 95% isodose.](1748-717X-8-193-4){#F4} In the head-and-neck case, plan versions a and c had acceptable qualities. Version b, with 50 segments distributed over 36 beam directions, gave worse quality than both version a (50 segments with 18 beam directions, giving more opportunity of modulation for each beam) and version c (36 beams with 72 segments), even though the DVH is similar for all six plans (both 6 MV and FFF 7 MV, Figure [5](#F5){ref-type="fig"}). Version c had best coverage and least extension of the 80% isodose outside the PTV, which is why we chose to present this plan (FFF 7 MV, which has comparable quality to the 6 MV version c plan, compared with a "standard" IMRT plan using 7 beams and 50 segments (Figure [6](#F6){ref-type="fig"}, Table [2](#T2){ref-type="table"})). ![Dose volume histogram (DVH) of head-and-neck patient plans. The thick dashed line shows the DVH for a standard IMRT plan generally used at our institution. The thick line corresponds to the FFF 7 MV plan version c, displayed in Figure [6](#F6){ref-type="fig"}. Thin lines correspond to the other plan versions a,b,c (6 MV and 7 MV) -- these plans are not analyzed in detail here and are therefore not distinguished in the figure.](1748-717X-8-193-5){#F5} ![Example dose distributions of head-and-neck treatment plan. The left panel shows our standard IMRT plan (also displayed in Figure [5](#F5){ref-type="fig"} as thick lines in the DVH), the right panel shows version c converted plan (FFF 7 MV), both in different CT slices.](1748-717X-8-193-6){#F6} Conversion and dosimetric verification -------------------------------------- The conversion of the plans to mARC format -- performed as described above -- yielded functional plans with no further modifications, which could be imported and irradiated without incidences. There was no difference in handling or feasibility between IMRT plans created in Philips Pinnacle and Prowess Panther. For each plan version, we perform dosimetric testing of the original step-and-shoot plan versus the converted mARC plan. Absolute dosimetric verification of both the step-and-shoot and mARC plans by point dose measurements showed deviations below 5% local dose, which is within the specifications allowed at our institution for IMRT verification. The absolute dosimetric verification of mARC plans deviated from step-and-shoot plans by no more than 1%. Planar dose distributions of both plan varieties (step-and-shoot vs. mARC) were measured in an acrylic phantom using GafChromic film. The agreement between both treatment modalities is excellent; an example is shown in Figure [7](#F7){ref-type="fig"} for prostate plan version c -- this is the version with least hybrid fields and most arclets, for which the difference between both plans is maximum. For this plan, 97.3% of the points pass the criteria of maximum 5% deviation in dose and 3 mm distance to agreement. ![Dosimetric verification of mARC vs. step-and-shoot original plan. Top: Gamma distribution of GafChromic film measurement of the converted mARC plan (version c), compared with the original step-and-shoot plan. Pass criteria are 5% deviation in local dose and 3 mm distance to agreement, which are satisfied by 97.3% of measurement points. Bottom left: Example left-right profile, bottom right: target-gantry profile measured in both films (orange line: original step-and-shoot plan, blue line: mARC converted plan).](1748-717X-8-193-7){#F7} A comparison of the 3D-dose distribution of the step-and-shoot plans with the converted plans was performed using the PTW Octavius phantom with 729 2D-Array. For all measurements, between 93% and 100% of the points satisfied the pass criteria (again 5% deviation in dose and 3 mm distance to agreement). In principle, validation of the treatment planning system is already proven by the fact that dosimetric verification for IMRT plans was checked in the commissioning phase \[[@B9]\] and in routine quality assurance, and since the deviation of the mARC plans from the step-and-shoot plans is negligible. However, we demonstrate these measurements for completeness. Besides, in the clinical routine it is more practical to perform routine mARC verification measurements directly (mARC plan vs. planning system) rather than the two-step approach of verifying the IMRT dose distribution and then comparing this with the converted plan. Verification of the mARC dose distribution before treatment is carried out using the PTW Octavius phantom with 729 2D-Array, which measured the irradiated mARC three-dimensional dose distribution with the treatment planning system. For all plans, over 95% of the points pass the criteria of 5% deviation in local dose and 3 mm distance-to-agreement (example shown in Figure [8](#F8){ref-type="fig"}), which provides an independent verification of the dose distribution of the mARC plan. ![Dosimetric verification of mARC vs. treatment planning system. Gamma distribution of Octavius 729 2D-Array measurement of the converted mARC plan (version c), compared with the dose distribution exported from the Pinnacle TPS. One transversal and one sagittal slice are shown. Pass criteria are 5% deviation in local dose and 3 mm distance to agreement, which are satisfied by over 95% of measurement points in all slices.](1748-717X-8-193-8){#F8} Treatment times were measured for all plans, an overview is given in Table [3](#T3){ref-type="table"}. At our institution, standard prostate PTV plans take between 5 and 8 minutes to irradiate, whereas mARC plans take about 5 minutes for the flat 6 MV beam (with a maximum dose rate of 300 MU/min) and around 3 minutes for the flattening-filter-free 7 MV beam (using a maximum dose rate of 2000 MU/min). Even for the original IMRT plan with 19 beam directions, which was not adapted to a near-rotational setting, the treatment time was noticeably reduced by conversion to mARC (8:40 min to 5:15 min for 6 MV and 7:20 min to 3:35 min for FFF 7 MV, respectively). ###### Treatment time comparison Prostate patient ---------------------------------------------------- -------------- ---------- Standard IMRT plans at our institution (6 MV flat) 5-8 minutes Converted mARC plans 6 MV FFF 7 MV Version a 05:15 03:35 Version b 05:00 03:30 Version c 05:05 02:55 Head-and-neck patient Standard IMRT plans at our institution (6 MV flat) 7-11 minutes Converted mARC plans 6 MV FFF 7 MV Version a 04:45 04:50 Version b 05:45 05:20 Version c 08:20 07:35 Treatment time (in minutes) with the different versions converted mARC for 6 MV and FFF 7 MV energies, compared with times usually observed at our institution for IMRT plans. Depending on the number of gantry directions and amount of segments, standard head and neck IMRT plans usually require between 7 and 11 minutes radiation time at our institution. The converted plans lie within the range of 5 to 8 minutes, depending on energy and beam configuration. Clearly, a trade-off exists between treatment time and plan quality. The best dose distributions were found for the double arc treatment, which takes longest. The version c plans do not present a large advantage over standard plans from the point of view of treatment times (7--8 minutes), but have a very good dose distribution; if the conditions on the dose distribution are relaxed, faster treatment (4--5 minutes) becomes possible. This is a medical decision faced both for arc and IMRT treatment; in all cases, some time can be saved by switching from step-and-shoot plans to mARC. Discussion ========== We have explained a conversion algorithm which operates on any step-and-shoot IMRT plan to output a converted mARC plan. Even though only two patient examples are shown for a proof of principle, we have applied this technique in the clinic for a number of prostate patients. For all these patients, the 3D dose distribution was examined in detail before treatment, and with negligible dosimetric deviation similar to those presented in this manuscript. The implementation presented here offers the practical advantage that it can be applied to all treatment planning systems capable of IMRT optimization and DICOM export, which evidently expands the applicability of the mARC technique to a large range of Siemens customers who rely on TPS other than Prowess. While in general every TPS can thus be used for the creation of mARC plans, it must be mentioned as a precaution that not every TPS is easily capable of achieving good optimization results for a large number of beams (compare \[[@B10]\]). It is not trivial to achieve good plans for all kinds of target structures and beam choices. For example, Pinnacle will more easily create good head and neck plans with 11 beams and 50 segments than using 36 beams with 36 (or even 50 segments), so special care must be taken in the choice of objectives to ensure good plan quality. In fact, it is well known in arc treatments that for complicated target volumes, one arc may not be enough, but two or three non-coplanar arcs are needed (e.g., \[[@B11],[@B12]\]). The demands posed by a scenario as this may exceed the optimization capabilities of many non-arc-oriented TPS. However, simpler arc treatments can more easily be implemented and already provide large relief to both the patients and clinical schedule. Furthermore, the mARC conversion can be applied not only to create perfect mARC plans, but provides an intermediate planning option -- a "poor man's hybrid arc" similar to our example using 19 beams. These plans can be optimized adequately by most IMRT-planning TPS; after conversion into "mARC" their treatment time will be noticeably improved. For Pinnacle, we have obtained sufficiently good quality prostate plans with 30 beam angles, and high-quality head-and-neck plans with 72 segments from 36 directions, converted into two coplanar arcs. These plans result in stable and fast mARC treatment. For the creation of good-quality IMRT plans, it might also be tested to first split beams with two or more segments as demonstrated for prostate plan versions b and the head-and-neck plan versions a and c, and then re-optimizing again after the splitting, similar to the VMAT-approach described by \[[@B9]\]. An additional practical advantage of our implementation is that a step-and-shoot back-up plan is created for all mARC treatments (simply the original plan). At our institution, there are three matched linacs with 6 MV energy, but only one with mARC facility. In case of malfunction of the mARC linac, the mARC plan need not be recomputed and re-optimized as a step-and-shoot alternative version, but the original plan can be simply shifted to another linac, with no change in applied dose. In implementations similar to the Pinnacle planning system, where IMRT is performed with a maximum number of segments distributed over all beam angles, the method offers the possibility to investigate beam-angle-optimization, which is a non-trivial problem in modern IMRT planning (for some recent approaches to this problem, compare, e.g., \[[@B13]-[@B15]\]). In our approach, the directions for hybrid arcs need not be pre-defined by the user, but are determined by the inversion algorithm in a way that the TPS assigns segments to those beam directions where they are "most needed". Conclusions =========== We present an in-house conversion algorithm for converting IMRT step-and-shoot plans from any treatment planning system (Philips Pinnacle and Prowess Panther, in our example) into mARC plans. Creation of good quality IMRT plans using a large number of beams with few segments requires particular attention depending on the tumor location, but can be achieved by good design of beam geometry and inversion objectives, as is demonstrated for a clinically treated example. The converted mARC plans can be irradiated by the Siemens Artiste with no difference to "real" mARC plans. The feasibility and dosimetric equivalence is demonstrated for the example of a prostate and a head-and-neck patient, using sets of different plans. Competing interests =================== The authors declare that they have no competing interests. Authors' contributions ====================== NL conceived the idea and general principle of the conversion algorithm. YD programmed the algorithm, implemented it in practice, performed all planning and plan comparison, and drafted the manuscript. Both YD and FN carried out the dosimetric verification measurements and discussed all issues related to the planning techniques and manuscript structure. All authors participated in the design of the study and read and approved the final manuscript. Acknowledgements ================ We are indebted to three anonymous reviewers for their constructive comments and suggestions, which have improved this manuscript. Patrick Melchior provided information on planning criteria. We are grateful to Martin Thorwart for his help in installing tcsh and awk and many helpful hints in scripting.	{ "pile_set_name": "PubMed Central" }
1. Virus-Like Particles (VLPs)---Brief Overview {#sec1-vaccines-06-00037} =============================================== Virus-like particles (VLPs) are supramolecular complexes formed by viral proteins that self-assemble into capsids resembling real viruses, albeit non-infectious due to the lack of viral genome packaged within the particle. Such non-infectious particles can be naturally formed during infection or genetically engineered and produced in large scales in laboratories. Most VLPs are characterized by stability, together with uniform and repetitive structure that allow diverse applications in the field of biomedical science. Since their early discovery five decades ago, VLPs led to enormous progress in the field of vaccine development with seven vaccines approved for humans and several others in pre-clinical phase or in clinical trials \[[@B1-vaccines-06-00037],[@B2-vaccines-06-00037]\]. Most VLPs are formed by proteins naturally forming nucleocapsids and consist of several copies of monomers in quasi-equivalent conformations forming icosahedral or helical (rod-shaped) structures. The final shape of VLPs usually imitates the symmetry of the original parental virus but can vary according to the nucleic acid content or biophysical context during the assembly phase \[[@B3-vaccines-06-00037]\]. The virus-like property of VLPs is a potent geometric pathogen-associated structural pattern (PASP) facilitating their engagement with innate and adaptive immune cells. The exterior and interior facets of VLPs can be functionalized and modified to enhance their immunogenicity and stability or to present heterologous antigens ([Figure 1](#vaccines-06-00037-f001){ref-type="fig"}). The viral capsids have evolved to preferentially package their own viral genome. Nucleocapsid-derived VLPs retain this property of packaging nucleic acids, and since the viral genome is absent, VLPs will package nucleic acids from host cells during the expression process. Interestingly, these naturally packaged nucleic acids can be substituted by other anionic polymers to modulate the immune response by engagement of different pattern recognition receptors (PRRs) \[[@B4-vaccines-06-00037],[@B5-vaccines-06-00037],[@B6-vaccines-06-00037]\]. The outer surface of VLPs has also been functionalized with heterologous antigens by chemical crosslinking, genetic fusion or peptide splicing techniques aiming to enhance the immunogenicity of coupled antigens \[[@B6-vaccines-06-00037],[@B7-vaccines-06-00037]\]. 2. VLPs Interaction with the Innate Immune System {#sec2-vaccines-06-00037} ================================================= VLPs---especially those packaging nucleic acids---are capable of inducing a comparable humoral immune response to viruses. This is considered a major advantage of VLPs over attenuated virus-derived vaccines as it guarantees safety of the vaccines whilst retaining their immunogenicity \[[@B8-vaccines-06-00037]\]. The immune system in general recognizes and interacts with viral capsid-derived VLPs on the basis of two major characteristics, size and surface geometry \[[@B2-vaccines-06-00037]\] as discussed next. 2.1. Drainage of VLP into the Lymphatoid Organs {#sec2dot1-vaccines-06-00037} ----------------------------------------------- The lymphatic system regulates the balance of fluid in the body as well as the trafficking of immune cells and antigens to lymphoid organs, in particular lymph nodes (LNs) \[[@B9-vaccines-06-00037]\]. The efficient draining of antigens and particles from the periphery to the primary and secondary lymphoid organs guarantees the encounter of such particles with the relevant immune cells that will initiate the immune response. Thus, draining properties of antigens have an important impact on its overall immunogenicity \[[@B10-vaccines-06-00037],[@B11-vaccines-06-00037]\]. The diameter of the blind-ended structure of the initial lymphatic vessels is about 10--60 μm. The vessels are lined up with single flattened endothelial cells forming valves that ensure the unidirectional movement of fluids through and along the lymphatic vessels \[[@B12-vaccines-06-00037]\]. Furthermore, the initial lymphatic vessels are highly permeable due to the presence of button-like junctions or gaps between the endothelial cells \[[@B13-vaccines-06-00037]\]. Particulate materials can directly diffuse through the 200 nm pores of the lymphatic vessels' walls with an optimal size between 30--40 nm \[[@B14-vaccines-06-00037]\]. VLPs range between 20--200 nm in size, which allows them to freely drain into the lymphatic system. The kinetics of free drainage of viral capsid-derived VLPs has been studied extensively and previous data has shown that VLPs with average size of 30 nm can be detected in mice footpads for at least 40 min post injection and at the draining LN 2 h post injection \[[@B14-vaccines-06-00037]\]. Recently, with the advancement of imaging techniques it has been shown that 30 nm VLPs can accumulate at the popliteal draining LN in less than 10 min post subcutaneous injection in murine footpads \[[@B15-vaccines-06-00037]\]. In addition to free lymphatic drainage, VLPs can also be actively transported via specialized cells such as skin-derived macrophages and DCs, mainly dermal DCs (dDCs) and Langerhans cells (LCs). These cells, especially upon activation, can efficiently squeeze through the button-like junctions of the endothelial cells lining lymphoid vessels and migrate through the lymph along a CCL19/CCL21 gradient towards the draining LN. Previous studies have shown that skin-derived DCs---CD11c^hi^CD40^hi^CD8^−^ and CD11c^int^CD40^hi^CD8^−^ DCs---can successfully uptake VLPs after intradermal injection. Both identified subsets of skin-derived DCs have also been shown to effectively cross-prime cytotoxic T-cells following VLPs uptake \[[@B16-vaccines-06-00037]\]. Even though, CD8^+^ DCs are usually much more potent in cross-presenting VLP-derived antigens \[[@B17-vaccines-06-00037]\]. In summary, the size of antigen influences the immunogenicity largely by two mechanisms: first, by allowing the appropriate draining from periphery to secondary lymphoid organs and second by facilitating the interaction with APCs due to their repetitive surface. VLPs size distribution falls within the aforementioned optimal range, which explains in part its remarkable immunogenicity. 2.2. Trafficking of VLPs Within Draining LN {#sec2dot2-vaccines-06-00037} ------------------------------------------- In general, particles ranging between 20--200 nm upon reaching the LN are distributed throughout the different areas of the draining LN. Previous studies have shown that VLPs are detected at the sub-capsular, medullar and cortical regions of LNs as fast as 2 h post subcutaneous injection. At a later stage, VLPs are also detected deeper in B-cell follicles bound to follicular dendritic cells (fDCs) ([Figure 2](#vaccines-06-00037-f002){ref-type="fig"}). At 48 h post injection VLPs were distributed in the sub-capsular area, cortex, and para-cortex of the draining LN and some were inside B-cell follicles \[[@B14-vaccines-06-00037]\]. Upon arrival at the draining LN, several LN residing myeloid cells and B-cells will carry out the active transportation of free VLPs from the sub-capsular sinuses to B-cell follicles \[[@B18-vaccines-06-00037],[@B19-vaccines-06-00037],[@B20-vaccines-06-00037],[@B21-vaccines-06-00037]\]. The pattern of distribution of VLPs allows access and interaction of particles with different cell types in the secondary lymphoid organs, which contributes to the immune response generated against those particles. The predominant subsets of APCs actively up taking viral capsid-derived VLPs in the popliteal LN post subcutaneous injection have been recently classified. These cells include the sub-capsular sinus (SCS) macrophages CD11b^+^F4/80^+^, different subsets of conventional dendritic cells (cDCs) including CD8^+^CD11c^+^, CD8^−^CD11b^+^, and CD8^+^CD11c^+^. B-cells characterized by CD45R/B220^+^ were less efficient in up taking VLPs \[[@B15-vaccines-06-00037]\] but are likely responsible for their transport to fDCs in B-cell follicles \[[@B21-vaccines-06-00037]\]. Medullary DCs within lymphoid organs would also participate in the transportation of antigens by binding them to their surface receptors such as SIGN-R1 (CD209b) \[[@B20-vaccines-06-00037]\]. Other studies have shown that SIGN-R1 and DC-specific ICAM3-grabbing non-integrin receptors (DC-SIGN) have also a role in antigen transportation and presentation. In addition, DCs can also capture VLPs and present VLP-derived peptides in the paracortical T-cell zone resulting in the activation of T and B-cells in the extra-follicular area \[[@B22-vaccines-06-00037]\]. Initial studies have shown that skin-derived DCs which have encountered antigens in the skin preferentially colonize specific areas within the draining LN. dDCs are known to migrate to the paracortex underneath B-cell zone, whereas LCs migrate and colonize the inner-paracortex \[[@B23-vaccines-06-00037]\]. Therefore, actively transported VLPs by skin-derived DCs either dDCs or LCs would follow similar migration pattern to T-cell zone and thereby are considered to be key initiators of T-cell immunity. A study suggested that skin-derived DCs act as initial transporters of herpes simplex virus (HSV) to non-migratory LN resident CD8^+^ DCs for effective CTL priming \[[@B24-vaccines-06-00037]\]. As described above, the draining and trafficking of VLPs within the immune system and subsequent interaction with immune cells such as DCs, macrophages, and B cells allow the initiation of both humoral and cellular immune responses. Details of the immune response to VLPs used as vaccines will be discussed in the following sections. 2.3. VLPs and Innate Humoral Immune Response {#sec2dot3-vaccines-06-00037} -------------------------------------------- The highly repetitive surface structure of viral capsid-derived VLPs facilitates their interaction with components of the innate humoral immune system that in turn mediates opsonization and phagocytosis by APCs ([Figure 3](#vaccines-06-00037-f003){ref-type="fig"}). The natural pentameric IgM antibody can efficiently bind to VLPs via low affinity/high avidity interactions \[[@B2-vaccines-06-00037],[@B21-vaccines-06-00037]\]. It has been shown that even if the interaction between a single viral coat sub-unit with a single recognizing element on the multimeric molecule is weak, the overall avidity of the interactions of the antibodies with the entire VLP would be considerably stronger thanks to the repetitive surface of the particles \[[@B21-vaccines-06-00037]\]. The pentameric IgM alone is not an efficient direct activator of the opsonization mechanism. However, IgM contributes to the activation of C1q molecule upon binding. Activation of C1q consequently leads to the activation of the classical complement cascade initiated by the formation of C1-complex consisting of C1q and serine proteases C1r and C1s molecules. C1q can also bind directly to the surface of VLPs causing a conformational change in C1q molecule leading to the activation of classical complement cascade. In our hands, however, both deficiency in IgM as well as C1q led to a complete failure of VLP-deposition on fDCs \[[@B21-vaccines-06-00037]\]. Such findings indicate that natural IgM binds to VLPs followed by recruitment of C1q. Interestingly, complement may not always enhance activation of innate cells. Papaya Mosaic Virus VLPs for example induce marked production of IFN-α in pDCs and complement inhibits rather than promotes this process \[[@B25-vaccines-06-00037]\]. The pentraxin protein family including the short pentraxin C-reactive protein (CRP), serum amyloid P (SAP) and long pentraxin (PTX3) may also recognize the repetitive structure of VLPs due to high avidity interactions. SAP, for example, has been shown to bind to influenza A viral particles \[[@B26-vaccines-06-00037]\]. This recognition may also result in high affinity binding to C1q molecule and effective initiation of the classical complement cascade \[[@B2-vaccines-06-00037]\]. More research, however, is necessary to study the interaction of the innate humoral immune system and VLPs. The hallmark of the immune response induced by VLPs is the induction of high levels of long lasting humoral responses. In contrast to soluble antigens, the highly organized and repetitive structure of VLPs promotes crosslinking of BCRs that surpasses the activation threshold and bypasses the initial need of T-cell help. As a consequence, any antigen exposed on the surface of VLPs in this same organized and repetitive array will benefit from this viral fingerprint and induce high levels of humoral responses. Another important feature of the humoral response elicited by VLPs is the promotion of isotype switching dependent on direct TLR7 or TLR9 signaling on B-cells \[[@B27-vaccines-06-00037],[@B28-vaccines-06-00037]\]. Upon BCR-mediated endocytosis of VLPs, the nucleic acid that is often packaged within VLPs reaches the endocytic compartment where it can interact and activate endosomal TLRs, the downstream response leads to isotype switching of antibodies to IgG2 in mice and IgG1 in humans, the isotypes with higher effector function. Furthermore, intranasal administration of VLPs in murine models resulted in strong B-cell responses and germinal center formation in the spleen. Roaming B-cells in the lungs have been shown to bind VLPs through their BCR and shuttle them to B-cell follicles within the spleen \[[@B29-vaccines-06-00037]\]. 2.4. Efficient Presentation of VLPs by Both MHC Pathways {#sec2dot4-vaccines-06-00037} -------------------------------------------------------- Following the uptake by professional APCs, processed VLPs will be subsequently presented on MHC molecules. In general, exogenous antigens (endocytosed or phagocytosed) are presented on MHC-II molecules to trigger the activation of T~h~ CD4^+^ cells following endosomal processing \[[@B30-vaccines-06-00037]\]. Endogenous antigens on the other hand are loaded on MHC-I molecules for CD8^+^ CTLs priming. The classical source of MHC-I antigens is newly synthesized proteins which are misfolded and degraded in the proteasomes. Nevertheless, it has been found that both presentation pathways are not strictly separated \[[@B31-vaccines-06-00037]\] as exogenous antigens can also reach MHC-I presentation pathway in a process called cross-presentation \[[@B32-vaccines-06-00037],[@B33-vaccines-06-00037],[@B34-vaccines-06-00037]\]. Cross-presentation can be TAP dependent or independent. In TAP dependent pathways the antigen will be taken by endosomes and leak to the cytosol afterwards (endosome to cytosol pathway). In TAP independent pathway, peptide load MHC-I in the endosomes directly as an exchange process under acidic conditions, such a process may be referred to as direct endosomal loading pathway \[[@B34-vaccines-06-00037],[@B35-vaccines-06-00037],[@B36-vaccines-06-00037]\]. The ability of exogenous VLPs to be presented on MHC-II molecules to generate protective IgG antibody titers via activating T~h~ cells has been previously documented to be highly efficient and biologically relevant. Cross-presentation on MHC-I molecules has also been studied extensively and has been documented for tumor cells, necrotic cells, apoptotic cells, viruses, and VLPs \[[@B34-vaccines-06-00037],[@B37-vaccines-06-00037],[@B38-vaccines-06-00037],[@B39-vaccines-06-00037]\]. The efficiency of VLPs presentation via MHC-I vs. MHC-II molecules in vivo was quantified \[[@B34-vaccines-06-00037]\]. This was done by using immunodominant MHC-I or MHC-II antigens derived from lymphocytic choriomeningitis virus (LCMV) linked to VLPs. The results indicated that cross presentation on MHC-I molecules for VLPs containing antigens was only 1 to 10-fold less efficient than classical presentation on MHC-II molecules. This finding indicates that VLPs as exogenous particles can be loaded efficiently onto both MHC molecules \[[@B34-vaccines-06-00037]\]. It has been reported that exogenous antigens can reach MHC-I molecules in CD8^+^ DCs only while CD8^−^ DCs present them on MHC-II molecules \[[@B40-vaccines-06-00037]\]. However, this dichotomy hypothesis was not supported in other studies which have shown that both CD8^+^ and CD8^−^ DCs can efficiently present VLP-derived antigens on MHC-II molecules, however only CD8^+^ DCs were able to cross-present VLPs derived peptides. This implies that VLPs presentation on MHC-II molecules to CD4^+^ T-cells is not restricted to a specific DCs subset but cross-presentation to CD8^+^ T-cells is usually restricted to a CD8^+^ DCs subset \[[@B17-vaccines-06-00037]\]. 2.5. Packaging VLPs with Innate Immune-Modulators {#sec2dot5-vaccines-06-00037} ------------------------------------------------- Many VLPs, including Norovirus, Papaya Mosaic Virus, Human Papilloma Virus and Qβ-derived VLPs activate innate cells through various means \[[@B25-vaccines-06-00037],[@B36-vaccines-06-00037],[@B41-vaccines-06-00037]\]. An important reason for the ability of RNA virus nucelocapsid derived VLPs to activate DCs and B-cells is the fact that they package RNA. Even though these VLPs lack infectious genome, they usually package nucleic acids (RNA) during the assembly process in host cells. Moreover, VLPs can be efficiently packaged in vitro with a range of charged polyanionic sequences such as RNA with specific secondary modification or CpGs, for instance. Nucleic acids are recognized and can activate PRRs, which modulates and alters the adaptive immune response \[[@B42-vaccines-06-00037]\]. VLPs are reportedly inefficient at inducing strong antigen-specific cytotoxic response, however the addition of other certain classes of PRR ligands such as non-methylated CpGs can efficiently activate TLR9 leading to robust T~h~1 and CTL response \[[@B43-vaccines-06-00037]\]. Non-methylated CpGs are classified into three different categories according to their structure and the immune response they induce. In general, CpGs of all three classes can activate inflammatory cytokines as well as type-I IFN at different levels upon stimulation of TLR9 in plasmacytoid dendritic cells (pDCs) \[[@B44-vaccines-06-00037]\]. Class A CpGs are characterized by the presence of phosophodiester poly G sequence at the 5′ and 3′ ends as well as an internal palindrome sequence in the central portion. Class A CpGs are capable of producing large amounts of type-I IFN in particular by pDCs \[[@B45-vaccines-06-00037]\]. In contrast, class B CpGs characterized by a full phosphorothioate backbone can induce type-I IFN production to a lesser extent than class A. However, class B CpGs is capable of stimulating pro-inflammatory cytokines, such as IL-12 \[[@B46-vaccines-06-00037]\]. The homologous TLR7/8 expressed within the endosomal compartments of APCs can recognize guanine nucleotide-homologues such as loxoribine, R848, imidazouinoline components, and uridine/guanosine rich ssRNA \[[@B47-vaccines-06-00037]\]. Free RNA is usually subjected to degradation by the RNase enzyme in the extracellular spaces. Accordingly, the possibility that free RNA will reach the endosomal compartments of APCs to interact with and activate endosomal TLRs is quite low. Therefore, packaging ssRNA into VLPs protects it from extracellular RNase and results in efficient activation of TLR7/8. TLR3 is mainly expressed in cDCs but not pDCs and recognizes dsRNA. The synthetic RNA sequence poly I:C, leads to a type-I IFN response and production of proinflammatory cytokines \[[@B48-vaccines-06-00037]\]. It has been reported that VLPs formulated with Poly I:C can lead to an improved CTL response \[[@B49-vaccines-06-00037]\]. In general, recognition of ssRNA, dsRNAm or CpGs packaged into VLPs by endosomal TLRs will result in the activation of signaling pathways that are essential for the expression of different genes required to initiate the inflammatory responses ([Figure 4](#vaccines-06-00037-f004){ref-type="fig"}). The process starts by ligand-induced TLR dimerization to bring TIR domains in close proximity and facilitate the recruitment of protein kinases. This will activate several major transcription factors such as NF-κB, AP-1, IRF3, and IRF7 \[[@B50-vaccines-06-00037],[@B51-vaccines-06-00037],[@B52-vaccines-06-00037],[@B53-vaccines-06-00037]\]. NF-κB and AP-1 will stimulate the expression of several inflammatory genes including cytokines (TNF and IL-6), chemokines (CCL2 and CXCL8), some endothelial adhesion molecules such as E-selectin and other costimulatory molecules including CD80 and CD86. On the other hand, IRF3 and IRF7 are responsible for the expression of type-I IFN α/β genes. TLR7 and TLR9 are MyD88-dependent and IRF-independent signaling pathways, and both are capable of activating NF-κB and IRFs transcription factors. In contrast, TLR3 acts through the activation of TRIF which will induce IRF3 responsible for the expression of type-I IFN \[[@B52-vaccines-06-00037],[@B53-vaccines-06-00037]\]. Apart from TLRs, other PRRs and nucleic acid sensors have also been reported to play a role in the immune response elicited by VLPs. RIG-I and the stimulator of IFN genes protein (STING) induces significant levels of type-I IFN upon viral infection mediating autocrine and paracrine signaling \[[@B54-vaccines-06-00037]\]. Activation of STING results in the activation of different transcription factors including STAT6 and IRF3 \[[@B55-vaccines-06-00037]\]. Although STING and RIG-I mediated responses are often associated with intermediates of viral replication, there are evidences that envelope derived VLPs may fuse with cell-membranes which can be sensed by the innate immune cells in murine and human cells. This fusion can activate a STING-dependent signaling pathway which induces the formation of Tank Binding Kinase Complex protein (TBK1) which also involves IRF3. This pathway will finally activate the expression of type-I IFN as well as other IFN stimulating genes (ISGs) \[[@B56-vaccines-06-00037]\]. Recently, it has been shown that by including cyclic di-nucleic (CDNs) acids into VLPs, the CTL response against tumors was improved in a cGAS and STING dependent manner in comparison to the effect observed with the addition of the CDNs alone \[[@B57-vaccines-06-00037]\]. This highlights the benefits of packaging of nucleic acid into VLPs, which has consistently shown to improve pharmacokinetics of the adjuvants and the observed immune response. 2.6. Important Considerations on VLP Based Vaccines {#sec2dot6-vaccines-06-00037} --------------------------------------------------- VLPs had been consistently successful at inducing protective immune responses in various pre-clinical models and clinical trials both for infectious and non-infectious diseases. The field had started employing naturally occurring VLPs during infection as vaccines, as exemplified by the Hepatitis B vaccine \[[@B58-vaccines-06-00037]\] and has now expanded to well stablished platform VLPs such as the well described Qβ \[[@B59-vaccines-06-00037]\], AP205 \[[@B41-vaccines-06-00037]\], MS2 \[[@B60-vaccines-06-00037]\] to name a few, that had been applied as vaccines against diseases ranging from nicotine dependency \[[@B61-vaccines-06-00037]\] to asthma \[[@B62-vaccines-06-00037]\] and hypertension \[[@B63-vaccines-06-00037]\]. The success of this approach relies on presenting heterologous antigens to the surface of model VLPs and in this manner, conferring the highly immunogenic viral fingerprint to those antigens. There are two main methods for presentation of heterologous antigens: chemical crosslinking or genetic fusion. The choice of method will impact the quality of the immune response as it will affect the valency of decoration of the particle with the antigen and the overall stability of the vaccine. For that reason, small changes in the crosslinking method must be closely evaluated as it can directly affect immunogenicity. A second important consideration for VLPs is the expression system of VLPs. During the assembly phase VLPs will randomly pack host derived components that can have an unbeknownst impact in the immune response. Thus, the same VLP can exert different immune responses depending on the manufacturing process. This is an important consideration when comparing results across different groups and when scaling up the manufacturing process. 2.7. Challenges for VLP-Based Vaccine Development {#sec2dot7-vaccines-06-00037} ------------------------------------------------- As is common to other vaccine development platforms, there are several challenges that should be considered for successful development of VLP-based vaccines. There are some risks that are specific to the VLP-platform technologies and some risks that are related to the specific vaccine candidates. Challenges specific to VLP-based vaccine platforms - Even though several VLP-based vaccines are on the market, some more recent candidates struggle with stability. In addition, no vaccine that displays foreign epitopes has made it to the market so far. Hence, real-life, market PoC for such vaccines is missing. While there is no a priori reason that this should not be possible, it may still be perceived as a potential risk. - Most if not all nucleocapsid VLPs derived from RNA viruses package RNA from the production host cells. This may need an additional Quality Control effort. - If epitopes are to be fused into VLPs, this can create substantial problems, as VLPs may not assemble anymore. Challenges for individual vaccines - The selected epitope may not be protective - Induced immune responses may be too low - The selected indication may sound interesting but does not attract interest from industry and/or the end-customer 3. Conclusions {#sec3-vaccines-06-00037} ============== The highly immunogenic properties of VLPs are a direct consequence of the viral fingerprint retained from the parental viruses. Such properties can be summarized by size, surface geometry, and ability to package nucleic acids. VLPs immunogenicity can also be extended by presenting heterologous antigens on their surface. All author contributed to the writing and editing of the manuscript. The researchers involved in this publication were supported by the Qatar Foundation, the National Center of Research (CNPq, Brazil) and the Swiss Cancer League (LFS-4132-02-2017). The authors declare no conflicts of interest. VLPs Virus-Like Particles PRR Pattern Recognition Receptor PASP Pathogen-Associated Structural Pattern TLR Toll-Like Receptor LN Lymph Node SCS Sub-Capsular Sinus DC Dendritic Cell cDC Conventional Dendritic Cell pDC Plasmacytoid Dendritic Cell dDC Dermal Dendritic Cell fDC Follicular Dendritic Cell LC Langerhans Cell APC Antigen-Presenting Cell CTL Cytotoxic T-Lymphocyte CRP C Reactive Protein PTX3 Pentraxin-Related Protein MHC-I Major Histocompatability Class I MHC-II Major Histocompatibility Class II LCMV Lymphocytic Choriomeningitis Virus dsRNA Double Stranded RNA ssRNA Single Stranded RNA TNF Tumor Necrosis Factor IFN Interferon NF-κB Nuclear Factor kappa-light-chain-enhancer of Activated B-Cells AP-1 Activator Protein 1 IRF Interferon Regulatory factor CCL2 Chemokine (C-C motif) Ligand 2 CXCL8 Chemokine (C-X-C motif) Ligand 8 MyD88 Myeloid Differentiation Primary Response Gene 88 TRIF TIR-Domain-Containing Adapter-Inducing Interferon-β ![A schematic diagram of nucleocapsid-derived VLPs illustrating some modifications that can enhance their immunogenicity. (1) packaging of nucleic acids and (2) displaying heterologous proteins/epitopes or functional molecules on the outer surface.](vaccines-06-00037-g001){#vaccines-06-00037-f001} ![Distribution of VLPs in the draining LN. VLPs arrive in the LNs by natural drainage in a cell-free manner or transported by APCs. Upon reaching the sub-capsular sinus of the draining LN, VLPs will be phagocytized by different APCs and preferentially reach the B and/or T-cell zones. VLPs endocytosed in the periphery by APCs preferentially reach the T-cell zone in the draining LN for T-cell priming.](vaccines-06-00037-g002){#vaccines-06-00037-f002} ![VLPs and innate humoral immune response. The repetitive surface of viral capsid-derived VLPs facilitates their interaction with molecules and components of the innate humoral immune system. (A) IgM binds the surface of VLPs via low affinity/high avidity interaction. This binding activates C1q molecules resulting in the initiation of the classical complement cascade. (B) More speculative, pentraxin protein family members including CRP, amyloid P SAP and PTX3 may also recognize and bind the repetitive surface structure of VLPs resulting in the activation of C1q molecule and the initiation of the classical complement cascade.](vaccines-06-00037-g003){#vaccines-06-00037-f003} ![Packaging VLPs with innate immune-modulators. Viral capsid-derived VLPs packaging different nucleic acids can be efficiently internalized by APCs. The packaged nucleic acids will be released in the endosomal compartment of the cell following degradation of VLP-protein shell. ssRNA is TLR7/8 ligand, CpGs is TLR9 ligand and dsRNA is TLR3 ligand. Ligand induced TLR-dimerization activates transcription factors such as NF-κB, AP-1, and IRFs and induces the expression of inflammatory genes and type-I IFN genes.](vaccines-06-00037-g004){#vaccines-06-00037-f004}	{ "pile_set_name": "PubMed Central" }
*Background.* Scrub typhus is a mite-borne rickettsial disease caused by Orientia tsutsugamushi in endemic areas, and a public health concern for a population of over a million new cases emerging and a billion people at risk of infection. Although doxycycline remains the standard therapy, some reported effectiveness of levofloxacin as alternative regimen when treatment with doxycycline fails. However, there are clinical studies that quinolone is not effective in patients with scrub typhus. To clarify these discrepant results, we evaluated the genotype and mutation in gyrA associated with quinolone resistances. *Methods.* This prospective observational study enrolled patients admitted to a tertiary hospital with scrub typhus from 2008 to 2012. 27 patients enrolled, and we obtained the gyrA gene of Orientia tsutsugamushi from 27 samples. With blood samples, we sequenced the quinolone resistance-determining region (QRDR), the target of fluoroquinolones, by nested polymerase chain reaction targeting the 56-kDa antigen gene. *Results.* The genotype of samples were as followed; 7 Je-cheon strain, 5 Taguchi strain, 4 Boryoung strain, 3 Kanda strain, 3 Karp strain, 3 Pa-joo strain, 2 Ikeda strain. Irrespective of genotype, all 27 samples had the Ser83Leu mutation in the QRDR domain of gyrA gene, which is known to be associated with quinolone resistance. *Conclusion.* Mutation of gyrA gene was identified in all DNA sample of O. tsutsugamushi, despite of genotype which enrolled for our study. These data provide evidence that O. tsutsugamushi has intrinsic resistance to fluoroquinolones, explaining treatment failures with such antibiotics in scrub typhus. Therefore fluoroquinolones should not be used for the treatment of scrub typhus, especially in severe cases. *Disclosures.* All authors: No reported disclosures. [^1]: Session: 194. Global Infectious Diseases [^2]: Saturday, October 11, 2014: 12:30 PM	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#S1} ============ Neutrophils are polymorphonuclear granulocytes comprised of enzyme-containing granules. Neutrophils generate from the bone marrow and account for 50%--70% of circulating leukocytes in humans and 10%--25% in mice^\[[@R1],[@R2]\]^. Under acute inflammation, particularly as a result of bacterial infection, neutrophils are the first leukocytes to respond, migrate to the site of inflammation, and kill microorganisms through phagocytosis, degranulation and generation of neutrophil extracellular traps (NETs)^\[[@R3]--[@R5]\]^. These cells have long been thought of as short-lived cells of the innate immune response. However, recent research evidence has demonstrated that neutrophils persist beyond acute inflammation to initiate and perpetuate chronic inflammation. The onset of inflammation increases the lifespan of neutrophils in circulation, anywhere from 12 h to several days^\[[@R6]\]^. Pro-inflammatory cytokines, such as tumor necrosis factor-α (TNF-α) and interferon-γ, inhibit neutrophil apoptosis^\[[@R7]\]^. Neutrophils also change phenotypes as inflammation persists. For instance, Neutrophils isolated from various inflammatory states show altered cell densities in gradient centrifugation^\[[@R8],[@R9]\]^. The expression of granulocyte marker CD66b in neutrophils increases in patients with rheumatoid arthritis (RA), and decreases with the treatment by anti-inflammatory glucocorticoids, suggesting neutrophils may alter functions during chronic inflammation^\[[@R10]\]^. Glucocorticoids, a class of drugs often prescribed to patients with autoimmune disease, also inhibit apoptosis of neutrophils and increase the neutrophilic production of reactive oxygen species (ROS)^\[[@R11]\]^. Tissue-specific autoimmune disease such as RA, multiple sclerosis, and type I diabetes are generated from unknown etiology and impact the quality of life of patients through sustained chronic inflammation driven by innate and adaptive immune responses. Neutrophils bridge the innate and adaptive immune response in autoimmune disease. This review focuses on the functional heterogeneity of neutrophils in autoimmune disease and the contribution of these roles to chronic inflammation, in the context of RA. RA {#S2} -- RA is an autoimmune disease, in which the body generates antibodies against its own tissues. RA is characterized by tissue-specific autoimmune-mediated chronic inflammation that affects multiple joints and results in destruction of cartilage and bone loss^\[[@R12]\]^. Risk factors of RA are multi-dimensional and include genetic defects, infections, and environmental influences^\[[@R12]\]^. Population studies of RA estimate that genetics account for about 50% of RA disease susceptibility^\[[@R13]--[@R15]\]^. The most consistent predictor of susceptibility to RA is the link between specific major histocompatibility (MHC) II-associated alleles of the DRB1 gene that encode human leukocyte antigen (HLA)-DR4 and the onset of RA^\[[@R16]\]^. All of the susceptible alleles encode a conserved amino acid sequences on MHC II^\[[@R17]\]^. RA patients with these susceptible HLA-DR4 alleles and subsequent expression of the conserved amino acid sequence on MHC II, develop more autoantibodies associated with RA-related joint breakdown than patients without these alleles^\[[@R18]\]^. Activated antigen-presentation cells (APCs), including dendritic cells and macrophages, upregulate MHC II surface molecule expression, which activates antigen-specific T cells and B cells and initiates the adaptive immune responses in autoimmune disease. Bacterial infections may trigger the onset of RA. A population-based study in Sweden determined that 45% of inflammatory arthritis patients had an infection prior to the onset of early arthritis^\[[@R19]\]^. In Chlamydia-induced reactive arthritis, the microbe primes neutrophils through toll-like receptor signaling, which activates the cell to clear the infection. However, activated neutrophils can also infiltrate the joint and cause chronic inflammation^\[[@R20]\]^. Neutrophils may employ the same mechanisms in the onset of RA after an infection. Although many microbial factors have been found in synovial fluid of RA patients, there is no clear agreement that these microbial factors are the causative agents of RA^\[[@R21]--[@R23]\]^. Environmental factors, such as cigarette smoke, make provoke the development of RA in genetically susceptible populations^\[[@R24],[@R25]\]^. Smokers with the HLA-DR shared epitope allele were 3 times more likely to test positive for rheumatoid factor (RF) than non-smokers with HLA-DR shared epitope allele^\[[@R25]\]^. Smoking increases the activation and migration of neutrophils^\[[@R26],[@R27]\]^. Dysregulated neutrophils as a result of cigarette smoking generate a systemic inflammatory environment that is associated with autoimmune disease, such as RA and systemic lupus erythematosus^\[[@R28],[@R29]\]^. Neutrophils account for the majority of inflammatory cells in the synovia of human RA patients, and the joints of collagen-induced arthritis (CIA) mice, a murine model of RA^\[[@R30],[@R31]\]^. In both humans and mice, neutrophils accumulate at the pannus-cartilage interface, where much the destruction to both bone and cartilage takes places^\[[@R32]--[@R34]\]^. Thus, an in-depth understanding of neutrophil function in arthritis is vital to the prevention and treatment of this disease. Neutrophil identification {#S3} ------------------------- The surface marker granulocyte receptor-1 (Gr-1) was previously used to identify neutrophil populations in murine models^\[[@R35]\]^, as it was thought only mature granulocytes express Gr-1. Later researchers found that Gr-1 antibodies cross-react and bind two Ly-6 family member proteins, Ly6G and Ly6C^\[[@R36],[@R37]\]^. While neutrophils express Ly6G, monocytes, memory T cells, and some dendritic cell subsets express Ly6C^\[[@R36],[@R38]--[@R40]\]^. Studies utilize anti-Gr-1 antibodies to deplete neutrophils and attribute the presence of neutrophils to the ability of cancers cells to acquire metastatic phenotypes^\[[@R41],[@R42]\]^. Although neutrophils make up the majority of Gr-1^+^ cells in these studies, the increased metastasis after administration of anti-GR-1 antibodies cannot be attributed solely to neutrophil depletion, as Ly6C^+^ monocyte are also depleted. In CIA, the depletion of neutrophils with anti-GR-1 antibody concluded that neutrophils were necessary for the onset and maintenance of disease, as this treatment prevented the onset of disease and ameliorated established disease^\[[@R31]\]^. Macrophages express Gr-1 and also infiltrate the CIA joint, so this method fails to delineate the roles of neutrophils from macrophages in CIA. Currently, studies utilize antibody clone IA8 to specifically bind Ly6G and deplete neutrophils without effect on other leukocyte populations^\[[@R43]\]^. In general, current murine studies identify neutrophils as CD11b^+^Ly6G^+^Ly6C^+^ cells, and monocytes as CD11b^+^Ly6G^−^Ly6C^+^ cells. Human neutrophils are defined as CD14^lo/neg^CD15^+^CD16^hi^CD33^+^CD11b^+^CD15^+^CD66b^+^, and human monocytes are defined as CD14^+^CD16^+^HLA^−^ DR^+^CD66^−^ cells^\[[@R44]\]^. Neutrophil effect on T cells {#S4} ---------------------------- Neutrophils, unlike dendritic cells and macrophages, are not defined as APCs. APCs express antigen to T cells via MHC II molecules and stimulates T cell activation with the aid of co-stimulatory molecules CD80 and CD86. In autoimmunity, autoantigen is presented and antigen-activated T cells quickly expand and migrate to the site of inflammation and induce tissue inflammation through the production of pro-inflammatory cytokines^\[[@R45]\]^. In genetically susceptible individuals, the conserved amino acid sequence on MHC II may manipulate the antigen presentation process and cause activation of autoreactive T cells^\[[@R46]\]^. In the absence of autoimmune disease, neutrophils isolated from the peripheral blood of healthy controls produce MHC II mRNA, but do not express the cell surface molecule^\[[@R47]\]^. Exposure of these healthy neutrophils to synovial fluid from an RA patient elicits surface expression of both MHC II and costimulatory molecules^\[[@R48]\]^. Contact with T cells induces neutrophilic expression of MHC II and costimulatory molecules^\[[@R48]\]^. Since T cells are abundant in the RA joint, neutrophils continually express MHC II and costimulatory molecules and act as APCs. This further activates T cells and forms a vicious feedback loop that promotes chronic inflammation and tissue damage in the joints^\[[@R48]--[@R50]\]^. The continuous activation of T cells in the joint advances joint destruction through the production of pro-inflammatory cytokines and activation of autoreactive B cells^\[[@R51],[@R52]\]^. Recently, studies have linked neutrophil functions to Th17 cells. Th17 cells, which produce IL-17, are potent proinflammatory mediators and have been implicated in the pathogenesis of autoimmune disease^\[[@R53]--[@R56]\]^. IL-17 can induce tissue inflammation by stimulating the recruitment of neutrophils. In the RA joint, IL-17 activates fibro-blast-like synoviocytes (FLS), macrophages, and osteoblasts^\[[@R53]--[@R58]\]^. FLS activation produces the potent neutrophil chemoattractant IL-8^\[[@R59],[@R60]\]^. Activation of macrophages produces TNF-α^\[[@R61]\]^, and the combination of Il-17 and TNF-α in the joint stimulates synovial endothelial cells to produce more neutrophil chemoattractants^\[[@R62],[@R63]\]^. Reciprocally, neutrophils help sustain Th17 cells in the joint through the secretion of Th17 chemokines CCL20 and CCL2^\[[@R60]\]^. Human neutrophils purified from the synovial fluid of RA patients express high levels of these chemokines^\[[@R60]\]^. Despite the presence of other chemokines in the joint, Th17 cells preferentially migrate toward CCL20^\[[@R64]\]^. Mice that lack the CCL2 receptor (CCR2^−/−^) develop exacerbated CIA. Furthermore, Th17 and neutrophil populations expand in the lymph nodes and joints of these mice^\[[@R65]\]^. Therefore, CCL20 likely plays a more prominent role in Th17 migration than CCL2. The expansion of both neutrophils and Th17 cells in mice with exacerbated arthritis demonstrates the importance of both these cell types to the onset and maintenance of autoimmune arthritis. The reciprocal signaling between Th17 cells and neutrophils causes accumulation and activation of these cells, cultivating an inflammatory microenvironment in the joint^\[[@R58],[@R60],[@R66]\]^. Although reducing neutrophils in the joint is beneficial to limit sustained T cell activation, systemic neutropenia can cause infection. Some therapeutics intervene with proinflammatory events that upregulate neutrophils, instead of depleting neutrophil function. A phase II trial that utilizes the anti-IL-17 monoclonal antibody drug, secukinumab, improves symptoms in 46% of RA patients after 16 wk of treatment, and can safely maintain these improvements through week 52^\[[@R67],[@R68]\]^. Infection rates of patients on these drugs were 31.9% through week 52. Most infections were mild, but, interestingly, were not associated with neutropenia^\[[@R68]\]^. Simultaneously targeting neutrophils may improve IL-17 treatments, as activation of Th17 cells would decrease in addition to just decreasing the IL-17 effector molecule. However, as with many autoimmune therapies, this may leave the patient susceptible to infection. Neutrophil effect on B cells {#S5} ---------------------------- One of the major features of autoimmune disease is the presence of autoantibodies in circulation. In RA patients, various autoantibodies against cartilage components, chaperones, enzymes, nuclear proteins, and citrullinated proteins have been identified. However, the clinical significances and pathogenic roles of these antibodies are largely unknown, except for RF, anti-citrullinated protein antibodies (ACPA), and anti-collagen antibodies, which are all associated with joint inflammation^\[[@R69]--[@R72]\]^. Neutrophils stimulate the activation, proliferation, differentiation, and antibody-production of B cells through the production of the B cell stimulating factor BAFF (also known as BLyS)^\[[@R73]--[@R75]\]^. Peripheral blood neutrophils from both RA patients and healthy control patients express BAFF as a membrane bound molecule^\[[@R76]\]^. However, TNF-α in the RA joint releases surface-bound BAFF from neutrophils and increases the concentration of soluble BAFF^\[[@R76]\]^. High levels of soluble BAFF in the serum of RA patients correlates with high concentrations of autoantibodies^\[[@R77]\]^. Neutrophils in the joint act on B cells similarly to splenic marginal zone neutrophils, which induce antibody production and immunoglobulin class switching through the production of the B cell stimulants BAFF, APRIL, and IL-21^\[[@R78]\]^. These stimulants drive the formation of splenic germinal centers that support the proliferation and differentiation of B cells. The blockage of BAFF decreases the size and disorganizes splenic germinal centers^\[[@R79],[@R80]\]^. Structures similar to splenic germinal centers develop in the synovia of some RA patients^\[[@R81]\]^. A therapy that reduces soluble BAFF released from neutrophils may prevent synovial germinal center formation and reduce autoantibody production^\[[@R82]\]^. Thus, soluble BAFF from neutrophils plays a critical role in facilitating an environment, both systemically and locally, that activates B cells and perpetuates autoantibody formation. The release of ROS by neutrophils^\[[@R83],[@R84]\]^ generates advanced glycation end-products (AGE) through oxidant-induced alteration of the structures of lipids, DNA, and proteins^\[[@R83]--[@R86]\]^. High levels of AGE in the sera of RA patients correlates with a high disease severity and high levels of inflammation markers^\[[@R87]\]^. B cells recognize ROS-modified structures as foreign molecules and produce autoantibodies against these structures^\[[@R88],[@R89]\]^. ROS modifies type II collagen, the main structural component in human articular cartilage and induces the production of autoantibodies to ROS-modified collagen^\[[@R90]--[@R94]\]^. In an in vitro study, serum from RA patients could only bind type II collagen after exposure to ROS produced by neutrophils^\[[@R95]\]^. Autoantibodies form immune complexes in the joint, which induce neutrophil infiltration and activation through complement-mediated pathways^\[[@R96],[@R97]\]^. B cells and neutrophils, therefore, work in concert to maintain inflammation in the RA joint. Treatment with rituximab, a monoclonal anti-CD20 antibody that depletes B cells, decreases the severity of RA most efficiently in RA patients with high levels of autoantibodies^\[[@R98],[@R99]\]^. Some patients experience late-onset neutropenia up to 12 mo following rituximab treatment^\[[@R99]\]^. The exact mechanism that causes late-onset neutropenia is not known. One popular hypothesis suggests B cells compete with neutrophils for resources in the developmental niche of the bone marrow as the B cells repopulate after rituximab treatment^\[[@R98]\]^. Neutrophils and NET formation {#S6} ----------------------------- Neutrophils form NETs through a unique model of cell death known as NETosis^\[[@R100]\]^. The formation of NETs requires activated neutrophils to lose integrity of intracellular membranes prior to that of the plasma membrane^\[[@R101]\]^. In the first steps of NET formation, granules containing cytotoxic antimicrobial proteins decay, and chromatin condenses as the nuclear membrane collapses. The plasma membrane then invaginates, ruptures, and releases NETs comprised of intracellular antimicrobial contents into the extracellular space^\[[@R102]\]^. In the context of infection, the antimicrobial proteins trap and kill infiltrating microbes in the extracellular space^\[[@R100],[@R103]\]^. NETs release citrullinated histones and proteins into the extracellular space^\[[@R104],[@R105]\]^. Citrullination of a protein or histone is a post-translational modification (PTM) that converts arginine residues to citrulline^\[[@R104],[@R105]\]^. This process changes the structure and antigenicity of proteins and histones, as the adaptive immune response can recognize PTM as non-self^\[[@R106]\]^. In RA, ACPA are of particular interest as an increase in ACPA correlate with an increased disease severity^\[[@R107],[@R108]\]^. An increased propensity for neutrophils to die via NETosis correlates with increased levels of ACPA in the serum of RA patients^\[[@R105]\]^, which suggests NETs are a major source of autoantigen in RA. In fact, proteins extracted NET-induced peripheral blood neutrophils react with sera from RA patients. Sera from RA patients react specifically with citrullinated histone H4 ^\[[@R109]\]^. The discovery of a conserved citrullinated antigen associated with the onset of RA opens up a novel avenue of therapeutic intervention. The enzyme PAD4 controls the citrullination of histone H4^\[[@R109]\]^. Interruption of PAD4 function could decrease citrullination of histone H4 and subsequent autoantibody production that is crucial to the development of RA. NETosis also stimulates FLS to produce the proinflammatory cytokines IL-6, IL-8, and the Th17-associated chemokine CCL20^\[[@R105]\]^. Production of IL-6 and IL-8 aid in the polarization of CD4+ T cells to Th17 phenotypes, while CCL20 traffics Th17 cells to the of inflammation. Thus, NET stimulation of FLS shapes a microenvironment favorable to sustained inflammation associated with Th17 cells. Neutrophils as myeloid-derived suppressor cells {#S7} ----------------------------------------------- Myeloid-derived suppressor cells (MDSCs) are a heterogeneous group of immature myeloid cells derived from the bone marrow under pathologic conditions that suppress T cell functions^\[[@R110]\]^. Murine MDSCs are divided into two subsets based on surface expression of Ly6G and Ly6C; the Ly6G^+^Ly6C^+^CD11b^+^ granulocytic-MDSC (G-MDSC) subset and the Ly6G-Ly6C^+^CD11b^+^ monocytic-MDSC (M-MDSC) subset^\[[@R111]\]^. The counterparts of M-MDSC subset in humans have been identified as CD11b^+^CD14^+^CD15^−^HLA^−^DR^−^Lin^−^, while human G-MDSC subset are identified as CD11b^+^CD15^+^CD14^lo^HLA^−^DR^−^ Lin^−\[[@R111]\]^ ([Table 1](#T1){ref-type="table"}). M-MDSCs are immature monocytes that suppress T cell functions through the secretion of inducible nitric oxide synthase (iNOS)^\[[@R112]--[@R114]\]^. G-MDSCs are neutrophil-like cells that inhibit T cell function through the production of arginase-1^\[[@R114]\]^. First described in murine tumor models, MDSCs accumulate in the lymphoid tissues of tumor-bearing mice^\[[@R115]\]^ and significantly increase in the circulation of cancer patients compared to healthy controls^\[[@R116]--[@R118]\]^. MDSCs promote tumor growth by inhibiting T-cell mediated immune surveillance and cytotoxic effects on tumor cells^\[[@R119]\]^. Recently, evidence indicates MDSCs also contribute to the regulation of autoimmune disease by inhibiting CD4^+^ T cell proliferation and differentiation^\[[@R120],[@R121]\]^. Neutrophil-like G-MDSCs are believed to be immunosuppressive in autoimmunity. In the experimental autoimmune encephalomyelitis (EAE) murine model of multiple sclerosis, G-MDSCs express high levels of programmed cell death 1 ligand 1 (PD-L1), a costimulatory molecule that negatively regulates T cell proliferation. G-MDSCs inhibit autoantigen-priming of Th1 and Th17 cells in a PD-L1-dependent manner^\[[@R122]\]^. In CIA, MDSCs that include mostly neutrophil-like G-MDSCs suppress both T cell proliferation and CD4^+^ T cell differentiation into Th17 cells, mainly through the production of arginase-1^\[[@R123]\]^. The depletion of MDSCs increases inflammation and disease severity, while the subsequent adoptive transfer of MDSC ameliorates arthritis^\[[@R123]\]^. Furthermore, based on our published^\[[@R124]\]^ and unpublished data, M-MDSCs as well as G-MDSCs not only suppress T cell functions, but also inhibit B cell proliferation in the context of CIA. The use of CD11b^+^Gr-1^+^ cells in most functional MDSC studies makes it difficult to confirm G-MDSCs immunosuppressive effects in autoimmune arthritis Additionally, utilization of Ly6G to recognize G-MDSCs cannot differentiate G-MDSCs from neutrophils. Similar issues exist in human MDSCs, as mature human neutrophils express many of the same surface receptors as G-MDSCs and cannot be distinguished without functional analysis^\[[@R125]\]^. Further studies comparing G-MDSC phenotypes vs neutrophil phenotypes are needed to determine the effect of these cells on the adaptive immune response. Current evidence indicates G-MDSCs and neutrophils act in opposing manners on the adaptive immune response^\[[@R125]\]^. Uncovering mechanisms that lead to the plasticity of G-MDSCs and neutrophils in autoimmune disease could lead to cell-based therapies that convert pro-inflammatory cells to immunosuppressive cells. Neutrophils and potential therapeutics {#S8} -------------------------------------- As one of major contributing factors in the establishment of chronic inflammation in RA, neutrophils serve as potential therapeutic targets. Some current therapies for RA interfere with the functions of neutrophils. For example, anti-TNF-α therapies reduce IL-33 receptor expression on neutrophils and subsequently decrease neutrophil migration. Neutrophils from RA patients treated with anti-TNF-α therapies do not respond to IL-33-mediated chemotaxis^\[[@R126]\]^. Impaired chemotaxis of neutrophils may lead to a decrease in inflammation and disease severity. The wide array of effector proteins produced by neutrophils, such as BAFF, could also become therapeutic targets for RA. NETs may serve as a novel therapeutic target for RA and other NET-associated autoimmune diseases. In addition to targeting enzymes associated with PTM, the use of DNAse to breakdown the extracellular DNA and histones associated with NETs has been suggested as a potential therapeutic in NET-associated diseases^\[[@R127]\]^. The study of NETs in RA may also reveal currently unknown citrullinated proteins that contribute to RA pathogenesis and could serve as therapeutic targets. In CIA, the adoptive transfer of MDSCs decrease T and B cell proliferation and decreases the severity of arthritis^\[[@R123],[@R124]\]^. The discovery of a mechanism that induces the G-MDSC phenotype from neutrophil-like cells could be crucial for the therapy of autoimmune disease. Manipulation of this mechanism will drive neutrophil-like cells toward an immunosuppressive MDSC phenotype that impedes the super-active adaptive immune response, and thereby reduces chronic inflammation in autoimmunity. An in-depth understanding of the contributions of the joint microenvironment to the various neutrophil phenotypes, and subsequent neutrophil functions, may aid in the development of neutrophil-based RA therapies. Similar to many autoimmune therapies, inhibition of neutrophil functions leaves the patients susceptible to infection. For this reason, the inhibition of specific enzymes, such as PAD4, may be a promising therapeutic intervention. If PAD4 can be neutralized, this could curb autoantibody production without completely diminishing the antimicrobial function of neutrophils. CONCLUSION {#S9} ========== Neutrophils are involved in the onset and progression of RA in a complex capacity. Neutrophils engage in several reciprocal signaling events with both B and T cells, which promote a microenvironment conducive to sustained inflammation. The formation of NETs increases the production of ROS and ACPA, which are hallmarks of RA. Neutrophils can act as pro-inflammatory cells influencing chemotaxis and immune cells signaling, but can also have a MDSC phenotype that suppresses the immune response. Since neutrophils affect many aspects of the adaptive immune response and drive chronic inflammation, the disruption of the signals between neutrophils and the adaptive immune response can serve as therapeutic targets for RA. Supported by The National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health (NIAMS−NIH), No. 1R01AR063132. Author contributions: Both authors contributed to this paper. Conflict-of-interest statement: The authors declare no conflict of interest. ###### Surface markers used in identification of myeloid-derived suppressor cells Human Mouse ----------------------- ----------------------------------------------------------------- -------------------------- MDSC CD11b^+^ CD33^+^ CD11b^+^ Gr-1^+^ M-MDSC CD11b^+^CD14^+^CD15^−^ HLA^−^DR^−^Lin^−^ CD11b^+^ Ly6C^+^ Ly6G^−^ G-MDSC Cd11b^+^CD15^+^CD14^lo^ HLA^−^DR^−^Lin^−^ CD11b^+^ Ly6G^+^ Ly6C^+^ Progenitor neutrophil CD14^lo/neg^ CD15^+^ CD16^lo/neg^ Mature neutrophil CD14^lo/neg^ CD15^+^ CD16^hi^ CD33^+^ CD11b^+^ CD15^+^ CD66b^+^ CD11b^+^ Ly6G^+^ MDSC: Myeloid-derived suppressor cell; M-MDSC: Monocytic-MDSC; G-MDSC: Granulocytic-MDSC.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#s0005} =============== Epiduroscopy or spinal endoscopy \[interventional endoscopy spinal surgery (IESS)\] is a minimally invasive surgical procedure in the spinal epidural space, which through the use of a directable flexible endoscope, allows for the direct visualization of the epidural space with diagnostic as well as therapeutic possibilities. Development of epiduroscopy techniques began in the early 1930s, and Pool in 1937 performed the first epiduroscopy ([@b0095]). The technique was reassessed in the 1990s, and various groups began using it for the treatment of spinal pain of diverse origins through the caudal approach, as it is currently known ([@b0115]). The most important advances in this field have been achieved through the development of various tools used for working within the lumbar epidural space, such as the Fogarty balloon modified for use in the epidural space, called Resaloon®, and the subsequent development of an ablation system using the quantum molecular resonance (QMR®) technology, called Resaflex®. This is a minimally invasive device used with the epiduroscope for the treatment of adhesions in case of epidural fibrosis or for the treatment of ligamentum flavum in case of lumbar canal stenosis due to ligamentum flavum hypertrophy ([@b0085], [@b0100]). The use of epiduroscopy represents one line of action for cases of chronic refractory lumbar pain and pain in the lower limbs due to post-laminectomy syndrome as well as for cases of stenosis of the central or foraminal canal before the use of more aggressive procedures, such as reoperation or neurostimulator or intrathecal pump implantation. Thus, the main indications for the use of this technique include post-laminectomy syndrome and lumbar canal stenosis ([@b0055], [@b0045], [@b0120]). Degenerative stenosis of the lumbar canal has been estimated to affect 400,000 individuals in the United States ([@b0055]), with prevalence of absolute and relative stenosis estimated at 19.4% and 47.2% in patients older than 60 years, respectively ([@b0050]). This constitutes the most frequent indication for spinal surgery in individuals older than 65 years ([@b0020]). Although the causes underlying this syndrome are multiple and the etiology is sometimes multifactorial, ligamentum flavum hypertrophy remains one of the main causes ([@b0055]). Clinical symptoms presented by patients, depending on their type of stenosis (central or foraminal), constitute changes in the cauda equina, bilateral radicular pain, alterations in saddle sensitivity, or motor alterations in the legs, which become worse with walking and prolonged standing. In most cases, the symptoms disappear when patients are resting, sitting, or lying in the prone position and are relieved by trunk flexion ([@b0015]). Although epiduroscopy using the new QMR® technology may be useful in patients with lumbar canal stenosis (to reduce hypertrophy of the ligamentum flavum), it is not a risk-free technique. Complications as transient neurological symptoms, dural puncture, visual disturbances, urinary and/or faecal incontinence are possible in such patients ([@b0135]). The need to perform epiduroscopies as safely as possible arises to avoid potential complications. Therefore, we propose the use of intraoperative neurophysiological monitoring. In this article, we present the first cases in the literature involving the use of intraoperative neurophysiological monitoring to avoid or reduce the rate of complications in patients with lumbar canal stenosis as well as to contribute information to the etiopathogenesis of this type of complications. 2. Methods {#s0010} ========== We present the first cases of the use of intraoperative neurophysiological monitoring during therapeutic epiduroscopy in patients with clinical symptoms of canal stenosis to reduce the possible complications of the technique. In all cases, the monitoring was conducted after obtaining the informed consent, and the study was registered at ClinicalTrials.gov with number NCT03863067 and approved by the Ethics Committee (Comité de Ética de la Investivgación de la Fundación Jiménez Díaz, Madrid, Spain). In all patients, the technique was performed as described by other authors, together with the use of the latest safety tools that are currently available ([@b0010]). 2.1. Technique {#s0015} -------------- Epiduroscope (Resascope®) is a system designed for diagnosis and surgical endoscopic treatment of the epidural space (IESS) ([Fig. 1](#f0005){ref-type="fig"}A). Resascope® and its components have been specially designed for a percutaneous approach through the sacral hiatus of the epidural canal with an external diameter of 3.3 mm and two open internal canals with a diameter of 1.25 mm and length of 30 cm.Fig. 1A. Resascope® B. Resaloon® C. Resaflex® D. Resablator®. Resaloon® is an inflatable balloon catheter for adhesiolysis during epiduroscopy ([Fig. 1](#f0005){ref-type="fig"}B) and Resaflex® comprises a flexible electrode for surgical lysis ([Fig. 1](#f0005){ref-type="fig"}C) with the QMR® technology ([Fig. 1](#f0005){ref-type="fig"}D). Resaflex® is a minimally invasive device, which through Resascope® enables to reach and coablate of scar tissue and treat epidural fibrosis in patients with post-laminectomy syndrome. It can also be used to perform ligamentum flavum ablation in patients with ligamentum flavum hypertrophy and canal stenosis ([Fig. 2](#f0010){ref-type="fig"}).Fig. 2L4--L5 dissection and laminectomy of a cadaver to visualize ligamentum flavum ablation with Resaflex® using macroscopy and epiduroscopy. The technology applied for IESS operates through the emission of a main waveform with a very well-defined harmonic at 4 MHz, followed by others of 8, 12, and 16 MHz with decreasing intensities. The combination produces a series of energy quanta calibrated for biological tissues, whose intensity is sufficient to break the tissue bonds without increasing the temperature and denaturing the structure of the fibrinogen protein in fibrin ([@b0100], [@b0105], [@b0110]). In all cases, other non-invasive and invasive treatments were used prior to the intervention, following the SAFE principle published by [@b0065]), and epidurolysis was performed before the epiduroscopy. Nuclear magnetic resonance and electrophysiological examinations were conducted for all patients before performing the technique in order to verify the diagnosis of the pathology, to confirm the existence of the ligamentum flavum hypertrophy, and to rule out the existence of any anatomical variant that may contraindicate the technique. The diameter of the epidural canal at the entrance to the sacral hiatus was measured to confirm that the approach could be applied without any complications to all cases. The technique was always performed in the operating room with strict aseptic measures and under conscious sedation. Intravenous cefazolin (2 g) was used as antibiotic prophylaxis 1 h before the procedure, which was performed on an ambulatory basis. After preparation of the field and cleaning with chlorhexidine, a caudal approach was used with the help of a fluoroscope, with the patient in the prone position. The midline of the sacral hiatus was located with an AP projection of the radiological C-arm, and the sacral hiatus was approached in the lateral projection. The entry point in the sacral hiatus and the floor of the sacral canal were infiltrated with local anesthetics using a needle (25G, 0.5 × 16 mm). The caudal epidural space was accessed using an 18G Tuohy needle and confirmed with two radiological projections, one AP and another lateral; no contrast was used in any case. The needle was removed after confirming the correct epidural position while keeping the guide inside the canal. From this point on, the Seldinger technique was used. The next steps in the technique were aimed at dilating the entrance for introducing the Resascope®. The entrance was enlarged with a \#11 cold knife to place the dilator and introducer (9 Fr) (Kit introductor vascular TERUMO CORPORATION), through which the Resascope® was placed in the caudal epidural space. Subsequently, 0.9% saline solution was used to distend the canal and to visualize the epidural structures better. Serum was heated to 37 °C in all cases to obtain better tolerance by the patients, and a constant infusion rate was maintained to prevent serum boluses, thus avoiding rise in pressure. 2.2. Neurophysiological monitoring {#s0020} ---------------------------------- Neurophysiological monitoring was conducted through continuous electromyographic recording (free-running EMG). For this, paired monopolar needle electrodes were inserted into the muscles depending on different lumbosacral roots studied. In our case, monitoring was carried out from L3 to S4. The following areas were monitored achieve these measurements: L3--L4 (quadriceps), L5 (anterior tibial/long peroneus), S1--S2 (lateral and medial gastrocnemius), S3--S4 (external anal sphincter), all bilaterally ([Fig. 3](#f0015){ref-type="fig"}).Fig. 3Intraoperative neurophysiological monitoring. A 32-channel Eclipse Intraoperative Neurophysiological Monitoring System (Medtronic) of was used for recording and analysis. Neurophysiological control enables identification of the manipulated roots (their mechanical manipulation determines the generation of nerve action potentials, which are translated into muscle action potentials in the free-running EMG) as well as of electromyographic activities that could be correlated with harmful maneuvers (generation of high-frequency electromyographic activities, such as the appearance of neuronal discharges, as neurotonic activity or trains). Free-running EMG was chosen over SEP -- Somatosensory Evoked Potentials -- and MEP-Motor Evoked Potentials because the patient was under conscious sedation, and these kind of monitorization need high intensity electric stimulation, which may be difficult to tolerate in patients who are not under general anesthesia. 3. Results {#s0025} ========== We present the first cases of intraoperative neurophysiological monitoring in patients diagnosed with lumbar canal stenosis who were subjected to epiduroscopy. Bilateral lumbosacral root monitoring was performed in all patients through continuous EMG recording, and spontaneous activity was verified while the patients were awake to check for proper functioning of the roots studied. Of the 12 cases studied, 8 did not present potentially pathological discharges ([Table 1](#t0005){ref-type="table"}), whereas 4 presented with electrophysiological abnormalities during the intervention, although all were completely resolved. Of the 4 patients with neurophysiological alterations, neurotonic activity was recorded in the left S1 territory in 1 patient during maneuvers for opening the working field with Resaloon® at the level of right L5--S1, which may correspond to pathological activation, with subsequent recovery when the procedure was discontinued in the contralateral area. Potentially injurious neurotonic activity in the right L5 root, which lasted 200--300 ms, was detected in another patient, coinciding with the use of Resaflex®, in ligamentum flavum at the level of the right lateral recess L5. Another patient presented with neurotonic activity, which suggested potentially harmful maneuvers in the neural tissue in the right L5 territory, when working on the same level with Resaflex®. Finally, depolarization of non-damaging mechanical causes was detected in roots S1--S4 in another patient, although neurotonic activity was observed in the right S1 territory, when working at the level of right L4--L5 ([Fig. 4](#f0020){ref-type="fig"}).Table 1Results of intraoperative neurophysiological monitoring.PatientEpiduroscopy levelEpiduroscopy volume (ml)EMG pre dataEMG intra data1L4, L5, L5, and S1 central180Subacute neurogenic changes in L3 and L4 left and chronic changes in L5 and S1 left compatible with radicular damage to these levels, mild, showing denervative activity at rest in L3 and L4 left but no loss of motor unitsNo neurotonic activity2L5 and S1 right190Chronic neurogenic changes in L5--S1 right and L4--S1 left with moderate degree of higher left affection. Compatible with radiculopathy of L5, S1 bilateral, and L4 left chronic motorNeurotonic activity in S1 left territoty while working with Resaloon® at level L5 and S1 right3L5 and S1 left190NormalNo neurotonic activity4L5 and S1 bilateral180S1 bilateral radiculopathyNo neurotonic activity5L4 and L5 right240Chronic neurogenic changes of axonal character in L5 to L5 right of moderate grade and L5 left of moderate levelNeurotonic activity in L5 right, lasting 200--300 ms6L4, L5, and S1 left210Loss of motor units from L3--L4 to L5S1 above all L4, L5 leftNo neurotonic activity7L4--L5, Y, and L5--S1 right200Chronic radiculopathy at L5 rightNeurotonic activity in L5 right territory, coinciding with Resaflex® use in recess L5 right8L4 and L5 right300Sub-chronic neurogenic changes of axonal character in territory L2, L3 bilateral of mild grade, compatible with radiculopathy L2 and L3 sub-chronic motor bilateralNeurotonic activity in territory S1 right for activity at level L4, L5 right with Resaflex®9Fibrosis grade IV, which prevents the passing of epidurosocpe over S280Chronic neurogenic changes in territory L5 and S1 bilateral compatible with radicular chronic damper moderate in L5 and mild in S1No neurotonic activity10L4 and L5 right160Chronic neurogenic changes in territory L5 right and S1 bilateralNo neurotonic activity11L5 and S1 left220Chronic neurogenic changes in territory L3--S1 right and L5 leftNo neurotonic activity12L5 and S1 left190Chronic radiculopathy in L5 leftNo neurotonic activityFig. 4A: Normal; B: Neurotonic discharges in the right L5 territory; C: Neurotonic discharges in the left S1 territory; D: Neurotonic discharges in the right S1 territory. All potentially harmful activities cease when momentarily interrupting the procedure, with all patients presenting normal neurological examination at the end of the procedure. 4. Discussion {#s0030} ============= Epiduroscopy is mainly indicated for lumbar canal stenosis and post-laminectomy syndrome. Although the procedure is safe and complications are typically minor, safety measures must be applied to reduce the incidence of complications considering the risk of subsequent potential major complications ([@b0070]). The complications of epiduroscopy can be determined by the technique itself or by distension of the epidural space with saline solution ([@b0075]). Distension of the epidural space can elicit two effects: affectation of the local perfusion and increase in pressure in the space that can be transmitted by the cerebrospinal fluid (CSF) ([@b0080]). The primary complications arising from the technique include pain at the puncture site, dural puncture with or without post-puncture headache, partial catheter shearing, paresthesia, transient subarachnoid block, and infection. Other types have also been described, such as nerve root injury, epidural bleeding and hematoma, meningitis or arachnoiditis, and abscesses ([Table 2](#t0010){ref-type="table"}) ([@b0135]), although majority of the studies have been performed with percutaneous epidurolysis and complications have been comparable to those described in epiduroscopy ([@b0130], [@b0090], [@b0145]). Severe problems with vision after epiduroscopy have also been described, possibly due to increased pressure in CSF due to the formation of saline solution boluses during the procedure, as well as retinal hemorrhages and blindness, although with complete recovery in 79% of the cases according to [@b0025]) and [@b0035]). All these complications of the procedure are generally minor and can be prevented with careful technique and proper handling of the amount and rhythm of the liquid infused ([@b0030]).Table 2Main complications arising from IESS.CategoryAdverse eventRelated to HardwareAdditional surgery due to malfunction of the device: extensionAdditional surgery due to malfunction of Resaloon®Additional surgery due to the use of Resaflex®Additional surgery due to rupture/fracture of the deviceBiologicalHematomaInfectionSeromaSkin erosionComplications at the lesion sitePersistent pain and/or numbness in the spinal areaPain related to trauma or inflammation of the sideAllergic reaction to surgical materials ((sutures, antibiotics, and anesthesia)Changes in subcutaneous tissue at the site of introduction of the deviceRelated to Resaflex®Inexplicable changes in headaches (with regard to intensity, type, or frequency)Decrease or loss of motor or muscular-skeletal controlNerve lesionPuncture or rupture of the dura mater Complications regarding sphincter control in the form of faecal and urinary incontinence related to bladder or faecal dysfunction have been described as a consequence of a possible nerve injury at the level of the sacral plexus, which can be increased when performing large-volume injections in the epidural space ([@b0060]). We did not find publications that can explain whether the reasons for this type of complications are the volume used, caudal entry into the epidural space with damage to the sacral plexus, or the use of Resaflex®. The lesion of the plexus does not seem logical at the entrance due to caudal access to the epidural space, although lesions may be formed by traversing the posterior to the anterior space, at the level of S3. This cause could be corroborated by means of electrophysiological studies. No changes were detected in neurophysiological monitoring conducted after this complication in patients who developed this type of complications. Conversely, using a new system within the coablative radio-frequency generator, sensory and motor stimulations were performed before each use of Resaflex® to avoid direct damage to the roots at which the procedure was performed; therefore, the location at which this tool was employed and safety measures taken before the procedure ([@b0010]) do not seem the most possible causes of these complications. Despite these safety measures, in our experience, there had been 5 cases of urinary incontinence from among more than 250 procedures, 4 of which fully recovered at 3--5 months after the procedure. Finally, the most possible cause for this type of complication could be increase in pressure and ensuing transient ischemia in the sacral plexus or cauda equina. To reduce the possible causes, utmost care must be taken when accessing the epidural space without the use of large volumes, as recommended in the literature ([@b0105], [@b0005]) as well as when infusing without boluses to avoid all complications arising from this type of procedure. To avoid this type of complications, we have started to perform intraoperative neurophysiological monitoring, which allows us to detect the presence of pathological discharges at different roots. These may indicate a potential lesion, so ceasing the maneuver over the nerve root, we can avoid the establishment of a neurogenic lesion. In fact, the existence of pathological neurogenic discharges on the side contralateral to the location at which Resaloon® was being used was detected in only 1 patient. This is an indication that increase in pressure on the contralateral side may be the origin of this type of complication and not necessarily either the volume used or the use of elements that can damage the sacral plexus or the passage of the epiduroscope through the caudal epidural canal. In 3 of the 12 patients in whom the monitoring was performed, neurotonic activity was detected in the roots when working with Resaflex®, 1 of which was detected at a lower level than that being treated, although in this case the activity did not seem to be pathological. Increase in pressure in the epidural space, given the distensibility of the system, is transmitted via the caudal, cranial, and foraminal spaces ([@b0140]), and although studies by [@b0040]) have shown an increase in the said distensibility of the epidural space associated with age due to the loss of epidural fat, the fact of working in patients with canal stenosis (anteroposterior distance \<10 mm or cross-sectional area \<100 mm^2^) ([@b0125]) may justify the appearance of transient ischemia (due to increased pressure transmitted to CSF) in roots where the vascular supply may be compromised and trapping that may justify urinary or faecal incontinence after the technique. This causes urinary incontinence in patients, which is difficult to manage and predict because there is no intraoperative symptom that we can correct. Thus, we have started using a technique with monitoring to identify patients at a risk of such complications who would not be otherwise discovered. We could detect changes in the surgical field as well as in roots below and/or contralateral to the surgical field through intraoperative neurophysiological monitoring, which would have otherwise gone unnoticed; however, these changes disappeared with complete recovery after stopping the procedure, which allowed us to avoid possible prolonged or even permanent complications in postoperatively. 5. Conclusions {#s0035} ============== Monitoring all types of constants throughout history has always been initially controversial because of the added cost to the technique and the lack of evidence of its benefit. Intraoperative neurophysiological monitoring is a known security measure that may help to reduce potential neurological insults and can optimize outcomes during epiduroscopy. Moreover, it enables the surgeon to work for long periods and discover changes within the epidural space where we work in real time. Advances in patient safety must be considered in all interventional procedures that we perform to improve results and reduce the incidence of complications. Acknowledgments {#s0045} =============== None. Conflict of interest {#s0050} ==================== David Abejon is a consultant for 10.13039/100008497Boston Scientific, PRIM, Cardiva, 10.13039/100001316Abbott, and 10.13039/100004374Medtronic. Eva M. Monzón and Pedro Moreno have no conflicts of interests. Funding {#s0055} ======= This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Numerous studies showed the adverse clinical and prognostic consequences of iron deficiency (ID) in patients with chronic heart failure (HF).[@ehz680-B1] Despite the significant prevalence of ID in HF, its pathophysiology and aetiology are not well-understood. Suggested mechanisms for ID in HF are poor dietary iron intake, drug interactions, (occult) gastrointestinal blood loss due to antiplatelet drugs and anticoagulants, and hepcidin-induced iron entrapment due to chronic low-grade inflammation.[@ehz680-B5] While ID is present in approximately half of all HF patients, its prevalence seems highest in females.[@ehz680-B1]^,^[@ehz680-B4] It is currently unclear which factors are driving this sex difference. In the present study, we identified determinants of ID in a large international cohort of worsening HF patients and sought to find sex-specific clinical and biochemical predictors of ID. Moreover, we established a cardiovascular biomarker profile of patients with ID. Methods ======= Study population ---------------- We included HF patients from the BIOSTAT-CHF study (A systems BIOlogy Study to TAilored Treatment in Chronic Heart Failure). This cohort has been described in full detail elsewhere.[@ehz680-B6] In short, the BIOSTAT-CHF study included patients either hospitalized for HF or presenting with worsening HF in the outpatient setting. Patients were eligible to participate with a left ventricular ejection fraction (LVEF) of ≤40% or, alternatively, brain natriuretic peptide or N-terminal pro B-type natriuretic peptide (NT-proBNP) levels of \>400 ng/L or \>2000 ng/L, respectively. Additionally, patients had to receive suboptimal evidence-based HF treatment (i.e. ≤50% of target dose of angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, and/or beta-blockers). After study inclusion, treating physicians were encouraged to up-titrate these drugs during a 3-month treatment optimization phase. The BIOSTAT-CHF study was conducted in accordance with the Declaration of Helsinki. All patients provided written informed consent prior to any study-related activities. Of all 2516 patients enrolled in the BIOSTAT-CHF cohort, serum for iron status analysis was available in 2357 (93.7%) patients. Laboratory measurements ----------------------- Iron parameters were assessed from venous blood. Blood samples were centrifuged at 2500 g for 15 min (4°C) and stored at −80°C afterwards. Samples were never thawed before laboratory analyses. The following blood markers reflecting iron metabolism were assessed on a Roche modular cobas 8000 using standard methods: serum iron, ferritin, and transferrin. Transferrin saturation (TSAT) was calculated as follows: \[72.17 \* iron (mg/dL)\]/transferrin (mg/dL).[@ehz680-B9] Renal function was expressed as the estimated glomerular filtration rate (mL/min/1.73 m^2^), calculated using the Chronic Kidney Disease Epidemiology collaboration equation (CKD-EPI). Serum NT-proBNP levels were determined using an immunoassay based on electrochemiluminescence (Elecsys, Roche Diagnostics, Mannheim, Germany). Serum hepcidin levels were measured using a competitive ELISA as described previously.[@ehz680-B10] Serum soluble transferrin receptor (sTfR) levels were measured using immunonephelometry on a BNII Nephelometer (Siemens AG, Erlangen, Germany). To establish a biomarker profile for patients with and without ID, 92 cardiovascular-related biomarkers from the Olink Cardiovascular III panel were measured, which were selected based on literature search on bioinformatics (e.g. Uniprot and DisGeNET) and by consulting experts in the cardiovascular field. All biomarkers were measured by Olink Proteomics (Uppsala, Sweden) using the Proximity Extension Assay technology, as previously described.[@ehz680-B11] Results are reported as Normalized Protein eXpression (NPX) on a log~2~ scale. Definitions and study endpoints ------------------------------- Anaemia was defined as a haemoglobin level \<12 g/dL in women and \<13 g/dL in men as per WHO standards.[@ehz680-B12] Iron deficiency was defined as a TSAT \<20%, as proposed by Grote Beverborg et al.[@ehz680-B13] This definition has been validated against the gold standard test for ID (bone marrow iron staining) in HF patients and has previously been used.[@ehz680-B4]^,^[@ehz680-B13] Daily protein intake of patients was estimated using spot urinary nitrogen and body mass index.[@ehz680-B14] Median follow-up of the study was 21 months. The primary endpoint of this study was the composite of all-cause mortality and first HF rehospitalization. Secondary endpoints included all-cause mortality and first HF rehospitalization. Statistical analyses -------------------- Data are presented as mean ± standard deviation (SD) when normally distributed, as median and interquartile range (IQR) when non-normally distributed or as percentage when categorical. Baseline characteristics were compared using the Student's t-test (normally distributed variables), the Mann--Whitney U test (non-normally distributed variables), and the χ^2^ test (categorical or binary variables). All baseline characteristics were stratified by iron status and sex. After baseline analyses, skewed variables were natural log~2~-transformed to obtain normal distributions. To identify independent predictors of ID, univariable and multivariable logistic regression models were constructed. Associates of ID with a univariable P-value of ≤0.1 were entered into the multivariable logistic regression models. Final multivariable models were established using backward elimination based on the significance of each variable. Bootstrap analyses with 1000 repeats (using the 'swboot' package in Stata) were performed to evaluate the robustness of the final models. Variables selected \>700 times were considered robust predictors. As a sensitivity analysis, a regularization approach was performed using multivariable lasso regression. The final model was checked for multicollinearity by calculating the variance inflation factor. Restricted cubic splines (three knots for all variables) were constructed for better visualization of the predictive value of continuous parameters on ID. Multivariable interaction analyses for sex were performed with each predictor of ID. The differential Olink biomarker expression pattern in iron-deficient patients was visualized using a volcano plot, displaying the magnitude in change of each biomarker (log~2~-fold change) against the significance of the difference in biomarker expression \[negative log~10~ of the P-value (Mann--Whitney U test)\]. Biomarkers in the top left or right of the plot are of interest (large magnitude fold change and high statistical significance). False discovery rate was controlled by correcting the P-values according to the Benjamini--Hochberg procedure (false discovery rate of 0.05). Biomarkers that were significantly up- or down-regulated in ID and had an absolute log~2~-fold change of \>0.25 were entered in uni- and multivariable logistic regression analyses with ID as dependent variable. Kaplan--Meier curves were constructed to determine the prognostic consequences of ID. Differences in survival rates were tested using the log-rank Mantel--Cox test. The influence of ID on outcome was further assessed with univariable and multivariable Cox proportional hazard regression models. In the multivariable models, adjustment was made for the BIOSTAT prediction models as described elsewhere.[@ehz680-B8] Additionally, the prognostic consequences of continuous TSAT levels were analysed using univariable and multivariable fractional polynomial analyses. The proportional hazards assumption was evaluated using Schoenfeld residuals and was applicable to all variables included in the outcome models. Missing predictor values were five times imputed as previously described.[@ehz680-B8] Survival analyses were performed in all five imputation models and the results were averaged in agreement with Rubin's rules. The degree of missingness of all variables used in the model development is depicted in [Supplementary material online, Table S6](#sup1){ref-type="supplementary-material"}. A two-sided P-value \<0.05 was considered statistically significant, while for interaction testing a P-value \<0.1 was used. Data were analysed using Stata version 15.1 (StataCorp LLC, College Station, TX, USA) and R version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria). Results ======= Baseline characteristics ------------------------ Baseline characteristics of the present study cohort, stratified by iron status, are depicted in Table [1](#ehz680-T1){ref-type="table"}. Mean age (±SD) of all patients was 69 ± 12.0 years, median LVEF (+IQR) was 30% (25--36), and 37.2% of all patients were in New York Heart Association (NYHA) functional class III or IV. The overall prevalence of ID was 61.6% (n = 1453) with highest prevalence in females (71.1% vs. 58.3%, P \< 0.001). According to the 'conventional' definition of ID in HF patients (i.e. ferritin \<100 µg/L or ferritin 100--300 µg/L with a TSAT \<20%), the prevalence of ID was even higher (n = 1632; 69.2%). Iron-deficient patients were older, had more comorbidities (including anaemia) and more signs of fluid overload compared to patients without ID (all P \< 0.05). Furthermore, patients with ID had lowest estimated protein intake, more inflammation, and highest rate of proton-pump inhibitor and P2Y~12~ inhibitor use (all P \< 0.001). HF aetiology and type of HF event (i.e. new-onset HF or worsening HF) was comparable between patients with and without ID. ###### Baseline characteristics for the total cohort, stratified by iron status Variables Total cohort No ID ID P-value ----------------------------------- ------------------- ------------------- ------------------- ----------- N 2357 904 1453 Clinical parameters Age (years) 68.9 ± 12.0 68.1 ± 12.1 69.3 ± 11.9 0.016 Females (%) 616 (26.1) 178 (19.7) 438 (30.1) \<0.001 BMI (kg/m^2^) 27.9 ± 5.5 27.8 ± 5.3 27.9 ± 5.6 0.65 Estimated protein intake (g/day) 55.0 ± 11.2 56.8 ± 12.2 53.9 ± 10.5 \<0.001 Ischaemic aetiology 1069 (46.2) 395 (44.6) 674 (47.1) 0.24 LVEF (%) 30 (25--36) 30 (25--35) 30 (25--37) 0.93 HFrEF 1707 (80.9) 687 (82.6) 1020 (79.8) 0.034 HFmrEF 271 (12.8) 107 (12.9) 164 (12.8) HFpEF 132 (6.3) 38 (4.6) 94 (7.4) Previous hospitalization for HF 736 (31.2) 278 (30.8) 458 (31.5) 0.70 NYHA functional class III/IV 1417 (61.8) 455 (51.6) 962 (68.3) \<0.001 Systolic blood pressure (mmHg) 125 ± 22 124 ± 20 125 ± 23 0.36 Heart rate (b.p.m.) 80 ± 19 77 ± 18 82 ± 20 \<0.001 Peripheral oedema 1165 (59.5) 354 (48.8) 811 (65.7) \<0.001 Elevated JVP 518 (33.5) 156 (26.4) 362 (37.8) \<0.001 Hepatomegaly 333 (14.2) 118 (13.1) 215 (14.8) 0.24 Orthopnoea 818 (34.8) 231 (25.6) 587 (40.5) \<0.001 6MWT (m) 316 (225--393) 345 (250--416) 300 (210--374) \<0.001 KCCQ (overall score) 49 ± 22 56 ± 22 45 ± 22 \<0.001 Comorbidities Atrial fibrillation 1063 (45.1) 393 (43.5) 670 (46.1) 0.21 Diabetes mellitus 759 (32.2) 238 (26.3) 521 (35.9) \<0.001 COPD 406 (17.2) 133 (14.7) 273 (18.8) 0.011 Renal disease 649 (27.5) 177 (19.6) 472 (32.5) \<0.001 Device therapy 582 (24.7) 215 (23.8) 367 (25.3) 0.42 Laboratory Haemoglobin (g/dL) 13.2 ± 1.9 13.9 ± 1.8 12.8 ± 1.8 \<0.001 Anaemia^a^ 778 (36.2) 181 (23.0) 597 (43.8) \<0.001 Haematocrit (%) 40.0 ± 5.4 41.6 ± 5.2 39.1 ± 5.2 \<0.001 Mean corpuscular volume (fL) 90 ± 9 92 ± 9 89 ± 8 \<0.001 Iron (mg/dL) 45 (28--73) 78 (62--101) 34 (22--45) \<0.001 Ferritin (µg/L) 103 (50--193) 142 (80--240) 78 (39--162) \<0.001 Transferrin (mg/dL) 200 (160--250) 200 (170--240) 210 (160--250) 0.17 Transferrin saturation (%) 17 (11--25) 27 (23--33) 12 (9--16) NA sTfR (mg/L) 1.5 (1.2--2.1) 1.3 (1.0--1.7) 1.7 (1.3--2.3) \<0.001 Hepcidin (nmol/L) 6.3 (2.2--16.5) 8.4 (4.4--20.0) 4.6 (1.4--13.1) \<0.001 CRP (mg/L) 13.0 (5.8--26.4) 8.0 (3.5--17.2) 16.9 (8.4--32.1) \<0.001 Leucocytes (10^9^/L) 7.8 (6.4--9.6) 7.5 (6.3--9.1) 8.0 (6.6--9.8) \<0.001 AST (U/L) 25 (19--35) 26 (20--35) 25 (19--35) 0.28 ALT (U/L) 25 (17--38) 26 (18--40) 24 (16--37) 0.005 γ-GT (U/L) 55 (28--109) 53 (29--107) 56 (27--110) 0.89 Alkaline phosphatase (µg/L) 85 (65--118) 83 (63--117) 86 (66--120) 0.26 Total bilirubin (µmol/L) 14 (10--21) 14 (10--20) 14 (10--22) 0.14 Sodium (mmol/L) 140 (137--142) 140 (137--142) 139 (137--142) \<0.001 Potassium (mmol/L) 4.2 (3.9--4.6) 4.3 (4.0--4.6) 4.2 (3.9--4.6) 0.012 NT-proBNP (ng/L) 4305 (2360--8329) 3300 (1833--6767) 4812 (2688--8991) \<0.001 Creatinin (µmol/L) 101 (82--128) 97 (80--122) 104 (84--133) \<0.001 eGFR (mL/min/1.73 m^2^) 60 (44--79) 64 (48--83) 57 (43--76) \<0.001 Albumin (g/L) 32 ± 9 34 ± 8 31 ± 9 \<0.001 Urea (mmol/L) 11.4 (7.6--18.2) 10.3 (7.1--16.4) 12.0 (7.8--19.0) \<0.001 Medication Loop diuretics 2346 (99.5) 902 (99.8) 1444 (99.4) 0.17 Beta-blockers on target dose 128 (5.4) 54 (6.0) 74 (5.1) 0.36 ACEi/ARB on target dose 314 (13.3) 128 (14.2) 186 (12.8) 0.35 Aldosterone antagonist 1259 (53.4) 514 (56.9) 745 (51.3) 0.008 Proton-pump inhibitors 825 (35.0) 260 (28.8) 565 (38.9) \<0.001 Antiplatelets P2Y~12~ inhibitors 363 (15.4) 108 (11.9) 255 (17.5) \<0.001 Acetylsalicylic acid 1168 (49.6) 443 (49.0) 725 (49.9) 0.67 Anticoagulants 915 (38.8) 349 (38.6) 566 (39.0) 0.87 Vitamin K antagonists 898 (38.1) 341 (37.7) 557 (38.3) 0.77 DOACs 17 (0.7) 8 (0.9) 9 (0.6) 0.46 6MWT, 6-min walk test; ACEi, angiotensin-converting enzyme inhibitor; ALT, alanine aminotransferase; ARB, angiotensin receptor blocker; AST, aspartate aminotransferase; BMI, body mass index; COPD, chronic obstructive pulmonary disease; CRP, C-reactive protein; DOAC, direct-acting oral anticoagulant; eGFR, estimated glomerular filtration rate; γ-GT, gamma-glutamyltransferase; HFmrEF, heart failure with mid-range ejection fraction; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; ID, iron deficiency; JVP, jugular venous pressure; KCCQ, Kansas City Cardiomyopathy Questionnaire; LVEF, left ventricular ejection fraction; NT-proBNP, N-terminal pro B-type natriuretic peptide; sTfR, soluble transferrin receptor. Anaemia was defined as a haemoglobin level \<12 g/dL in women and \<13 g/dL in men. Similar in females and males, patients with ID had more severe HF at baseline (e.g. peripheral oedema, orthopnoea, and highest NT-proBNP levels), showed more signs of inflammation, had highest prevalence of anaemia and lowest estimated protein intake (all P \< 0.01, see [Supplementary material online, Table S1](#sup1){ref-type="supplementary-material"}). Numerically, iron-deficient males had most comorbidities and highest rate of proton-pump inhibitor and P2Y~12~ inhibitor use, although no significant interaction between these variables and sex was present. A higher prevalence of atrial fibrillation in iron-deficient patients was observed in females, but not in males (P for interaction, 0.028). Determinants of iron deficiency ------------------------------- Univariable and multivariable logistic regression prediction models for ID are shown in Table [2](#ehz680-T2){ref-type="table"}. Independent determinants of ID were lower estimated protein intake, higher heart rate, presence of peripheral oedema and orthopnoea, history of renal disease, lower haemoglobin, higher C-reactive protein (CRP), lower serum albumin, and use of P2Y~12~ inhibitors (all P \< 0.005). The c-statistic of this model was 0.76. None of the determinants had a significant interaction with sex. All determinants in the final model remained highly selected in additional bootstrap analyses. Additional Lasso regression analysis of the multivariable model confirmed our findings and selected haemoglobin and CRP as best predictors for ID. The variance inflation factors of variables in the multivariable model were not suggestive of multicollinearity (range of factors 1.02--1.27). Restricted cubic splines showing the association between ID and estimated daily protein intake, serum levels of CRP and albumin, and haemoglobin are displayed in Figure [1](#ehz680-F1){ref-type="fig"}A--D. Sex-specific restricted cubic splines for these determinants can be found in [Supplementary material online, Figure S1A--D](#sup1){ref-type="supplementary-material"}. ![(A) Restricted cubic spline of the association between estimated protein intake and the prevalence of iron deficiency. (B) Restricted cubic spline of the association between C-reactive protein and the prevalence of iron deficiency. (C) Restricted cubic spline of the association between serum albumin and the prevalence of iron deficiency. (D) Restricted cubic spline of the association between haemoglobin and the prevalence of iron deficiency. The solid lines indicate estimates of the prevalence of iron deficiency across continuous levels of estimated protein intake, C-reactive protein, serum albumin, and haemoglobin, fitted using logistic regression analysis. The dashed lines indicate 95% confidence intervals.](ehz680f1){#ehz680-F1} ###### Univariable and multivariable logistic regression prediction models for iron deficiency Variables Univariable Multivariable -------------------------------------------- ------------------- --------------- --------- ------- ------------------- ------- --------- Clinical parameters Age (per 5 years) 1.04 (1.01--1.08) 2.40 0.016 0.464 Sex (female vs. male) 1.76 (1.44--2.15) 5.58 \<0.001 --- 1.42 (1.13--1.79) 2.99 0.003 BMI (per 5 kg/m^2^) 1.02 (0.95--1.10) 0.60 0.551 0.368 Estimated protein intake (per 10 g/day) 0.80 (0.74--0.86) −5.74 \<0.001 0.405 0.87 (0.79--0.94) −3.32 0.001 Ischaemic aetiology (yes vs. no) 1.11 (0.93--1.31) 1.17 0.241 0.687 LVEF (per 5%) 1.03 (1.00--1.08) 1.71 0.087 0.008 Male 0.98 (0.93--1.02) −0.95 0.340 Female 1.11 (1.02--1.19) 2.54 0.011 Previous HF hospitalization (yes vs. no) 1.04 (0.87--1.24) 0.39 0.695 0.261 NYHA functional class III/IV (vs. I/II) 2.02 (1.70--2.40) 7.95 \<0.001 0.899 Systolic blood pressure (per 5 mmHg) 1.01 (0.99--1.03) 0.85 0.395 0.726 Heart rate (per 5 b.p.m.) 1.08 (1.05--1.10) 6.10 \<0.001 0.157 1.06 (1.04--1.09) 3.83 \<0.001 Peripheral oedema (yes vs. no) 1.97 (1.67--2.34) 7.90 \<0.001 0.649 1.36 (1.12--1.66) 3.03 0.002 Elevated JVP (yes vs. no) 1.69 (1.35--2.12) 4.59 \<0.001 0.735 Hepatomegaly (yes vs. no) 1.16 (0.91--1.48) 1.23 0.220 0.116 Orthopnoea (yes vs. no) 1.96 (1.64--2.36) 7.28 \<0.001 0.927 1.33 (1.07--1.66) 2.59 0.010 Comorbidities Atrial fibrillation (yes vs. no) 1.11 (0.94--1.31) 1.25 0.211 0.016 Male 1.02 (0.84--1.23) 0.19 0.850 Female 1.69 (1.17--2.44) 2.82 0.005 Diabetes mellitus (yes vs. no) 1.56 (1.30--1.88) 4.80 \<0.001 0.539 COPD (yes vs. no) 1.34 (1.07--1.68) 2.54 0.011 0.165 Renal disease (yes vs. no) 1.98 (1.62--2.41) 6.76 \<0.001 0.474 1.62 (1.28--2.04) 4.07 \<0.001 Device therapy (yes vs. no) 1.08 (0.89--1.31) 0.81 0.419 0.133 Laboratory Anaemia^a^ (yes vs. no) 2.49 (2.07--3.00) 9.64 \<0.001 0.201 Haemoglobin (per g/dL) 0.75 (0.71--0.78) −11.77 \<0.001 0.216 0.79 (0.74--0.83) −7.87 \<0.001 CRP (per doubling) 1.48 (1.40--1.57) 13.66 \<0.001 0.805 1.39 (1.30--1.48) 10.14 \<0.001 AST (per doubling) 0.97 (0.87--1.09) −0.45 0.653 0.740 ALT (per doubling) 0.94 (0.86--1.02) −1.44 0.150 0.467 γ-GT (per doubling) 0.99 (0.84--1.24) −0.18 0.859 0.967 Alkaline phosphatase (per doubling) 1.06 (0.89--1.26) 0.65 0.516 0.186 Total bilirubin (per doubling) 1.08 (0.98--1.18) 1.57 0.116 0.942 Sodium (per mmol/L) 0.97 (0.94--0.99) −3.19 0.001 0.162 Potassium (per mmol/L) 0.81 (0.70--0.94) −2.73 0.006 0.182 NT-proBNP (per doubling) 1.23 (1.16--1.30) 6.95 \<0.001 0.028 Male 1.26 (1.18--1.35) 6.91 \<0.001 Female 1.08 (0.96--1.22) 1.25 0.211 Creatinin (per doubling) 1.35 (1.15--1.58) 3.74 \<0.001 0.027 Male 1.69 (1.40--2.05) 5.40 \<0.001 Female 1.09 (0.78--1.53) 0.52 0.604 eGFR (per doubling) 0.79 (0.69--0.91) −3.39 0.001 0.162 Albumin (per 5 g/L) 0.81 (0.77--0.86) −8.09 \<0.001 0.776 0.93 (0.87--0.98) −2.46 0.014 Urea (per doubling) 1.26 (1.15--1.38) 4.90 \<0.001 0.387 Medication Loop diuretics (yes vs. no) NA Beta-blockers on target dose (yes vs. no) 0.84 (0.59--1.21) −0.92 0.359 0.716 ACEi/ARB on target dose (yes vs. no) 0.89 (0.70--1.13) −0.94 0.346 0.472 Aldosterone antagonist (yes vs. no) 0.80 (0.68--0.94) −2.64 0.008 0.745 Proton-pump inhibitors (yes vs. no) 1.58 (1.32--1.88) 4.99 \<0.001 0.197 P2Y~12~ inhibitors (yes vs. no) 1.57 (1.23--2.00) 3.64 \<0.001 0.693 1.64 (1.24--2.16) 3.47 0.001 Acetylsalicylic acid (yes vs. no) 1.04 (0.88--1.22) 0.42 0.674 0.225 Anticoagulants (yes vs. no) 1.01 (0.86--1.20) 0.17 0.866 0.156 ACEi, angiotensin-converting enzyme inhibitor; ALT, alanine aminotransferase; ARB, angiotensin receptor blocker; AST, aspartate aminotransferase; BMI, body mass index; CI, confidence interval; COPD, chronic obstructive pulmonary disease; CRP, C-reactive protein; eGFR, estimated glomerular filtration rate; γ-GT, gamma-glutamyltransferase; JVP, jugular venous pressure; LVEF, left ventricular ejection fraction; NT-proBNP, N-terminal pro B-type natriuretic peptide. Anaemia was defined as a haemoglobin level \<12 g/dL in women and \<13 g/dL in men. Biomarker profile of iron deficiency ------------------------------------ Median log~2~ levels of the 92 cardiovascular biomarkers from the Olink Cardiovascular III panel are depicted in [Supplementary material online, Table S2](#sup1){ref-type="supplementary-material"}. In patients with ID, the following biomarkers were significantly up-regulated with largest magnitude of change: fatty acid binding protein 4 (FABP4), growth differentiation factor 15 (GDF15), NT-proBNP, osteopontin (OPN), ST2 protein (ST2), tumour necrosis factor receptor 1 (TNF-R1), and transferrin receptor protein 1 (TR). Only paraoxonase 3 (PON3) and tartrate-resistant acid phosphatase type 5 (TR-AP) were strongly and significantly down-regulated in ID (Figure [2](#ehz680-F2){ref-type="fig"}). After correcting for the determinants for ID originating from Table [2](#ehz680-T2){ref-type="table"}, only PON3, TR-AP, ST2, NT-proBNP, and TR remained significantly associated with ID (all P \< 0.05; [Supplementary material online, Table S3](#sup1){ref-type="supplementary-material"}). ![Biomarker expression profile (volcano plot) in iron-deficient heart failure patients compared to patients without iron deficiency. The volcano plot shows the difference in cardiovascular biomarker expression in patients with and without iron deficiency. Each dot represents one of the 92 biomarkers of the Olink Cardiovascular III panel. On the x-axis, the log~2~-fold change in biomarker expression is depicted (positive log~2~-fold change is higher biomarker expression in patients with iron deficiency; negative log~2~-fold change is lower biomarker expression in patients with iron deficiency), while the y-axis shows the magnitude of the biomarker expression difference as −log~10~ of the P-value. Red dots are biomarkers with a significant up- or down-regulation in patients with iron deficiency (corrected for a false discovery rate of 5%); green dots indicate biomarkers with an absolute log~2~-fold change of \>0.25. Most biomarkers were significantly up-regulated in patients with ID (n = 39, 42.4%), while 15 biomarkers had significantly lower expression (16.3%). FABP4, fatty acid binding protein 4; GDF15, growth differentiation factor 15; NT-proBNP, N-terminal prohormone brain natriuretic peptide; OPN, osteopontin; PON3, paraoxonase 3; ST2, ST2 protein; TNF-R1, tumour necrosis factor receptor 1; TR, transferrin receptor protein 1; TR-AP, tartrate-resistant acid phosphatase type 5.](ehz680f2){#ehz680-F2} Prognostic consequences of iron deficiency ------------------------------------------ During a median follow-up of 21 months, overall rates of mortality and first HF hospitalization were 26.9% (n = 615) and 24.5% (n = 578), respectively. Event rates were comparable between males and females (see [Supplementary material online, Figure S3A and B](#sup1){ref-type="supplementary-material"}). Kaplan--Meier estimator curves (for all-cause mortality and the composite endpoint of all-cause mortality and first HF hospitalization) and cumulative incident curves (for first HF hospitalization), stratified by iron status, are shown in [Supplementary material online, Figure S2A--C](#sup1){ref-type="supplementary-material"}. Iron deficiency was a significant predictor for all endpoints (all P \< 0.05). No interaction was observed between iron status and sex on all endpoints. Univariable and multivariable Cox proportional hazard regression analyses for all endpoints are depicted in [Supplementary material online, Tables S4 and S7](#sup1){ref-type="supplementary-material"}. Iron deficiency remained independently associated with the primary composite endpoint of all-cause mortality and first HF rehospitalization after correcting for the BIOSTAT prediction model \[hazard ratio (HR) 1.30, 95% confidence interval (CI) 1.12--1.50; P = 0.0005\] and the logistic regression prediction model for ID (HR 1.25, 95% CI 1.06--1.46; P = 0.007). In a multivariable fractional polynomial analysis, lower TSAT levels were associated to an increased risk of all-cause mortality (see [Supplementary material online, Figure S4](#sup1){ref-type="supplementary-material"}). Finally, in the prognostic models including haemoglobin (all-cause mortality and the composite endpoint), we compared the prognostic power of haemoglobin and ID (defined as TSAT \< 20% and using TSAT as a continuous variable). Exchanging haemoglobin for either ID or TSAT did not alter the prognostic power of both models (see [Supplementary material online, Table S8](#sup1){ref-type="supplementary-material"}). Discussion ========== In a large cohort of patients with worsening HF, we identified the following independent determinants of ID: female sex, lower estimated protein intake, higher heart rate, presence of peripheral oedema and orthopnoea, history of renal disease, lower haemoglobin, higher CRP levels, lower serum albumin levels, and antiplatelet use. None of these factors had a significant interaction with sex. The adverse prognostic consequences of ID are independent of these identified predictors. Finally, we provided a biomarker profile of patients with ID, in which predominantly pro-inflammatory markers seem up-regulated. Determinants of iron deficiency ------------------------------- Our observational data suggest factors which may be involved in the aetiology of ID in HF. These determinants are depicted in [Take home figure](#ehz680-F3){ref-type="fig"} and are discussed below. ![Determinants of iron deficiency in heart failure. Several graphical elements in this figure are provided by Freepik and DinosoftLabs from [www.flaticon.com](http://www.flaticon.com).](ehz680f3){#ehz680-F3} ### Sex difference In the final prediction model for ID, we could not identify significant interactions with sex. Instead, female sex was an independent predictor for ID, which has been reported in other studies as well.[@ehz680-B1]^,^[@ehz680-B4]^,^[@ehz680-B15] Several mechanisms might be underlying this association. First, to exclude menstrual blood loss as a confounding factor, we performed a sensitivity analysis in which we excluded premenopausal women (i.e. age \<52 years; [Supplementary material online, Table S5](#sup1){ref-type="supplementary-material"}). This yielded a nearly identical prediction model for ID. Second, female patients in our cohort had a higher prevalence of HFpEF, a HF subtype which has been linked to highest prevalence of ID.[@ehz680-B16] ### Reduced estimated protein intake As depicted in Figure[1](#ehz680-F1){ref-type="fig"}A, the prevalence of ID rapidly increases in patients with lower estimated protein intake. Additionally, we identified lower serum albumin levels as an independent predictor of ID (Figure[1](#ehz680-F1){ref-type="fig"}C). These findings might suggest a poor nutritional status as an aetiological pathway of ID. Several risk scores for estimating malnutrition have included serum albumin levels as predictor, also in HF patients.[@ehz680-B17] Although we did not study dietary iron intake per se, daily protein intake should provide a fair estimation of daily dietary iron intake, as a significant amount of dietary iron intake is provided by protein-rich food, such as meat (haem iron), nuts and legumes (non-haem iron).[@ehz680-B21] It should be acknowledged that we only estimated total daily protein intake using surrogate markers; we did not have data on the exact daily protein intake, nor exact daily iron intake. ### Venous congestion Heart failure is associated with right-sided venous congestion which leads to increased gastrointestinal wall thickness and malabsorption.[@ehz680-B22] Intestinal wall oedema due to right-sided congestion might also negatively influence nutrient absorption, including dietary iron. In the multivariable prediction model for ID, we showed that peripheral oedema, as an indicator of right-sided congestion, was an independent predictor of ID. Consequently, malabsorption due to venous congestion may also play a role in the aetiology of ID in HF. ### Antiplatelet drugs Iron-deficient patients had higher prevalence of antiplatelet drug use compared to patients without ID. These patients might be more prone to (sub)clinical gastrointestinal blood loss, for example due to gastrointestinal malignancies or angiodysplasia, which might eventually lead to ID due to iron loss. A recent study by Meijers et al.[@ehz680-B26] revealed that HF patients may be at risk for incident cancer, including colorectal cancer, possibly due to circulating cardiac and inflammatory markers. We did not find differences in vitamin K antagonist use in patients with and without ID. Due to the very low use of direct-acting oral anticoagulants (DOACs) in the present cohort (n = 17, 0.7%), it was not possible to study this drug group in relation to ID. Given the conflicting gastrointestinal bleeding risk of DOACs compared to vitamin K antagonists, the prevalence of anticoagulant-related bleeding as a cause of ID might change in the future as DOAC prescriptions become more common in general practice. ### Chronic inflammation Iron-deficient patients in the present study had higher levels of inflammatory markers compared to patients without ID, whereas hepcidin levels were lowest in iron-deficient patients. This is an interesting finding, as hepcidin levels are expected to be elevated in inflammatory states. Despite the pro-inflammatory state in iron-deficient patients, there must be mechanisms lowering hepcidin levels in these patients. It seems conceivable that these patients have ID due to iron unavailability, malabsorption or loss rather than inflammation, which might explain lower hepcidin levels. In this hypothesis, the influence of chronic inflammation on iron status via the inflammation-hepcidin-iron axis seems limited: systemic iron status itself seems to dictate hepcidin release over inflammatory status. This hypothesis has been postulated by Weber et al.[@ehz680-B27] in their study comprising 60 stable chronic HF patients with anaemia. They showed that iron-deficient patients have lower hepcidin levels despite increased levels of the pro-inflammatory cytokine tumour necrosis factor-α (TNF-α). Our study yields comparable results using CRP and pro-inflammatory biomarkers from the Olink Cardiovascular III panel. Furthermore, some pro-inflammatory cytokines directly influence iron status independently of hepcidin, for example TNF-α.[@ehz680-B28] Biomarker profile of iron deficiency ------------------------------------ As shown in Figure [2](#ehz680-F2){ref-type="fig"}, many biomarkers from the Olink Cardiovascular III panel were significantly up- or down-regulated in ID. After correcting for HF severity, renal function and predictors for ID, several biomarkers remained independently associated with iron status. Of these, paraoxonase 3 (PON3) and TR-AP are both down-regulated in ID. PON3 is a liver-derived, HDL-bound protein, which has several antiatherogenic and antioxidative properties. In animal studies, overexpression of PON3 seemed protective against atherosclerosis and cardiac hypertrophy, while PON3-deficient mice show mitochondrial and fatty acid oxidation dysfunction.[@ehz680-B29] Second, TR-AP is predominantly an osteoclast-derived, iron-containing protein reflecting bone turnover rate and is expressed by activated macrophages. In vitro studies show that the expression of TR-AP is regulated by iron status.[@ehz680-B33]^,^[@ehz680-B34] TR-AP knockout mice display altered osteoclastic function leading to mild osteopetrosis and an increased proinflammatory response.[@ehz680-B35]^,^[@ehz680-B36] Unfortunately, both PON3 and TR-AP are poorly studied in human (patho)physiology. Although our study shows an independent link between PON3, TR-AP, and iron status, the clinical significance of the biomarker expression pattern in ID need to be elucidated. Clinical implications --------------------- Our data confirm the adverse prognostic consequences of ID, which are independent of established predictors of outcome. While ID is frequently observed in HF patients, the aetiology is often unknown. However, it is essential to explore the underlying cause(s), since some of them are treatable and reversible.[@ehz680-B37] For example, if ID is caused by gastrointestinal blood loss, the underlying cause of this blood loss (e.g. malignancy, angiodysplasia or antiplatelet use) needs to be detected and treated. When ID is caused by the use of antiplatelets or anticoagulants, their use should be reconsidered, especially in patients without a direct treatment indication. Finally, when poor nutritional status is causing ID, this should be treated as well. Strengths and limitations ------------------------- To our knowledge, this is the largest cohort with clinical and biochemical parameters available, providing more knowledge on the drivers of ID in HF patients. However, an important limitation of this study is its observational character, making it challenging to directly study aetiological and pathophysiological mechanisms. Nevertheless, we hope our data encourage more studies on the determinants identified in our study. As iron indices were only measured at a single time point, differences in iron status over time could not be studied. Second, there were no data available on the presence or absence of recent blood loss (e.g. blood donations, surgery), which could possibly affect iron status. Third, protein intake was estimated using a formula based on urinary urea and body mass index, and not directly measured. Finally, the majority of patients in BIOSTAT-CHF is male. Therefore, we cannot rule out that the prediction models for ID in males had more statistical power to identify independent determinants of ID compared to the models in female patients. Funding ======= The BIOSTAT-CHF study was supported by the European Commission (FP7-242209-BIOSTAT-CHF). Conflict of interest: The University Medical Center Groningen, which employs several authors, has received research grants and/or fees from AstraZeneca, Abbott, Bristol-Myers Squibb, Novartis, Roche Diagnostics, Trevena, and Thermofisher GmbH. N.G.B. received personal fees from Vifor Pharma. K.D. received honoraria and/or research support from device companies Biotronik and Sorin, Boston Scientific St Jude, and Medtronic, and pharmaceutical companies Abbott, Amgen, Astra Zeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, GSK, Leo, Merck, Novartis, Otsuka, Pfizer, Roche, Sanofi, and Servier. S.D.A. received grants from Abbott Vascular and Vifor Pharma, and consultancy fees from Bayer, Boehringer Ingelheim, Brahms, Cardiorentis, Janssen, Novartis, Relypsa, Servier, Stealth Peptides, Vifor Pharma, and ZS Pharma. C.C.L. received consultancy fees and/or research grants from Amgen, Astra Zeneca, MSD, Novartis, and Servier. D.J.v.V. received board membership fees or travel expenses from BioControl, Cardiorentis, Novartis, Johnson & Johnson, Vifor Pharma, Zoll Medical, CorviaMedical and Arca. A.A.V. received consultancy fees and/or research grants from Alere, Amgen, Bayer, Boehringer Ingelheim, Cardio3Biosciences, Celladon, GSK, Merck/MSD, Novartis, Roche Diagnostics, Servier, Singulex, Sphingotec, Stealth Peptides, Trevana, Vifor Pharma, and ZS Pharma. P.v.d.M. received consultancy fees and/or grants from Novartis, Servier, Vifor Pharma, Astra Zeneca, Pfizer and Ionis. H.H.v.d.W. and L.L.N. have nothing to disclose. Supplementary Material ====================== ###### Click here for additional data file.	{ "pile_set_name": "PubMed Central" }
1. Introduction =============== Non-alcoholic fatty liver disease (NAFLD) was first described in 1980 and has recently become one of the most prevalent liver diseases in developing as well as in developed countries \[[@B1-ijms-15-09924],[@B2-ijms-15-09924],[@B3-ijms-15-09924],[@B4-ijms-15-09924],[@B5-ijms-15-09924]\]. NAFLD represents a spectrum ranging from simple steatosis to hepatic inflammation, hepatocyte ballooning, presence of Mallory-Denk bodies (MDBs) and fibrosis, which is referred as non-alcoholic steatohepatitis (NASH) in case of proven alcohol abstinence \[[@B6-ijms-15-09924],[@B7-ijms-15-09924],[@B8-ijms-15-09924],[@B9-ijms-15-09924],[@B10-ijms-15-09924]\]. The distinction between these diseases is important as steatosis is less likely to evolve into severe liver related complications \[[@B11-ijms-15-09924],[@B12-ijms-15-09924]\], whereas NASH might progress to liver cirrhosis and hepatocellular carcinoma (HCC) \[[@B8-ijms-15-09924],[@B12-ijms-15-09924]\]. NAFLD is strongly associated with obesity, which is frequently accompanied by the metabolic syndrome (MetS), reduced glucose tolerance and type 2 diabetes mellitus (T2DM) \[[@B13-ijms-15-09924],[@B14-ijms-15-09924]\]. A well-established theory says that the development from steatosis to NASH follows a "two-hit" concept where the first hit is the sensitization of the liver to injurious effects and the second one the cause of inflammation and fibrosis \[[@B15-ijms-15-09924]\]. However, this theory has been put into question based on insights on the interaction between insulin resistance, adipose tissue inflammation and adipokines. Therefore a modification, the "multiple-parallel-hit" hypothesis has arisen \[[@B16-ijms-15-09924]\], meaning that after the first hit, which is insulin resistance in combination with many associated metabolic dysfunctions the liver is at risk for other hits resulting in NASH and liver fibrosis. Both hypotheses have in common that simple steatosis is believed to be the basis for the development of NASH \[[@B17-ijms-15-09924]\]. 2. Epidemiology and Risk Factors ================================ Many studies have been published from developed nations like the United States, Europe and Australia, which show a prevalence of NASH in 20% to 30% of the population \[[@B18-ijms-15-09924],[@B19-ijms-15-09924],[@B20-ijms-15-09924],[@B21-ijms-15-09924]\]. Asia, which was commonly thought to be a low risk area shows a prevalence range between 15% and 30% \[[@B22-ijms-15-09924],[@B23-ijms-15-09924],[@B24-ijms-15-09924]\]. In one study from Korea 51% of the tested patients suffered from NAFLD \[[@B25-ijms-15-09924]\]. Demographic analysis indicates how ethnicity influences the extent of NASH incidence. In Hispanic populations the prevalence of NAFLD and NASH is the most common followed by Non-Hispanic Whites and African Americans \[[@B21-ijms-15-09924],[@B26-ijms-15-09924],[@B27-ijms-15-09924]\]. The lowest rate of NAFLD reported was in American Indian and Alaska-Native populations, which ranges from 0.4% to 2% \[[@B28-ijms-15-09924],[@B29-ijms-15-09924]\]. Children from these ethnicities are predisposed in a similar way. The highest incidence was observed in Hispanics and Asians \[[@B30-ijms-15-09924],[@B31-ijms-15-09924]\], but up to now it is not known how far the genetics and racial differences compared to the environmental factors reflect the incidence of NAFLD. A clear gender difference concerning amount and distribution of body fat is also recognized. Men are more likely to develop NAFLD based on their tendency to store fat around organs in the abdominal cavity \[[@B32-ijms-15-09924]\]. One Asian study showed on 26,527 patients that the prevalence of NAFLD was 31% in men compared to 13% in women \[[@B33-ijms-15-09924]\]. A study from India demonstrated that the majority of those with NAFLD are men \[[@B34-ijms-15-09924]\]. These results may correlate with the elevated aminotransferase levels and the presence of NASH and hepatic fibrosis in males \[[@B26-ijms-15-09924],[@B35-ijms-15-09924]\]. One reason for this inequality in fat distribution is the twice as high hepatic lipase activity in men, which may contribute to a lower high density lipoprotein (HDL) cholesterol level and heightened levels of dense low density lipoprotein (LDL) particles \[[@B36-ijms-15-09924]\]. This distinction in hepatic lipase activity can be traced back to the suppression of androgenic steroids \[[@B37-ijms-15-09924]\]. In addition, it has been shown that female steroids, especially estrogen, promote the utilization of fat as an energy source, as well the non-visceral fat, which makes women less likely to accumulate abdominal fat \[[@B38-ijms-15-09924],[@B39-ijms-15-09924],[@B40-ijms-15-09924]\]. There are several risk factors known which are likely to be required for the development of obesity, visceral adiposity, insulin resistance, steatosis and fibrosis: genetic predisposition, ethnicity, age, gender and lifestyle. Most patients diagnosed with NAFLD have these metabolic risk factors. In 57% to 98% \[[@B41-ijms-15-09924],[@B42-ijms-15-09924],[@B43-ijms-15-09924]\] of overweight people NAFLD is diagnosed, whereas simple steatosis ranges between 30% and 37% \[[@B44-ijms-15-09924],[@B45-ijms-15-09924]\]. Compared to non-obese patients, being overweight increases the risk to develop fibrosis three-fold \[[@B13-ijms-15-09924]\]. Especially visceral obesity shows a higher correlation to fatty liver than normal body mass. This increased fat accumulation in the abdomen causes higher lipolytic activity and thus increased plasma free fatty acid (FFA) levels \[[@B46-ijms-15-09924],[@B47-ijms-15-09924]\] that are toxic for hepatocytes and mitochondria, leading to apoptosis and inflammation. Insulin resistance is an indicator of the risk to develop NAFLD \[[@B44-ijms-15-09924]\]. Recently, several studies have shown that adipose tissue insulin resistance (Adipo-IR) could be a predictor of the histological changes of the liver and, for patients already suffering from NAFLD, a possible predictor of fibrosis progression \[[@B48-ijms-15-09924],[@B49-ijms-15-09924]\]. 3. Genomics and Epigenomics =========================== Several molecular mechanisms are believed to be associated with the pathogenesis of NASH, but their influence on NAFLD still needs to be determined \[[@B50-ijms-15-09924],[@B51-ijms-15-09924]\]. The hedgehog (hh) pathway plays a critical role in the immune response \[[@B52-ijms-15-09924]\] and is involved in natural killer T cell enrichment leading to a fibrogenetic hepatic response, which is observed in NASH \[[@B53-ijms-15-09924],[@B54-ijms-15-09924]\]. miRNAs, DNA methylation patterns, histone modifications and ubiquitination have therefore been investigated. miRNAs are proven to be important regulators of cellular processes \[[@B55-ijms-15-09924],[@B56-ijms-15-09924]\]. In a recent study, 46 out of 474 miRNAs were differentially expressed in NASH patients compared to a control group \[[@B57-ijms-15-09924]\]. The most under-expressed one was miRNA-122, being involved in lipid and cholesterol metabolism as well as in adipocyte differentiation. Further, an up-regulated expression level of miRNA-335 was determined in murine liver and white adipose tissue. miRNA-335 is involved in increased body, liver and white adipose tissue weight and in heightened cholesterol- and hepatic triglyceride levels \[[@B58-ijms-15-09924]\]. Also the miRNA-34 family, being a direct transcriptional target of p53, is dysregulated in NAFLD \[[@B59-ijms-15-09924]\]. Especially miRNA-34a expression, is involved in apoptosis, increases in correlation to the severity of disease \[[@B60-ijms-15-09924]\]. NAFLD appears to be multifactorial, including environmental as well as genetic factors \[[@B27-ijms-15-09924]\]. There are several genetic variants known to be involved in energy balance. The most prominent one is rs738409, a SNP in the patatin-like phospholipase domain-containing 3 (PNPLA3) gene \[[@B61-ijms-15-09924]\]. PNPLA3 is a membrane-bound protein with lypolytic and lipogenic activities being expressed in hepatocytes and adipocytes \[[@B62-ijms-15-09924],[@B63-ijms-15-09924]\]. Other genetic variants being associated with NAFLD involve genes related to oxidative stress, inflammation and fibrogenesis, such as super oxide dismutase 2 (SOD2) \[[@B64-ijms-15-09924]\]. Controversially, even with a proven association of PNPLA3 with NAFLD \[[@B62-ijms-15-09924],[@B63-ijms-15-09924]\], there is no effect on very low density lipoprotein (VLDL), LDL, HDL, insulin resistance, total and circulating cholesterol levels, which are accounted as primary risk factors \[[@B65-ijms-15-09924],[@B66-ijms-15-09924]\]. The PNPLA3 genotype distribution was reported to differ between patients affected by NAFLD and NASH. This indicates, that NASH might genetically differ from the other spectrum of the disease complex, thus PNPLA3 being associated with hepatic necroinflammation \[[@B61-ijms-15-09924]\]. 4. Diagnosis ============ Several diagnostic methods are available to evaluate NAFLD. The liver biopsy remains the gold standard, but as it is invasive, it cannot be used for population-based studies. Thus, several non-invasive methods have been introduced including imaging techniques like magnetic resonance imaging (MRI) and the measurement of serum markers such as alanine aminotransferase (ALT) and aspartate aminotransferase (AST). In 45%--100% of NAFLD cases the disorder remains asymptomatic \[[@B11-ijms-15-09924]\] and 55%--79% of NAFLD patients have inconspicuous transaminase levels \[[@B67-ijms-15-09924]\]. Therefore aminotransferases in combination with the measurement of glucose, triglyceride, cholesterol and lipoprotein serum levels are used in combination with anthropometric parameters such as body mass index (BMI) and fat distribution and other information, like gender, age, lifestyle habits and family disease history \[[@B2-ijms-15-09924],[@B68-ijms-15-09924]\]. Liver ultrasonography is a good tool to estimate the degree and extent of steatosis by using a series of ultrasonographical characteristics \[[@B69-ijms-15-09924]\]. Computed tomography (CT) is more specific to ultrasonography, but because of the radiation exposure it is only used for research studies in adults. An alternative would be magnetic resonance tomography (MRT), which is currently tested with different phase-shift imaging methods to heighten the accuracy of the total liver-fat content measurement \[[@B70-ijms-15-09924],[@B71-ijms-15-09924],[@B72-ijms-15-09924]\], but due to the high costs it remains primarily investigational. Transient elastography (TE) is an ultra-sound based method to evaluate fibrosis. Recently, a study demonstrated that TE is a valuable tool for the detection and staging of fibroses even in children \[[@B73-ijms-15-09924]\]. Another study showed that TE has a sensitivity of 91% and a specificity of 75% in detecting stage 3 or higher fibrosis in NAFLD patients \[[@B2-ijms-15-09924]\], but here the failure rate increases in correlation to body weight as well. An improvement of this technique would be magnetic resonance elastography, measuring the stiffness of the entire organ, in contrast to TE, which measures only around 50 cm^2^ \[[@B74-ijms-15-09924]\]. Another promising technique to predict fibrosis in NAFLD patients are the FibroTest (FT) and the Actitest (Biopredictive Paris, France, FibroSURE in the US patented artificial intelligence algorithm USPTO 6,631,330). In combination with the fibrosis index which includes a various amount of markers, like alanine aminotransferase (ALT), apolipoprotein A1, α2-macroglobulin (A2M) and others, this combination showed a sensitivity of 77% and a specificity of 98% by diagnosing fibrosis in 170 NAFLD patients \[[@B75-ijms-15-09924]\]. In summary, liver biopsy remains the gold standard to characterize liver diseases. It is currently the only tool to assess the degree of inflammation as well the degree and stage of fibrosis, but the appearance of new promising diagnostic techniques such as the FT, the Actitest or magnetic resonance elastography, will hopefully replace usage of liver biopsy. 5. Molecular Pathology ====================== Progression of steatosis is mainly mediated by cytokines and a panel of different enzymes which fulfill important roles in lipogenesis and lipolysis \[[@B76-ijms-15-09924],[@B77-ijms-15-09924],[@B78-ijms-15-09924]\]. In this context, visceral fat seems to play a role in NAFLD progression due to its ability to actively secret a multitude of different adipokines such as tumor necrosis factor alpha (TNFα), interleukin-6 (IL-6), macrophage chemoattractant protein (MCP)-1 and resistin, which promote insulin resistance and type 2 diabetes mellitus (TD2M) \[[@B79-ijms-15-09924],[@B80-ijms-15-09924],[@B81-ijms-15-09924]\]. Obesity and increased levels of inflammatory cytokines are well established as tightly related to each other \[[@B82-ijms-15-09924],[@B83-ijms-15-09924]\]. Different studies on obese patients and obese murine models showed a correlation between visceral fat, insulin resistance and an increased predisposition for NAFLD and NASH \[[@B84-ijms-15-09924],[@B85-ijms-15-09924],[@B86-ijms-15-09924]\]. The protein hormone adiponectin and the cytokine TNFα have extremely important functions in this disease. Adiponectin is synthesized by visceral adipocytes and plays an important role regarding insulin tolerance, because it directly activates 5\'-adenosine monophosphate-activated protein kinase (AMPK), glucose consumption and fatty acid oxidation \[[@B87-ijms-15-09924]\]. Measurement of adiponectin serum levels from visceral adipose subjects of different ages revealed significantly decreased levels of adiponectin representing an inverse relation to the rate of body fat \[[@B88-ijms-15-09924],[@B89-ijms-15-09924],[@B90-ijms-15-09924]\]. The reduction of adiponectin seems to have a major impact on the development of hepatic steatosis and NASH caused by its direct antagonistic effect on TNFα, one of the most important cytokines in mediating inflammation. As opposed to this, human NAFLD patients have been demonstrated to display increased serum levels of TNFα \[[@B89-ijms-15-09924],[@B91-ijms-15-09924]\]. It appears that the ratio between adiponectin and TNFα might be essential for the progression of NASH \[[@B92-ijms-15-09924]\]. TNFα functions antagonistically to adiponectin and is secreted by hepatocytes and adipocytes. As a master regulator of inflammation, TNFα is involved in both regulation of other cytokines like IL-6 and in the NF-κB signaling cascade \[[@B93-ijms-15-09924],[@B94-ijms-15-09924]\]. 6. Treatment and Prevention =========================== Because of the incomplete understanding of the molecular pathogenesis of NAFLD, the current therapy focuses rather on preventing risk factors or treatment of side effects accompanying obesity. These interventions target lifestyle changes, including diet and physical exercises mainly in combination with pharmacotherapy. This includes next to NAFLD treatment, improvement of the MetS, T2DM and related cardiovascular diseases. The most important lifestyle modification is weight loss, which has been shown to improve serum aminotransferases and BMI, but also a decrease in hepatic steatosis \[[@B95-ijms-15-09924],[@B96-ijms-15-09924],[@B97-ijms-15-09924]\]. One study of biopsy-proven NASH patients had shown that patients who lost 5% of their body weight had improved insulin sensitivity and steatosis. Moreover, it could be demonstrated that only a weight reduction of 9% leads to significant positive changes concerning inflammation, ballooning and the NAFLD activity score \[[@B98-ijms-15-09924]\]. However, rapid weight loss (24% body weight in 8 weeks) increases portal inflammation and fibrosis as well \[[@B99-ijms-15-09924]\]. An important mediator for hepatic injury in NASH is oxidative stress \[[@B100-ijms-15-09924],[@B101-ijms-15-09924],[@B102-ijms-15-09924]\], therefore an antioxidant therapy may slow the progression of steatosis and NASH. A prominent antioxidant is Vitamin E which when used in dosages ranging from 400 to 1200 units per day over a time period of 2--4 months leads to a normalization of obese children \[[@B103-ijms-15-09924]\]. The Pioglitazone, Vitamin E, or Placebo for Nonalcoholic Steatohepatitis (PIVENS) trial, the largest study evaluating the effect of Vitamin E in nondiabetic adults with NASH demonstrated a significant clinical and biochemical improvement \[[@B104-ijms-15-09924]\]. Metformin which can also applied in combination with Pioglitazone reduces glucose production and increases insulin sensitivity in patients suffering from T2DM \[[@B105-ijms-15-09924],[@B106-ijms-15-09924]\], but subsequent randomized clinical trials were not able to proof significant differences in liver histology \[[@B107-ijms-15-09924],[@B108-ijms-15-09924]\]. Therefore metformin as therapy for NASH patients is not recommended. Thiazolidinediones (TZDs) increase hepatic insulin sensitivity and improve glucose and lipid utilization in T2DM \[[@B105-ijms-15-09924]\]. One study with 22 NASH patients showed improved inflammation, ballooning and fibrosis, but 67% of the patients gained weight \[[@B109-ijms-15-09924]\]. Another double-blind study recognized a significant improved steatosis, ballooning and inflammation in 73% of the NASH patients \[[@B110-ijms-15-09924]\]. Long-chain omega-3 fatty acids showed a putative decrease in steatosis and markers of inflammation, and improvement of insulin sensitivity in experimental studies \[[@B111-ijms-15-09924]\]. Pentoxifylline, a xanthine derivative, and promising in several studies, could be worth for further testing \[[@B112-ijms-15-09924]\]. At this time it should be concluded that there is no drug treatment available that is proven to cure fatty liver without side effects. 7. Murine Non-Alcoholic Fatty Liver Disease (NAFLD) Models and Their Potential Human Relevance ============================================================================================== Inhibitor of nuclear factor kappa-B kinase subunit beta (IKKβ) and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) have been shown to be crucial in promoting inflammation and insulin resistance in different studies investigating murine transgenic and knockout models fed with high fat diet (HFD) \[[@B113-ijms-15-09924],[@B114-ijms-15-09924],[@B115-ijms-15-09924]\]. The use of the IKKβ blocker aspirin or of other salicylates had a markedly positive effect on insulin resistance in patients with TD2M as well as in insulin resistant mouse models \[[@B113-ijms-15-09924]\]. Generally, salicylates seem to have a potentially positive influence on the development of NASH and its concomitant effects. Therefore, it would be certainly a useful drug, but the side effects of salicylates such as liver toxicity and anticoagulation have to be seen critical and are currently under debate. Besides TNFα, IL-6 is another important adipocytokine linked to NAFLD and obesity. Similarly to TNFα, enhanced amounts of IL-6 are also secreted by the visceral fat of obese subjects \[[@B81-ijms-15-09924]\]. In the work by Cai et al. \[[@B115-ijms-15-09924]\] increased levels of IL-6 in NAFLD patients were associated with IKKβ mediated NF-κB activation. This study demonstrated a significantly positive influence of an anti-IL-6 antibody on the progression of insulin resistance \[[@B115-ijms-15-09924]\]. Overall, HFD leads to a considerable activation of TNFα and IL-6. Nevertheless, inhibition of TNFα via TNFR1 or of IL-6 through antibodies or blocking of IKKβ activity protects against hepatic steatosis. Moreover, ablation of Fas (CD95), a further member of the tumor necrosis factor receptor (TNFR)-superfamily, in murine adipocytes has a similar positive effect on insulin resistance during HFD as demonstrated for TNFα and IL-6 \[[@B116-ijms-15-09924]\]. One trigger of inflammation is caused by endotoxins/lipopolysaccharides (LPS) from gut bacteria. LPS is one element of the outer membrane of gram-negative bacteria. It is released into the gut and circulates through the portal blood directly into the liver \[[@B78-ijms-15-09924]\]. There, LPS activates Toll-like receptor 4 (TLR 4), a member of the interleukin-1 receptor/toll-like receptor superfamily. TLRs are located in the liver on the membrane of Kupffer cells, hepatocytes, hepatic stellate cells, biliary epithelial cells, hepatic dendritic cells and liver sinusoidal epithelial cells \[[@B117-ijms-15-09924]\]. The work by Cani and colleagues showed that LPS has the potential to promote insulin resistance. The authors set C57Bl6/J mice on a HFD for 4 weeks and measured 2--3 times higher LPS blood plasma concentrations compared to normal chow fed mice. An increased growth rate of LPS-releasing microbiota in the gut was reported. Interestingly, similar results were generated upon administration of subcutaneously injected LPS ("induction of endotoxemia") for 4 weeks \[[@B118-ijms-15-09924]\]. Other murine models have validated that NASH induced by high fructose or methionine-choline deficient (MCD) diet also leads to increased expression rates of Tlr4 \[[@B119-ijms-15-09924]\]. Conversely, deletion of Tlr4 in mice by depletion of Kupffer cells through chlodronate administration, or generated by a single point mutation fed with MCD diet, showed less hepatic steatosis in comparison to the wild-type controls \[[@B120-ijms-15-09924]\]. These results give evidence that TLR4 and Kupffer cells might have a critical function in mediating steatohepatitis and other concomitant effects like fibrosis. It is under debate if downstream effector molecules such as reactive oxygen species (ROS), chaperone proteins or transcription factors could also play a role in NASH development. Next to the study by Cani, Yoshimoto and colleagues \[[@B121-ijms-15-09924]\] recently published that the gut microbiota has a great impact on HCC development in case of obesity. Genetic obesity seems to directly influence the composition of gut microbiota, thus promoting enhanced secretion of the metabolite deoxycholic acid (DCA), which has DNA damaging properties. Increased levels of DCA stimulate hepatic stellate cells (HSCs) to release proinflammatory and tumor-promoting factors through the liver. In their obesity mouse model, application of a cancer-promoting chemical in these mice increased their susceptibility to develop HCC. Similar results such as activated HSCs in the vicinity of HCCs could be obtained in human NASH patients, underlining the impact of obesity and gut bacteria composition \[[@B121-ijms-15-09924]\]. Based on the results of various mouse models and the medical diagnosis of patients, it is debated if NASH/NAFLD is the result of multiple hit events ([Figure 1](#ijms-15-09924-f001){ref-type="fig"}) \[[@B119-ijms-15-09924]\]. The interplay of oxidative stress and increased cytokine release and enhanced hepatocyte death are mainly responsible for this disease. In the course of NASH, several signaling pathways are activated which leads to cell death such as apoptosis, autophagy, necrosis, necroptosis and maybe pyroptosis \[[@B119-ijms-15-09924]\]. A main feature of hepatic steatosis is an unbalanced accumulation of triglycerides (TG). If the energy intake is higher than the usage, an increased storage of TG in the liver has to follow. Several proteins, such as fatty acid transport proteins (FATPs), fatty acid translocase (also known as CD36) and fatty acid binding proteins (FABPs) regulate the uptake of fatty acids into hepatocytes \[[@B122-ijms-15-09924]\]. A second promoter of steatohepatitis is an increased synthesis of fatty acids. Hyperinsulinemia is the main driver for de novo lipogenesis and is often an accompanying symptom of NAFLD. Lipogenesis is regulated by transcription factors which are initiated by insulin such as sterol regulatory element binding protein (SREBP)-1c \[[@B123-ijms-15-09924]\]. SREBP-1c controls transcription of other important proteins, which are associated with lipogenesis and TG synthesis, e.g., fatty acid synthase (FAS), stearoyl-CoA desaturase (SCD) 1 and acetyl-CoA carboxylase (ACC) \[[@B124-ijms-15-09924]\]. Next to insulin, LPS and TNFα as well as the endoplasmic reticulum stress response leads to activation of mature SREBP-1c in the liver. Despite contrary results in humans it seems that SREBP-1c promotes NAFLD progression, as increased expression levels of SREBP-1c and its target enzymes FAS and ACC were determined to be elevated in NAFLD patients \[[@B125-ijms-15-09924],[@B126-ijms-15-09924],[@B127-ijms-15-09924],[@B128-ijms-15-09924]\]. In a murine model of ob/ob mice suffering from obesity, insulin resistance, hyperinsulinemia and SREBP-1c levels were enhanced \[[@B129-ijms-15-09924],[@B130-ijms-15-09924]\]. Carbohydrate response element-binding protein (ChREBP) is another important transcription factor that controls lipogenesis. In contrast to SREBP-1c, ChREBP is activated by increased glucose levels and is involved in the metabolic change of carbohydrates to TG \[[@B131-ijms-15-09924]\]. Thereby it mediates initiation of transcription of different genes associated with lipogenesis (FAS and ACC), gluconeogenesis (glucose-6-phosphatase) and glycolysis \[liver pyruvate kinase (LPK)\] \[[@B132-ijms-15-09924]\]. ![Schematic illustration of non-alcoholic fatty liver disease (NAFLD)/non-alcoholic steatohepatitis (NASH) progression. A lifestyle with a balanced diet, moderate alcohol consumption and regular exercise supports the maintenance of an intact liver function. The intake of high energy food, less exercise and/or enhanced drinking of alcohol (light blue rectangle and arrow) leads to increased amounts of triglycerides (TG) in the liver which are stored as lipid droplets (yellow dots). Over time the degree of fatty change might reach an extent of hepatic steatosis. Concomitant effects like insulin resistance, obesity, type 2 diabetes mellitus (T2DM) and hyperlipidemia have a promoting effect (red arrows). It is under debate if the development of steatosis and NAFLD/NASH is the result of a "two" or "multiple hit event'' (red rectangle). Adipocytokines, endoplasmic reticulum (ER) stress, endotoxins from gut bacteria, reactive oxygen species (ROS) and epigenetic modifications (dark blue rectangle and arrows) are potential mediators promoting the development from hepatic steatosis to steatohepatitis and liver fibrosis and/or cirrhosis. Under physiological conditions fatty acid metabolism is tightly controlled by antagonists such as adiponectin and antioxidants (green rectangle and arrows) which are involved in insulin tolerance, glucose consumption and fatty acid oxidation as well as in elimination of ROS.](ijms-15-09924-g001){#ijms-15-09924-f001} Analysis of different ChREBP knockout mouse models showed less liver steatosis, decreased amounts of SCD1, ACC, FAS and LPK. Moreover, reduction of obesity, insulin resistance and a positive progression of the metabolic syndrome were obtained as well \[[@B133-ijms-15-09924],[@B134-ijms-15-09924]\]. The third transcription factor playing a role in NAFLD is the Liver X Receptor (LXR). LXR in association with retinoid X receptor (RXR) controls the expression of genes involved in cholesterol metabolism, SREBP-1c and ChREBP as well as their target genes ACC and FAS \[[@B135-ijms-15-09924],[@B136-ijms-15-09924],[@B137-ijms-15-09924],[@B138-ijms-15-09924],[@B139-ijms-15-09924],[@B140-ijms-15-09924],[@B141-ijms-15-09924]\]. Metabolites such as glucose and glucose-6-phosphate are activators of LXR. The role of AMPK in NAFLD is not clearly defined. Generally, AMPK takes part in the energy homeostasis of the cell regarding katabolic processes as fatty acid oxidation and glycolysis and anabolic processes like fatty acid and amino acid synthesis \[[@B142-ijms-15-09924]\]. AMPK has an inhibitory effect on ACC, SREBP-1c and ChREBP77. Experiments with ethanol-fed mice, ethanol treated rat hepatoma cell lines and hepatocytes showed a reduction in AMPK activity correlating with a rising activation of ACC, SREBP-1c and hepatic steatosis progression \[[@B143-ijms-15-09924],[@B144-ijms-15-09924]\]. Further experiments with the AMPK activators metformin or 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR) in rat hepatoma cells and hepatocytes activate AMPK, thereby inhibiting the blocking function of ethanol on ACC and SREBP-1c \[[@B143-ijms-15-09924]\]. Nevertheless, AMPK does not seem to have an important regulatory function in NAFLD progression. Mice fed a high sucrose diet revealed hepatic steatosis, but no differences in the activation level of AMPK were noticed \[[@B145-ijms-15-09924]\]. Moreover, transgenic mice over-expressing constitutively active AMPKα1 in the liver revealed reduced weight gain, white fat mass and blood glucose, which was linked to reduced expression levels of SREBP-1c and its target genes \[[@B146-ijms-15-09924]\]. In a study with rats fed with a HFD for 16 weeks, a reduction of 60% in the activation of AMPK was found. Administration of the AMPK activator Resveratrol for 10 weeks in these animals resulted in less hepatic steatosis and a positive effect on insulin resistance \[[@B147-ijms-15-09924]\]. A further mediator of NAFLD might be a member of the peroxisome proliferator activated receptor family, PPARγ. PPARγ is a nuclear receptor that builds a complex with RXR to promote transcription of its target genes. PPARγ is expressed in a smaller amount in the liver and is predominantly detected in adipocytes. PPARγ is involved in the regulation of adipocyte differentiation, fatty acid uptake and glucose metabolism \[[@B77-ijms-15-09924]\]. Ethanol has the ability to prevent PPARγ expression \[[@B148-ijms-15-09924]\]. PPARγ promotes SREBP-1c activation leading to enhanced expression of lipoprotein lipase in adipocytes \[[@B149-ijms-15-09924],[@B150-ijms-15-09924]\]. Mice with liver-specific PPARγ knockout were reported to rarely develop hepatic steatosis accompanied by hyperlipidemia, triglyceride clearance, and muscle insulin resistance \[[@B151-ijms-15-09924]\]. A study with human patients with a dominant-negative PPARγ mutation revealed that these patients suffered from the MetS and NAFLD \[[@B152-ijms-15-09924]\]. Moreover keratin 8 and 18 was shown to contribute to the development of steatohepatitis \[[@B153-ijms-15-09924],[@B154-ijms-15-09924]\]. 8. Conclusions ============== Despite intensive research on steatosis and steatohepatitis, fully satisfying treatment options are currently not available. The best prevention is a life style with balanced nutrition, avoiding excessive alcohol consumption and including sufficient exercise. NALFD/NASH patients suffer from obesity, T2DM and hyperlipidemia promoting the development of fatty liver and increased inflammation resulting in a high susceptibility to develop liver fibrosis, cirrhosis and HCC. Estimation of certain diagnostic parameters revealed a high variability within different human cohorts, which makes it challenging to find a common strategy in diagnosis and treatment. Work on different murine models led to the identification of different key players, which have a promoting or protecting function in AFLD/ASH or NAFLD/NASH. To which extent they might be translated to the human situation is still questionable. Our expanding knowledge of epigenetics will help us to get a clearer understanding of the molecular mechanisms behind AFLD/ASH and NAFLD/NASH in the near future. Certain studies could show that miRNAs and SNPs might have an important regulatory function in NAFLD/NASH progression. ASH and NASH show almost identical morphological features. As the clinical presentation is not characteristic, liver biopsy still represents the diagnostic gold standard. Both diseases reveal morphological hallmarks such as steatosis, hepatocellular injury with hepatocytic ballooning, apoptosis, necrosis, inflammation and fibrosis. Differentiation of the two diseases is only possible by confirmation or exclusion of an alcohol abuse. In the case of NASH, obesity is the most constantly associated cause. Ballooning of hepatocytes is linked to a disturbance of the normal keratin-intermediate filament cytoskeleton, being found in both conditions. In some studies on the pathogenic mechanisms of ASH and NASH, altered keratin 8 and 18 could be established as major players for Mallory-Denk body (MDB) formation. In case of an impaired proteolytic activity, protein aggregation can occur \[[@B155-ijms-15-09924]\]. In addition to abstinence from excessive alcohol intake in the case of ASH, experimental models and patients with steatohepatitis have been demonstrated to strongly depend on cholesterol and sphingolipids, in particular on ceramide during the progression from steatosis to steatohepatitis and insulin resistance. Cholesterol accumulation and its transfer to mitochondria render fatty liver more vulnerable to following "hits", comprising pro-inflammatory cytokines, in a signaling cascade with ceramide generation by acidic sphingomyelinase (ASMase). Therefore, cholesterol and/or ASMase might serve as novel therapeutic targets in ASH and NASH \[[@B156-ijms-15-09924]\]. Support of the multiple-hit hypothesis of NASH is increasing, and the development of new diagnostic techniques is emerging. Therefore the approach to find a list of diagnostic, prognostic as well as predictive parameters, including serum markers, genetic variants, imaging techniques and lifestyle habits will be expanded and promising approaches are on the horizon. K.B.: design of manuscript, manuscript writing, drawing of the figure, critical review and revision of the manuscript; T.H.: design of manuscript, manuscript writing, critical review and revision of the manuscript; and J.H.: design of manuscript, manuscript writing, critical review and revision of the manuscript, funding. The authors declare no conflict of interest.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-materials-13-01361} =============== Roller compacted concrete (RCC) material has been widely applied in the construction of infrastructure such as hydraulic structures since the early 1970s due to its advantages of cost-effectiveness and rapid construction \[[@B1-materials-13-01361],[@B2-materials-13-01361],[@B3-materials-13-01361]\]. Recently, a series of 100--300 m world-class high RCC dams have been under construction or built. During construction, the RCC mixture is paved and spread using a bulldozer and then compacted to a layered structure by using a vibratory roller. RCC, as a special type of concrete material, has a different mixture from traditional concrete, for example, using less water and more fly ash to replace Portland cement \[[@B1-materials-13-01361]\]. During their service life, these structures may be subjected to different types of loads, such as blast and impact loads. Therefore, it is necessary to understand the dynamic mechanical properties of RCC materials as well as the dynamic performance of the RCC dams. Predicting the performance of RCC structures subjected to explosive and impact loads through full-scale tests is often beyond affordability. Numerical simulation has proven to be an effective method for understanding the dynamic performance of RCC dams subjected to blast and impact loads. In a numerical simulation, a sound constitutive model selected to represent the dynamic mechanical properties of RCC under a high strain rate is vital to ensure the validity of the numerical simulation \[[@B4-materials-13-01361]\]. Numerical analyses have shown that the dynamic constitutive relationship of the material and the compressive strength, ultimate deformation and elastic modulus can greatly influence the accurate modelling of the structural dynamic response \[[@B5-materials-13-01361],[@B6-materials-13-01361]\]. For describing the dynamic mechanical behaviour of concrete-like materials, several comprehensive constitutive models have been studied with the consideration of the effect of hydrostatic pressure, strain rate strengthening and strain softening, such as the Holmquist--Johnson--Cook (HJC) model \[[@B7-materials-13-01361]\], the Riedel--Hiermaier--Thoma (RHT) model \[[@B8-materials-13-01361]\], the Taylor--Chen--Kuszmaul (TCK) model \[[@B9-materials-13-01361]\], the continuous smooth cap (CSC) model \[[@B10-materials-13-01361]\] and the Karagozian & Case (K&C) model \[[@B11-materials-13-01361]\]. Among these, the K&C model, RHT model and HJC model are established based on the plasticity and damage theories in DYNA3D software \[[@B12-materials-13-01361]\], while the CSC model is based on visco-plastic and damage theories considering the coupled volumetric and shear behaviour \[[@B13-materials-13-01361]\]. The K&C model and RHT model use three yield surfaces in terms of the initial elastic yield surface, the failure surface and the residual surface to describe the strength property. All of these dynamic constitutive models refer to many parameters for describing the complicated mechanical behaviour, making it difficult to determine a suitable and effective model for RCC. The HJC model considers the strength yield surface, the damage accumulation and the strain rate effect, which contains fewer parameters to be determined, and the model can present the dynamic compressive behaviour of concrete well. Though the parameters of the original HJC model can be obtained from the literature \[[@B14-materials-13-01361]\], the automatically generated parameters cannot avoid errors when the original HJC model is used to evaluate the blast resistance of RCC structures subjected to underwater explosion because the material behaviour of RCC deviates from the regular behaviour of normal concrete. For example, the properties and stress--strain model are both different for RCC from normal concrete \[[@B15-materials-13-01361],[@B16-materials-13-01361]\]. The composition of the materials and the difference in the stress state of the structure would cause a significant distinction in the meso-stress field, deformation and failure forms \[[@B4-materials-13-01361]\]. For the brittleness of RCC, many factors may affect the validity and accuracy of the experimental results \[[@B5-materials-13-01361]\]. Appropriate parameters or a sound dynamic constitutive model for RCC are almost impossible to find in existing studies, and parameter calibrations of the HJC model for RCC materials are important based on the experimental observations and unique characteristics of RCC materials. To this end, sound constitutive models are still lacking that can be used to describe the dynamic properties of RCC materials and predict the dynamic behaviour of RCC structures. The present study aims to provide an appropriate constitutive relationship for RCC under a dynamic compressive state. In this paper, modifications are made for the parameters of the strength surfaces and the dynamic increase factor for compression. A modified HJC material model representing the dynamic properties of RCC under a high strain rate and its corresponding parameters are established and validated for the improvement of a numerical simulation of the dynamic response of RCC compared to the test data. Furthermore, sensitivity analyses of the parameters that control the yield surface are carried out to understand the influence of each individual parameter. 2. Experimental Observations for RCC Materials {#sec2-materials-13-01361} ============================================== The experimental observations before the modifications to the HJC model for RCC are briefly introduced in this section, combining the uniaxial compressive test results, the triaxial compressive test results and the dynamic compressive test results. 2.1. Uniaxial and Triaxial Compressive Results {#sec2dot1-materials-13-01361} ---------------------------------------------- Uniaxial and triaxial compressive data for RCC materials are crucial to define the failure surface in the modified HJC model for RCC materials. The RCC mixture in this study is made of the mortar matrix, aggregates and additive. The mortar matrix is a mixture of water, cement, sand, fly ash, a water reducing agent and an air-entraining agent. The water-cement ratio (W/C) is set to 0.50. The fly ash content and the sand ratio are 60% and 31% by weight, respectively. The details of the RCC mixture and the preparation process of the specimens are given in our previous studies \[[@B17-materials-13-01361],[@B18-materials-13-01361],[@B19-materials-13-01361],[@B20-materials-13-01361]\], and details of the RCC mixture are given in [Table 1](#materials-13-01361-t001){ref-type="table"}. Uniaxial and triaxial compression tests were performed at the State Key Laboratory of Hydraulic Engineering Simulation and Safety, Tianjin University. The uniaxial compressive tests on the ∅100--200 mm specimens were conducted using an electro-hydraulic servo-controlled loading test machine. The testing machine delivers a constant crosshead movement with a loading rate of 1.20 mm/min, corresponding to a quasi-static strain rate of 1 × 10^−4^/s. The 90-day uniaxial compressive strength $f_{c}^{\prime}$, Young's modulus and critical strain of the RCC were measured as 20.86 MPa, 2.68 GPa and 0.76%, respectively. The triaxial compressive tests were also conducted using the ∅100--200 mm specimens. The confining pressure of the cell can be up to 100 MPa, while the loading mode can be switched between displacement control and load control. The confining pressure $\sigma_{3}$ can be directly measured by a pressure transducer inside the intensifier. The axial load $F_{d}$ is applied by an axial actuator and measured by an in-vessel load cell. The actual axial stress $\sigma_{1}$ can be expressed by Equation (1):$$\sigma_{1} = F_{d}/A_{s} + \sigma_{3}$$ where $A_{s}$ is the cross-sectional area of the specimens. An axial extensometer and circumferential extensometer were employed to measure axial displacement and circumferential deformation, respectively. In this way, the total circumferential deformation along the whole perimeter of the cylinder was obtained, which was utilized to derive the average lateral strain $\varepsilon_{3}$. Under uniaxial compression, a displacement control mode was used to obtain the stress--strain curve, with a rate of 0.002 mm/s. The confining pressure was first applied to the target value in load control mode and then kept constant. After the target $\sigma_{3}$ was obtained, the control mode was switched to displacement control and deviatoric stress was applied until the fracture of the RCC specimens occurred, with a rate of 0.002 mm/s. A total of seven levels of confining pressure ($\sigma_{3}$) were employed to cover the response under both moderate and high confinement, 0, 5, 10, 15, 20, 25 and 30 MPa, corresponding to confinement ratios of approximately 0.00, 0.34, 0.67, 1.01, 1.34, 1.68 and 2.01, respectively. For each confinement level, at least three reasonable test results were obtained. The test results are shown in [Table 2](#materials-13-01361-t002){ref-type="table"}, including the peak additional axial stress ($\Delta\sigma_{p}$), the peak axial strain ($\varepsilon_{1p}$) and lateral strain ($\varepsilon_{3p}$). These results are from the average of the specimens. [Figure 1](#materials-13-01361-f001){ref-type="fig"} shows the axial deviatoric stress versus the strain curves for RCC samples under confining pressures. Equation (2) shows the relationship between the normalized peak axial strain ratio ($\varepsilon_{1p}/\varepsilon_{cp}$) and the confinement ratio ($\sigma_{3}/f_{c}^{\prime}$), in which $\varepsilon_{cp}$ denotes the peak axial strain of RCC under uniaxial compression. A linear relationship between $\varepsilon_{1p}/\varepsilon_{cp}$ and $\sigma_{3}/f_{c}^{\prime}$ with a correlation coefficient of 0.994 ($R^{2} = 0.994$) can be expressed as follows:$$\frac{\varepsilon_{1p}}{\varepsilon_{cp}} = 1 + 4.74\frac{\sigma_{3}}{f_{c}^{\prime}}$$ 2.2. Split Hopkinson Pressure Bar (SHPB) Experimental Results {#sec2dot2-materials-13-01361} ------------------------------------------------------------- The ∅50--25 mm specimens were prepared for dynamic compressive tests. The SHPB test system shows in [Figure 2](#materials-13-01361-f002){ref-type="fig"}. In the SHPB test, the speed of the custom-shaped punch and the input wave strain rate of the incident bar are controlled by adjusting the gas chamber pressure. Two assumptions needed to be satisfied in the SHPB experiment \[[@B21-materials-13-01361]\]: (1) The stress wave in the pressure bar is in a one-dimensional state. (2) The stress is uniform throughout the specimen. For the ∅50--25 mm specimens, the maximum size of the coarse aggregate is 10 mm to satisfy the requirement of sample size for SHPB test, the medium aggregate size is 5--10 mm, the small aggregate size is 5 mm, and the composition (%) by weight is 40:30:30. [Figure 3](#materials-13-01361-f003){ref-type="fig"} plots some typical stress--strain curves of RCC specimens subjected to various strain rates. The compressive strengths of the specimens are 38.15 MPa and 47.16 MPa under average strain rates of 91 s^−1^ and 136 s^−1^, which are 54% and 126% higher than the quasi-static uniaxial compressive strength of 20.86 MPa. The peak stress and initial elastic modulus increase with increasing strain rate, which means that RCC material is also strain rate sensitive, as it consists of strain rate effects associated with the changing damage modes and viscosity effects of confined air and water in the pores and exhibits structural effects owing to the lateral inertia confinement, which is similar to the dynamic behaviour of conventional concrete in the literature \[[@B22-materials-13-01361],[@B23-materials-13-01361]\]. However, the RCC material exhibits a more obvious plasticity plateau at a higher strain rate, which indicates that RCC material exhibits a dynamic strengthening effect and better ductility with increasing strain rate. 3. The Modified HJC Model {#sec3-materials-13-01361} ========================= The HJC model is often used for describing the characteristics of normal concrete, although there is a difference between normal concrete and RCC in preparation technology, considering that various indicators are close for these two kinds of concrete and their consistent use in engineering; the modified HJC model is established based on the original model \[[@B7-materials-13-01361]\], combining new experimental observations for RCC materials in [Section 2](#sec2-materials-13-01361){ref-type="sec"}. From the description above, the strength yield surface is determined by the triaxial compressive experimental data, and the strain rate effect should be considered, which can be obtained from the SHPB experiment of RCC materials. 3.1. The Original HJC Model {#sec3dot1-materials-13-01361} --------------------------- The original HJC model is introduced briefly before the parameter modifications. Holmquist et al. \[[@B24-materials-13-01361]\] proposed the HJC model, which considers the influence on yield stress by the damage of the material, strain rate effect and hydrostatic pressure and can effectively describe the compressive behaviour of concrete-like materials under medium and high strain rates. [Figure 4](#materials-13-01361-f004){ref-type="fig"} shows the equation of the yield surface, damage model, hydrostatic pressure and volumetric strain curve of concrete. ### 3.1.1. Yield Function {#sec3dot1dot1-materials-13-01361} The yield surface, shown in [Figure 4](#materials-13-01361-f004){ref-type="fig"}a, of the HJC model can be expressed as follows:$$\sigma* = \left\lbrack {A\left( {1 - D} \right) + BP^{N}} \right\rbrack\left( {1 + C\ln\overset{˙}{\varepsilon}} \right)$$ where $\sigma* = \sigma/f_{c}{}^{\prime}$ is the normalized equivalent stress, $\sigma$ is the realistic equivalent stress ($\sigma* \leq \sigma_{\max}$, $\sigma_{\max}$ is the maximum normalized strength of concrete), $f_{c}{}^{\prime}$ is the quasi-static uniaxial compressive strength, $D$ is the damage degree ($0 \leq D \leq 1.0$). $P* = p/f_{c}{}^{\prime}$ is the normalized hydrostatic pressure ($p$ is the realistic hydrostatic pressure, shown in [Figure 4](#materials-13-01361-f004){ref-type="fig"}c). $\overset{˙}{\varepsilon}* = \overset{˙}{\varepsilon}/{\overset{˙}{\varepsilon}}_{0}$ is the dimensionless strain rate ($\overset{˙}{\varepsilon}$ is the realistic strain rate, and ${\overset{˙}{\varepsilon}}_{0}$ is the reference strain rate). $A$, $B$, $N$ and $C$ are constants, where $A$ is the normalized cohesive strength, $B$ is the normalized pressure hardening, $N$ is the pressure hardening exponent and $C$ is the strain rate coefficient. The strain rate effect is included in Equation (3), and the determination of $C$ is important to describe the strain rate effect in the constitutive model \[[@B25-materials-13-01361]\]. This relationship will be fitted according to the SHPB experimental data of RCC materials. ### 3.1.2. Damage Evolution {#sec3dot1dot2-materials-13-01361} The damage is described by the equivalent plastic strain and plastic volumetric strain for the HJC model, and the damage evolution equation is $$D = {\sum\frac{\Delta\varepsilon_{p} + \Delta\mu_{p}}{\varepsilon_{p}^{f} + \mu_{p}^{f}}}$$ $$\varepsilon_{p}^{f} + \mu_{p}^{f} = D_{1}\left( {P* + T} \right)^{D_{2}}$$ where $\Delta\varepsilon_{p}$ and $\Delta\mu_{p}$ are the equivalent plastic strain increments and the plastic volumetric strain increments of the elements in a computational cycle. Here, $\varepsilon_{p}^{f}$ and $\mu_{p}^{f}$ are the equivalent plastic strain and plastic volumetric strain at an ordinary pressure, respectively. $T = T/f_{c}{}^{\prime}$ is the maximum tensile pressure that the material can bear, and $T$ is the tensile strength. $D_{1}$ and $D_{2}$ are the damage constants of the concrete material. When $P* = - T$, the concrete can no longer bear any plastic strain. The parameter $\varepsilon_{f\min}$ is the amount of plastic strain before fracture, such that $D_{1}\left( {P + T} \right)^{D_{2}} \geq \varepsilon_{f\min}$. [Figure 4](#materials-13-01361-f004){ref-type="fig"}b shows the damage model. ### 3.1.3. Equation of State {#sec3dot1dot3-materials-13-01361} The relationship between the hydrostatic pressure and the volumetric strain of the concrete is expressed by the segmental state equation shown in [Figure 4](#materials-13-01361-f004){ref-type="fig"}c. Stage OA* describes the linear elastic stage during which $p < p_{c}$, and both the loading section and unloading section can be expressed by:$$p = K_{e}\mu$$ where $K_{e}$ is the bulk modulus, $p_{c}$ and $\mu_{c}$ are the crushing pressure and crushing volumetric strain in the uniaxial compressive experiment, respectively. Stage AB is a plastic transition stage during which $p_{c} \leq p < p_{l}$, and plastic deformation occurs as the voids of concrete are compressed. Loading section:$$p = p_{l} + \frac{\left( {p_{l} - p_{c}} \right)\left( {\mu - \mu_{l}} \right)}{\mu_{l} - \mu_{c}}$$ where $p_{l}$ is the locking pressure, and $\mu_{l}$ is the locking volumetric strain. Unloading section:$$p - p_{\max} = \left\lbrack {\left( {1 - F} \right)K_{e} + FK_{l}} \right\rbrack\left( {\mu - \mu_{\max}} \right)$$ $$F = \frac{\mu_{\max} - \mu_{c}}{\mu_{l} - \mu_{c}}$$ where $K_{l}$ is the plastic volumetric modulus, $p_{\max}$ and $\mu_{\max}$ are the maximum volumetric pressure and volumetric strain before unloading in which the holes of the concrete are expelled and damage occurs accompanied by cracks. Stage BC is a fully compacted stage during which $p > p_{l}$, when the pressure reaches $p_{l}$, the holes are crushed completely. The relationship between $p$ and $\mu$ is represented by a cubic polynomial. Loading section:$$p = k_{1}\overline{\mu} + k_{2}{\overline{\mu}}^{2} + k_{3}{\overline{\mu}}^{3}$$ $$\overline{\mu} = \frac{\mu - \mu_{l}}{1 + \mu_{l}}$$ where $\overline{\mu}$ is the amended volumetric strain, and $k_{1}$, $k_{2}$ and $k_{3}$ are constants. Unloading section:$$p - p_{\max} = K_{l}\left( {\overline{\mu} - {\overline{\mu}}_{\max}} \right)$$ There is no hole in this stage, and the concrete is crushed completely. 3.2. Modified Strength Yield Surface {#sec3dot2-materials-13-01361} ------------------------------------ The parameters representing the strength yield surface for concrete-like materials were mostly taken as the original parameters revealed by Holmquist et al. \[[@B24-materials-13-01361]\], which was acquired from concrete with a compressive strength of 48 MPa, which is not reasonable for RCC material. The strength parameters of RCC in the HJC model contain $A$, $B$, $N$ and $\sigma_{\max}$, which can be determined from the triaxial compression experimental data based on plastic yield surface theory \[[@B26-materials-13-01361]\]. Based on the triaxial compressive experimental data derived from this study and data from previous studies \[[@B27-materials-13-01361],[@B28-materials-13-01361],[@B29-materials-13-01361]\], some parameters on the strength yield surface have been calibrated according to the strength theory of materials mechanical and curve fitting method. [Figure 5](#materials-13-01361-f005){ref-type="fig"} shows the yield surface ascertained by experimental data of RCC and normal concrete. It can be seen that $A$ is significantly smaller than the original value of 0.79, which is coincident with the conclusion given by Malvar et al. \[[@B11-materials-13-01361]\]. Finally, the equation of the strength yield surface can be expressed as shown below, ignoring the effects of damage and strain rate:$$\sigma = 0.23 + 1.84P^{0.88}$$ 3.3. Modified Strain Rate Effect {#sec3dot3-materials-13-01361} -------------------------------- Many factors affect the compressive strengths directly obtained from the RCC experimental tests, and attention has been paid by researchers \[[@B23-materials-13-01361],[@B30-materials-13-01361],[@B31-materials-13-01361]\], especially on the dynamic strength increment, which is a combination of the strain rate effect and the lateral inertial confinement. [Figure 6](#materials-13-01361-f006){ref-type="fig"} shows the relationship between the $DIF_{\overset{˙}{\varepsilon}}$ (compressive dynamic increase factor) and the average strain rate of normal concrete from SHPB experimental results in various publications \[[@B32-materials-13-01361],[@B33-materials-13-01361],[@B34-materials-13-01361]\], and the data points in [Figure 6](#materials-13-01361-f006){ref-type="fig"} represent the experimental result of RCC whose time change law is similar to the CEB (Comite Euro-International du Beton) model. The findings are also consistent with those of other researchers \[[@B1-materials-13-01361],[@B35-materials-13-01361]\]. The material strength enhancement using the DIF~TOT~, directly obtained from laboratory tests, will overestimate the true dynamic material strength. Therefore, the lateral inertia confinement effects need to be removed to derive the true dynamic material strength at high strain rates. In our previous study \[[@B1-materials-13-01361]\], numerical simulations were carried out to quantify the lateral inertia confinement effect on the concrete strength increment with respect to strain rate. The same approach is adopted here. After the lateral inertia confinement, the contribution to the strength increment is removed. The empirical formulas for the $DIF_{\overset{˙}{\varepsilon}}$ of RCC material in terms of compressive strength are derived below and will be adopted in the modified HJC model. $${DIF_{\overset{˙}{\varepsilon}} =}\left\{ \begin{array}{lr} {1.2619 + 0.0524\log\overset{˙}{\varepsilon}} & {\overset{˙}{\varepsilon} < 50s^{- 1}} \\ {3.4326 - 1.7692\log\overset{˙}{\varepsilon} + {0.3151\log}^{2}\overset{˙}{\varepsilon}} & {\overset{˙}{\varepsilon} \geq 50s^{- 1}} \\ \end{array} \right.$$ 3.4. Parameter Calibration for the Modified HJC Model {#sec3dot4-materials-13-01361} ----------------------------------------------------- The basic parameters in the HJC model for RCC, which are $\rho$ (density), $G$ (shear modulus) and $f_{c}{}^{\prime}$ (quasi-static compressive strength), can be determined by the quasi-static experiment. The elastic modulus $E = 2G(1 + \nu)$ and the bulk modulus $K_{e} = E/3(1 - 2\nu)$ can also be obtained, and $\nu$ is Poisson's ratio. The maximum tensile hydrostatic pressure $T$ can be determined by the ACI equation \[[@B36-materials-13-01361]\] below:$$T = 0.62\sqrt{f_{c}{}^{\prime}}$$ The equation of the state is mainly drawn from the Hugoniot curve, and the state of RCC would not reach the third stage, so the parameters $k_{1}$, $k_{2}$ and $k_{3}$ are taken as the original values. $p_{c}$ and $p_{l}$ can refer to the formula proposed by Holmquist et al. \[[@B24-materials-13-01361]\] as well. Other parameters are the crushing volumetric strain $\mu_{c} = p_{c}/K_{e}$, locking volumetric strain $\mu_{l} = \frac{\rho_{{}_{grain}}}{\rho_{0}} - 1$, which can be easily obtained, and $\rho_{grain}$ which is the compacted density of concrete. Holmquist et al. \[[@B24-materials-13-01361]\] assumed that the damage parameter is independent of the material strength, so $D_{1}$, $D_{2}$ and $\varepsilon_{f\min}$ are the same as the original parameters. For concrete-like material, crushing is the main failure form that can be described by fracture mechanics theory and applied to the simulation analyses. The stress--strain relationship and the cracks distributed in the concrete are considered, and different methods are employed to simulate the behaviour of the concrete under high strain rate loading \[[@B37-materials-13-01361],[@B38-materials-13-01361]\]. The proper failure criterion should be included for describing the failure and erosion of concrete, and many kinds of failure criteria can be used to determine the status of the material; the maximum principal strain failure criterion and shear strain failure criterion are the most suitable factors for describing the failure status of concrete-like materials \[[@B39-materials-13-01361]\]. Therefore, combining the specimen failure process and morphology shot by a high-speed camera, the maximum principal strain failure criterion is adopted, which means that when $\varepsilon_{1} \geq \varepsilon_{\max}$, the element is in failure and deleted from the model and loses its carrying capacity, in which $\varepsilon_{1}$ is the maximum principal stress, and $\varepsilon_{\max}$ is the maximum principal stress at failure \[[@B40-materials-13-01361]\]. [Table 3](#materials-13-01361-t003){ref-type="table"} lists the modified HJC parameters of RCC (C20). 4. Numerical Verification {#sec4-materials-13-01361} ========================= To evaluate the performance of the modified HJC model in predicting and reproducing the dynamic response of concrete structures under impact loading, the SHPB test of concrete is simulated using LS-DYNA finite element software, and the test results of the existing experimental tests on ∅50--25 mm concrete specimens with the SHPB test device are selected to validate the accuracy of the numerical simulation. 4.1. Numerical Model {#sec4dot1-materials-13-01361} -------------------- In the SHPB test simulation, the geometric model consists of four parts, i.e., the concrete specimen, the impact bar, the incident bar and the transmission bar are shown in [Figure 7](#materials-13-01361-f007){ref-type="fig"}. Considering the symmetry of the SHPB system, a quarter of the simulation specimen is modelled to significantly reduce the model size and computation time. The mesh size effect of shock wave peak pressure at different scaled distances were analysed and compared and a convergence test was carried out to determine the mesh size. The incident bar and transmission bar are both 2.0 m long, and they are dispersed to meshes of 0.005 m. The specimen is dispersed to meshes of 0.00125 m according to our previous work \[[@B41-materials-13-01361]\] using "explosive radius and the mesh size ratio" as a criterion for determining mesh size. The non-reflecting boundary condition is applied for the two sides of the model that are perpendicular to each other. The eroding surface is applied to describe the contacting characteristics between the pressure bars/specimen surfaces, and the physical crushing and fracture can be captured well \[[@B42-materials-13-01361]\]. The pressure bars, which are made of high strength steel springs, are assumed to be elastic with the main parameters as follows: the density is equal to 7800 kg·m^3^, the Young's modulus is equal to 200 GPa and the Poisson's ratio is 0.25. The RCC specimen is simulated by the modified HJC model with parameters determined in [Section 3](#sec3-materials-13-01361){ref-type="sec"}. Two monitoring units are set at the surface of the bar centre (1^\#^ is at the incident bar and 2^\#^ is at the transmission bar) to gather data, and the location of gage points 1^\#^ and 2^\#^ is suitable for acquiring the stress--strain relationship of the specimen. F1 and F2 are the gage points on the front surface of the specimen, M1 and M2 are the gage points on the middle surface of the specimen, and B1 and B2 are the gage points on the back surface of the specimen. 4.2. Stress--Strain Curve {#sec4dot2-materials-13-01361} ------------------------- The stress--strain relationship of the stress/strain monitoring elements in specimens and three-wave method results calculated by the simulation are compared in [Figure 8](#materials-13-01361-f008){ref-type="fig"}. Using a custom-shaped punch can achieve a loading with a single frequency half-sine wave to the pressure bar and according to the strain wave measured during the SHPB experiment, a half-sine wave with peak values of 75, 125, 175 and 225 MPa is applied severally on the front surface of the incident pressure bar and is the boundary input in the simulation. [Figure 8](#materials-13-01361-f008){ref-type="fig"}a--c shows the stress--strain curves under different strain rates, which correspond to the SHPB experiment. Taking the stress history and strain history of the three elements at the front surface, middle surface and back surface of the specimen (elements should avoid nearing the specimen edge) eliminates the time to acquire the stress--strain curves as the dashed lines show. The full line represents the stress--strain relationship obtained by the three-wave method. The results with three strain rates of, 48, 99, and 134 s^−1^, which correspond to the loading stress amplitudes of 75, 125 and 175 MPa, are shown. Before the maximum stress is achieved, the curve of the typical elements and the reconstructed curve agree well. After the damage begins to develop, there is a certain degree of discretisation. For the typical elements, the descent stage of the curve is not smooth because of the evolution of the crushing damage, and the higher the loading strain rate is, the more discrete the stress--strain curves are. The stress--strain curve can be divided into four sections: the linear elastic ascending stage (OA), the initial damage stage (AB), the severe damage evolution stage (BC) and the structural failure stage (CD). The stress--strain relationships of the reconstructed element and the typical elements are well fitted. Differences among the elements on the three surfaces appear when the specimen is at the initial damage stage, which is caused by the stress wave propagation and its reflection and refraction in the specimen, and this discrepancy deviates from the reconstructed curve. The SHPB experiment reflects the average behaviour of the specimen, which reduces the effect of the heterogeneity of the concrete material, so the existence of a low level of stress heterogeneity, as mentioned above, is allowed \[[@B43-materials-13-01361]\]. Before the stress reaches the peak value, damage occurs in some parts of the specimen. The discreteness of the stress--strain curves reflects the difference in the damage evolution caused by the transverse Poisson effect intensifying the multidimensional stress state, and the oscillation in the later stage is mainly brought out by the element deletion. The average stress value of the three elements is larger than the reconstructed stress, mainly because the typical element is located on the specimen centre, which is not severely damaged. Above all, using a 50 mm-diameter SHPB apparatus with a custom-shaped punch that can produce a half-sine strain wave and without consideration of the defect of the specimen, the stress uniformity state can be reached, and the valid range of the stress--strain curve is from the linear elastic ascending stage to the severe evolution of the damage stage. In the last stage, cracks pass through the whole specimen, and the specimen severely deforms and fragments. The stress--strain curve at this stage not only represents the behaviour of the concrete material but also reflects the structural characteristics. [Figure 9](#materials-13-01361-f009){ref-type="fig"} shows the stress--strain curves under different strain rates obtained from the SHPB experiment, whose average strain rates are 50, 91 and 136 s^−1^, respectively, corresponding to the amplitudes of the loading pressures of 75, 125 and 175 MPa in the simulation. The solid lines depict the result from the SHPB experiment, while the dashed lines represent the stress--strain relationship calculated by the three-wave method, and the mechanical behaviour shown by the curves is consistent with the analysis above. It can be seen that the strain rate strengthening effect increased with the enhancement of the loading strain rate, and the arc curvature during the severe evolution of the damage stage decreased with the strain rate increment, which indicates a greater adequacy of the damage evolution. Comparing the stress--strain relationships of the RCC specimen between the SHPB experimental result and the numerical simulation via loading with a half-sine pressure wave with different amplitudes on the incident bar, it can be proved that using the parameters of the modified HJC model determined above to describe the dynamic compressive mechanical behaviour of the RCC specimen is effective and feasible. 4.3. Failure Criterion and the Damage Mode {#sec4dot3-materials-13-01361} ------------------------------------------ The failure criterion is crucial to the failure mode and other responses of the structure subjected to the impact and explosion loadings. The influence of the failure criterion in the modified HJC model on the specimen failure mode is shown in [Figure 10](#materials-13-01361-f010){ref-type="fig"}, and the simulated result is compared with the experimental result. The erosion failure criterion of MAT_ADD_EROSION is introduced to control the failure of the elements, which mainly affects the unloading section. The experimental result shows that the specimen retains a core and that the failure develops around the surrounding area. Adopting the maximum principal strain failure criterion can better depict the specimen failure than the shear strain failure criterion, and $\varepsilon_{\max} = 0.05$ is recommended for RCC due to a better simulation results for the experimental failure pattern illustrated in [Figure 10](#materials-13-01361-f010){ref-type="fig"}. [Figure 11](#materials-13-01361-f011){ref-type="fig"} shows the failure process of the RCC specimen captured by a high-speed camera during the SHPB experiment. Cracking occurs in the specimen surface immediately after the stress wave is applied. Then, the specimen is damaged near the end of the transmission bar, which further spreads to the entire specimen and scatters. Finally, the specimen is destroyed and loses its load carrying capacity. [Figure 12](#materials-13-01361-f012){ref-type="fig"} shows the simulated failure processes of the RCC specimens under nearly the same average strain rate of 48 s^−1^. The failure develops slowly, and the damage degree is low, which first occurs at the edge of the front surface and back surface. Then, as the stress wave spreads along the z-axis direction, the damage grows continuously, and for the reflection and the refraction of the stress, an increasing number of elements fail until the cracks run through the whole specimen. For a relatively low loading strain rate, the specimen damage pattern exhibits the phenomenon of core retention. This damage process coincides with the pattern captured by the high-speed camera, proving that using the numerical simulation method to reproduce the indoor SHPB experiment for qualitative analysis is feasible. [Figure 13](#materials-13-01361-f013){ref-type="fig"} shows the accumulated amount of failed elements of the specimen, which is scattered to 8640 elements under three loading strain rates. It is shown that at the strain rates of 48, 99 and 134 s^−1^, element deletion starts at step 15, step 10 and step 155, respectively, and more elements fail at higher strain rates. The loading stress wave arrives at the specimen at approximately 350 $\mathsf{\mu}s$ (step 150), but the element failure is delayed, which is obvious at 800 $\mathsf{\mu}s$ (step 200). After 400 $\mathsf{\mu}s$, the number of failed elements grows sharply, especially at the strain rate of 99 and 134 s^1^, until most of the elements fail, and the specimen loses its load supporting capability. 5. Parameter Studies {#sec5-materials-13-01361} ==================== As is known, there are many parameters, including concrete's compressive and tensile strengths, modulus of elasticity and density, etc., that may influence on the numerical simulations. To illustrate the effect degree and differences between modified strength yield surface and the original one, four parameters ($A$, $B$, $N$ and $C$) in the yield function (Equation (3)) have been selected to make numerical simulations of RCC under impact loading. In this section, the sensitivity of these parameters to the modified HJC model is analysed by varying one parameter by ± 25 percent of the value in [Table 3](#materials-13-01361-t003){ref-type="table"} while the others are kept as the baseline. The loading stress wave has a peak value of 175 MPa, and the computational model is the same as that described in [Section 4.1](#sec4dot1-materials-13-01361){ref-type="sec"}. The influences of the normalized parameters on the compressive properties are shown in [Figure 14](#materials-13-01361-f014){ref-type="fig"}a,b comparing the reconstructed stress--strain curves. The peak value of stress increases with the increasing increment of $A$ while the peak strain remains unchanged; the rising step of the stress--strain curve becomes steeper, and the descending stage becomes slower. $A$ is the normalized cohesive strength in the yield equation, and a smaller $A$ corresponds to a reduction in the cohesive strength in terms of a decrease in the peak stress. When considering the damage, $A$ is the coefficient of the damage term that controls the proportion of the damage in the yield surface, and a smaller $A$ indicates a less apparent influence on the yield surface, so the descending stage of the stress--strain curve is slower. [Figure 14](#materials-13-01361-f014){ref-type="fig"}b shows that the variation in $B$ has no effect on the elastic stage of the stress--strain curve; however, the peak stress increases with the increasing increment of $B$, and the gradient of the yield increase stage becomes larger. The peak value of strain and the slope of the descending segment are insensitive to the value of $B$. $B$ controls the proportion of the pressure in the yield equation, and its variety only affects the peak stress, but the overall shape of the curve is invariant. Although $A$ and $B$ share similar controlling effects on the peak value of the stress, their physical meanings are distinct. [Figure 14](#materials-13-01361-f014){ref-type="fig"}c shows the stress--strain curves with different exponents of the normalized hydrostatic pressure, which has an obvious influence on the gradient of the yield ascending curve. With the increment of $N$, the yield ascending stage becomes slower, and the width of the curve becomes narrower, indicating a process of softening. A reasonable and accurate value of $N$ will determine the shape of the stress--strain curve. [Figure 14](#materials-13-01361-f014){ref-type="fig"}d presents the reconstructed stress--strain curves with different strain rate effects. The CEB model considers the dynamic increment, which contains the strain rate effect and the lateral confinement effect, so the stress--strain wave under this condition has the largest peak stress. Li et al. \[[@B32-materials-13-01361]\] eliminated the influence of the lateral confinement effect as well as the modified HJC model in the present study, resulting in lower strength increments, but the strain rate effect on RCC is more obvious than that on normal concrete. Adopting different formulas for the dynamic increment factor that governs the strain rate influence factor in the yield equation will change the magnitude of the peak stress but has a trivial influence on the overall shape of the stress--strain curve. 6. Conclusions {#sec6-materials-13-01361} ============== Based on the original HJC model and combining the SHPB experimental data and the triaxial compressive test results of RCC, a modified constitutive model is developed to describe the dynamic compressive properties of RCC. The main contributions and conclusions can be drawn as follows:(1)The triaxial compressive test and SHPB experimental results of RCC are provided as the basis of the modified HJC model. Then, the strength yield surface and the strain rate effect represented in the HJC model are modified by combining the triaxial compressive test results of concrete (including RCC and normal concrete for the limitation of the effective experimental data) and the SHPB experimental data of RCC, while the maximum principal strain failure criterion is adopted to describe the crushing and fracture properties of RCC. Parameters in the modified HJC model are obviously different from the values in the original model, and these corrections contribute to a more credible and accurate constitutive model to predict the dynamic mechanical characteristics of RCC.(2)The improved simulation method is established by the parameters obtained from the modified HJC model. The custom-shaped punch is replaced by loading a half-sine stress wave at the front surface of the incident bar, and contact of the eroding surface to the surface is employed, which shows great consistency with the SHPB experiment.(3)An improvement in the simulation with the parameters obtained from the modified HJC model is demonstrated compared to the test data. Stress--strain curves from the SHPB simulation results agree well with the experimental results, and the typical element at different parts of the specimen shows a similar stress--strain law to that of the results from the three-wave method.(4)The influence of the failure criterion, normalized parameters, pressure hardening exponent and strain rate effect on the dynamic compressive characteristics of RCC are analysed. The parameters referred to on the yield surface mainly affect the ascending section of the stress--strain curve, while the failure criterion basically controls the unloading section. Adopting the maximum principal strain failure criterion can better depict the specimen failure than the shear strain failure criterion, and $\varepsilon_{\max} = 0.05$ is recommended for RCC so that the failure process of the RCC specimen simulated via the modified HJC model is consistent with the results captured by the high-speed camera. The results of the dynamic mechanical behaviour of RCC specimens obtained from the SHPB experiment and simulation are reliable and effective and agree well with each other. Conceptualization, C.W. and S.Z.; methodology, R.S.; resources, X.C.; writing---original draft preparation, R.S. and P.W.; writing---review and editing, G.W.; funding acquisition, S.Z. All authors have read and agreed to the published version of the manuscript. This research was funded by open Foundation from the State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, grant number 2017SGG02, and the National Natural Science Foundation of China, grant number 51979188 & 51779168. The authors declare no conflict of interest. ![Axial deviatoric stress versus strain curves for RCC samples under confining pressures.](materials-13-01361-g001){#materials-13-01361-f001} ![Split Hopkinson pressure bar (SHPB) test system (a) schematic diagram of SHPB test system, (b) specimen.](materials-13-01361-g002){#materials-13-01361-f002} ![Typical stress--strain curves of RCC specimens.](materials-13-01361-g003){#materials-13-01361-f003} ###### Original Holmquist--Johnson--Cook (HJC) model. (a) Equation of the yield surface. (b) Damage model of concrete. (c) Hydrostatic pressure and volumetric strain curve of concrete. ![](materials-13-01361-g004a) ![](materials-13-01361-g004b) ![Determinations of the yield surface by triaxial compression data.](materials-13-01361-g005){#materials-13-01361-f005} ![The true compressive strength $DIF_{\overset{˙}{\varepsilon}}$ of RCC material.](materials-13-01361-g006){#materials-13-01361-f006} ![Numerical model of SHPB systems. (a) SHPB system and gage points at the bars (unit: cm). (b) Local model. (c) Specimen and gage points.](materials-13-01361-g007){#materials-13-01361-f007} ![Stress--strain curves with different strain rates. (a) stress--strain curves with strain rate of 48 s^−1^, (b) stress--strain curves with strain rate of 99 s^−1^, (c) stress--strain curves with strain rate of 134 s^−1^.](materials-13-01361-g008){#materials-13-01361-f008} ![Stress--strain curves of the experiment and simulation with solid lines depicting the experimental curves and dashed lines representing numerical curves.](materials-13-01361-g009){#materials-13-01361-f009} ![Comparative failure patterns of the specimens after being impacted with the test results.](materials-13-01361-g010){#materials-13-01361-f010} ![Failure process of a 50 mm diameter specimen. (a) about 308 μs, (b) about 498 μs, (c) s about 698 μs. (d) about 898 μs, (e) about 1098 μs, (f) about 1298 μs.](materials-13-01361-g011){#materials-13-01361-f011} ![RCC specimen failure process and the accumulated damage.](materials-13-01361-g012){#materials-13-01361-f012} ![Failure element trend correlation during the simulation.](materials-13-01361-g013){#materials-13-01361-f013} ![Effect of strength parameters on the stress--strain curve: (a) effect of parameter A on the stress--strain curve, (b) effect of parameter B on the stress--strain curve, (c) effect of parameter N on the stress--strain curve, (d) effect of parameter C on the stress--strain curve.](materials-13-01361-g014){#materials-13-01361-f014} materials-13-01361-t001_Table 1 ###### Mixture and material properties of Roller Compacted Concrete (RCC). ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- W/C Sand Ratio\ Fly Ash Content\ Water Reducing Agent (%) Air Entraining Agent (%) Material Consumption (kg/m^3^) Air Content\ Wet Density (kg/m^3^) (%) (%) (%) ------ ------------- ------------------ -------------------------- -------------------------- -------------------------------- -------------- ----------------------- ----- ------ ----- ------ 0.50 31 60 0.8 0.05 88 70 106 672 1507 3.8 2453 ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Note: W/C: water to cement ratio materials-13-01361-t002_Table 2 ###### Triaxial compression results of RCC. ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ $\mathbf{\mathbf{\sigma}_{3}\left( {MPa} \right)}$ $\mathbf{\mathbf{\sigma}_{3}/\mathbf{f}_{\mathbf{c}}^{\prime}}$ Peak Additional Axial Stress\ Peak Axial Strain\ Peak Lateral Strain\ $\mathbf{\Delta\mathbf{\sigma}_{\mathbf{p}}\left( {MPa} \right)}$ $\mathbf{\mathbf{\varepsilon}_{1\mathbf{p}}/\%}$ $\mathbf{\mathbf{\varepsilon}_{3\mathbf{p}}/\%}$ ---------------------------------------------------- ----------------------------------------------------------------- ------------------------------------------------------------------- -------------------------------------------------- -------------------------------------------------- 0 0 14.9 0.411 −0.237 5 0.34 26.1 0.625 −0.490 10 0.67 37.3 1.578 −0.760 15 1.01 41.9 2.270 −1.279 20 1.34 50.3 3.078 −2.228 25 1.68 54.1 3.630 −2.849 30 2.01 60.1 4.516 −3.478 ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ materials-13-01361-t003_Table 3 ###### HJC model parameters for RCC (C20). Fundamental Parameters Yield Surface Damage Parameters EOS Parameters ------------------------ --------------- ------------------- ---------------- ------------- ------ ----------- ---------- $\rho$ 2350 kg·m^−3^ $A$ 0.23 $D_{1}$ 0.04 $p_{c}$ 7.0 MPa $G$ 10.63 GPa $B$ 1.84 $D_{2}$ 1.00 $\mu_{c}$ 0.005 $f_{c}{}^{\prime}$ 20.68 MPa $N$ 0.88 $e_{f\min}$ 0.01 $p_{l}$ 800 MPa $T$ 2.80 MPa $\sigma_{\max}$ 7.00 $\mu_{l}$ 0.12 $k_{1}$ 85 GPa $k_{2}$ −171 GPa $k_{3}$ 208 GPa	{ "pile_set_name": "PubMed Central" }
![](hosplond73052-0022){#sp1 .372}	{ "pile_set_name": "PubMed Central" }
Table 1 has been corrected for improved readability. The publisher apologizes for the error. Please see the complete, corrected Table 1 here. 10.1371/journal.pone.0152449.t001 ![](pone.0152449.t001){#pone.0152449.t001g} ------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------- −36. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .+1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .+45 pRc/CMV AGGTCTATATAAGCAGAGCTCTCTGGCTAACTAGAGAACCCACTGCTTAACTGGCTTATCGAAAT natural CMV AGGTCTATATAAGCAGAGCTCGTTTAGTGAACCGTCAGATCGCCTGGAGACGCCATCCACGCTGTTTTGACCTCCATAGAA SCP2 AGGTCTATATAAGCAGAGCTCGTTTAGTGAACCGTCAGATCGCCTGGAGACGTCGAGCCGAGTGGTTGTGCCTCCATAGAA SCP3 AGGTCTATATAAGCAGAGCTCGTTTAGTGAACCGTCAG [TC]{.ul}CGCCTGGAGAC[C]{.ul}TCGAGCCGAGTGGT [C]{.ul} GTGCCTCCATAGAA . . . . .TATA-box. . . . . . . . . . . . . . . . . . . . . .Inr. . . . . . . . . . . . . . . . . .MTE. . . . .DPE. . . . . . . . . . . . . . ------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------	{ "pile_set_name": "PubMed Central" }
Arterial embolization may be used effectively for connective tissue tumours: to reduce bleeding during surgical excision, or for palliative treatment. Case one {#sec1-2042533313476685} ======== A 59-year-old woman (see [Table 1](#table1-2042533313476685){ref-type="table"}) presented with a very large mass in the right gluteal region which had grown rapidly over the past five months. The mass was painful and overlying skin was tightly stretched with areas of bruising; see [Table 1](#table1-2042533313476685){ref-type="table"} for imaging and histological details. The tumour was considered surgically inoperable owing to size, risk of bleeding and likelihood of recurrence. Table 1.Summary of clinical details for patients undergoing selective arterial catheterization and embolization.CaseSexAgeTumourIndication for embolizationAngiographic outcomeFunctional outcomeComplications1Female59High-grade sarcoma of right gluteus (25 × 20 × 18 cm) with six pulmonary metastasesPalliationMultiple feeding vesselsReduced tumour sizeNone2Male31Metastatic phaeochromocytoma of right distal humerus and right proximal tibiaFacilitation of surgerySignificantly reduced tumour vascularityUncomplicated surgeryNone3Male61Metastatic renal tumour of left distal femurFacilitation of surgeryAngiographic stasisUncomplicated surgeryNone The patient therefore underwent selective arterial catheterization and embolization (SACE) via a left femoral approach. Pre-embolization angiography revealed the mass was supplied by branches of the posterior division of the internal iliac and lateral circumflex femoral (LCF) arteries. The superior gluteal artery and its branches were embolized using polyvinyl alcohol (PVA) particles until a pruned tree appearance was obtained; two metallic coils were also deployed. Ascending branches of the LCF artery were then catheterized and embolized using PVA particles and metallic coils. Owing to multiple feeding vessels, embolization of all vessels at a single procedure was considered impossible. However, successful palliation of symptoms was achieved, and the patient died later of disease progression. Case two {#sec2-2042533313476685} ======== A 31-year-old man with multiple skeletal metastases from phaeochromocytoma presented with painful pathological fractures of the right humerus and tibia. Radiographic findings are given in [Figure 1](#fig1-2042533313476685){ref-type="fig"}. Figure 1.Radiograph showing metastatic tumour of the right distal humerus with an undisplaced pathological fracture through the large lytic area, and further lytic areas both proximal and distal to the lesion. In addition, a metastatic tumour was present at the right proximal tibia, with a pathological tibial plateau fracture of the right knee. The patient underwent pre-operative SACE (via a left femoral approach) in order to reduce the risk of bleeding at surgery. Pre-embolization angiography revealed highly vascular tumours. Embolization of the humeral tumour was carried out using coils, PVA particles and Gelfoam, leading to immediate reduction in vascular filling of around 30% ([Figure 2A,B](#fig2-2042533313476685){ref-type="fig"}). Similarly, embolization of the tibial tumour by Embozene spheres and coils immediately reduced vascular filling by around 60%. Figure 2.Hypervascular metastatic tumour of the right distal humerus with pathological fracture. (A) Angiographic image immediately before embolization. (B) Angiographic image following embolization with coils, PVA particles and Gelfoam, with immediate reduction in vascular filling. SACE enabled subsequent uncomplicated open reduction and internal fixation (ORIF) of the humeral fracture ([Figure 3](#fig3-2042533313476685){ref-type="fig"}), followed by radical excision of the tibial tumour together with bone grafts and ORIF ([Figure 4](#fig4-2042533313476685){ref-type="fig"}). The patient subsequently underwent radiotherapy to both tumours. Hence, SACE facilitated surgery with reduced risk of peri- or post-operative haemorrhage, and SACE itself was uncomplicated. The patient remains alive with good local tumour control at follow up of 10 months. Figure 3.Radiograph following open reduction and internal fixation of the right humeral fracture. Figure 4.Radiograph following radical excision of the right tibial tumour, bone graft insertion with open reduction and internal fixation. Case three {#sec3-2042533313476685} ========== A 61-year-old man presented with a painful hypervascular metastatic renal tumour in the left femur, with extensive bony erosion and risk of imminent fracture. SACE was performed, using PVA particles to branches of the superficial and deep femoral arteries, to reduce the risk of bleeding during surgery. This produced immediate angiographic stasis, and enabled uncomplicated intramedullary nail insertion. The patient recovered well from surgery with normal use of the left lower limb, but died 18 months later of disease progression. Discussion {#sec4-2042533313476685} ========== Arterial embolization has been reported in oncological practice for several purposes: to decrease tumour size before excisional surgery, to permit surgery by reducing peri-operative haemorrhage, to manage acute haemorrhage, and for palliative care and tumour reduction.^[@bibr1-2042533313476685],[@bibr2-2042533313476685]^ However, experience with arterial embolization in the management of tumours is still early,^[@bibr1-2042533313476685]^ and few reports exist in the literature for its use in the management of connective tissue tumours.^[@bibr1-2042533313476685],[@bibr3-2042533313476685][@bibr4-2042533313476685]--[@bibr5-2042533313476685]^ For sarcomas, as for other connective tissue tumours, substantial improvements have been made in reducing mortality and morbidity; this has been achieved through a multi-disciplinary approach combining treatment modalities and ensuring local control and effective chemotherapy.^[@bibr6-2042533313476685],[@bibr7-2042533313476685]^ Local tumour control refers to the 'physical elimination of all malignant clones and/or tumour stem cells' within the primary tumour and identified metastases.^[@bibr3-2042533313476685]^ For sarcomas in particular, the preferred method of achieving local control involves early surgery.^[@bibr8-2042533313476685]^ However, various factors may mean that surgery is not feasible, including the presence of multiple disease sites or the morbidity associated with the operation.^[@bibr3-2042533313476685]^ Arterial embolization is an important adjunctive treatment to facilitate surgery in certain situations. As demonstrated in this case series, pre-operative embolization may reduce the risk and severity of peri-operative and post-operative haemorrhage and therefore may reduce the threshold of safe surgical intervention. In patients where surgery is completely precluded, embolization may still provide useful clinical improvement and palliation, particularly for tumour reduction and cessation of bleeding or discharge. In this case series, immediate angiographic stasis was achieved by one procedure in one case, using PVA particles administered to branches of the gluteal or femoral arteries. In another case, tumour vascularity was significantly reduced by embolization using a combination of PVA particles, gelatine microspheres, Gelfoam and metallic coils. Angiographic success was associated with functional improvements in these two patients. In the final case, the presence of multiple feeding vessels meant that embolization of all vessels was not possible with a single procedure. There were no complications of embolization or surgery. Conclusion {#sec5-2042533313476685} ========== Arterial embolization may be used effectively in the management of both primary and metastatic tumours involving connective tissue. It has a definite role in palliative tumour reduction where surgery is inappropriate, and should also be considered before surgical excision to reduce the risk of severe bleeding in vascularized tumours (e.g. renal metastases, phaeochromocytomas and certain sarcomas). DECLARATIONS {#sec6-2042533313476685} ============ Competing interests {#sec7-2042533313476685} ------------------- None declared Funding {#sec8-2042533313476685} ------- Publication costs were funded by the Department of Orthopaedics, Manchester Royal Infirmary, Manchester, UK. No other funding was obtained for this study Ethical approval {#sec9-2042533313476685} ---------------- Written informed consent for publication was obtained from the patient Guarantor {#sec10-2042533313476685} --------- AP Contributorship {#sec11-2042533313476685} --------------- AP and OK had the idea for the paper. All authors assisted in data collection and analysis. OK wrote the paper. All authors helped revise the paper and approved the final version Provenance {#sec12-2042533313476685} ---------- This article was submitted by the authors and peer reviewed by James Skipworth	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-sensors-16-00769} =============== From the technological and societal point of view, the 2015--2025 decade has been identified as the "Wearable Era" \[[@B1-sensors-16-00769],[@B2-sensors-16-00769],[@B3-sensors-16-00769]\]. This decade is characterized by the implementation and large diffusion of new miniaturized and wearable products supporting innovative sensing and feedback functions and related services with potentially enormous impacts on our lives. Most of these systems collect data and information from our body and also support a basic data processing to provide the user with immediate feedback about his/her status and/or lifestyle. For this reason, they are often considered health-related devices. Wearable Health Systems (WHS) or Wearable Biomedical Systems (WBS) are integrated systems on body-worn platforms such as wrist-worn devices or even biomedical clothes, offering pervasive solutions for continuous health status monitoring though non-invasive biomedical, biochemical, and physical measurements \[[@B4-sensors-16-00769]\]. Thanks to their continuous miniaturization, they support a non-intrusive measurement of user parameters. The non-intrusive monitoring paradigm is when the user forgets the ongoing process. This means that, even if wearing these devices, he/she is not aware of the continuous monitoring, but he/she can retrieve data and related information whenever he/she wants, thanks to mobile technology. Several fields of application are under exploration and exploitation, and among them the main ones are medicine, life style monitoring, and sport analysis (including, in this sector, fitness). Currently, we are undertaking the realization and diffusion of the first generation of these systems. They are small devices that have usually one or limited sensors, and a plurality of devices to be worn is needed for a comprehensive approach ([Figure 1](#sensors-16-00769-f001){ref-type="fig"}a). Research and innovation is facing the evolution of these systems, where our garments represent the worn platform in which to embed sensors and processing units. This technological advance makes use of smart fabrics and yarns to empower our natural interface with the external world, i.e., our clothes, to provide new and high-level functions. This represents the 2.0 generation of WHS ([Figure 1](#sensors-16-00769-f001){ref-type="fig"}b). The application of WBS in medicine would offer the opportunity to monitor patients easily and over extensive periods of time, thus implementing the telemedicine definition: to offer affordable and interactive healthcare, anyplace, anytime to anyone. This will allow the implementation of the strategy of Western countries of offering high quality services with reduced healthcare costs, which is the ultimate requirement and the future of healthcare. Clinical applications are probably most critical due to their requirements in terms of signal quality and device certification. However, by matching these requirements, it is possible to simplify the systems for high level applications turning them into reliable solutions for lifestyle or sport monitoring functions \[[@B1-sensors-16-00769],[@B2-sensors-16-00769],[@B3-sensors-16-00769],[@B4-sensors-16-00769],[@B5-sensors-16-00769],[@B6-sensors-16-00769],[@B7-sensors-16-00769],[@B8-sensors-16-00769],[@B9-sensors-16-00769],[@B10-sensors-16-00769],[@B11-sensors-16-00769],[@B12-sensors-16-00769],[@B13-sensors-16-00769],[@B14-sensors-16-00769]\]. Designing and implementing wearable systems and specifically sensing garments might seem like a simple activity; however, a plurality of disciplines is needed to interact and integrate to develop the right solution. Several requirements at different levels have to be taken into account. Sometimes, even obvious elements can become critical in an integrated approach; for example, washability is an obvious requirement for a garment, but this can irreversibly deteriorate an integrated sensor. A simplified list of requirements to follow could be as follows: sensors and actuators, materials, communication (usually wireless), power supply and management, processing techniques (onboard/on external devices, off line/real-time), Graphical User Interfaces (GUIs), algorithms for signal processing, connectors, washability and the related stability of sensors, body positions and wearability (tasks, sensors, anthropometry, etc.), elasticity, and adherence to the supporting platform (garments or belts, an adhesive patch, or other). Here, the authors discuss the most significant aspects of designing a sensing wearable platform, specifically in relation to the implementation and integration of textile sensors or other sensors for physiological signal monitoring. The discussion is on two levels: the technology requirements, focusing in particular on the sensor design and materials, and the design specifications, where specific focus on anthropometry and garment features are presented. 2. Materials and Methods {#sec2-sensors-16-00769} ======================== When designing wearable systems, two main sets of specifications can be outlined: the technological features and the design requirements. The main issues related to technology are the identification of the sensing principle and the corresponding sensor implementation, the selection of the right materials and their properties, data storage, data processing and related algorithms, data transmission (bit rate, distance, etc.), and the amount of power to be supplied in accordance with the desired functioning time, processing, etc. These are very technical issues but strongly and mutually influence the design issues. In fact, the design factors include the physical features of the device such as shape, the form factor, the softness/rigidness of the case and related materials, and other design-related issues in relation to anthropometry and gender, GUIs, positions on the body and/or into the garments, wearability, elasticity, and adherence to the body fixing element or of the garments. 2.1. Technological Specifications of Wearable Systems {#sec2dot1-sensors-16-00769} ----------------------------------------------------- The main goal of a wearable system is to collect information about the user wearing it. These data could be of a different nature and level of importance (the highest level and the corresponding required quality is obviously needed for medical applications/services). Usually, non-invasive biosignals are measured and stored in the systems or transmitted to a body gateway. These biosignals are measurable and can be collected on the body surface (usually the skin) through specific sensors. The first requirement is to define the physical nature of the signal to be measured and the corresponding sensor \[[@B15-sensors-16-00769]\]. The main categories of signals are: bioelectric signals, such as Electro-Cardio-Gram (ECG), Electro-Myo-Gram (EMG), Electro-Encephalo-Gram (EEG), Electro-Oculo-Gram (EOG), Electro-Neuro-Gram (ENG), and others;thermal parameters, such as body temperature in specific points and its map on a specific portion of the body surface;mechanical measurements, such as the kinematics of body segments and the pressure at the human/object interface (e.g., the human back and a mattress);optical signals, such as the SpO~2~ assessment and other NIRS measurements;chemical analyses, such as the determination of the composition of the sweat and that of the inhaled/exhaled air. [Table 1](#sensors-16-00769-t001){ref-type="table"} describes the sensor to be used in the measurement of each bio-signal and its typology, while a simplified map of the sensing point to the body for each signal is reported in [Figure 2](#sensors-16-00769-f002){ref-type="fig"}. The basic sensors to measure bioelectric signals are electrodes. The standard solution is represented by adhesive, non-polarized, silver/silver chloride (Ag/AgCl) electrodes \[[@B15-sensors-16-00769],[@B16-sensors-16-00769]\], which are small discs with a stainless male snap button in their center and that are directly placed on the body in specific positions according to the measurement setup and requirements. The snap is reached by a wire that, on the other side, is connected to the measuring device. In the vision of the 2.0 wearable generation, these electrodes are transformed into textile electrodes (sometimes also called textrodes). They can be implemented by knitting or weaving or by integrating the electrically conductive yarns with other textile techniques (e.g., embroidering) to create sensing areas in specific positions of the clothes. The other essential requirement is that these sensing zones have to be in contact with the skin to assure the proper signal measurement. They are a category of polarized electrodes. In general, non-polarized electrodes are better than polarized electrodes in terms of their rejection of motion artifacts and their response to defibrillation currents. Different solutions are available to implement textrodes, specifically in terms of textile yarns with electrical properties; these yarns can be classified into two main categories \[[@B4-sensors-16-00769],[@B8-sensors-16-00769],[@B16-sensors-16-00769]\]: (a)metal yarns, i.e., yarns containing conductive fibers such as stainless steel, copper, and silver mixed with natural or synthetic fibers;(b)yarns containing electro-conductive fibers such as polymeric or carbon-coated threads. [Table 2](#sensors-16-00769-t002){ref-type="table"} summarizes the materials used in textrodes with pros and cons. The adoption of a specific material (and the yarn-related properties) is also a unique choice for the textile solution in terms of production technique and final characteristics of the sensing garment. In fact, the selection of the fibers determines elasticity, washability, and the stability of the sensor properties to the chemical agents used for cleaning. An improper stability could affect sensors' efficacy and reliability, and this factor is also crucial in the definition of the proper duration of the product and its life cycle. The choice of the material is important in order to define the proper measurement set-up and system. For example, a capacitance sensor can be built using dielectric materials ([Figure 3](#sensors-16-00769-f003){ref-type="fig"}). In this case, capacitance increases with applied force. Another type of pressure sensor is a piezo-resistive element. Instead of using a dielectric material, it is possible to insert a piezo-resistive substance between the two textile plates/contacts. Therefore, force produces a deformation of the material and its structure or conductive content. Usually, resistance decreases with the increase in the stimulation signal ([Figure 4](#sensors-16-00769-f004){ref-type="fig"}), but this effect depends on the used material or its construction (e.g., yarn directions with respect to the elongation direction that can produce a positive or negative coefficient into the resistance-strain function that corresponds to the gauge factor). Therefore, this sensor measures the applied force on its surface (which can be also used for measuring the interface pressure distribution on a wider area in a configuration of a mat of sensors) and requires a very different circuitry in accordance to the sensing principle \[[@B17-sensors-16-00769],[@B18-sensors-16-00769],[@B19-sensors-16-00769]\]. Applications of these textile pressure sensors are multifold. One example is provided by an innovative system for sleep monitoring that integrates some of them into a sensorized blanket; another product is aimed at monitoring walking or running performance and/or foot disorders by putting these textile capacitive sensors into insoles or socks. This system is capable of measuring---nay, of estimating, because they are related only to some specific points of the foot---the pressure points under the foot during gait or exercise. Other solutions or concepts that integrate the pressure sensors into monitoring systems are dedicated to posture analysis, for example, by being used in sensorized seat cushions (or simply their coverings) \[[@B17-sensors-16-00769],[@B18-sensors-16-00769],[@B19-sensors-16-00769],[@B20-sensors-16-00769],[@B21-sensors-16-00769]\]. The material that the sensors are made of also influences the property of a skin-sensor contact and interface \[[@B8-sensors-16-00769],[@B16-sensors-16-00769],[@B17-sensors-16-00769],[@B18-sensors-16-00769],[@B19-sensors-16-00769]\]. From the operational point of view, proper sensing requires good conductance and a stable skin-sensor contact. Standard silver/silver chloride adhesive electrodes use Ag/AgCl gel to establish a good conductive contact with the skin, consequently improving the output signal. However, many patients experience a certain discomfort since this gel may cause skin irritation and softening, especially in the case of prolonged application. These inconveniences impose restrictions on the use of this kind of electrode for long-term monitoring. For textrodes, some preparations are required to have good signal quality: skin cleaning, scrubbing the external layer of dead skin cells, and removing the presence of hair, especially in male subjects. Textrodes are usually dry sensors, but they are capable of receiving, and sometimes retaining, perspiration and sweat to increase local contact conductivity. In fact, in textile electrodes, the presence of sweat is demonstrated to be useful because it is a conductive solution, thus improving the conductivity of the electrical contact with the skin and the signal quality revealed through the electrode \[[@B8-sensors-16-00769],[@B16-sensors-16-00769],[@B17-sensors-16-00769],[@B22-sensors-16-00769]\]. Sometimes, before the perspiration effect arises, in order to have an immediate good sensing capability, wetting the textile electrode before wearing the smart garment could immediately increase the quality of the skin-electrode contact. In some garments, this effect could be obtained with the insertion of small sponges placed under the sensing surface (in a small pocket) or filling the space underneath the sensor with hydrophilic yarns; this process also produces a convex sensor surface that helps to increase the skin contact pressure. In this way, the sensor can stay in a more stable position on the skin itself \[[@B19-sensors-16-00769]\]. Thanks to these design solutions and the effect obtained (usually after a few minutes) as a result of the local sweat (perspiration), electro-gel in textrodes was usually found to be unnecessary. Indeed, textrodes also represent a possible solution to the limitations related to the other possible effects such as the production of cutaneous events due to the material in contact with the skin (tolerance and irritation, and/or reactions or modifications due to sweat and perspiration). One of the main advantages of textrodes is that they do not irritate the skin, even in prolonged application. In any case, these situations have to be properly investigated before the sensor implementation or application to subjects. In fact, not all the materials are suitable for textrodes implementation: carbon black has been shown to produce similar skin effects, while silver or stainless steel textile electrodes are safer \[[@B23-sensors-16-00769]\]. Instead, the main disadvantage of textile electrodes (in particular with respect to the standard adhesive electrodes) is the unstable skin/electrode contact due to the non-adhesive surface contact. This means that a dedicated design of the garment and/or higher performances on the electronic circuit are required. For example, the analogue front-end of the sensing circuitry needs to be redesigned and optimized for better signal amplification and noise reduction \[[@B9-sensors-16-00769],[@B22-sensors-16-00769]\]. In relation to the signal quality, in a previous work, we analyzed the comparison between standard electrodes and textrodes \[[@B24-sensors-16-00769]\]. We analyzed the reliability of textile electrodes compared to standard adhesive electrodes using a two-channel portable electro-cardiograph in two conditions: (a) in vitro, by using a ECG signal generator; and (b) in vivo, registering the signals of 10 subjects during a set of standardized activities of daily living, such as walking, standing and sitting down, and stair climbing. The experimental setup consisted of a very close set of four electrodes paired in twos (1 was textile and 1 was a traditional adhesive Ag/AgCl electrode) to verify that the quality of the signal and the computed parameter in time and frequency were comparable. Results demonstrated a very good reliability. Textile structures containing electro-conductive material could be used as textile strain gauges when they are carefully engineered and characterized. A strain gauge exhibits a percentage change in resistance that is directly proportional to the applied strain ([Figure 4](#sensors-16-00769-f004){ref-type="fig"}). In this case, they are suitable for measuring mechanically induced elongations (from motion of body structures). With proper calibration, they can provide satisfactory accuracy in the measurements of breathing parameters or joint movements ([Figure 5](#sensors-16-00769-f005){ref-type="fig"}). The following equations describe the process: where S is the strain, L is the actual length, L~0~ is the length at rest, R is the actual electrical resistance of the textile elastic element, R~0~ is the electrical resistance of the textile elastic element at rest, and S~g~ is the gauge factor, i.e., the coefficient to convert the strain to dR/R~0~. As a possible solution for sensor characterization, we also defined an experimental setup to identify the relationship between the electrical resistance and the sensor elongation. The experimental setup usually pairs a displacement measurement system with a four-wire resistance measurement circuit, as described in detail in \[[@B25-sensors-16-00769]\]. In that case, a set of two realizations of textile strain gauges was tested obtaining the calibration elongation-resistance. The curve increased exponentially according to the different yarn distribution with respect to the example in [Figure 4](#sensors-16-00769-f004){ref-type="fig"}a. Thus, through characterization, the elongation textile sensor can be used for reliable signal measurements. [Figure 5](#sensors-16-00769-f005){ref-type="fig"} displays the experience of using these textile strain gauge sensors to measure breathing parameters. We have applied two textile strain gauges along the circumference of the human trunk---one at the hearth level, and one on correspondence of the abdominal compartment. Simultaneously with the textile sensors monitoring, the pneumo-tachograph (i.e., the gold standard technique) recorded breathing parameters. As shown in [Figure 5](#sensors-16-00769-f005){ref-type="fig"}b, the correspondence of the two signals---the temporal parameters (breathing rate, inspiration time, expiration time)---perfectly match, while the computation of the current volume suffers from some underestimation (average difference −6%) in the peaks. This demonstrates how a simple textile structure in the garment could be very useful and reliable for monitoring this vital sign. By integrating this technique with textrodes for ECG measurement and adding an Inertial Movement Unit (IMU) (for body movement measurement, i.e., actigraphy) into the electronic frontend for signal acquisition, a simple but almost complete compact system for basic monitoring was obtained. This solution can be easily applied to medical and sport monitoring applications \[[@B11-sensors-16-00769],[@B12-sensors-16-00769],[@B13-sensors-16-00769],[@B23-sensors-16-00769]\]. 2.2. Design Specifications of Wearable Systems {#sec2dot2-sensors-16-00769} ---------------------------------------------- A basic requirement for the measurement process is a good, stable contact between the body and the sensor for the proper sensing time. Ideally, in the case of monitoring, it should be in continuous close contact with the body and without movements of the sensors over the skin. This last event produces artifacts definitely corrupting the signal pattern and quality. For this reason, anthropometric factors and gender considerations, that are also related to body sizes and shapes and have implications on user preferences, should be among the first ones to be addressed in terms of the design of wearable monitoring systems. Anthropometry is the measurement of physical characteristics of the human body and their differences in relation to age, gender, lifestyle, and ethnicity as inter-subject variability factors, or even time. For example, stature varies in the same subject with time and according to different phenomena \[[@B26-sensors-16-00769]\]. In a short time, the height of people usually follows a circadian cycle: subjects are about 0.7%--1% higher in the morning than in the evening. Over long periods, for example, stature varies according to growth or the ageing processes: there is a large increase in stature during growth in childhood, and a moderate decrease in the late adult age. Also, living in external environments could influence anthropometry (and therefore garment design): e.g., in microgravity (that is an extreme condition) when a stature increase of about 4% is reached and kept after a few days of permanence in space, as well as the fluid shift produces variations of volumes of different body districts. Gender issues are quite obvious: men and women have different requirements not only for their different dimensions, but for shape, fit, and physiology, as well as for the daily task these users undertake. The functional aspect is crucial for identifying the best places for sensors for different applications. This is crucial, for example, in sports activities, where the contribution of the upper limbs and their synergic, antagonistic, or synchronized movement could produce different signal artifacts. For example, the most used electrode configuration for heart electrical activity monitoring is the two-sensor setup: it requires two electrodes placed on the right and on the left side of the chest, approximately in correspondence to the 10th rib ([Figure 6](#sensors-16-00769-f006){ref-type="fig"}). When the subject performs wide and/or strong upper limb movements, EMG artifacts are produced by pectoral muscles. The EMG signal is superimposed to the electrical activity of the heart and this could significantly affect the quality and the reliability of the ECG signal recording. Thus, an equivalent but transversal lead (i.e., bipolar electrode set-up) should be used to minimize the incidence of these artifacts ([Figure 7](#sensors-16-00769-f007){ref-type="fig"}); therefore, a specific design of the sensing garment is required. A study and design of the proper elasticity of the garment at different levels of the trunk, and in specific body areas where the sensors are present, is needed. The specific and modulated elasticity in the different parts of the garment is needed to assure good sensor-skin contact and to minimize their movements over the skin during the recording \[[@B23-sensors-16-00769]\]. This phenomenon, also known as a skin motion artifact in biomechanical analysis with passive markers and opto-electronic motion capture systems, applies to wearable monitoring as well, and its avoidance or minimization is crucial for a good signal quality for most of the measuring time. Again, this possible critical issue can be solved by keeping a close and stable contact of the sensors with the skin. This requirement is fundamental also in relation to the small value of the measurand, e.g., bioelectric signals having a peak-to-peak amplitude ranging from 0.5 mV up to 2 mV. An example of this is reported in [Figure 7](#sensors-16-00769-f007){ref-type="fig"}, which shows an ECG signal collected through textrodes. The signals are collected with the same positioning of the textile electrodes, but the only variable factor is the resting position of the upper limbs ([Figure 7](#sensors-16-00769-f007){ref-type="fig"}a). This is compared to the same setup, but with movement in the upper limbs in the horizontal plane at shoulder level ([Figure 7](#sensors-16-00769-f007){ref-type="fig"}b). In this second condition, it is clearly shown that the ECG pattern is completely corrupted by the artifacts. The sensorized T-shirt has high elasticity, and its size was greater than the one worn by the subject. All these factors, which may seem very simple to address, are equally crucial with respect to the technological factors in wearable applications. For this reason, the study and the design of the garment is crucial for its elasticity and body fitting, in particular for obtaining a good adherence to the garment in the region of the cloth and of the body where the sensors are placed. Modulated elasticity is obtained by proper choice of elastic yarns and/or by garment design. Comfort and wearability of the garments itself are not to be forgotten at the same time. In fact, wearability is crucial for weak users (e.g., elderly, patient under rehabilitation, etc.) for autonomous operations and repeatable sensor inter-session repositioning to achieve a reliable monitoring. Finally, but with the same importance and priority, aesthetical requirements should address user preferences that are crucial for the acceptance and usability of the systems (garments, devices, etc.). This can determine the final success of the wearable monitoring approach. Design and aesthetics include the analysis of several factors including the user's age, the target application, possible competitors, and fashion trends. The personal preferences are also particularly important because the sensing garments are usually the underwear, and there is a requirement that the sensors be in close contact with the skin. Thus, color, material (cotton vs. technical fabrics), fitting, and the choice of substituting or integrating it with one's own underwear (this specifically applies to elderly women that often prefer a sensorized belt to be worn in parallel and in addition to their traditional bra, despite the availability of a new integrated solution) are fundamental aspects to be considered in the system design. Gemperle et al. \[[@B27-sensors-16-00769]\] analyzed the wearability requirements for hardware systems that have been formalized in a 13-point guideline here reported: (1) placement (where on the body it should go); (2) form language (defining the shape); (3) human movement (consider the dynamic structure); (4) proxemics (human perception of space); (5) sizing (for body size diversity); (6) attachment (fixing forms to the body); (7) containment (considering what's inside the form); (8) weight (as its spread across the human body); (9) accessibility (physical access to the forms); (10) sensory interaction (for passive or active input); (11) thermal (issues of heat next to the body); (12) aesthetics (perceptual appropriateness); and (13) long-term use (effects on the body and mind). 3. Discussion {#sec3-sensors-16-00769} ============= In the technology-based medicine era, the innovation and development of new technologies represent a winning strategy to create solutions for more accurate, personalized, and continuous healthcare services \[[@B1-sensors-16-00769],[@B2-sensors-16-00769],[@B3-sensors-16-00769],[@B4-sensors-16-00769]\]. Personal Health Systems (PHS) were introduced in the late 1990s to support the generation of innovative healthcare services (e.g. remote ambulatory monitoring, home care, etc.) for a personalized medicine. PHS empower and put the individual citizen/patient at the center of the healthcare process. They empower citizens/patients because they allow them to have more responsibility in managing their own health thanks to the amount of data and information they can provide directly to the patients themselves and/or to their related care providers. The main benefits of PHS for citizens and health authorities are a significant improvement in the quality of care for the individuals themselves because continuous non-intrusive monitoring allows for a light process, often delivered at home, and, secondly, the reduction of healthcare costs thanks to these efficient and affordable (both in use and in economics) technologies \[[@B1-sensors-16-00769],[@B2-sensors-16-00769],[@B3-sensors-16-00769],[@B4-sensors-16-00769],[@B17-sensors-16-00769],[@B18-sensors-16-00769],[@B28-sensors-16-00769],[@B29-sensors-16-00769]\]. WBS are a specific category of PHS. They are integrated systems that can be embedded into a wearable platform (sometimes and optimally in the sense of clothing) or a net of devices attachable to the human body for continuous monitoring of biomedical, biochemical, and/or physical parameters. In this way, it is possible to achieve early detection of anomalies. This patient awareness can produce an increased feeling of safety and confidence, which is perceived as an increase in quality of life. In addition, this kind of ambulatory monitoring can allow patients to engage in normal activities of daily life, rather than staying at home or close to specialized medical services. Thus, WBS represent an extraordinary opportunity to support the provision of a remote primary and secondary prevention, to obtain early diagnosis and management of several diseases (in particular, cardiovascular and/or respiratory diseases, metabolic pathologies, and the assessments of physical rehabilitation treatments), and to support elderly and disabled people. WBS can be used to measure a plurality of signals from the human body; the main and most frequently used ones are heart rate (or even ECG tracks), electro-encephalography, respiration, blood gases saturation, and body movements. They represent a good balance between system complexity and user compliance for a light or general monitoring of the health status through the following main physiological parameters. A good wearable device for monitoring could include 1 or 2 ECG leads, breathing parameters (respiratory rate but also in/expiration times and volumes), and should accommodate human activity and posture (through one or more three-axial accelerometers or IMUs). From the results that we have presented and discussed in this paper, in designing smart sensing garments and WHS, we can identify three main requirements: aesthetics, function, and technology. They do not stand alone; they are strongly interdependent. When facing a new project, the basic questions are the task or pathology to be used, and from this first aspect, one must identify the signal of interest and the corresponding sensors. Then, the technological choices and requirements are to be defined accordingly. At the same time, from the identification of the task or pathology, the aesthetical requirements for wearability and elasticity descend. These also influence the definition of shapes, body position, and size of the WHS in strong correlation with the technological choices. This decision tree is shown in [Figure 8](#sensors-16-00769-f008){ref-type="fig"}. The most frequently used sensors for bioelectric signals (e.g., ECG, EMG, and EEG) are the surface electrodes, whose textile version can be integrated into garments to replace the standard adhesive silver/silver chloride (Ag/AgCl) electrodes, which are not suitable for long-term continuous monitoring because of skin irritation problems that are shown in a relevant number of subjects. Usually, the last issue to be considered when designing WHS is the connection between the electronic device and the textile component (the garment or the supporting element). The essential requirement is the presence of a constant, stable, and rigid link between the two parts (specifically for IMUs attached to the body to monitor movements). This will ensure the possibility of obtaining a good signal quality during its recording. If the number of contacts is low (2 or 3 maximum) so that standard pins can only produce a weak link, a specific design of the device including a connecting snap on its case can be achieved. In this way, a connection through stainless steel nickel free snap buttons could be provided to solve this issue. However, if the number of required contacts is higher (five or more), the snaps solution becomes difficult to adopt because of the low acceptance expressed by users; therefore, a specifically designed or dedicated connector, possibly integrating flexible PCB or flexible/rigid PCB technology, could be the optimal choice to deal with this complexity in matching components. In fact, even if this is an expensive solution, it still represents the best way to optimize the interface between textile and hardware components without compromising the flexibility and easy wearability of the overall system. As a result of the previous discussion, a basic methodological approach can be proposed. The process of designing a wearable monitoring system can be described as the balance of three elements: function, technology, and aesthetics. They are strongly interconnected and interdependent. The design path and related choices can start from each point; however, the definition of the system's specification should follow the identification of different components. They are summarized and represented in [Figure 8](#sensors-16-00769-f008){ref-type="fig"}. The function requirements or specifications are related to monitoring needs; the basic requirements are: -human parameters to be monitored and related signals for the application or pathology, identifying the main signal and computed parameters;-typology and numbers of sensors required for the monitoring;-a monitoring paradigm---continuous signal transmission or delayed data transfer;-real-time data processing and analysis or off-line operations, secondary parameter computation and storage on-board or remote solutions;-direct patient participation or other caregivers operations, and, indirectly but obviously, who the subjects to be monitored are and their features (age, sex, body build, etc.), which are also pieces of information very relevant for aesthetical specifications. Secondary function requirements are related to the production and maintenance of the sensors and the devices. Moving towards human parameters, the identification of the signals and their related functions drives the choice of the sensors and the preferable position on the body; at the same time, it gives the essential requirements for hardware and software design or selection. In the case of the design of the technological solution, this will also match the needs in terms of shape, dimensions, and position on the body. User Interfaces (UIs) are the physical interfaces between the device and the user (buttons, switches, fixing supports, etc.) or the sensors (connectors) and provide input to garment design and system usability. If standard or commercial devices or systems are to be adopted, the technological features only apply to garment design for connection with the embedded sensors or to the building of the proper supports for the device itself (pockets or rigid element for fixing the system to the body in respect of the maximization of the non-intrusive monitoring approach). The garment design relates to the selection of materials (type of yarns and related production techniques) as well as design and customization of the sensing textile, also in consideration of wearability and un-wearability. Motor disorders or age can influence the freedom of movement or specific needs for these tasks, thus determining the acceptance of the solution. The system usability is linked to the wearability but also implies other features such as GUI and maintenance. Elasticity and adherence to the body and minimization of the sensors displacement over the skin is crucial, both from the technological point of view and for the ergonomics of the garment. In designing the garment, other secondary but relevant issues to be considered are: -integration of textile wiring through conductive yarns connecting the sensor to the measuring device;-sweat and thermal management to assure a good quality and then garment acceptance to the users;-specific requirements for washing, ironing, and sterilization or disinfection should be considered if applicable in special applications (e.g., monitoring in an Intensive Care Unit or in the case of the presence of exposed or open wounds). This design paradigm is just an example. The design path could follow different directions following the different inter-relations described in [Figure 8](#sensors-16-00769-f008){ref-type="fig"}. The important conclusion is to face all the requirements and to make the proper decisions. A similar approach could also be imagined in the case of wearable devices not including a textile component, but a simple device worn on a specific body part (e.g., as a bracelet, necklace, earring, shoe insole, etc.). 4. Conclusions {#sec4-sensors-16-00769} ============== In this paper we have presented some issues related to the design of WBS systems with specific reference to requirements related to technology and ergonomic or human factors. However, this method is not a rule. It is a proposal deriving from our experiences in designing smart sensing garment both medical applications (in monitoring mother and fetus in pregnancy, in preterm babies or newborn in the first two hours of life, in cardiologic adults, in elderly at home or in hospital during rehabilitation exercises) and in sport activities (adolescents in their physical education at school or monitoring their lifestyle during the whole day, adults while running, playing soccer and skyrace) \[[@B6-sensors-16-00769],[@B7-sensors-16-00769],[@B8-sensors-16-00769],[@B10-sensors-16-00769],[@B11-sensors-16-00769],[@B12-sensors-16-00769],[@B13-sensors-16-00769],[@B14-sensors-16-00769],[@B17-sensors-16-00769],[@B18-sensors-16-00769],[@B20-sensors-16-00769],[@B21-sensors-16-00769],[@B22-sensors-16-00769],[@B23-sensors-16-00769],[@B28-sensors-16-00769],[@B29-sensors-16-00769]\]. In this process, co-design activities are proposed and carried out for better matching user preferences and aesthetics requirements that significantly vary in relation to age and human anthropometry. Wearable systems are rapidly spreading internationally because they are easy-to-use. They use our natural interface with respect to the external world to integrate sensing or actuation capabilities. This fact is crucial because it allows for the actual implementation of the non-intrusive monitoring paradigm, i.e., a process that does not affect user behavior and is completely transparent to him/her during his/her own daily activities. Only this methodological approach to be pursued by wearable systems can lead to the innovative wearable 2.0 generation systems. We expect that this development and diffusion will continue in the next 5 to 10 years and then will become stable because the new generations of garments will directly include these features. Lifestyle monitoring and light medical monitoring are becoming essential for the new trends in healthcare and wellness programs. However, long and hard work is still necessary to exploit the sector and also work through legal/privacy issues for services or data management that have to be properly regulated. Sport is another very important and huge market: performance and training analysis together with injury prevention or optimization for the functional recovery after it, are activities supported in a very promising way by wearables. At the same time, we cannot forget that a new frontier is arising: the "insideable" systems. Electronics miniaturization and new biocompatible materials are opening this innovative era. However, this will be another story. This work has been partially funded by the European Commission: FP7-ICT-2013.5.1---Grant Agreement n° 610727. The Authors would like to thank all the partners of the PEGASO consortium for their proactive collaboration in the project. G.A. was in charge of the original idea of the paper: he revised the related literature and started proposing, discussing and reviewing the technological and design issues of wearable systems. He contributed in writing the first draft of the paper and he revised the final version. C.E.S. contributed in proposing, discussing and reviewing the design issues of wearable systems. He contributed in writing the related parts of the paper and he revised the final version. P.P. contributed in proposing, discussing and reviewing the technological issues of wearable systems. He contributed in writing the related parts of the paper and he revised the final version. The authors declare no conflict of interest. ECG Electro-Cardio-Gram EEG Electro-Encephalo-Gram EMG Electro-Myo-Gram ENG Electro-Neuro-Gram EOG Electro-Oculo-Gram GUI Graphical User Interface IMU Inertial Movement Unit NIRS Near Infra-Red Spectroscopy PCB Printed Circuit Board PHS Personal Health Systems POF Plastic Optical Fiber SpO 2 Peripheral Capillary Oxygen Saturation UI User Interface WBS Wearable Biomedical Systems WHS Wearable Health Systems ![The comparison of the two generations of Wearable Health Systems (WHS). (a) Wearables 1.0: the garments and other accessories are the supporting platform of a set of devices for monitoring human functions; (b) Wearables 2.0: sensors and related electronics components and integrated into the garments (adapted from \[[@B5-sensors-16-00769]\]).](sensors-16-00769-g001){#sensors-16-00769-f001} ![A simplified map of the bio-signals to be measured from the human body through wearables systems and the corresponding sensing point. ECG: Electro-Cardio-Gram, EMG: Electro-Myo-Gram, EEG: Electro-Encephalo-Gram.](sensors-16-00769-g002){#sensors-16-00769-f002} ![A model of capacitance sensor implemented with two layers of conductive fabric and a spacer that could be a 3D textile.](sensors-16-00769-g003){#sensors-16-00769-f003} ![A model of a textile strain gauge and the related description. (a) The modification of the yarn configuration inside the fabric while stretching it; a wider superficial contact area is obtained among the different conductive yarns, thus resulting in a decrease in the electrical resistance. (b) The typical graphical representation of the mathematical equation that expresses the relationship between elongation and electrical resistance of the textile strain gauge.](sensors-16-00769-g004){#sensors-16-00769-f004} ![An example of an application of textile strain gauges in the monitoring of respiratory acts and related parameters. (a) A sensorized T-shirt for monitoring respiration through textile strain gauges at thorax and abdominal level; (b) The comparison of the signals from the textile sensor and the pneumo-tachograph (gold standard technique); the blue line shows the gold standard data measured during normal breathing at resting in sitting posture, and the red line shows the textile strain gauge signal in the same condition. A \~5% air volume overestimation at peaks is shown.](sensors-16-00769-g005){#sensors-16-00769-f005} ![The electrode setup for measuring heart electrical activity. (a) The standard sensor positions for Electro-Cardio-Gram (ECG) measurements in the reduced five-lead configuration; (b) Transversal ECG sensors setup to avoid pectoralis muscles artifacts (C7 = in correspondence with the 7th cervical vertebra, XP = in correspondence of the xiphoid process).](sensors-16-00769-g006){#sensors-16-00769-f006} ![The effect of movement artifacts on bioelectrical signal recordings: an example of the different signal quality with the "standard" ECG lead implemented by two sensors on the left and right costal chest and in correspondence with the 10th rib (RL and LL in [Figure 6](#sensors-16-00769-f006){ref-type="fig"}). (a) 4 s of the typical ECG pattern recorded during quiet activity in sitting posture (typing). (b) A corrupted 4-s ECG signal with no recognizable peaks during upper limb movements (frontal abduction/adduction) in the same posture.](sensors-16-00769-g007){#sensors-16-00769-f007} ![A methodology and the corresponding decision tree for designing smart sensing garments.](sensors-16-00769-g008){#sensors-16-00769-f008} sensors-16-00769-t001_Table 1 ###### The basic signals and the derived parameters to be measured by wearable sensing systems, and the identification of the sensor to be used and its possible typologies. ECG: Electro-Cardio-Gram, EMG: Electro-Myo-Gram, EEG: Electro-Encephalo-Gram. No. Signal ^1^ Parameter (s) Sensor Typology ----- -------------------- -------------------------------------------- ----------------------------------------------- ----------------------------------------------------------------- 1 ECG Electrical heart activity, Heart Rate electrode Adhesive, textile, plate 2 EEG Electrical Brain activity electrode Plate, textile prototypes 3 EMG Electrical muscle activity electrode Adhesive, textile prototypes 4 Respiration Breathing rate, Volumes, respiratory times Strain gauge; Electrode for impedance measure Hardware probe, adhesive, textile sensor 5 Blood gas SpO~2~, CO~2~, Heart rate LED/optical Hardware probe, POF for signal transmission 6 Blood pressure Systolic/diastolic values, Heart rate Cuff Hardware System 7 Interface pressure Contact pressure Piezoresistive; capacitive Piezoresistive ink, capacitive sensor both electric and textile 8 Resistance GSR, Body impedance Electrodes for Impedance measure Hardware System 9 Temperature Temperature Piezoresistive Hardware probe (thermistor) ^1^ Note: Movement is measured with miniaturized hardware system applied to each body district. sensors-16-00769-t002_Table 2 ###### Materials for textile electrodes and related properties. Material Merits Demerits ---------------------------------- ----------------------------------------------------------- ------------------------------------------------------------------------- Conductive rubber High conductivity, easy to shape, cheap Poor flexibility and permeability to air and liquid Silver-coated polymer foam High conductivity, easy to shape, flexible, antibacterial Poor washability and permeability to air and liquid, possible oxidation Metal-coated or sputtered fabric High conductivity, fabric material Poor washability, possible oxidation Woven metal fabric Controlled conductivity, fabric material Difficult to handle, skin irritation, low elasticity Woven conductive polymer fabric Fabric material, elasticity Low conductivity Carbon yarn High mechanical resistance, high thermal insulation Average conductivity, skin irritation, low elasticity Stainless steel yarn High conductivity, no skin interaction Low elasticity, high weight	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-jcm-07-00342} =============== Considerable debate is ongoing regarding whether multiple sclerosis (MS) is initiated in the periphery, via activation of immune cells that subsequently enter the central nervous system (CNS) and cause damage, or within the CNS via primary myelin/oligodendrocyte damage. The latter could result from mutations in molecules encoding key components of myelin or oligodendrocytes. Whereas genome-wide association studies have not suggested significant linkages to any myelin or oligodendrocyte components in MS, they have not ruled out the possibility that mutations in genes encoding these components might occur at low frequencies in MS patients. Point mutations in PLP1---the X chromosome gene encoding the most abundant myelin protein, myelin proteolipid protein (PLP)---occur at a low frequency in the population, and typically result in development of Pelizaeus-Merzbacher disease (PMD) or spastic paraplegia 2 (SPG2) in men, whereas female carriers are usually symptom-free \[[@B1-jcm-07-00342]\]. However, several reports have been published of individuals with disease who meet the diagnostic criteria for MS and who carry PLP1 point mutations \[[@B2-jcm-07-00342],[@B3-jcm-07-00342]\]. A recent study showed that mice expressing some of these PLP1 point mutations develop spontaneous myelin and axonal damage that is dependent on the presence of B and T lymphocytes, suggesting that PLP1 point mutations could induce an immune-mediated disease indistinguishable from MS \[[@B4-jcm-07-00342]\]. The above data support the idea that damage to cells within the CNS could potentially be the initiating event in some MS cases. The exact mechanism by which these mutations could cause disease is still unknown; however, in animal models of PMD, PLP1 mutations led to apoptosis of oligodendrocytes \[[@B5-jcm-07-00342],[@B6-jcm-07-00342]\]. Similarly, cells transfected with PLP1 sequences that contain known PMD mutations also undergo apoptosis due to accumulation of PLP in the endoplasmic reticulum (ER) and the induction of the unfolded protein response (UPR) \[[@B7-jcm-07-00342],[@B8-jcm-07-00342],[@B9-jcm-07-00342],[@B10-jcm-07-00342]\]. The UPR is a cellular mechanism responsible for ensuring that incorrectly formed proteins are not expressed by cells; a clear marker of UPR activation is the translocation of the protein CCAAT/enhancer binding protein homologous protein (CHOP) from the cytoplasm of the cell to the nucleus \[[@B11-jcm-07-00342],[@B12-jcm-07-00342]\]. In cells consistently transcribing misfolded proteins at levels difficult to overcome, the UPR initiates apoptosis in an effort to remove the defective cell \[[@B13-jcm-07-00342]\]. It has been suggested that, at least in some MS patients, the initial pathological damage in the CNS is apoptosis of oligodendrocytes \[[@B14-jcm-07-00342]\]. The finding of PLP1 mutations in patients with MS-like disease therefore raises the question of whether PLP1 mutations may induce oligodendrocyte apoptosis and be a cause of the onset of disease in some patients. Of the two patients thus far described with PLP1 mutations and MS, one was a 10-year-old male with relapsing-remitting MS (RR-MS) \[[@B2-jcm-07-00342]\], and the other was a female with primary progressive MS (PP-MS) \[[@B3-jcm-07-00342]\]. In male patients who only have one copy of the X chromosome, it is likely that any signs of disease caused by PLP1 mutations would occur at a young age. In women, X chromosomes containing mutations are usually inactivated, but the inactivated X chromosome can change (referred to as skewing), particularly as women age \[[@B15-jcm-07-00342],[@B16-jcm-07-00342]\]. Therefore, it is more likely that PLP1 mutations could be a potential cause of MS symptoms in female patients who have skewed X inactivation, or who develop disease at a later age. This is of potential interest particularly for female patients with PP-MS, who typically develop disease at a later age than women who develop RR-MS \[[@B17-jcm-07-00342]\]. In the current study, we sequenced the protein-coding exons of PLP1 in 22 female PP-MS patients who developed MS after the age of 40. One patient was found to carry a novel PLP1 mutation. We then assessed in vitro evidence of cell damage (altered trafficking of PLP, and activation of CHOP and the UPR) in cells transfected with plasmids containing the mutant PLP1 sequences occurring in the patient we identified and in the two other reported MS-related PLP1 mutation patients, in order to determine whether the mutations found in the patients were potentially pathogenic. 2. Experimental Methods {#sec2-jcm-07-00342} ======================= 2.1. Patients and Controls {#sec2dot1-jcm-07-00342} -------------------------- Blood samples (5 mL) were collected from 42 healthy women and from 22 female patients with PP-MS after informed consent was obtained. Demographic details of patients and controls are shown in [Table 1](#jcm-07-00342-t001){ref-type="table"}. This study was approved by the Human Ethics Committees of the Royal Brisbane and Women's Hospital (HREC/11/QRBW/28) and the University of Queensland, Australia (20110000711). 2.2. DNA Extraction and PLP1 Sequencing {#sec2dot2-jcm-07-00342} --------------------------------------- DNA was extracted using a NucleoSpin^®^ Blood XL DNA extraction kit (Macherey-Nagel, Düren, Germany), according to the manufacturer's protocol. Primers (Life Technologies, Mulgrave Australia) to amplify exons 2 to 7 of human PLP1 were designed from the human PLP1 mRNA sequence using the online tool National Center for Biotechnology Information (NCBI) Primer-Blast (<http://www.ncbi.nlm.nih.gov/tools/primer-blast>). DNA was sequenced at the Australian Genome Research Facility (AGRF), the University of Queensland, Brisbane, Australia. Sequences were analysed using Sequence Scanner v1.0 software (Life Technologies, Carlsbad, CA, USA). 2.3. PLP1-Containing Vectors {#sec2dot3-jcm-07-00342} ---------------------------- Plasmids containing wild type (wt) or mutant PLP1 sequences encoding the 7 exons of PLP were constructed by the Protein Expression Facility (University of Queensland, Brisbane, Australia). Gene inserts were first cloned into the pOPIN-(n)-mCherry vector, in order to couple the inserts with the mCherry tag gene sequence to allow subsequent identification of transfected cells. Finally, the gene inserts were inserted into the pcDNA™ 3.3-TOPO TA (Life Technologies, Carlsbad, CA, USA) vector. DNA vector constructs were amplified and plasmid DNA was extracted using a MaxiPrep Plasmid DNA extraction kit (Qiagen, Hilden, Germany), as per the manufacturer's instructions. To confirm that the sequences of the gene inserts were correct, the plasmid DNA preparations were sequenced at the Australian Genome Research Facility Ltd. (Brisbane, Australia). 2.4. Transfection of Cos-7 Cells {#sec2dot4-jcm-07-00342} -------------------------------- Transfection via electroporation was performed on dissociated Cos-7 cells using the Nucleofection System (Lonza, Basel, Switzerland) together with the Amaxa II transfection device (Lonza, Basel, Switzerland), following the manufacturer's recommended protocol. The transfected cells were plated in 16-well chamber slides (Thermo Fisher, Melbourne, Australia) at a concentration of 10^4^ cells/well. Each slide included non-transfected cells and pulse-only cells as controls. 2.5. Immunocytochemical Analysis of Transfected Cells {#sec2dot5-jcm-07-00342} ----------------------------------------------------- Cells were fixed in 4% paraformaldehyde (Merck, Darmstadt, Germany) for 15 min at room temperature, and permeabilized for 15 min with 0.05% saponin (Sigma, St Louis, MO, USA) and 0.5% bovine serum albumin (BSA; Sigma, St Louis, MO, USA) in dH~2~O (saponin-BSA). Primary antibody (rabbit anti-PLP antibody; 1:2000; Thermo Fisher, Melbourne, Australia) or mouse anti-CHOP antibody (1:200; Abcam, Cambridge, UK) in saponin-BSA was then added for 2 h at room temperature. After 3 washes, cells were labelled with the appropriate fluorescein isothiocyanate (FITC)-conjugated secondary antibody (donkey anti-rabbit immunoglobulin (Ig), 1:300, (Santa Cruz Biotechnology, Dallas, TX, USA) or goat anti-mouse Ig (1:500; Abcam, Cambridge, UK)) in saponin-BSA solution for 2 h at room temperature in the dark. After 3 washes, the cells were then labelled with 4′,6-diamidino-2-phenylindole (DAPI) (Sigma; 1:20,000 dilution in 1.5M NaCl) for 15 min at room temperature in the dark, washed 3 times in phosphate-buffered saline (PBS), and coverslips were mounted. To assess protein trafficking out of the ER, cells were stained for 30 min at room temperature with ER-ID Green Assay Kit (Enzo Life Sciences, Farmingdale, NY, USA) per the manufacturer's recommended protocol. Using this protocol, the ER fluoresced green. Cells were photographed using an Eclipse 50i fluorescent microscope (Nikon Instruments, Tokyo, Japan) with an Olympus DP70 camera (Tokyo, Japan). Photomicrographs of multiple fields were taken using blue (to detect nuclei), red (to detect mCherry), and green (to detect FITC) filters. The photos were merged using Photoshop 7.0 (Adobe Systems, San Jose, CA, USA). Images were analyzed using Image J software. 2.6. Statistical Analyses {#sec2dot6-jcm-07-00342} ------------------------- Two-way analysis of variance (ANOVA) with Dunnett's multiple comparison test was used to compare transfection efficiency, percentages of cells trafficking PLP out of the ER, and intensity of CHOP expression. 2.7. HLA Binding Predictions {#sec2dot7-jcm-07-00342} ---------------------------- The human leukocyte antigen (HLA) class I and class II binding predictions were produced on November 14, 2016 using the Immune Epitope Database (IEDB---[www.iedb.org](www.iedb.org)) analysis resource tools \[[@B19-jcm-07-00342],[@B20-jcm-07-00342],[@B21-jcm-07-00342]\]. 3. Results {#sec3-jcm-07-00342} ========== 3.1. Sequencing of PLP1 in Human Subjects {#sec3dot1-jcm-07-00342} ------------------------------------------- There is little information available regarding mutations or polymorphisms in PLP1 in healthy individuals, aside from relatives of patients with PMD. Therefore, exons 2 to 7 of PLP1 were sequenced in genomic DNA from a group of 42 healthy women and 22 women with PP-MS. Exon 1 was not sequenced, as it only encodes a single amino acid (methionine), which is not incorporated into the mature protein. The different genotypes of the single nucleotide polymorphism GAT \> GAC (encoding the silent mutation D202D), which were previously reported in PLP1, occurred at approximately the same frequency in healthy control and MS patients ([Figure 1](#jcm-07-00342-f001){ref-type="fig"}a and [Table 2](#jcm-07-00342-t002){ref-type="table"}). No other mutations were observed in the healthy control group. However, one PP-MS patient had a mutation in exon 2 (CTG \> GTG), which would result in a coding change from leucine to valine at amino acid 30 (L30V) ([Figure 1](#jcm-07-00342-f001){ref-type="fig"}b). This patient presented with a history of gradually increasing right lower limb weakness for 10 years and a burning sensation in the left lower limb for several years. Neurological examination revealed bilateral lower limb weakness, pathologically brisk right knee jerk, bilateral extensor plantar responses, and decreased pain and temperature sensation in the left T7 to S5 dermatomes. Magnetic resonance imaging of the brain and cervical spine demonstrated multiple deep cerebral white matter lesions, as well as lesions in the medulla and upper cervical spinal cord. She had oligoclonal immunoglobulin G (IgG) bands restricted to the cerebrospinal fluid and met the 2010 Revised McDonald Criteria for a diagnosis of MS \[[@B22-jcm-07-00342]\]. She had one daughter; however, analysis of the daughter's DNA showed that she did not carry the exon 2 mutation. 3.2. Effects of Point Mutations in PLP1 on PLP Expression in Cos-7 Cells {#sec3dot2-jcm-07-00342} -------------------------------------------------------------------------- We used a Cos-7 cell transfection/expression system to assess the effects of PLP1 mutations on overall cell health and trafficking of PLP through the cells, as previously described \[[@B23-jcm-07-00342]\]. Constructs were prepared that carried PLP1, one of the five mutations listed in [Table 3](#jcm-07-00342-t003){ref-type="table"}, or an empty vector, coupled to a mCherry tag to allow identification of transfected cells. Cells were transfected into Cos-7 cells. Initial studies were completed to show that mCherry expression co-localised with anti-PLP antibody staining ([Figure 2](#jcm-07-00342-f002){ref-type="fig"}a). Thereafter, the number of cells that were mCherry positive, compared to the number of nuclei present, was used to assess the number of transfected cells 48 h after transfection ([Figure 2](#jcm-07-00342-f002){ref-type="fig"}b). Approximately 40% of cells transfected with the tagged-wtPLP vector expressed PLP by 48 h after transfection ([Figure 2](#jcm-07-00342-f002){ref-type="fig"}b). In comparison, there were significantly fewer PLP+ cells in the cultures transfected with the PMD/SPG2-related PLP1 mutant sequences, L30P, or H139Y. Of the three MS-associated mutations tested, only the L30R mutation caused a significant reduction in the percentage of PLP+ cells. This variation in expression was not due to effects on gene transcription, as analysis of mRNA from these cells using quantitative real-time polymerase chain reaction (qRT-PCR) showed no significant differences in any of the cells expressing mutant PLP1 compared to the wild type (data not shown). 3.3. Effects of Point Mutations in PLP1 on the Trafficking Pattern of PLP in Cos-7 Cells {#sec3dot3-jcm-07-00342} ------------------------------------------------------------------------------------------ Cos-7 cells transfected with PLP1 sequences containing PMD-associated point mutations were previously reported to accumulate PLP in the endoplasmic reticulum (ER) and it does not traffic out to the cell membrane \[[@B7-jcm-07-00342]\]. Cos-7 cells transfected with mCherry-tagged wt- or mutant-PLP1 were therefore investigated for the distribution of PLP within the cells by co-labelling cells with ER-ID to show the ER ([Figure 3](#jcm-07-00342-f003){ref-type="fig"}a). At 24 h post-transfection, almost all cells transfected with wt-PLP1 expressed mCherry throughout the cells ([Figure 3](#jcm-07-00342-f003){ref-type="fig"}b,d). Some of the staining outside of the ER had a vesicular appearance (arrowheads in [Figure 3](#jcm-07-00342-f003){ref-type="fig"}b), consistent with previous reports \[[@B8-jcm-07-00342]\]. The percentage of PLP-expressing cells in which PLP had moved out from the ER and/or showed vesicular staining was also determined in cells transfected with the various PLP1 mutants. Compared to wt-PLP1 transfectants, the amount of PLP that moved out of the ER in cells transfected with all of the mutant PLP1 sequences significantly decreased, although the transfectants having the greatest effect on transport out of the ER were those with mutations at L30, particularly L30P ([Figure 3](#jcm-07-00342-f003){ref-type="fig"}c,d). 3.4. Investigation of the Induction of the Unfolded Protein Response in PLP1-Transfected Cos-7 Cells {#sec3dot4-jcm-07-00342} ------------------------------------------------------------------------------------------------------ The accumulation of mutant PLP in the ER of an oligodendrocyte leads to the activation of the UPR, a natural feedback system that is in place to help cells clear incorrectly synthesised proteins. Upon UPR activation, CHOP expression is upregulated and ER-localised CHOP translocates to the nucleus to activate genes involved in UPR regulation \[[@B26-jcm-07-00342],[@B27-jcm-07-00342],[@B28-jcm-07-00342]\]. We therefore investigated CHOP expression as a measure of induction of the UPR by the PLP1 mutants. As a positive control for the study of CHOP expression and activation of the UPR in our Cos-7 system, cells were incubated for 6 h with 1:1000 Brefeldin A (BFA, eBiosciences, San Diego, CA, USA), as prolonged exposure of cells to BFA was previously shown to activate the UPR \[[@B28-jcm-07-00342]\]. All cells upregulated CHOP in the nuclei following exposure to BFA ([Figure 4](#jcm-07-00342-f004){ref-type="fig"}a). Cos-7 cells transfected with the PLP1 mutants and cultured for 24 h were also assessed for upregulation of nuclear CHOP. Compared with normal untransfected Cos-7 cells, there was no significant increase in CHOP intensity for pulsed (no DNA control; data not shown) or wild-type PLP transfected cells. As anticipated from our trafficking pattern study, L30P- and L30R-transfected cells showed the most significant increase in CHOP expression in the nucleus compared to both normal cells (p \< 0.0001) and in cells transfected with wt-PLP1 ([Figure 4](#jcm-07-00342-f004){ref-type="fig"}). In these cells, the intensity of nuclear CHOP expression was clearly linked to the ability of the cell to traffic PLP out of the ER. Cells that were unable to traffic PLP out of the ER ([Figure 4](#jcm-07-00342-f004){ref-type="fig"}b) had a much higher intensity of nuclear CHOP staining than did cells expressing PLP throughout the whole cell ([Figure 4](#jcm-07-00342-f004){ref-type="fig"}c), indicating that protein accumulation and activation of the UPR is associated with a failure of cells to traffic PLP. CHOP nuclear expression in L30V-, R136W-, and H139Y-transfected cells also significantly increased compared with normal cells, but L30V-, R136W-, and H139Y-transfected cells were not significantly different compared with cells transfected with wt-PLP1 ([Figure 4](#jcm-07-00342-f004){ref-type="fig"}d). 3.5. L30V Mutation Could Potentially Change Immune Responsiveness to PLP {#sec3dot5-jcm-07-00342} ------------------------------------------------------------------------ The above results suggest that the L30V mutation identified in the patient in our study had a relatively mild effect on PLP expression and trafficking. Even if PLP1 mutations were not sufficiently severe to cause widespread apoptosis of oligodendrocytes in patients, another way in which they could potentially be related to the development of MS would be to change the ability of the peptide to bind to the HLA molecules carried by the patient, thereby changing a non-immunogenic PLP peptide to an immunogenic and potentially pathogenic epitope. The HLA type of this patient was HLA-A2, 29; B44; DRB1\07:01, 13:02; DQA1\01:02, 02:01; DQB1\03:03, 06:04. For each of the HLA-A, HLA-B, and HLA-DRB1 alleles, in silico analysis was used to investigate the potential binding of PLP peptides containing residue 30 as either a leucine or valine, using the IEDB analysis resource ([www.iedb.org](www.iedb.org)) \[[@B19-jcm-07-00342],[@B20-jcm-07-00342],[@B21-jcm-07-00342]\]. The 50% concentrations of peptides with either L or V at residue 30 that were predicted to inhibit binding of a control peptide to the HLA allele of interest (the half maximal inhibitory concentration (IC~50~) values) are shown in [Table 4](#jcm-07-00342-t004){ref-type="table"}. The lower the number in the table, the stronger the binding, with peptides with an IC~50~ less than one predicted to be very high binders. Those with an IC~50~ between 1 and 10 being moderate binders, and those between 10 and 50 being relatively weak binders. In most cases, the predicted binding of a peptide with either L30 or V30 was similar for any specific HLA allele, which is not surprising as L`→`V is a relatively conservative substitution. However, for some peptides, a change from L30 to V30 would significantly increase the binding of the peptide to a HLA molecule. For example, peptide FFGVALFCG was predicted to be an extremely weak binder to DRB1\07:01 (IC~50~ \> 50 µM), whereas FFGVAVFCG was predicted to bind with extremely high affinity to that allele (IC~50~ = 0.9 µM), and would therefore have a much higher potential to generate an immune response in the patient. 4. Discussion {#sec4-jcm-07-00342} ============= The initiating events that lead to MS are still unknown, and there has been considerable debate on the relative merits of "outside-in" (i.e., autoimmune) vs. "inside-out" (i.e., primary CNS damage) models for development of MS \[[@B29-jcm-07-00342]\]. Multiple genes, mostly of low penetrance, have been linked to susceptibility to the development of MS, and the nature of those genes strongly suggests that autoimmunity drives the development of disease in the majority of patients \[[@B30-jcm-07-00342],[@B31-jcm-07-00342]\]. However, not all patients follow the same pathway to disease development. In the current study, we identified a female patient with disease meeting the diagnostic criteria for PP-MS who carried a point mutation in PLP1 that resulted in an amino acid substitution at residue 30 of PLP (L30V). We subsequently showed that this and another two PLP1 mutations that were described in people who meet the criteria for MS \[[@B2-jcm-07-00342],[@B3-jcm-07-00342]\] were all able (to varying degrees) to affect trafficking of PLP out of the ER and induce the UPR ([Figure 5](#jcm-07-00342-f005){ref-type="fig"}). These findings suggest that these PLP1 mutations would all have the potential to cause damage to oligodendrocytes, and that they could therefore represent an alternative pathway for MS development. This conclusion is supported by the recent finding that mice carrying the L30R and R136W mutations show neuroinflammation and clinically relevant axonal degeneration, neuronal loss, and brain atrophy by one year of age \[[@B4-jcm-07-00342]\]. In those mice, inactivation of the recombination activating gene 1, which is necessary for maturation of B and T lymphocytes, resulted in a less severe clinical phenotype, suggesting that the adaptive immune response plays an important role in the disease development, similar to what is observed in MS. The region of the PLP1 gene encoding residue 30 appears to be a hotspot for mutation, as not only did it feature in the new patient identified in our study, but also in one of the other previously described MS patients (L30R mutation) \[[@B3-jcm-07-00342]\]. The L30P mutation has also been reported in cases of severe PMD \[[@B24-jcm-07-00342]\]. The concept of mutation hotspots in PLP1 has been previously described in PMD patients, and there are several amino acids at which multiple mutations have been reported \[[@B1-jcm-07-00342],[@B32-jcm-07-00342]\]. The L30V mutation identified in our study (which, of the three mutations at L30, showed the fewest effects on PLP expression or trafficking or on induction of the UPR) is a conservative mutation that would not be expected to have dramatic effects on the hydrophobicity of the first transmembrane domain in which it resides, as both L and V are hydrophobic amino acids with similar structures and neutral side chains. In contrast, an L30R mutation in the first transmembrane domain of PLP would be expected to change the hydrophobicity, overall charge, and/or secondary structure of the transmembrane helix, and could have a deleterious effect on the membrane-spanning potential of the helix and greatly disrupt the PLP structure. The L30P substitution would result in a rigid bend being imposed on the polypeptide and would be expected to affect the secondary structure of the transmembrane helix. The patients carrying the L30R and L30V mutations were both women and both developed clinical signs and symptoms consistent with PP-MS at an older age. The late onset of clinical symptoms and the progressive nature of their disease could potentially be explained by the dynamics of X-inactivation in women. Female carriers of transmembrane and extracellular PLP1 PMD mutations, which tend to cause the most severe cases of PMD in men, have been shown to be less likely to show any clinical manifestations of disease than women carrying mutations in the intracellular domains of PLP \[[@B1-jcm-07-00342],[@B3-jcm-07-00342]\]. This is likely due to skewed X inactivation of the mutant X chromosome, as it has previously been reported that expression of X-linked genes in women can favor the gene on one X chromosome over another in order to avoid producing mutated proteins \[[@B15-jcm-07-00342]\]. However, there is also evidence that suggests that a woman's X-inactivation pattern can change over time, even in the presence of mutant genes \[[@B33-jcm-07-00342]\]. One study reported that women with PP-MS exhibit a greater degree of X-inactivation skewing than healthy women and those with other MS subtypes \[[@B34-jcm-07-00342]\]. These findings, taken together, could be one explanation for the late onset of neurological symptoms seen in the patients with L30R and L30V mutations. We predict that if PLP1 mutations are a potential initiating factor in MS, then female patients who develop disease at an older age would be the most likely female patients in whom such mutations would be found. However, there is another recent report of two women (mother and daughter) who both carried a novel nonsense PLP1 mutation at codon 210 T \> G \[[@B35-jcm-07-00342]\] ([Figure 5](#jcm-07-00342-f005){ref-type="fig"}). Although disease in these patients mimicked PP-MS, clinical signs of disease were present from adolescence in both. Of interest, there was evidence of skewing of X inactivation in both patients, supporting the idea that heterozygous PLP1 mutations in women can lead to disease if skewed X inactivation occurs, regardless of whether this occurs due to aging or to some other process. Since PLP1 is located on the X chromosome, it is likely that any men with PLP1 mutations would develop disease at a very early age. The nature of the mutation would determine the clinical phenotype of the disease (i.e., PMD, SPG2, or MS-like). The R136W mutation \[[@B2-jcm-07-00342]\] occurs in exon 3B, which is deleted in the alternatively-spliced isoform of PLP known as DM20. Thus the 10-year-old boy who carried this mutation would be able to produce normal DM20. DM20 is expressed prior to PLP during ontogenesis and may have a role in the development of new oligodendrocytes \[[@B36-jcm-07-00342],[@B37-jcm-07-00342]\]; however, in post-natal CNS, the ratio changes and PLP becomes more abundant than DM20 (PLP:DM20 ratio 3:1 in the CNS) \[[@B38-jcm-07-00342]\], with the presence of both isoforms critical for successful myelination in adolescence \[[@B39-jcm-07-00342]\]. In in vitro studies, the presence of the correct DM20 sequence was shown to help mutant PLP traffic out of the ER \[[@B40-jcm-07-00342]\]. In the 24 or 48 h in vitro assays in this paper, the L30V and R136W variants showed fewer detrimental effects on the transfected Cos-7 cells than the L30R mutation. However, both the L30R and R136W mutations, when expressed as transgenes in mice, resulted in significant damage within the CNS. Thus, over time, even mild mutations such as R136W can cause sufficient oligodendrocyte damage so that disease ensues. Since the severity of the in vitro effects seen in cells expressing the L30V mutation lay between that seen in cells expressing the L30R and the R136W mutations, the L30V mutation identified in the current study could likely result in damage within the CNS. Direct oligodendrocyte damage might not be the only way in which PLP1 mutations could induce MS. As shown in [Table 4](#jcm-07-00342-t004){ref-type="table"}, several peptides containing the L30V mutation were predicted to bind with higher affinity to some of the HLA molecules carried by the patient than the native peptide, thus creating de novo epitopes and potentially inducing/activating a new set of autoreactive T cells in this patient. The generation of such responses would be dependent on the presence of specific proteases in the patient that would allow processing of the peptides shown in [Table 3](#jcm-07-00342-t003){ref-type="table"}, and on the presence of T cells in the patient's T cell repertoire that could recognize the new epitopes. The results of this study support the idea that mutations in myelin-related genes could contribute to the development of MS in a small proportion of patients. Furthermore, they suggest two possible mechanisms by which PLP1 mutations could cause MS: PLP1 mutations could drive an inside-out disease process by directly damaging oligodendrocytes; however, they could initiate an outside-in process, causing the expression of neoantigens that could be targeted by the immune system. Both processes could occur simultaneously. Genes on the X chromosome have been somewhat ignored in genetic studies that have been undertaken in MS, but PLP1 should possibly be considered for further investigation, particularly when MS develops in juvenile men or in older women. Conceptualization, N.C.C. and J.M.G.; Formal analysis, N.C.C. and J.M.G.; Funding acquisition, J.M.G.; Investigation, N.C.C., J.Y., A.A. and J.M.G.; Methodology, N.C.C. and J.Y.; Project administration, N.C.C. and J.M.G.; Resources, M.P.P.; Supervision, J.M.G.; Visualization, J.M.G.; Writing---original draft, N.C.C. and J.M.G.; Writing---review & editing, N.C.C., J.Y., A.A., M.P.P. and J.M. This research was funded by Multiple Sclerosis Research Australia grant number 10053. The authors declare no conflict of interest. ![Polymorphisms/mutations in proteolipid protein (PLP): (a) Exon 4 polymorphism: the three panels on the left show the DNA sequences of exon 4 of individuals who are homozygous at T (top), homozygous at C (center), or heterozygous at this position. (b) Exon 2 mutation: the top panel shows the wildtype sequence seen in a healthy woman, and the bottom panel shows a C \> G mutation found in a female patient with primary progressive multiple sclerosis (PP-MS). This mutation results in a coding change from CTG (leucine) to GTG (valine).](jcm-07-00342-g001){#jcm-07-00342-f001} ![Expression of PLP in Cos-7 cells. (a) Anti-PLP antibody co-localizes with expression of mCherry in transfected Cos-7 cells after 24 h in culture. (b) Percentages of cells transfected with the different PLP1 constructs that expressed PLP after 48 h in culture. Bars show mean ± SE of four experiments. The black bar shows cells transfected with the wildtype construct, the red bars are those transfected with constructs based on known Pelizaeus-Merzbacher disease/spastic paraplegia type 2 (PMD/SPG2) mutations, and the blue bars are cells transfected with constructs containing the MS-related mutations. \* p \< 0.05, \\\\ p \< 0.0001.](jcm-07-00342-g002){#jcm-07-00342-f002} ![Mutations at L30 cause PLP to be retained within the endoplasmic reticulum (ER). (a) Cells transfected with empty vector show labelling with the ER-ID stain (green), but no labelling for PLP, as expected. (b) After 24 h of culture of cells transfected with the wild-type (wt)-PLP vector, mCherry-tagged PLP (red) trafficked out of the ER. (c) In contrast, almost all PLP was retained within the ER (indicated by co-localisation of PLP and ER labelling) in cells transfected with the L30P mutant. Blue = 4′,6-diamidino-2-phenylindole (DAPI), red = mCherry-tagged PLP, and green = ER-ID stain. (Size bar = 50 µm). (d) Summary of the percentage of cells expressing the different PLP1 mutations that can traffic PLP out of the ER. The bars show mean ± SE of at least 20 cells analyzed for each group, and are color-coded as in [Figure 2](#jcm-07-00342-f002){ref-type="fig"}b. \\ p \< 0.01, \\\* p \< 0.001, \\\\ p \< 0.0001.](jcm-07-00342-g003){#jcm-07-00342-f003} ![Induction of the unfolded protein response (UPR), as indicated by increased CCAAT/enhancer binding protein homologous protein (CHOP) expression in the nuclei following Brefeldin A (BFA) treatment. (a) CHOP staining in untreated (no BFA) cells and Cos-7 cells treated with BFA for 6 h. Nuclei are stained blue with DAPI. In the untreated cells, the level of BFA staining was very low. In the treated cells, the overall level of CHOP expression (green) was higher, and there was a concentration of CHOP in the nuclei. (b) Examples of CHOP accumulation in L30R transfected cells in which PLP was been retained in the ER (inset shows just CHOP and DAPI overlay). (c) CHOP did not accumulate in L30R transfected cells in which PLP was able to move out into the cell periphery. (d) Summary of the CHOP nuclear intensity in transfected Cos-7 cells 24 h after transfection. Bars show the interquartile range and whiskers the 10th--90th percentile of CHOP staining in the nuclei of at least 20 cells per transfection. Asterisks/text in blue above each bar show p-value compared to normal Cos-7 cells. n.s.: not significant, \* p \< 0.05, \\ p \< 0.01, and \\\\ p \< 0.0001.](jcm-07-00342-g004){#jcm-07-00342-f004} ![Diagram showing the orientation of PLP in the myelin membrane, the location of the mutations mentioned in this study, and, for those mutations tested in this study, a summary of their effects on Cos-7 cells, with the asterisks indicating the significance of the effects (\* p \< 0.05, \\ p \< 0.01, \\\* p \< 0.001, \\\\ p \< 0.0001).](jcm-07-00342-g005){#jcm-07-00342-f005} jcm-07-00342-t001_Table 1 ###### Demographics of female primary progressive multiple sclerosis (PP-MS) patients and controls. Group Number Mean Age (Years ± SE) Mean MS Duration (Years ± SE) EDSS ^1^ ------------------ -------- ----------------------- ------------------------------- ---------------- Healthy controls 42 37.8 ± 1.8 Not applicable Not applicable PP-MS 22 53.0 ± 2.7 7.8 ± 1.5 6.3 ± 0.3 ^1^ EDSS: Expanded Disability Status Scale, with scores ranging from 0 (no neurological impairment) to 10 (death from MS) \[[@B18-jcm-07-00342]\]; SE: standard error. jcm-07-00342-t002_Table 2 ###### Frequency of PLP1 polymorphisms in female patients with primary progressive multiple sclerosis (PP-MS) and in healthy female controls. Group Polymorphism -------------------- -------------- ------------ ----------- ------------------ PP-MS (n = 22) 12 (54.5%) 7 (31.8%) 3 (13.6%) 1 (L30V) het ^1^ Healthy (n = 42) 24 (57.0%) 16 (38.1%) 2 (4.8%) None ^1^ het = heterozygous. jcm-07-00342-t003_Table 3 ###### PLP1 constructs used in the study. Designation Mutation Description ------------- ----------- ---------------------------------------- ------------------------------------------------- Wildtype None None Wildtype PLP sequence L30P 89T \> C Leucine to proline at position 30 Severe PMD mutation \[[@B24-jcm-07-00342]\] L30R 89T \> G Leucine to arginine at position 30 MS-like disease mutation \[[@B3-jcm-07-00342]\] L30V 88C \> G Leucine to valine at position 30 MS-like disease (this study) R136W 406C \> T Arginine to tryptophan at position 136 MS-like disease mutation \[[@B2-jcm-07-00342]\] H139Y 415C \> T Histidine to tyrosine at position 139 Mild SPG2 mutation \[[@B25-jcm-07-00342]\] ^1^ MS: multiple sclerosis; PLP: myelin proteolipid protein; PMD: Pelizaeus-Merzbacher disease; SPG2: spastic paraplegia type 2; ^1^ This mutation was chosen because it is the closest mutation to residue 136 known for PLP1. jcm-07-00342-t004_Table 4 ###### Predicted half maximal inhibitory concentration (IC~50~) (µM) of potential peptides from the region surrounding residue 30 of PLP for binding to the HLA-DR, HLA-A, and HLA-B alleles carried by the patient when residue 30 is an L or a V. Starting Amino Acid Peptide DRB1\07:01 DRB1\13:02 A\02 A\29 B\44 --------------------- --------------- ------------- ------------- ------- ------- ------- ---------- ------ ------ ------ ------ 22 GLCFFGVA(L/V) 0.6* -- ^1^ 8.7 5.8 0.4 0.2 9.2 8.5 -- -- 23 LCFFGVA(L/V)F 0.5 0.5 15.7 14.4 33.2 9.2 1.4 6.3 12.4 18.4 24 CFFGVA(L/V)FC -- -- 12.9 11.8 17.6 7.2 0.5 1.8 15.7 24.7 25 FFGVA(L/V)FCG -- 0.9 -- -- 14.7 14.3 6.8 7.6 -- -- 26 FGVA(L/V)FCGC -- -- -- -- 21.7 10.2 -- -- -- -- 27 GVA(L/V)FCGCG -- -- -- -- -- 24.7 14.5 15.4 -- -- 28 VA(L/V)FCGCGH 1.8 3.7 -- -- -- 29.5 11.1 14.9 -- -- 29 A(L/V)FCGCGHE -- -- -- -- 16.7 26.8 15.2 11.2 -- -- 30 (L/V)FCGCGHEA -- -- 10.5 7.8 7.7 19.9 3.2 3.4 -- -- ^1^ indicates a very weak binding affinity of \>50 µM.	{ "pile_set_name": "PubMed Central" }
Introduction to Gene Drive {#sec1} ========================== Most genes are thought to spread and persist in populations because they do something useful for the organisms carrying them, increasing survival and/or reproduction, at least on average. That is, most genes spread in populations by positive Darwinian selection and are maintained in the face of recurrent mutation by purifying selection. Some features of a gene may be selectively neutral---such as which nucleotide is found at a particular silent site---but random drift by itself will not lead to long open reading frames and associated control sequences that produce complex proteins performing a particular function. There are other genes, a minority, which spread and persist not by natural selection, not by increasing organismal survival or reproduction, but instead by distorting their transmission from one generation to the next. For example, some genes manage to be transmitted to more than half of an individual's gametes, even when that individual only inherited the gene from one of its two parents. In this case, the frequency of the gene increases due to the process of gene transmission from one generation to the next, and it is this unequal genetic transmission that gives the gene its advantage. Genes or genetic elements showing such transmission ratio distortion, or "drive," include gametic and zygotic killers, meiotic drivers, homing endonuclease genes, B chromosomes, and transposable elements, each of which has evolved several or many times in different taxa.^[@ref1],[@ref2]^ Moreover, drive has been an important process affecting such genomic features as genome size, base composition, chromosome shape, repeat structure, distribution of recombination hotspots, and centromere structure.^[@ref1],[@ref3],[@ref4]^ Not only can driving genes spread without doing anything useful for the organisms carrying them, they can even spread if they cause some harm, as long as the effect of the transmission distortion is greater than the effect of the reduced survival and reproduction. For this reason, they are often called selfish genes, or selfish genetic elements.^[@ref1],[@ref2]^ And since they can be harmful to the organism, genes that suppress these genes can themselves spread by natural selection (analogous to the spread of genes suppressing any other parasite), which the selfish gene will then be selected to avoid, potentially leading to an arms race. Occasionally, features of a selfish genetic element may be co-opted to do something useful for the host. Classic examples include mating type switching in yeast, antibody diversification in vertebrates, and telomere maintenance in Drosophila (and in eukaryotes more generally)---the evolution of all of these operations has involved the domestication or co-option of functions of selfish genetic elements.^[@ref1],[@ref3]^ Synthetic Gene Drive Systems {#sec1-1} ============================ The fact that gene drive can lead to the spread of fitness-reducing traits makes it potentially useful for controlling disease vectors and other pests. Moreover, the spread can be rapid: in a closed, random mating population, a construct with 100% drive and no fitness effects can increase from 1% to 99% in the population in just nine generations---fast enough to be attractive for public health interventions. Discussions about how to exploit gene drive for pest control date back for decades, long before there was any mechanistic understanding of how they worked, particularly among medical entomologists looking for new ways to control disease vectors.^[@ref5]−[@ref10]^ However, classical genetic approaches were not sufficiently flexible to be able to construct a useful gene drive system. Now, with the recent progress in molecular biology, there is renewed interest in trying to make synthetic gene drive elements,^[@ref11],[@ref12]^ and preliminary discussions of potential uses have expanded to agriculture and conservation.^[@ref13]^ In broad outline, two types of intervention have been considered, either to reduce the size of the target population, or to leave numbers more-or-less intact and genetically modify the population such that it is less harmful (e.g., less able to transmit a pathogen). And three main molecular paradigms are being explored, the use of toxin--antidote systems, chromosomal rearrangements, or sequence-specific nucleases. Toxin-antidote systems and chromosomal rearrangements may be useful for introducing and spreading a new "cargo" gene through a population that makes the population less harmful (e.g., an effector gene that makes mosquitoes unable to transmit a pathogen).^[@ref14],[@ref15]^ Nuclease-based drive systems may also be used to introduce novel genes, or they may be used for population suppression.^[@ref16],[@ref17]^ Toxin-Antidote Systems and Chromosomal Rearrangements {#sec1-2} ----------------------------------------------------- Many naturally occurring gene drive systems act as if they produce a toxin and antidote, though often the molecular details are not known. For example, in mice heterozygous for the t-haplotype, and Drosophila heterozygous for Segregation Distorter, these elements somehow act during spermatogenesis to sabotage spermatids or sperm carrying the wild-type allele, with the result that each is transmitted to over 90% of the progeny (compared to the Mendelian 50%).^[@ref18],[@ref19]^ In Tribolium flour beetles, the medea gene acts in heterozygous females to somehow cause progeny that do not inherit the medea gene to die.^[@ref20],[@ref21]^ Though the underlying molecular mechanisms are not known, this example stimulated the development of a synthetic gene drive construct in Drosophila with the same logic.^[@ref22]^ The construct combined a microRNA-based repressor of myd88 (an important protein normally supplied by the mother into the embryo) with a zygotically expressed myd88 gene that was not affected by the microRNA and supplied the missing protein. As intended, this construct was able to increase in frequency over successive generations in experimental cage populations. Two other medea systems, using different components, have also been developed in Drosophila,^[@ref23]^ as have toxin--antidote systems that display maternal-effect lethal underdominance and threshold-dependent invasion into population cages.^[@ref15],[@ref24]^ Recent descriptions of natural toxin--antidote systems in plants, fungi, and nematodes^[@ref25]−[@ref31]^ may provide further insights into how these sorts of systems can be engineered. One way for toxin--antidote systems to work is by generating underdominant fitness interactions, in which the heterozygote is less fit than either of the two homozygotes. Underdominant interactions can also be generated with chromosomal rearrangements such as reciprocal translocations, and if these can be introduced at a sufficient frequency into a population (\>50% in the simplest scenario), then they can spread to fixation.^[@ref6]^ Strains of Drosophila with reciprocal translocations have been engineered, and these showed the expected frequency-dependent spread in lab populations.^[@ref32]^ Nuclease-Based Systems: Chromosome Shredding {#sec1-3} -------------------------------------------- In Aedes and Culex mosquitoes, there is a naturally occurring driving Y chromosome that, in some crosses, is transmitted to more than 90% of a male's progeny. First described in the 1960s,^[@ref5],[@ref33]^ there is still no good understanding of how it works at the molecular level, but cytologically it is associated with breaks of the X chromosome at male meiosis, perhaps having something to do with interrupted crossovers.^[@ref34],[@ref35]^ This observation led to the idea that cleavage of the X chromosome during male meiosis might lead to drive of the Y.^[@ref16]^ In Anopheles gambiae, the most important vector of malaria in Africa, the rRNA genes are found in a single cluster of hundreds of copies on the X chromosome, making it an ideal target,^[@ref36]^ and sure enough, production of a nuclease targeting this sequence during spermatogenesis can produce biased sex ratios, up to 95% males, using both an engineered meganuclease and a CRISPR-based nuclease.^[@ref37],[@ref38]^ A male-biased population sex ratio would be useful because males do not bite people and transmit disease, nor do they contribute as much materially to population productivity, and so total population size is also likely to decline.^[@ref39]^ The constructs reported to date do not yet constitute a fully functional gene drive system, because the nuclease genes have been inserted on an autosome, and so are themselves still transmitted in a Mendelian manner; the next step is to put them on the Y chromosome, which is challenging because it is highly repetitive and largely suppressed at meiosis, though some progress has been made.^[@ref40],[@ref41]^ Nuclease-Based Systems: Homing {#sec1-4} ------------------------------ Homing endonuclease genes (HEGs) are a class of natural occurring driving elements for which there is a good understanding about the molecular mechanisms, and these are both simple and general enough to potentially be worth exploiting. HEGs encode a nuclease that recognizes and cuts a sequence that typically occurs just once in the genome. The gene is in the middle of its own recognition sequence, disrupting it and protecting the chromosome it is on from being cut. Therefore, in heterozygotes, only the chromosome not containing the gene is cut; it is then repaired using the HEG-containing homologue as a template, with the result that the HEG is copied across to the chromosome where previously it was absent, converting a heterozygote into a homozygote.^[@ref42],[@ref43]^ This "homing" reaction simply requires a gene encoding a sequence-specific nuclease, with the gene inserted in the middle of its own recognition sequence, and the cell's DNA repair system takes care of the rest. HEGs occur naturally in many microbes but have not yet been identified in any insects or vertebrates. Important proof-of-principle experiments demonstrated that the homing reaction can occur in both Drosophila and Anopheles.^[@ref44]−[@ref47]^ These first experiments used meganucleases, and later experiments showed the reaction could also be catalyzed by zinc finger and TALE nucleases,^[@ref48]^ and, more recently, CRISPR-based nucleases.^[@ref49]−[@ref51]^ In principle, the homing reaction can be used for both population-wide knockout of target genes, such as those involved in survival, fertility, or pathogen transmission, or for population-wide knock-in of novel effector genes.^[@ref17],[@ref52],[@ref53]^ As with other forms of pest or disease control, due attention must be given to the possibility of resistance evolving.^[@ref54]−[@ref57]^ Prospects {#sec2} ========= There are a relatively small number of species for which genetic control methods, including gene drive, may be appropriate. Most prominent are those causing or transmitting diseases. Even now, more than 700 000 people die every year from vector-borne diseases, and there is an additional heavy burden of nonlethal morbidity.^[@ref58]^ Much of disease control is, ultimately, chemical, with efficacy largely determined by the degree to which production and delivery can be targeted and affordable. Vaccination can be among the most cost-effective of all health interventions because it uses the adaptive immune system to generate and deliver the active agents. The promise of genetic approaches---and gene drive in particular---is again to use biological processes--mating, meiosis, transcription, translation, etc.---for targeted, cost-effective delivery of appropriate chemicals (e.g., nucleases or antimicrobial peptides) that will substantially reduce disease transmission. Indeed, gene drive may take efficiency a step further, with a single release (perhaps with periodic "booster" releases) giving area-wide, population-level control. Important steps have been made toward realizing this potential, though there remains much more to do. The authors declare no competing financial interest. We thank the reviewers for useful comments. Funded by grants from the Bill & Melinda Gates Foundation and the Open Philanthropy Project.	{ "pile_set_name": "PubMed Central" }
Published: March 27, 2018 Introduction {#sec1} ============ Neural circuits within the hypothalamus regulate energy balance in response to peripheral nutrient-related cues ([@bib1], [@bib18]). Leptin-responsive Agouti-related protein (Agrp)-expressing neurons in the arcuate nucleus of the hypothalamus are activated during fasting or caloric deficit to drive an increase in food intake, while in the nutritionally replete or fed state, Pro-opiomelanocortin (Pomc) neurons are activated to reduce food intake ([@bib13], [@bib14]). In humans, loss-of-function mutations that disrupt the function of these neural circuits result in severe obesity, demonstrating their pivotal role in human energy homeostasis ([@bib29], [@bib40]). However, experiments in rodents ([@bib3], [@bib6]) and genetic studies in humans ([@bib22]) suggest that the neural mechanisms that regulate energy homeostasis are complex and that many molecular components of these circuits remain to be discovered ([@bib36]). One potential approach to identifying genes and pathways is to use transcriptomic analysis of key tissues and organs to identify changes in gene expression in response to a perturbation or genetic manipulation. The specificity of these approaches has been enhanced by recent technological developments that have enabled the labeling, sorting, and RNA sequencing of molecularly defined populations of neurons in the mouse brain. To this end, the recent detailed analysis of high-quality gene expression data from mouse Agrp and Pomc neurons has provided a framework for investigating the genes whose expression changes with fasting and feeding ([@bib11], [@bib23]). Although comparable studies of specific cell types are not feasible in humans, transcriptional analysis of hypothalamic tissue from people with molecularly defined subtypes of severe obesity has the potential to inform the discovery of neural mechanisms involved in energy balance. Here, we characterized the hypothalamic transcriptome of individuals with Prader-Willi syndrome (PWS), a genetic obesity syndrome caused by loss of expression of paternally expressed genes and noncoding RNAs on chromosome 15q11--q13 ([@bib12]). Results and Discussion {#sec2} ====================== RNA sequencing was performed on post-mortem hypothalamic tissue from four PWS patients and four age-matched controls from the University of Maryland Brain and Tissue Bank ([Figure S1](#mmc1){ref-type="supplementary-material"}). Although samples from controls matched for both age and obesity were not available, the body mass index (BMI) values of patients and controls were comparable ([Figure S1](#mmc1){ref-type="supplementary-material"}A). Principal-component analysis revealed segregation between PWS and control samples ([Figure 1](#fig1){ref-type="fig"}A). We identified 3,676 differentially expressed genes (DEGs) in PWS individuals compared with controls ([Table S1](#mmc2){ref-type="supplementary-material"}; Benjamini-Hochberg false discovery rate \[FDR\] \< 0.25; 658 with FDR \< 0.05). The most highly downregulated genes (FDR \< 5 × 10^−5^) were located in the PWS critical region ([Figure 1](#fig1){ref-type="fig"}B). A random subset of genes were validated by qRT-PCR ([Figure S1](#mmc1){ref-type="supplementary-material"}E). In the absence of high-quality hypothalamic tissue for replication, we compared our data with a previous high-density microarray study of hypothalamic gene expression in two PWS patients ([@bib17]) and found significant overlap of dysregulated genes ([Figures 1](#fig1){ref-type="fig"}C and [S1](#mmc1){ref-type="supplementary-material"}D; [Table S1](#mmc2){ref-type="supplementary-material"}). However, there was minimal overlap with datasets derived from PWS induced pluripotent stem cell (iPSC)-derived neuronal cell lines (data not shown); notably, we did not find reduced expression of the obesity-associated gene PCSK1 reported recently ([@bib10]).Figure 1Genome-wide Transcriptional Changes in PWS Hypothalamus(A) Principal-component (PC) analysis showing segregation of PWS and control hypothalamic samples.(B) Heatmap representing the top 45 most significantly DEGs shown as within-gene Z score (left) and rlog-normalized read counts (right).(C) Venn diagrams illustrating differentially down- and upregulated genes in PWS versus control samples in this study (discovery set) and overlap with genes from a previous study in PWS (replication set) ([@bib17]).(D) Heatmaps representing the expression of brain cell-type-specific genes among the DEGs displayed as within-gene Z score of rlog-normalized read counts.See also [Figure S1](#mmc1){ref-type="supplementary-material"} and [Table S1](#mmc2){ref-type="supplementary-material"}. To identify the cellular origin of DEGs, we ranked genes on the basis of their relative expression in single-cell transcriptomic data from neurons, astrocytes, microglia, oligodendrocytes, and endothelial cells ([Supplemental Experimental Procedures](#mmc1){ref-type="supplementary-material"}). We found that downregulated genes were enriched for neuronal markers (p = 3 × 10^−8^), while upregulated genes were enriched for microglial genes (p = 9 × 10^−5^) ([Figure 1](#fig1){ref-type="fig"}D). Further analysis using CIBERSORT ([@bib28]) also showed that PWS hypothalamic tissue was characterized by a reduction in neurons ([Figure S1](#mmc1){ref-type="supplementary-material"}F). Interestingly, this cellular transcriptomic profile aligns with that seen in autism ([@bib30]), in several neurodegenerative diseases, and in the aging brain ([@bib7], [@bib26]) ([Figure S2](#mmc1){ref-type="supplementary-material"}A), suggesting that fundamental mechanisms regulating neuronal maintenance may contribute to a range of human neurological diseases, including PWS. Overlap of the Human PWS Transcriptome with the Transcriptome of Agrp Neurons in Fasting {#sec2.1} ---------------------------------------------------------------------------------------- To identify potential candidate obesity genes, we compared PWS DEGs with genes expressed in hypothalamic Agrp and Pomc neurons in mice ([@bib11], [@bib23]) ([Supplemental Experimental Procedures](#mmc1){ref-type="supplementary-material"}). We found that expression of Agrp was increased 3-fold in PWS hypothalamus versus controls (p = 0.01), suggesting this potent orexigenic may play a role in the hyperphagia associated with PWS. Other upregulated genes were predominantly expressed in mouse Agrp neurons that signal hunger, while genes downregulated in PWS were relatively overrepresented in mouse Pomc neurons that signal the fed state (Fisher's exact test, odds ratio \[OR\] = 7.2, p = 2.3 × 10^−4^) ([Figures 2](#fig2){ref-type="fig"}A and [S2](#mmc1){ref-type="supplementary-material"}). A significant number of PWS upregulated genes were expressed in mouse Agrp neurons and upregulated in fasted animals (Fisher's exact test, OR = 5.3, p = 10^−12^; [Figure 2](#fig2){ref-type="fig"}B), suggesting that these genes represent a conserved signature of the neural response to fasting or food deprivation.Figure 2Dysregulated Gene Co-expression Modules in PWS Hypothalamus Converge with Fasting and Feeding Responses in Specific Hypothalamic Cell Types from Mice(A) Venn diagrams illustrating the number of DEGs that are down- and upregulated in PWS hypothalami compared with controls and their expression in Pomc, Agrp, and other neurons ([@bib11], [@bib23]). For comparison, the reference gene sets (Pomc, 261 genes; Agrp, 167 genes; other neurons, 1,589 genes) are included in [Figure S2](#mmc1){ref-type="supplementary-material"}A.(B) Number of PWS DEGs (up- or downregulated) that are expressed in Agrp neurons in the fasted versus fed state (q \< 0.05 in [@bib23]).(C) Gene co-expression modules among upregulated PWS DEGs. Hierarchical clustering of DEGs upregulated in PWS with log~2~ fold change \>1.5. The heatmap illustrates pairwise gene-gene correlation clustering (Pearson correlation, distance = 1-cor, Ward clustering). The sidebar (right) displays the overlap with genes previously reported upregulated (red) or downregulated (green) in Agrp neurons in the fasted versus fed state (q \< 0.05 in [@bib23]).See also [Figure S2](#mmc1){ref-type="supplementary-material"} and [Table S1](#mmc2){ref-type="supplementary-material"}. Using hierarchical cluster analyses of high-confidence DEGs (absolute log fold change \> 1.5), we identified sets of co-expressed genes and gene modules whose expression was upregulated in Agrp neurons in the fasted state ([Figure 2](#fig2){ref-type="fig"}C). We observed increased expression of ribosomal proteins involved in protein synthesis. This finding aligns with the upregulation of genes involved in endoplasmic reticulum (ER) protein translocation and Golgi trafficking seen in Agrp neurons in mice with fasting ([@bib23]) and may reflect increased production of neuropeptides for secretion. Several genes downregulated in PWS, and also in mouse Pomc neurons, were involved in synaptic transmission and neuronal maintenance and integrity. As loss-of-function mutations in some of these genes (SRPX2 and ZBTB16; [Table S1](#mmc2){ref-type="supplementary-material"}) are known to cause human neurological disorders, their reduced expression could contribute to both the obesity and the neurodevelopmental phenotype of PWS. A subset of co-regulated genes dysregulated in the PWS hypothalamus are expressed in Agrp neurons in fasting and are known to play a role in energy homeostasis and adipocyte biology in rodents (SOCS3, ANGPTL4, FOSL1, FOSL2, and STC2; [Table S1](#mmc2){ref-type="supplementary-material"}). Interestingly, bone morphogenic factor-3 (BMP3), whose expression is markedly decreased in mouse Agrp neurons in the fasted state (−17.7-fold, q = 2.0 × 10^−5^; [@bib23]), was found to be significantly decreased in the human PWS hypothalamus. These findings generate hypotheses that will need to be explored further. Characterization of the neurons in which these genes are expressed and the processes they regulate, as well as DEGs expressed in other transcriptionally distinct neuronal cell types, may provide insights into the mechanisms involved in human energy balance. Human PWS Hypothalamus Is Characterized by Downregulation of Genes Involved in Neuronal Function and Upregulation of Microglial Genes and Inflammatory Markers {#sec2.2} -------------------------------------------------------------------------------------------------------------------------------------------------------------- We found that downregulated DEGs were significantly enriched for genes involved in certain processes, namely, neurogenesis, neurotransmitter release, and synaptic function ([Figure 3](#fig3){ref-type="fig"}A). Using Ingenuity Pathway Analysis, we identified 11 potential regulators of clusters of downregulated DEGs ([Table S2](#mmc3){ref-type="supplementary-material"}), including the neurotrophin brain-derived neurotrophic factor (BDNF) and its receptor, TrkB (encoded by NTRK2). Putative BDNF/TrkB targets among the downregulated DEGs were predominantly related to synaptic processes ([Figure 3](#fig3){ref-type="fig"}B). This finding is intriguing, as BDNF is a major regulator of the development, maturation, and maintenance of neurons and a modulator of synaptic plasticity ([@bib35]). Moreover, in mice and humans, genetic disruption of BDNF and TrkB causes developmental delay, stereotyped behaviors, impaired pain sensation, hyperphagia, and severe obesity ([@bib19], [@bib41]), phenotypes that show some overlap with those seen in PWS. We also obtained several post-mortem brain samples for histology. Very few samples were of sufficient quality, limiting quantitative analysis, but fluorescence in situ hybridization of human hypothalamic tissue suggested that the number of cells expressing BDNF and NTRK2 mRNA was reduced in the ventromedial nucleus of the hypothalamus in PWS ([Figures 3](#fig3){ref-type="fig"}C and [S3](#mmc1){ref-type="supplementary-material"}). We measured levels of plasma BDNF (potentially derived from platelets) in patients with PWS versus age-matched obese controls, but we did not find a significant difference ([Figure S3](#mmc1){ref-type="supplementary-material"}G), in contrast to one previous study ([@bib21]). Possible explanations are that BDNF levels are known to vary considerably in plasma versus serum and among assays; additionally, plasma BDNF may not reflect BDNF expression in the brain.Figure 3Pathways Predicted to Be Affected by Changes in Gene Expression Seen in PWS Hypothalamus(A) A gene annotation network illustrating terms (Gene Ontology, Reactome, Key) enriched among downregulated DEGs. Nodes represent downregulated DEGs annotated with illustrated terms; edges join pairs of genes annotated with the respective term.(B) Ingenuity Pathway Analysis (IPA) regulator effects analysis indicates the inhibition of regulatory factors NTRK2, ADCYAP1, and BDNF (top) with predicted effects on target genes and processes. Phenotypes predicted to occur as a consequence of the gene expression changes are shown in blue (inhibited) or orange (enhanced).(C) Representative FISH images of BDNF and NTRK2 mRNA-expressing cells in the ventromedial nucleus of the hypothalamus and oxytocin mRNA-expressing cells in the paraventricular nucleus of the hypothalamus in PWS and control samples (BDNF \[n = 2 PWS, n = 2 controls\], NTRK2 \[n = 2 PWS, n = 1 control\], and oxytocin \[n = 2 PWS, n = 1 control\]).(D) A gene annotation network illustrating terms enriched among upregulated DEGs. Nodes and edges as in [Figure 2](#fig2){ref-type="fig"}A.(E) IPA upstream regulator analysis indicates inhibition of TNF/NFKb signaling.(F) Representative immunohistochemistry images of S100Beta- and GFAP-immunoreactive cells in the ventromedial nucleus of the hypothalamus in PWS and control samples.See also [Figure S3](#mmc1){ref-type="supplementary-material"} and [Table S2](#mmc3){ref-type="supplementary-material"}. A previous histopathological study of the PWS hypothalamus found a significantly reduced number of oxytocin neurons ([@bib37]), and clinical trials of intranasal oxytocin administration in PWS are ongoing ([@bib38]). In our study, we found a low level of oxytocin mRNA and a smaller number of cells immunoreactive for oxytocin in the paraventricular nucleus in PWS samples ([Figure 3](#fig3){ref-type="fig"}C), supporting the key role of oxytocin as well as BDNF in the neuropathology of PWS. Additional studies are needed to replicate these findings and to investigate the potential loss of other neuronal populations (including Pomc and Agrp neurons) within the hypothalamus in PWS. We found that upregulated genes in the PWS hypothalamus were enriched for cytokine signaling and inflammatory processes ([Figure 3](#fig3){ref-type="fig"}D; [Table S2](#mmc3){ref-type="supplementary-material"}). The most statistically significant predicted regulator of these genes was tumor necrosis factor (TNF)-alpha, which plays a critical role in systemic inflammation ([Figure 3](#fig3){ref-type="fig"}E; [Table S2](#mmc3){ref-type="supplementary-material"}). In the human hypothalamus, we studied expression of S100b (a glial-specific protein marker of neural damage) and GFAP (an astrocyte filament protein that plays a critical role in synaptic function and is reduced in neurodegenerative disorders but increased in brain injury). We found that S100b protein levels were increased and GFAP immunoreactivity was decreased in the PWS hypothalamus compared with controls ([Figure 3](#fig3){ref-type="fig"}F). These findings overlap with data from other neurodevelopmental conditions ([@bib20]). Further studies with larger sample sizes are needed to explore the potential relevance of these findings. Targeted Deletion of SNORD116 Affects Neuronal Differentiation, Proliferation, and Survival {#sec2.3} ------------------------------------------------------------------------------------------- Chromosomal deletions that cause PWS vary in size and thus can affect a number of genes and noncoding RNAs. None of the mouse models involving deletion of the homologous region fully recapitulate the human PWS phenotype ([@bib32]); as such, investigation of the molecular mechanisms that underlie the clinical phenotype has been challenging. The minimal genetic lesion associated with severe hyperphagia and obesity in PWS contains a cluster of noncoding small nucleolar RNAs (snoRNAs) referred to as the SNORD116 gene cluster ([@bib15], [@bib34]). Postnatal deletion of SNORD116 in the mediobasal hypothalamus has recently been shown to lead to increased food intake in mice ([@bib31]). To test whether loss of SNORD116 affects neuronal development and maintenance, as suggested by our transcriptomics analysis and in line with a rodent model ([@bib9]), we deleted a 57.4 kb genomic segment encompassing the SNORD116 cluster using CRISPR-Cas9 in a SH-SY5Y neuroblastoma human cell line ([Figure S4](#mmc1){ref-type="supplementary-material"}A). We found that SNORD116-deficient cells exhibited reduced neuronal differentiation, cell proliferation, and survival compared with wild-type cells ([Figures 4](#fig4){ref-type="fig"}A--4C). A higher proportion of SNORD116-deficient cells displayed neurites when treated with BDNF (mean 13%) compared with no treatment (mean 23%, p = 0.005, two-tailed t test), whereas no significant difference was observed within wild-type cells (28% with no treatment, 36% with BDNF; p = 0.2, two-tailed t test). Cumulatively, these data identify a transcriptomic signature in PWS consistent with marked hypothalamic neurodegeneration, which may be mediated in part by reduced expression of the neurotrophin BDNF and its receptor, TrkB. These data align with experiments in cortical neurons of the SNORD116 knockout mouse ([@bib9]). Neuronal loss is associated with a marked inflammatory response in the hypothalamus, which may be a primary defect, secondary to the neurodegenerative process or, as microglia have a role in synaptic development and function ([@bib4]), an inflammatory response to disordered synaptic plasticity in the PWS hypothalamus.Figure 4Deletion of SNORD116 Impairs Neuronal Differentiation, Proliferation, and Survival(A) Targeted deletion of SNORD116 (SNORD116del) affects the neuronal differentiation of SH-SY5Y cells, cultured for 7 days in retinoic acid (RA) in the absence (n = 5) or presence (n = 3) of BDNF. Left: representative images of wild-type (WT) and SNORD116del cells; right: quantification plot.(B) Cellular proliferation measured by EdU incorporation at day 7 (n = 3).(C) Cell survival measured by FACS at day 7 in culture (n = 6).(D) Overlap between in silico predicted SNORD116 gene targets and PWS differentially expressed and differentially spliced genes.All data are presented as mean ± SEM. Statistical significance was measured using two-tailed Student's t test (^∗^p \< 0.05, ^∗∗^p \< 0.01, ^∗∗∗^p \< 0.001; ns, non-significance). See also [Figure S4](#mmc1){ref-type="supplementary-material"} and [Tables S3](#mmc4){ref-type="supplementary-material"} and [S4](#mmc5){ref-type="supplementary-material"}. Predicted snoRNA Targets and Detection of Reduced Splicing Efficiency {#sec2.4} --------------------------------------------------------------------- SNORD116 and the closely related SNORD115 cluster belong to a group of orphan snoRNAs with presumed non-canonical functions. SNORD115 has been shown to regulate the post-transcriptional processing of a single pre-mRNA, the serotonin 2c receptor, through alternative splicing and RNA editing ([@bib25]). Using snoTARGET, we identified 588 predicted targets for snoRNAs within protein-coding genes ([Figure 4](#fig4){ref-type="fig"}D; [Table S3](#mmc4){ref-type="supplementary-material"}), some of which were differentially expressed in PWS hypothalamus ([Figure S4](#mmc1){ref-type="supplementary-material"}B). Further studies will be needed to test the functional significance of these findings. Interestingly, RNA-specific adenosine deaminase (ADARB1), a predicted target that is significantly downregulated ([Figure S4](#mmc1){ref-type="supplementary-material"}B), is involved in pre-mRNA editing of glutamate receptor subunit B and when deleted causes hyperphagia and obesity in mice ([@bib39]). As snoRNAs can modulate RNA splicing ([@bib42]), a process that plays a major role in human neuronal development, we performed a transcriptome-wide search for evidence of alternative splicing ([Supplemental Experimental Procedures](#mmc1){ref-type="supplementary-material"}). We found evidence of differential use of alternative splice variants in PWS samples compared with controls ([Table S4](#mmc5){ref-type="supplementary-material"}). Focusing on 180 loci with evidence of differential use of two alternative splice variants, the most frequently observed type of splice variant in PWS was retained introns ([Table S4](#mmc5){ref-type="supplementary-material"}; [Figure S4](#mmc1){ref-type="supplementary-material"}). Of note, we did not find evidence for differential splicing of the serotonin 2c receptor ([Figure S4](#mmc1){ref-type="supplementary-material"}C). Genes with putative differential splicing did not tend to be differentially expressed, consistent with decoupling of differential expression and splicing as seen in other disorders; exceptions included genes involved in microglial and inflammatory processes, which were among the top-ranked alternatively spliced genes ([Table S4](#mmc5){ref-type="supplementary-material"}). Motif searches within retained introns and 250 bp flanking regions indicated the presence of binding sites for canonical serine/arginine-rich splicing factors, and the presence of binding sites with predicted similarity to FUS splicing factor binding motifs ([Figure S4](#mmc1){ref-type="supplementary-material"}E). The FUS splicing factor regulates alternative splicing in the brain and has been previously linked to neurodegenerative diseases including amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD) ([@bib24], [@bib33]). In summary, in this study of the human hypothalamus in a small number of individuals with PWS, we identified a transcriptomic signature characterized by neuronal loss, altered neuro-plasticity, and neuroinflammation. Of note, several neuroimaging studies and case reports in PWS have identified structural abnormalities that would be consistent with a reduced number of neurons, such as reduced gray matter volume in a number of cortical areas and abnormal gyrification ([@bib27]). We identify a potential role for BDNF in PWS that requires further exploration and may have therapeutic relevance for this complex neuro-behavioral disorder. Additionally, we demonstrate that transcriptomic analysis of the human hypothalamus can generate testable hypotheses of potential relevance to the understanding of the neural circuits involved in human energy homeostasis. Experimental Procedures {#sec3} ======================= Human Samples {#sec3.1} ------------- Hypothalamic specimens used in the study were obtained at autopsy from control subjects with no reported clinical signs and patients with genetic diagnoses of PWS through the University of Maryland Brain Bank at the University of Maryland ([Figure S1](#mmc1){ref-type="supplementary-material"}A). All procedures were approved by the University of Cambridge Human Biology Research Ethics Committee (HBREC.2014.14). RNA Sequencing and Analysis {#sec3.2} --------------------------- Total RNA was prepared by tissue homogenization in Trizol reagent (Thermo Fisher Scientific) of about one-third of hypothalamus. Sequencing of RNA samples was performed by the University College London (UCL) Genomics core facility, using the TruSeq poly-A mRNA method (Illumina) and a HiSeq 2000 machine (Illumina). Differential expression, splicing, and pathway analysis are described in detail in [Supplemental Experimental Procedures](#mmc1){ref-type="supplementary-material"}, as is the validation of DEGs using qRT-PCR. In Silico Prediction of SNORD116 Gene Targets {#sec3.3} ----------------------------------------------- Genome-wide in silico prediction of SNORD116 targets was performed using snoTARGET software ([@bib5]) and RNA-cofold from the Vienna RNA package (<http://www.tbi.univie.ac.at/RNA/>). Cross-Species Comparison with Agrp and Pomc Neuronal Subtypes and Response to Food Deprivation {#sec3.4} ---------------------------------------------------------------------------------------------- Reference gene sets for broad neuronal subtype classifications were derived from [@bib11] as described in [Supplemental Experimental Procedures](#mmc1){ref-type="supplementary-material"}. Reference gene sets for fasting response in Agrp neurons were obtained from [@bib23] using a threshold of q \< 0.05 (unless otherwise stated) to define differential expression between fasting conditions. Immunohistochemistry and Fluorescence In Situ Hybridization {#sec3.5} ------------------------------------------------------------- Immunohistochemistry was performed as reported previously ([@bib8]) using the following primary antibodies: guinea pig anti-oxytocin (Peninsula Laboratories), rabbit anti-GFAP (Dako), and rabbit anti-s100beta (Abcam). Secondary antibodies were Alexa Fluor 488 donkey anti-guinea-pig IgGs or Alexa Fluor 488 goat anti-rabbit IgGs (Thermo Fisher Scientific). For the fluorescence in situ hybridization (FISH) experiments, sense and antisense digoxigenin-labeled riboprobes were generated from plasmids containing PCR fragments of BDNF and NTRK2 (generously provided by Dr. Baoji Xu, The Scripps Research Institute). Staining density and cell number were calculated using ImageJ analysis software (NIH). Full details are presented in [Supplemental Experimental Procedures](#mmc1){ref-type="supplementary-material"}. Cellular Studies {#sec3.6} ---------------- SH-SY5Y (ATCC CRL-2266) cells were used in all the cellular assays. We used a well-established protocol to differentiate SH-SY5Y cells into neurons with retinoic acid ([@bib16]). Full details on maintenance, neuronal differentiation, proliferation, and cell survival are presented in [Supplemental Experimental Procedures](#mmc1){ref-type="supplementary-material"}. SNORD116 Cluster Deletion Using CRISPR-Cas9 {#sec3.7} ------------------------------------------- We applied a cloning-free CRISPR protocol using gBlocks (gene fragments) encoding FE-modified single guide RNAs (sgRNAs) promoting enhanced stability ([@bib2]). Two gBlocks carrying the guide flanking the SNORD116 cluster on chr15q11.2 were nucleofected alongside GFP-expressing Cas9 plasmid PX458into the SH-SY5Y line. Fluorescence-activated cell sorting (FACS)-sorted cells were screened for successful editing using conventional PCR and confirmed by Sanger sequencing. Full details are presented in [Supplemental Experimental Procedures](#mmc1){ref-type="supplementary-material"}. Statistical Analysis {#sec3.8} -------------------- Statistical analyses were performed using GraphPad Prism version 6.0 for MacOS X. Data are represented as mean ± SEM. A two-tailed Student's unpaired t test was used, and p values \< 0.05 were considered to indicate statistical significance. Data and Software Availability {#sec4} ============================== The accession number for the RNA sequencing data reported in this paper is EGA: EGAS00001002901. Supplemental Information {#app2} ======================== Document S1. Supplemental Experimental Procedures and Figures S1--S4Table S1. DEGs in PWS Patients versus Controls Displayed by Ensembl GeneID, Related to Figures 1 and 2Grey shading indicates genes located in the PWS region. Column "ncRNA" indicates gene names for Ensembl gene biotypes lincRNA, miRNA, processed transcript, TEC ('to be experimentally confirmed'); or overlapping miRNAs. Also shown are overlaps with Falaleeva et al. (2015) and putative function of genes among co-expressed gene clusters overlapping with Henry et al. (2015) fasted versus fed in Agrp neurons.Table S2. IPA Upstream Regulator Analysis of Down, Up, or All DEGs, Related to Figure 3IPA Ingenuity Upstream Regulator Analysis reporting putative upstream regulators top-ranked by "Activation z-score," and p values (Fisher's Exact Test) for overlaps with putative target genes (see Supplemental Experimental Procedures).Table S3. List of Potential SNORD116 Targets Generated Using Software snoTARGET (Bazeley et al., 2008), Related to Figure 4Table S4. Differential Splicing in PWS versus Controls, Related to Figures 4 and S4List of differential splicing events displayed thresholded to FDR \< 0.25.Document S2. Article plus Supplemental Information The authors would like to thank the donors, their families, and the staff of the University of Maryland Brain Bank and the PWS patients and parents from the Dutch PWS cohort studies. Human tissue was obtained from University of Maryland Brain and Tissue Bank, which is a Brain and Tissue Repository of the NIH NeuroBioBank. We thank the core facility at UCL, the Cambridge National Institute of Health Research (NIHR) Biomedical Research Centre (BRC) core laboratory, and the Children's Hospital Los Angeles (CHLA) histology core. We also thank Professor Mark Lalande (University of Connecticut) for sharing the RNA-seq data from iPSC-derived neuronal cell lines. This work was supported by the Wellcome Trust (098497/Z/12/Z), the NIHR Cambridge Biomedical Research Centre, and the Bernard Wolfe Health Neuroscience Endowment (all to I.S.F.), the NIH (grants DK84142 and DK102780 to S.G.B.), the Foundation for Prader-Willi Research (to S.G.B.), a Society for Endocrinology early career grant (to E.G.B.), and a Royal Society research grant (RG160311 to E.G.B.). J.H. is funded by a Medical Research Council (MRC) studentship grant number (516700). Author Contributions {#sec5} ==================== E.G.B. and I.S.F. conceived and directed the study. E.G.B., K.L., K.K.L., J.H., and V.P. carried out the RNA sequencing and all downstream analysis. S.C. and S.G.B. performed the in situ hybridization and immunohistochemical experiments and analyses. N.P. and E.G.B. performed the functional experiments on SNORD116. G.S. and E.G.B. performed the in silico SNORD116 work. J.M.K., I.S.F., A.H.-K., L.D., and S.D. recruited patients and controls and contributed to the analysis of human samples. All authors analyzed and interpreted the results. E.G.B., K.L., and I.S.F. wrote the manuscript with contributions from all authors. Declaration of Interests {#sec7} ======================== The authors declare no competing interests. Supplemental Information includes Supplemental Experimental Procedures, four figures, and four tables and can be found with this article online at [https://doi.org/10.1016/j.celrep.2018.03.018](10.1016/j.celrep.2018.03.018){#intref0015}. [^1]: Lead Contact	{ "pile_set_name": "PubMed Central" }
Introduction {#S0001} ============ Dementia has been estimated to affect about 25 million persons worldwide and it is projected to nearly double every 20 years, reaching about 66 million in 2030 and about 115 million in 2050. The total number of new cases of dementia each year is nearly 7.7 million worldwide, implying one new case every 4 s \[[@CIT0001]\]. About 60--70% of cases of dementia are due to Alzheimer\'s disease (AD). The AD is clinically characterized by progression from episodic memory problems to a slow global decline of the cognitive function that leaves patients with end-stage AD bedridden and dependent on custodial care, with death occurring on average 9 years after diagnosis \[[@CIT0002]\]. Neuropathologically, AD is characterized by the presence of extracellular amyloid deposits composed of aggregated β-amyloid (Aβ) peptides and intracellular neurofibrillary tangles containing hyperphosphorylated, aggregated tau protein \[[@CIT0002]\]. To date, the etiological mechanisms underlying the neuropathological changes in AD remain unknown; on the other hand, inflammation is a fundamental protective immunological response, but when left unregulated, it can become a major cofactor in the pathogenesis of many chronic human diseases, including AD. The role of neuroinflammation in AD has been intensely investigated, in part because of its tremendous clinical implications \[[@CIT0002]\]. Prostaglandin-endoperoxide synthase 2 (PTGS2), also called cyclooxygenase 2 (COX2) (MIM\# 600262), has been shown to be expressed mostly in the central nervous system and inflammatory cells \[[@CIT0003], [@CIT0004]\]. The gene coding for PTGS2 has been mapped to 1q31.1 between two regions to which genetic linkage to AD has been reported \[[@CIT0005]\]. In addition, elevated PTGS2 levels are present in neurons from hippocampal pyramidal regions of the AD brain \[[@CIT0006]\]. A number of single nucleotide polymorphisms (SNPs) have been described in the promoter region of the PTGS2 gene that probably regulates its transcription, but only one polymorphism located at position --765 G/C, a putative stimulatory protein-1 binding site, has been shown to be functional \[[@CIT0007]\]. The C allele of COX2-765 polymorphism has been associated with decreased risk of AD \[[@CIT0007]\]; however, other findings seem to suggest that the distribution of COX2-765G/C polymorphism is similar between patients with AD and controls \[[@CIT0008]\]. Pyrosequencing is a DNA sequencing technology based on the sequencing-by-synthesis principle \[[@CIT0009]\], and the method is based on a real-time bioluminescence technique. Gharizadeh et al. analyzed in detail the sequence data of a large number of amplicons, using two techniques (pyrosequencing and Sanger dideoxy sequencing). Their results demonstrate that for short DNA sequences, pyrosequencing excels in reducing sample preparation time, providing ease-of-use and cost and labor savings \[[@CIT0010]\]. Also, the drawbacks of pyrosequencing are rapidly being addressed with the use of Sequenase, and automated software \[[@CIT0010]\]. In the present study, we evaluated the --765 G/C promoter variant of the PTGS2 gene by pyrosequencing technology in Sicilian patients with AD and in normal controls. Material and methods {#S0002} ==================== Patients {#S20003} -------- A total of 164 subjects, including 84 AD patients (41 males and 43 females; mean age: 75.9 ±7.8 years; age range: 56--95 years) and 80 normal subjects (37 males and 43 females; mean age: 77.3 ±8.5 years; age range: 56--95) were recruited at the IRCCS Associazione Oasi of Troina (Italy). The diagnosis of probable AD was made according to the recent National Institute of Aging (NIA) and Alzheimer\'s Association (AA) work group (NIA-AA) revision of the diagnostic criteria that enlarges the window for the detection of the disease to the early stages \[[@CIT0011]\]. Symptoms of cognitive impairment or family history of dementia were exclusion criteria for the controls. Written informed consent was obtained from the participants or from their families. Patients and controls were all born in Sicily and were of European origin. DNA extraction, polymerase chain reaction amplification and pyrosequencing {#S20004} -------------------------------------------------------------------------- DNA was isolated from a lymphocyte-enriched fraction of whole blood. The procedures for detecting the single nucleotide polymorphism (SNP rs 20417) of the --765 G/C promoter variant in the PTGS2 gene were based on polymerase chain reaction (PCR) amplification and pyrosequencing technology (PyroMark ID instrument; Biotage, Uppsala, Sweden) as previously described \[[@CIT0012]\]. Biotinylated PCR primer sequences for the amplification of PTGS2 promoter variant sites were selected according to the same previously published study \[[@CIT0012]\]. Five μl of DNA were added to produce 50 μl of PCR solution mixture that contained 0.2 mmol of each dNTP, 1.5 mmol/l MgCl ~2~, 1X PCR buffer, 20 pmol of each primer and 1.5 units of DNA Taq polymerase. The PCR was performed with an initial denaturation for 5 min at 95°C followed by 40 cycles of 30 s at 95°C, 30 s at 56°C, 30 s at 72°C and a final incubation for 10 min at 72°C. PCR products were resolved by agarose gel electrophoresis to confirm successful amplification. The biotinylated products were then immobilized to streptavidin-coated beads using a solution from a commercial PSQ TM 96 sample preparation kit (Biotage, Uppsala, Sweden). Beads (3 μl) were diluted in 15 μl of binding buffer with 10 μl of biotinylated PCR products, incubated for 10 min at room temperature and then transferred to a filter probe where the liquid was removed by vacuum filtration. DNA in the denaturation solution was separated, the templates were washed with washing buffer, transferred to a PSQ 96 SNP plate (Biotage, Uppsala, Sweden) and annealed with the sequencing primers described by Skarke et al. \[[@CIT0012]\], in annealing buffer at room temperature. Finally, the samples were analyzed using a PyroMark ID System and the SNP reagent kit (both purchased from Biotage, Uppsala, Sweden). Statistical analysis {#S20005} -------------------- Fisher\'s exact test was used for comparisons in 2 × 2 and 2 × 3 contingency tables. Statistical significance was accepted when the p value was lower than 0.05. The power of the study was calculated by the methods of Cohen \[[@CIT0013]\]. The χ^2^ test was use to determine whether there was a significant difference between the number of actual and expected genotypes in Hardy-Weinberg equilibrium. Results {#S0006} ======= No statistically significant difference between the observed and expected values in Hardy-Weinberg equilibrium was observed (p \> 0.05). The distribution of PTGS2 --765 G/C showed no significant difference between patients with AD and controls. PTGS2 --765 G/C genotype distributions in the cases and controls are presented in [Table I](#T0001){ref-type="table"}. In addition, no significant difference in the G and C allele frequency between AD patients and controls was observed (p \> 0.05) ([Table II](#T0002){ref-type="table"}). The power of the study is 28% calculated by the methods of Cohen \[[@CIT0013]\]. ###### Genotype frequencies of PTGS2 --765 G/C polymorphism in patients with Alzheimer\'s disease (AD) and controls Genotype Controls (%) AD patients (%) Value of p ---------- -------------- ----------------- -------------- GG 49 (61.2) 55 (65.5) 0.381 GC 24 (30.0) 26 (30.9) CC 7 (8.8) 3 (3.6) Total 80 (100) 84 (100) ###### Allele frequencies of PTGS2 --765 G/C polymorphism in patients with Alzheimer\'s disease (AD) and controls Allele Controls (%) AD patients (%) Value of p -------- -------------- ----------------- -------------- G 122 (76.2) 136 (80.9) 0.346 C 38 (23.8) 32 (19.1) Total 160 (100) 168 (100) Discussion {#S0007} ========== This preliminary study in Sicilian patients with AD showed that PTGS2 --765 G/C alleles and --765 C/C genotype seem not to be associated with the risk of developing AD. These data agree with the data obtained by other authors \[[@CIT0008]\]. The role of PTGS2 in neurodegenerative diseases is very controversial, and conflicting data exist in the literature \[[@CIT0007], [@CIT0008], [@CIT0014]\]. For instance, although some studies have demonstrated that the increase in PTGS2 activity greatly contributes to the progression of AD \[[@CIT0007]\], other investigators have suggested an entirely opposite effect \[[@CIT0008]\]. In this latter study, the C allele of PTGS2 --765 G/C promoter polymorphism was associated with decreased risk of AD, a finding which further supports the involvement of PTGS2 in AD etiology. Abdullah et al. reported that the C allele of COX-2 --765 promoter polymorphism is associated with decreased risk of AD in their sample of 168 AD cases (recruited from the Memory Disorder Clinics in Tampa and in Miami, Florida) \[[@CIT0007]\]. In fact, the rare C allele is associated with approximately 30% less expression of COX-2 compared to the more commonly occurring G allele \[[@CIT0015]\]. A study of Listi et al. suggests that the G allele of COX-2 --765 could be a risk factor for AD in a population from Northern Italy \[[@CIT0016]\]. Fehèr et al. indicate that the COX-2 G/G genotype is associated with AD and support the involvement of COX-2 in AD etiology \[[@CIT0017]\]. In addition, in the study of Tang et al., no significant difference in the distributions of COX-2 --765G/C polymorphism was observed between AD cases and controls \[[@CIT0018]\]. Our data do not seem to associate this polymorphism with AD, at least in the Sicilian population, but in light of the limited power of the study (28%) our conclusions should be evaluated with caution. In any case, further and broader studies are needed to confirm the role of PTGS2 --765 G/C polymorphism in AD. We plan to extend our preliminary analysis to a larger cohort of Sicilian patients with AD in the near future.	{ "pile_set_name": "PubMed Central" }
Identification of drugs to treat coronary heart disease (CHD) patients continues to be an important research area. Despite significant advances in treatment including statin therapy, CHD caused by atherosclerosis remains a major cause of morbidity and mortality in the United States.^[@ref1]^ Owing to the powerful effect of liver X receptor (LXR) agonists on reverse cholesterol transport (RCT)^[@ref2]^ and immune system modulation,^[@ref3],[@ref4]^ which culminate in reduced atherosclerosis lesions in animals,^[@ref5]−[@ref7]^ there have been many LXR agonist medicinal chemistry campaigns to treat atherosclerosis driven cardiovascular disease.^[@ref8]^ LXRα and LXRβ agonists increase RCT by induction of ATP binding cassette transporters ABCA1 and ABCG1,^[@ref9],[@ref10]^ which efflux cholesterol from cells to HDL particles, and transporters ABCG5 and ABCG8, which traffic cholesterol from liver to the feces and promote its excretion. The immune system effects of LXR receptors continue to be elucidated, and LXRs are involved in innate and acquired immunity processes.^[@ref3]^ LXRs regulate many additional pathways in lipid homeostasis and energy utilization, and as such LXR agonists have been suggested as therapeutic treatments for other diseases including several types of cancers,^[@ref11]^ skin conditions,^[@ref12]^ and heart failure.^[@ref13]^ While many LXR agonists have been reported including the well-studied TO-091317 (1)^[@ref14]^ and GW3965 (2),^[@ref15]^ a challenge in the field has been to develop agonists that maintain the positive effects described above while not causing increases in low-density lipoprotein cholesterol (LDL-C) and triglycerides (TG). The TG increases are primarily caused by LXR induction of the transcription factor sterol regulatory element binding protein 1c (SREBP1c) and the enzyme fatty acid synthase (FAS) in liver, leading to increased very low density lipoprotein (VLDL) production and secretion from liver,^[@ref14]^ as well as upregulation of hepatic angiopoietin-like protein 3 and down-regulation of apoA-V expression resulting in decreased lipolysis of circulating TG-rich lipoproteins.^[@ref16],[@ref17]^ Increased LDL-C has been reported after repeat dose treatment in hamsters and cynomolgus monkeys with LXR agonists.^[@ref18]^ The LDL-C increases may be caused by a combination of multiple mechanisms, including increased VLDL production, induction of cholesteryl ester transfer protein (CETP),^[@ref19]^ and induction of inducible degrader of LDL receptor (IDOL).^[@ref20]^ The lipid effects have been proposed to be driven by hepatic LXRα based on knockout mice,^[@ref21],[@ref22]^ influencing the field to optimize for LXRβ selectivity.^[@ref8]^ Several reports of LXR agonists with improved therapeutic windows compared to full pan-agonists have been described ([Figure [1](#fig1){ref-type="fig"}](#fig1){ref-type="fig"}). For instance, the agonist LXR-623 (3) has been reported to decrease LDL-C in cynomolgus monkeys, while inducing transporters such as ABCA1; however, the mechanism for the LDL-C decrease has not been reported.^[@ref23],[@ref24]^ Unfortunately neurological effects were observed in humans after 1 day of dosing 3.^[@ref25]^ Further supporting that efficacy can be achieved without lipid effects, AZ876 (4) was reported to reduce atherosclerosis plaques and improve heart failure outcomes in mice at doses that did not cause increased TGs.^[@ref26]^ We have previously reported that an improved therapeutic window can be achieved with agonist BMS-779788 (5).^[@ref27],[@ref28]^ ![Examples of LXR agonists reported in the literature.](ml-2016-00234b_0002){#fig1} Agonist 5 shows a preference for LXRβ in binding and functional assays and induces LXR target genes ABCA1 and ABCG1 in human whole blood with an EC~50~ value of 1.2 μM and 55% efficacy ([Table [1](#tbl1){ref-type="other"}](#tbl1){ref-type="other"}).^[@ref27]^ An improved lipid profile was observed in mouse and cynomolgus monkeys with 5 compared to TO-091317 (1), and agonist 5 had a good pharmacokinetic and safety profile so it was taken into human trials.^[@ref28]^ Key goals for optimization were to identify a molecule with lower LXRα agonist activity to minimize the TG effects, while improving the potency for on-target ABCA1 and ABCG1 induction in human whole blood. ###### SAR Optimization of Lead 5 ![](ml-2016-00234b_0004){#fx1} \# LXRβ /LXRα Binding K~i~, nM LXRβ EC~50~ nM (% eff) LXRα EC~50~ nM (% eff) ABCA1 HeLa EC~50~ nM (% eff) hWBA EC~50~ nM (% eff) ------- ------------------------------- ------------------------ ------------------------ ------------------------------ ------------------------ 5 14/68 250 220 33 1200 (72%) (38%) (50%) (55%) 6 11/16 170 99 18 380 (41%) (20%) (42%) (32%) 7 5/31 60 100 9 300 (50%) (39%) (54%) (72%) 8 13/74 170 110 12 300 (74%) (46%) (56%) (46%) 9 40/180 220 120 21 870 (93%) (47%) (68%) (47%) 10 10/53 72 76 8 57 (83%) (29%) (43%) (47%) Standard deviations are reported in the [Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf) when n \> 2. The imidazole agonists were prepared as reported previously^[@ref27]^ and as described in [Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf). Agonists were profiled in a suite of five LXR assays. The binding affinity was determined with full-length LXRα--RXRα and LXRβ--RXRα heterodimers.^[@ref29]^ Functional isoform activity was assessed using LXRα and LXRβ transactivation assays in CV-1 cells,^[@ref30]^ and in HeLa cells with endogenous LXRα and LXRβ receptors and an ABCA1 LXREx3 reporter. Compounds were tested for ABCA1 and ABCG1 induction in a human whole blood assay (hWBA). The hWBA potency was a key driver because we anticipated it would predict clinical efficacy. The hWBA EC~50~ value for ABCA1 gene induction is reported because we used that data to evaluate SAR; in general, ABCG1 EC~50~ values were within 2-fold and are reported for 5 and 15 in the [Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf). In pursuit of improving properties, several areas were explored simultaneously, including substitutions on the A and D rings ([Table [1](#tbl1){ref-type="other"}](#tbl1){ref-type="other"}). While many compounds were prepared with different R^1^ substitution patterns that had similar activities to 5 (structures and data not shown), the 2,6-dichloro substitution (6) achieved a 3-fold boost in hWBA potency with lower LXRα efficacy (20%) compared to 5 (38% LXRα efficacy). Substituted phenyl sulfones were prepared based on the LXRβ crystal structure obtained with agonist 5 that had a water channel where R^4^ is positioned.^[@ref27]^ Investigations to exploit this position led to the synthesis of the hydroxymethyl sulfones 7 and 8. Small gains in potency were observed in LXRα and LXRβ agonist assays, which translated to 300 nM EC~50~ values for hWBA ABCA1 induction. Since 7 was prepared first it was dosed to mice at 10 mg/kg, and good plasma exposure was observed ([Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf)); however, when dosed to cynomolgus monkeys, the observed clearance was higher than hepatic blood flow at 61 mL/min/kg ([Table [2](#tbl2){ref-type="other"}](#tbl2){ref-type="other"}). Cynomolgus monkeys were critical to compound progression because they were a key model to study the lipid effects. High clearance was also observed with other analogues containing this D-ring substitution (data not shown), so compounds were prepared to address the high clearance. ###### PK in Cynomolgus Monkeys after i.v. Dosing example 5 5 7 9 10 15 ---------------- ------- ------- ------- ------- -------- -------- dose (mg/kg) 1.0 0.2 0.25 0.2 0.2 3.0 Cl (mL/min/kg) 1.9 2.9 61 5.6 8.4 8.0 t~1/2~ (h) 7.4 8.9 0.8 5.5 5.6 12 n = 2 for 0.2 and 0.25 mg/kg doses, and n = 3 for 1 mg/kg doses. Standard deviations for 5 and 15 are in the [Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf). Anticipating that a secondary alcohol would have reduced clearance, 9 was prepared and found to show partial 47% activity at LXRα with an hWBA EC~50~ value of 870 nM. Analogue 10 was prepared with an adjacent electron withdrawing fluorine at R^5^ to try to slow metabolism of the hydroxymethyl R^4^ substituent. Compound 10 had LXRβ agonist potency of 72 nM with a dramatic improvement in hWBA potency to 57 nM (47% efficacy), which was 20-fold better than 5. While limited selectivity was observed in binding assays, 10 showed differential agonist activity with 29% LXRα efficacy and robust LXRβ 83% efficacy. Both 9 and 10 had improved clearance rates of 5.6 and 8.4 mL/min/kg in cynomolgus monkeys ([Table [2](#tbl2){ref-type="other"}](#tbl2){ref-type="other"}). Due to the significant potency improvement with 10 we progressed this series and did not test the closest comparator 8 in cynomolgus monkey PK to confirm the improved clearance was entirely due to the fluorine. With optimized D-ring substituents, we focused SAR exploration on the A and C aryl rings ([Table [3](#tbl3){ref-type="other"}](#tbl3){ref-type="other"}). Our goal was to identify the most potent analogues with LXRα efficacy ≤25% and robust LXRβ efficacy. The 2-fluoro substitution (11) at R^1^ had a similar profile to 10, although the % efficacy was modestly reduced across all agonist assays. Mirroring the observation with 6, the 2,6-dichloro analogue 12 had a 3--4-fold improvement in hWBA potency to 15 nM with limited 25% LXRα efficacy. Interestingly, the 2-Cl,3-F analogue 13 had very limited LXRα activity at 12% while maintaining a hWBA EC~50~ value of 76 nM. The R^3^ fluorine substitution (14) maintained similar activity assays compared to 10. When the fluorine R^3^ substitution was combined with the 2,6-diCl A-ring (15) very potent hWBA activity was observed with an EC~50~ value of 9 nM (26% efficacy). Although 15 has similar LXRα and LXRβ binding K~i~ values (19 and 12 nM, respectively), in agonist assays the compound achieved 88% efficacy toward LXRβ and only 20% efficacy toward LXRα compared to a full pan-agonist. When tested in antagonist mode, 15 was a potent LXRα antagonist with an IC~50~ value of 69 nM (83% inhibition); whereas no antagonism was observed in LXRβ assays up to 10,000 nM ([Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf)). Further SAR investigation with a 2-Cl,6-F R^1^ substitution pattern provided 16 with a hWBA potency of 41 nM (33%). Introduction of an R^3^ chlorine atom in 17 caused a decrease in efficacy in all four agonist assays with a potent hWBA EC~50~ value of 5 nM just above the limit of assay detection (16%). Analogue 18 with hydrogen at R^2^ had in vitro activity consistent with the gem-dimethyl analogue 12. The monomethyl analogue 19 was identified as a metabolite of 15, and upon synthesis the profile showed it to be a potent, partial LXR agonist as well. ###### Optimization of R^1^, R^2^, and R^3^ with (2-Fluoro-6-(methylsulfonyl)phenyl)methanol D-Ring ![](ml-2016-00234b_0005){#fx2} \# LXRβ/ LXRα/ Binding K~i~, nM LXRβ EC~50~ nM (%Eff) LXRα EC~50~ nM (%Eff) ABCA1 HeLa EC~50~ nM (%Eff) hWBA EC~50~ nM (%Eff) ------- -------------------------------- ----------------------- ----------------------- ----------------------------- ----------------------- 11 14/81 160 130 12 43 (68%) (13%) (15%) (28%) 12 6/38 42 30 3 15 (72%) (25%) (23%) (43%) 13 14/53 72 72 8 76 (68%) (12%) (16%) (34%) 14 18/9 50 57 2 46 (79%) (25%) (29%) (35%) 15 12/19 24 8 0.6 9 (88%) (20%) (29%) (26%) 16 14/70 20 11 1 41 (86%) (15%) (30%) (33%) 17 48/50 27 8 2 5 (51%) (6%) (12%) (16%) 18 11/75 140 69 5 42 (54%) (17%) (43%) (51%) 19 13/17 25 12 2 23 (67%) (18%) (9%) (17%) Standard deviations are reported in the [Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf) when n \> 2. During the characterization of our lead molecules the clinical single ascending dose PK results were available for 5. The human plasma t~1/2~ was 100--200 h, which was 5--10-fold longer than predicted by preclinical studies. Studies with clinical plasma samples showed high binding (i.e., \>99.9%) of 5 to plasma proteins. The volume of distribution in humans was at least 10× smaller than the projected values based on preclinical species, suggesting limited distribution of this molecule outside of systemic circulation. One hypothesis brought forward to explain the unexpected long half-life, limited volume of distribution and slow clearance in humans was tight binding to α-1-acid glycoprotein (α1 AGP), which was different between human and preclinical species (manuscript in preparation). A similar explanation has been proposed for the human PK of UCN-01 (7-hydroxystaurosporine).^[@ref31]^ Equilibrium dialysis with 5 demonstrated 99.9% binding to human α1 AGP. In contrast 5 has moderate binding of 97, 92, and 98% to human serum albumin (HSA) and rat and dog α1 AGP, respectively. Binding to human α1 AGP was measured for several agonists ([Table [4](#tbl4){ref-type="other"}](#tbl4){ref-type="other"}). Whereas the biphenyl sulfones 5 and 6 had \>99% binding to α1 AGP, introduction of the polar R^4^ hydroxymethyl group in 8, 10, 13, and 15 reduced α1 AGP binding, consistent with the hWBA potency improvement. Binding to HSA, a major plasma protein, did not differentiate the analogues with 94--98% bound. ###### Equilibrium Dialysis with Human α1 AGP and HSA example human α1 AGP (% bound) HSA (% bound) --------- ------------------------ --------------- 5 99.9 ± 0.0 97.2 ± 0.3 6 99.6 ± 0.2 97.9 ± 0.5 8 90.2 ± 4.2 96.1 ± 0.7 10 86.6 ± 0.9 95.5 ± 0.2 13 87.6 ± 1.3 94.4 ± 1.1 15 97.3 ± 0.1 96.6 ± 0.1 The average is reported with standard deviation (n = 3). The crystal structure of 15 complexed with the ligand binding domain of LXRβ has been determined to 2.4 Å resolution ([Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}). The complex crystallized in space group C2 with four independent subunits in the asymmetric unit, with subunits A and B forming a canonical dimer, as did C and D. Helix 12 from subunit A was bound in the coactivator binding pocket of subunit B. The binding mode of 15 is very similar to that of 5, and there do not appear to be any major changes to side chain positions. The sulfone interacts with the backbone Leu330 as was observed with 5. The hydroxymethyl interacts with Ser-278, Glu281, and a bound butane diol molecule from the crystallization solvent. The fluorine gives an improved shape complementarity to the pocket near the "D" ring of 15 compared to 5, likely providing some of the improved potency. The benzylic phenyl (A-ring) forms a pi-stacking interaction with Phe340. The second chlorine substituent causes the benzylic methyl groups to rotate compared to 5, improving the molecular shape complementarity to the LXR pocket. Whereas many LXR agonists have a direct interaction with His435 stabilizing helix 12^8^ that is important for LXR agonist activity, the carbinol hydroxyl group of 17 appears to interact with a water in the active site that looks to be positioned to H-bond with His435. While we are not able to give a definitive structural reason for the low LXRα agonist activity from the LXRβ structure, it is possible that the indirect interaction through a water molecule with His 435 provides some of the observed differences between LXRα and LXRβ agonist activity. ![LXRβ complexed with 15 to 2.4 Å resolution (PDB code: 5JY3).](ml-2016-00234b_0003){#fig2} Agonists 13 and 15 were nominated for further study because both compounds had robust LXRβ efficacy with low LXRα agonist efficacy (\<20%), which was anticipated to improve the separation of desired efficacy from TG and LDL-C effects. In addition, 15 was very potent in the hWBA. Analogues 13 and 15 were not active in 16 nuclear hormone receptor agonist assays (\>10 μM), except PXR with EC~50~ values of 3 μM (85% of full agonism) and 1 μM (108% of full agonism), respectively. When dosed in mice at 10 mg/kg, the C~max~ coverage was high compared to the hWBA potency ([Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf)). Given that LXR agonists could have deleterious effects in brain, as observed with LXR-623, the brain levels were measured and found to be low with 15 having a brain to plasma ratio of \<0.05. In cynomolgus monkeys, 13 and 15 displayed good bioavailability, moderate clearance rates, and 10--12 h plasma half-lives ([Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf)). While 15 was considered the lead compound due to exceptional hWBA potency coupled with low LXRα efficacy ([Table [3](#tbl3){ref-type="other"}](#tbl3){ref-type="other"}), both 13 and 15 were studied in cynomolgus monkeys for 14 days to investigate the ABCG1 dose response to the lipid effects compared to those of 1. The agonists showed robust induction of the RCT target gene ABCG1 in plasma at drug concentrations that were predicted by the cynomolgus monkey WBA potency (1 cynoWBA EC~50~ = 310 nM (100%); 13 cynoWBA EC~50~ = 52 nM (29%); 15 cynoWBA EC~50~ = 5 nM (32%)). ABCA1 had shown variable vehicle effects in multiple cynomolgus monkey studies, precluding its use as a pharmacodynamic biomarker. Both 13 and 15 had improved TG profiles compared to 1. Fourteen days of dosing 1 at 10 mg/kg (200 nM plasma concentration at 5 h) caused a 6-fold ABCG1 induction in blood cells with TGs elevated 140% over baseline values (p \< 0.05, ANOVA). After 14 days, the 1 and 3 mg/kg doses of 13 afforded 4- and 10-fold ABCG1 induction in blood cells with 85 and 310 nM plasma exposures, respectively. These doses yielded TGs of 2% and 58% above baseline (not significant). Comparatively the 0.1, 0.3, and 1 mg/kg doses of 15 provided 5 h plasma exposures of 7.5, 22, and 57 nM with 4.7-, 15-, and 11-fold ABCG1 induction on day 14. The TGs were elevated nonsignificantly 20, 8, and 10% over baseline, respectively. As anticipated, 15 provided robust ABCG1 induction at very low plasma drug concentrations, with little effect on plasma TGs. A full data set from this cynomolgus monkey study 15 is reported elsewhere.^[@ref32]^ In summary, we have identified a potent biphenyl imidazole series of LXR partial agonists containing a (2-fluoro-6-(methylsulfonyl)phenyl)methanol substituent. Importantly, agonist 15 induces ABCA1 and ABCG1 RCT targets in human whole blood at nanomolar drug exposures with robust LXRβ agonism and limited LXRα agonist activity ([Table [3](#tbl3){ref-type="other"}](#tbl3){ref-type="other"} and [Supporting Information](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf)). In cynomolgus monkeys this profile gave robust blood cell activity with limited elevations of TGs. Based on coupling the in vitro properties with an excellent pharmacokinetic profile and favorable lipid profile in cynomolgus monkeys, 15 (BMS-852927) advanced into clinical studies. In human trials the PK was well predicted by preclinical data; however, TGs and LDL-C were observed to be elevated after multiple days of dosing with a limited therapeutic window, indicating that the low LXRα efficacy was not sufficient to protect from deleterious lipid elevations.^[@ref32]^ The Supporting Information is available free of charge on the [ACS Publications website](http://pubs.acs.org) at DOI: [10.1021/acsmedchemlett.6b00234](http://pubs.acs.org/doi/abs/10.1021/acsmedchemlett.6b00234).LXR assay data with standard deviations, mouse and cyno PK data for 5, 7, 13, and 15, antagonist and WBA data for 5 and 15, synthesis methods, compound characterization, biology assays, crystallization protocol, and in vivo study protocols ([PDF](http://pubs.acs.org/doi/suppl/10.1021/acsmedchemlett.6b00234/suppl_file/ml6b00234_si_001.pdf)) Supplementary Material ====================== ###### ml6b00234_si_001.pdf ^§^ These authors contributed equally to this work. Bristol-Myers Squibb and Exelixis funded the described work. The authors declare no competing financial interest. We thank Akbar Nayeem for preparing [Figure [2](#fig2){ref-type="fig"}](#fig2){ref-type="fig"}; Amy Gallagher, Chris Bonagura, and Hangjun Zhan for LXRβ crystallization efforts; Bang-Chi Chen for help with large scale synthesis of 15; and Petia Shipkova, Robert Langish, Michael Witkus, Yingru Zhang, Michael Hicks, Cindy Li, and Rich Dalterio for analytical characterization of 13 and 15. α1 AGP : α-1-acid glycoprotein ABCA1, ABCG1, ABCG5, or ABCG8 : ATP binding cassette transporter apoA : apolipoprotein A CHD : coronary heart disease cyno : cynomolgus monkey HDL : high density lipoprotein HSA : human serum albumin hWBA : human whole blood assay LDL-C : low-density lipoprotein cholesterol LXR : liver X receptor PAMPA : parallel artificial membrane permeability assay PK : pharmacokinetics PXR : pregnane X receptor RCT : reverse cholesterol transport RXR : retinoid X receptor SAR : structure--activity relationship TG : triglygerides VLDL : very low density lipoprotein	{ "pile_set_name": "PubMed Central" }
GENOME ANNOUNCEMENT {#s1} =================== Pseudomonas oceani is a Gram-negative bacillus isolated from deep seawater ([@B1]). Pseudomonas oceani belongs to the Pseudomonas pertucinogena group ([@B2]). This group is comprised primarily of marine and aquatic strains. The taxonomically closest strains to P. oceani, P. aestusnigri ([@B3], [@B4]) and P. pachastrellae ([@B5]), are from marine isolates and are characterized by the smallest genomes in the Pseudomonas genus. The whole-genome shotgun sequence of P. oceani was performed on an Illumina platform in a combination of 300-bp paired-end reads. The Newbler Assembler version 2.7 software package (Roche) was used for de novo genome assembly. The draft genome size is 4,155,021 bp and contains 64 contigs, with an average contig length of 62,983 kb, a median coverage depth of 70×, and an average G+C content of 59.94 mol%. Prediction and annotation of the genome was performed using the NCBI Prokaryotic Genome Annotation Pipeline (<https://www.ncbi.nlm.nih.gov/genome/annotation_prok>). Analysis and comparison of the functional annotation was done using the KEGG Automatic Annotation Server (KAAS) ([@B6]). The genome has 3,849 genes, with a total of 3,790 coding sequences, 52 tRNA sequences, and 1 rRNA sequence identiﬁed in the chromosome. Flagellation and twitching motility genes have been found. Genes encoding type II and VI secretion systems were localized in the genome. Few metabolic characteristics are present in this group of species. The utilization of putrescine, a differentiating characteristic between P. oceani DSM 100277^T^ and P. pachastrellae CCUG 46540^T^ or P. aestusnigri CECT 8317^T^, is corroborated by the presence of the genes for transporters of mineral and organic ions, such as iron III, sulfate, molybdate, and putrescine. Other genes present are related to phosphate, phosphonate, and urea transport, together with those for transport of iron complex siderophores, microcin C, and zinc. The genes for the multidrug-resistance efflux pumps MdtABC and MexGHI-OpmD have been annotated. Glycolysis, Entner-Doudoroff, and pentose phosphate pathways for sugar assimilation are present. Genes encoding alkane 1-monooxygenase and rubredoxin are related to fatty acid degradation. Putative nitrate and nitrite reductase genes for the conversion of nitrate to ammonia have been annotated, but the original phenotypic test was given as negative ([@B1]). EnvZ-OmpR two-component signal transduction system genes involved in osmotic stress response are present, together with a gene encoding ectoine synthase. P. oceani has been annotated for 11 transposases, 4 of them belonging to the IS3 family. A cluster of 24 phage-related genes is present. Accession number(s). {#s1.1} -------------------- This whole-genome shotgun project has been deposited at DDBJ/ENA/GenBank under the accession number [PPSK00000000](https://www.ncbi.nlm.nih.gov/nuccore/PPSK00000000). The version described in this paper is version PPSK01000000. Citation García-Valdés E, Gomila M, Mulet M, Lalucat J. 2018. Draft genome sequence of Pseudomonas oceani DSM 100277^T^, a deep-sea bacterium. Genome Announc 6:e00254-18. <https://doi.org/10.1128/genomeA.00254-18>. This work was supported by project CGL2015-70925-P from the Spanish MINECO, with cofunding from Fondo Europeo de Desarrollo Regional (FEDER).	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== The Human Microbiome Project revealed that bacterial cells account for \~3% of total human body weight and are at an equal level in number to human somatic cells. While the bacterial communities in the human body contribute to health, their imbalance predisposes to a wide variety of diseases \[[@B1]\]. Lactobacillus species are classically known to dominate the vaginal cavity in premenopausal women \[[@B2]\]. Vaginal Lactobacilli play a role in the maintenance and homeostasis of the local microbial milieu by dropping pH through production of lactic acid. Meanwhile, the human uterine cavity has been long believed to be germfree. Recent studies, however, proved the presence of a microbiota in the uterine cavity, which is also characterized by Lactobacillus-dominant composition \[[@B3]--[@B5]\]. Moreover, it was demonstrated that the status of the Lactobacillus-dominant (90% or more) microbiota in the endometrial fluid (EF) was favorable for embryo implantation in the subsequent in vitro fertilization-embryo transfer (IVF-ET) treatment in infertile women. On the contrary, non-Lactobacillus-dominant microbiota is associated with a poor reproductive outcome including implantation failure and miscarriage \[[@B6]\], supporting the idea that endometrial microbial composition is a key determinant for a successful embryo implantation process. Repeated implantation failure (RIF) is an infertile condition recognized as serial failed conception following three or more transfer cycles with good-quality embryos \[[@B7]\]. RIF occurs in 15-20% of infertile couples undergoing an IVF-ET program \[[@B8]\]. RIF potentially originates in aberrant embryonic factors (such as chromosomal abnormalities, mitochondrial DNA quantity, and oxidative stress) \[[@B9]--[@B12]\], impaired endometrial receptivity (such as hydrosalpinx, endometrial polyps, distorted uterine cavity, and chronic endometritis (CE)) \[[@B13]--[@B16]\], and systemic factors (such as thrombophilic and immunological factors) \[[@B17], [@B18]\]. Despite the accumulating evidence that Lactobacillus species are essential for the integrity of both the vaginal and the uterine cavity environments, the relationship between the vaginal secretions (VS) microbiota and the EF counterpart within the same infertile individuals remains largely unknown. Using next-generation sequencing, we aimed to compare the diversity of the microbiota in the paired EF and VS samples and characterize their dysbiosis in patients with a history of RIF. 2. Materials and Methods {#sec2} ======================== 2.1. Subjects {#sec2.1} ------------- This was a preliminary analysis of an ongoing case-control study, which was approved by the Ethical Committee of the Institutional Review Board (Approval Number 2017-02) and registered on the University Hospital Medical Information Network-Clinical Trial Registration, Japan (UMIN000029449) on the 6th of October 2017. Under a given written informed consent, infertile patients with a history of RIF (RIF group, n = 28) and those undergoing the first IVF attempt (control group, n = 18) in high-volume centers (\>2000 oocyte-pickup cycles per year) were enrolled into the study. They had undergone infertility examinations including hysterosalpingogram, hysteroscopy, thyroid functions, and thrombophilic and immunological factors. According to Veeck\'s classification \[[@B19]\], morphologically good cleavage-stage embryos were defined as day 3, grade 1 or 2, seven-to-nine-cell embryos. According to Gardner\'s score \[[@B20]\], morphologically good blastocysts were defined as day 5 blastocysts with a score of 3BB or above. Serum human chorionic gonadotropin (hCG, Tosoh Co., Shunan, Japan) was measured on the eleventh day after transfer of day 3 embryos or on the ninth day after transfer of day 5 blastocysts. According to the manufacturer\'s instruction, a value less than 0.5 IU/L was regarded as a negative pregnancy test. RIF was defined as serial negative pregnancy tests following transfer of five or more morphologically good cleavage-stage embryos and/or blastocysts. 2.2. Sample Collection {#sec2.2} ---------------------- Endometrial biopsy samples were obtained in the proliferative phase (on days 6-12) of the menstrual cycle using a 3 mm width curette (Atom Medical, Tokyo, Japan). On days 6-8 after luteinizing hormone surge in the natural cycle, or hCG trigger in the oocyte-pickup cycle, or on day 5 following initiation of luteal support in the hormone replacement cycle, the paired EF and VS samples were obtained carefully avoiding contamination. In brief, the perineum was cleansed twice using sterilized cotton balls soaked in benzalkonium chloride solution. A bivalve speculum was inserted slowly into the vaginal cavity to visualize the uterine cervix sufficiently. The VS samples were obtained from the vaginal mucosa from all directions using an OMNIgene accessory swab (DNA Genotek Inc., Ottawa, ON, Canada) and solubilized into a collection tube containing stabilizing liquid (DNA Genotek Inc.). After removing the mucous, the vaginal cavity and cervix were cleaned twice using sterilized cotton balls soaked in benzalkonium chloride solution. A MedGyn Pipette IV (MedGyn Products Inc., Addison, IL, USA) was used for EF sample collection. Avoiding contact between the speculator and vaginal wall, a pipette was inserted slowly from the cervical os into the uterine cavity until it reached the fundus uteri. The EF samples were then carefully aspirated and soaked into another collection tube. 2.3. Histopathologic/Immunohistochemical Examinations for CE {#sec2.3} ------------------------------------------------------------ Endometrial biopsy samples were fixed overnight in 4% paraformaldehyde (in phosphate buffer, pH 7.3) and embedded in paraffin. The sections (4 μm thickness) on slide glasses were dewaxed in limonene (Falma Inc., Tokyo, Japan), rehydrated in a graded series of ethanol (in phosphate-buffered saline, pH 7.4), and subjected to microwave pretreatment in citrate buffer solution (pH 6.0) for 5 minutes for antigen retrieval and immersion in 3% hydrogen peroxide for 5 minutes for endogenous peroxidase activity blocking. After being washed, the sections were soaked in 10% fetal calf serum (SAFC Biosciences, Lenexa, KS, USA) for 10 minutes to minimize nonspecific antibody binding and incubated with the ready-to-use mouse monoclonal IgG antibody against human CD138 (a plasmacyte marker, B-A38; Nichirei, Tokyo, Japan) or control mouse IgG. After being washed three times, the immunoreactivity was developed using a LSAB kit (Dako, Kyoto, Japan). Following hematoxylin counterstaining, the sections were observed by an experienced gynecologic pathologist under a light microscope (400x magnification). Stromal CD138+ cells with a nucleic heterochromatin pattern were enumerated in 20 or more high-power fields. The endometrial stromal plasmacyte density index was calculated as the sum of the stromal CD138+ cell counts divided by the number of the high-power fields evaluated. CE was diagnosed as 0.25 or more ESPDI, as previously described \[[@B21]\]. 2.4. DNA Extraction and Sequencing {#sec2.4} ---------------------------------- Both the EF and VS samples were treated with proteinase K (Beckman Coulter Inc., Brea, CA, USA) containing 100 mg/mL lysozyme solution (Sigma-Aldrich, Darmstadt, Germany) and 100 mg/mL RNase A (Sigma-Aldrich). The genomic DNA was extracted using an Agencourt Genfind v2 Blood & Serum DNA Isolation Kit (Beckman Coulter Inc.). The double-stranded DNA concentration was quantified fluorometrically with a Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific Inc., Waltham, MA, USA). The variable region 4 (V4) hypervariable region of the bacterial 16S rRNA gene was amplified from the specimen DNA by using a modified primer pair, 515f (5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGGTGYCAGCMGCCGCGGTAA-3′) and 806rB (5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG-GGACTACNVGGGTWTCTAAT-3′), with Illumina Nextera XT (Illumina Inc., San Diego, CA, USA) adapter overhang sequences \[[@B22]\]. Polymerase chain reaction (PCR) was performed with 25 ng DNA, 200 μmol/L 4-deoxynucleotide triphosphates, 400 nmol/L of each primer, 2.5 U of FastStart HiFi polymerase, 4% of 20 mg/mL BSA, 0.5 mol/L betaine, and the appropriate buffer with MgCl~2~ supplied by the manufacturer (Sigma-Aldrich). Thermal cycling consisted of initial denaturation at 94°C for 2 minutes followed by 30 cycles of denaturation at 94°C for 20 seconds, annealing at 50°C for 30 seconds, extension at 72°C for 1 minute, and final extension at 72°C for 5 minutes. The amplicon mixture was purified using Agencourt AMPure XP (Beckman Coulter Inc.). Purified PCR samples were multiplexed using a dual-index approach with the Nextera XT Index Kit v2 according to the Illumina 16S Metagenomic Sequencing Library Preparation protocol. The indexing PCR was performed with a KAPA HiFi HotStart ReadyMix (Kapa Biosystems, Boston, MA, USA) in a 50 μL reaction volume, and purification was then performed with Agencourt AMPure XP beads. The final library was paired-end sequenced at 2 × 200 bp using a MiSeq Reagent Kit v3 on the Illumina MiSeq platform. The ZymoBIOMICS Microbial Community Standard (Zymo Research, Orange, CA, USA) containing a mixture of Pseudomonas, Escherichia, Salmonella, Lactobacillus, Enterococcus, Listeria, Bacillus, and two yeast species Saccharomyces and Cryptococcus was used as a positive control. UltraPure™ DNase/RNase-Free Distilled Water (Thermo Fisher Scientific Inc.) was used as a blank control. Using EA-Utils fastq-join \[[@B23]\], a median 291-base pair merged sequence length was obtained. The quality control of the merged sequence was performed using USEARCH v10.0.240 \[[@B24]\] to remove PhiX reads, truncate primer-binding sequences, and discard sequences with \<100 bp length and sequence quality \< Q20. Quantitative Insights Into Microbial Ecology (QIIME) 1.9.1 \[[@B25]\] was used with default parameters for quality filtering, chimera check, clustering sequences into operational taxonomic units (OTUs), and assignment of taxonomy. The sequences were clustered into OTUs by open-reference OTU picking strategy using the UCLUST method based on 97% sequence identity. Taxonomy was assigned to each OTU using the Ribosomal Database Project Classifier \[[@B26]\] with a 0.50 confidence threshold against the Greengenes database version 13_8 \[[@B27]\]. The following 15 bacterial taxa (Acidovorax, Acinetobacter, Chryseobacterium, Citrobacter, Elizabethkingia, Escherichia, Flavobacterium, Janthinobacterium, Leptothrix, Methylobacterium, Pseudomonas, Rhodococcus, Sphingomonas, Stenotrophomonas, and Yersinia), which are known as contaminants found in a blank control \[[@B28]--[@B30]\], were excluded from ES samples using QIIME. 2.5. Statistics {#sec2.5} --------------- α-Diversity including the Shannon index, Chao1 richness, and observed species were calculated at the 1,000-th sequence in QIIME. The unweighted and weighted UniFrac distances were used to inspect the phylogenetic-based β-diversity and principal coordinate analysis plot based on rarified sequences for 1,000 \[[@B31]\]. The plots of α- and β-diversity were generated in QIIME, and β-diversity between the groups was compared using the permutational multivariate analysis of variance (PERMANOVA) test. Pearson\'s correlation analysis was applied for comparison between the EF and the VS microbiota within the same individual. Fisher\'s exact test was conducted to compare taxon-relative abundances between the control and the RIF group. A p value less than 0.05 was regarded as statistically significant. 3. Results {#sec3} ========== 3.1. Characteristics of Infertile Patients {#sec3.1} ------------------------------------------ The demographics of the infertile patients enrolled were summarized in [Table 1](#tab1){ref-type="table"}. All the patients were from the Japanese population. There were no cigarette smokers and obese women (body mass index \> 30) in both groups. CE was detected in 6 of 28 (21.4%) patients in the RIF group and 2 of 18 (11.1%) patients in the control group. The prevalence of CE was higher in the RIF group than in the control group but did not reach a significant level (p = 0.38, relative risk 1.93, 95% CI 0.43 to 8.53). The association between CE and specific EF/VS microbiota was not found in this small cohort setting. 3.2. Sequencing Result of EF and VS Samples {#sec3.2} ------------------------------------------- The paired EF and VS samples were obtained from the RIF group (n = 28; 17 in the natural cycle, 4 in the oocyte-pickup cycle, and 7 in the hormone replacement cycle) and the control group (n = 18; 8 in the natural cycle, 4 in the oocyte-pickup cycle, and 6 in the hormone replacement cycle) and were subjected to sequencing. A total of 12,570,533 sequence reads were obtained with a mean 106,308 reads per sample (range, 5,969--297,391) in EF and 166,965 reads per sample (range, 60,484--535,057) in VS. The microbiota obtained in EF was a mean 26,725 OTU-assigned sequences per sample (range, 1,065--43,657), whereas the microbiota obtained in VS was a mean 37,712 OTU-assigned sequences (range, 18,232--43,936). 3.3. Comparison of EF versus VS Microbiota in Infertile Patients {#sec3.3} ---------------------------------------------------------------- Both the EF and the VS microbiota were highly correlated within the same infertile individual (average Pearson correlation coefficient for all subjects, r = 0.952). α-Rarefaction analysis demonstrated that the Shannon index was highly stable above 100 sequences, indicating that enough sequencing was conducted to analyze the diversity of both the EF and the VS microbiota ([Figure 1](#fig1){ref-type="fig"}). Assessment of the Shannon index revealed that the EF microbiota (mean ± SE, 1.104 ± 0.777) was more diverse (p = 0.020) than the VS microbiota (mean ± SE, 0.768 ± 0.540) at 1,000 reads in infertile patients. In addition, the number of the bacterial species observed in the EF microbiota (mean ± SE, 11.950 ± 5.262) was significantly higher (p \< 0.0001) compared to that in the VS microbiota (mean ± SE, 7.091 ± 2.865) in both groups ([Figure 1](#fig1){ref-type="fig"}). Richness of bacterial community measured by Chao1 richness was higher (p \< 0.001) in the EF microbiota (15.330 ± 6.214) compared with the VS microbiota (8.550 ± 3.494). Analysis of variance to partition UniFrac distance matrices between EF and VS revealed significantly different clustering (p = 0.001) ([Figure 2](#fig2){ref-type="fig"}). 3.4. Comparison of EF and VS Microbiota between the RIF Group and the Control Group {#sec3.4} ----------------------------------------------------------------------------------- The Shannon index of the EF microbiota in the RIF group (mean ± SE, 0.893 ± 0.567) was significantly lower (p = 0.02) than that in the control group (mean ± SE, 1.431 ± 0.931) ([Figure 3](#fig3){ref-type="fig"}). The Shannon index of the VS microbiota in the RIF group (mean ± SE, 0.654 ± 0.431) was comparable (p = 0.07) to that in the control group (mean ± SE, 0.946 ± 0.637). The unweighted UniFrac distance of the bacterial community in the EF microbiota showed a significant difference between the RIF group and the control group (p = 0.0089). Meanwhile, the unweighted UniFrac distance of the bacterial community in the VS microbiota was similar between the two groups (p = 0.38) ([Figure 4](#fig4){ref-type="fig"}). 3.5. Comparison of Bacterial Species in EF and VS Microbiota between the RIF Group and the Control Group {#sec3.5} -------------------------------------------------------------------------------------------------------- Lactobacillus-dominated EF microbiota, defined by \>90% Lactobacillus genus status, was observed at a higher rate in the RIF group (64.3%, 18/28) than in the control group (38.9%, 7/18), although the difference did not reach a significant level (p = 0.13, odds ratio 2.83, 95% CI 0.83-9.61) ([Figure 5](#fig5){ref-type="fig"}). Similar results were obtained from the VS microbiota where 67.9% (19/28) in the RIF group and 44.4% (8/18) in the control group represented Lactobacillus-dominated microbiota (p = 0.14, odds ratio 2.64, 95% CI 0.78-8.96). The detection rate of Gardnerella in the EF microbiota was 39.3% (11/28) in the RIF group and 27.7% (5/18) in the control group (p = 0.53, odds ratio 1.68, 95% CI 0.47-6.05). Burkholderia was not detected in any of the EF microbiota in the control group (0/18) but was detectable in 25% (7/28) of the RIF group (p = 0.032, odds ratio 12.91). There were no significant differences in the detection rate of the specific bacterial species in the VS microbiota between the two groups. 4. Discussion {#sec4} ============= Several investigators evaluated the bacterial communities in the endometrium and vagina using microbiota in infertile women with various causes \[[@B4], [@B6], [@B32]--[@B37]\]. To our best knowledge, this is the first study investigating the microbiota in the paired EF and VS samples in infertile patients with a history of RIF. We demonstrate that the bacterial species in EF and VS are similar within the same individual. However, the diversity measurements such as the Shannon index, observed species, and Chao1 richness indicate that EF has a higher α-diversity than VS. This finding was supported by the analysis of the UniFrac distance, which demonstrated that the bacterial communities were fairly different between EF and VS. Although the possible impact of the endometrial biopsy procedure on the subsequent EF/VS sample status cannot be fully denied, we found reassuring results that there was no endometrial thinning or hemorrhage on the day of the EF/VS aspiration. Interestingly, Burkholderia was not detectable in infertile women undergoing the first IVF-ET attempt but in a quarter of those with a history of RIF. Burkholderia is a genus of Proteobacteria, of which members include Burkholderia pseudomallei, a microorganism responsible for melioidosis \[[@B38]\], and Burkholderia cepacia, a pathogen causing serious pulmonary infections in patients with cystic fibrosis \[[@B39]\]. Burkholderia is usually resistant to multiple antibiotics \[[@B40]\]. The literature on this bacterium in the human female reproductive tract is scant. While some studies demonstrated that Burkholderia species are the common environmental contaminants which are frequently detectable in the uterine cavity of levonorgestrel intrauterine contraceptive system users \[[@B41]\], a case report suggests that Burkholderia may be one of the potential pathogens causing tuboovarian abscess \[[@B42]\]. The impact of Burkholderia on endometrial receptivity awaits further study. The human endometrium is regulated throughout the menstrual cycle under the influence of ovarian steroids. Previous reports found that the endometrial microbiota profiles were stable across the menstrual cycle, between the menstrual cycles, and during the shift from the prereceptive phase (LH+2) to the receptive phase (LH+7) in most women. Some fluctuation, however, was seen in the acquisition of the endometrial receptivity in a fraction (4 of 22) of the subjects \[[@B6], [@B33]\]. The strength of our study is that we fixed the endometrial sampling period to the window of implantation (on days 6-8 after natural luteinizing hormone surge or hCG trigger or on day 5 following initiation of luteal support in the hormone replacement cycle). While the proportion of the pathogens was at a similar level between the paired EF and VS samples, there was a marked variance between the individuals. One potential confounding factor for this variance is inclusion of three different types of the cycles (natural, hCG-triggered, and hormone replacement cycles). The limitation of this research is that the study design is cross-sectional. Given that the control group (infertile patients undergoing the first IVF-ET attempt) may include some prospective RIF cohort, longitudinal investigations are required. A potential bias is the contamination of the endocervical secretions and VS in the process of EF sampling \[[@B3], [@B4]\], although the EF microbiota was not suspected to be brought from the VS microbiota as some differences in the bacterial community were noted between the EF and VS samples within the same individuals. An association between CE and EF/VS microbiota is anticipated in infertile women with RIF, but we were unable to find it in this small sample size. Larger studies are required to evaluate the relationship between the female reproductive tract microbiota and CE. 5. Conclusions {#sec5} ============== To the best of our knowledge, this is the first study investigating the microbiota in the paired EF and VS samples in infertile women with a history of RIF, along with those undergoing the first IVF-ET attempt. This work will facilitate the understanding of the microbial etiology in the female reproductive tract of the infertile patients with RIF. Data Availability ================= The data used to support the findings of this study are included within the article. Disclosure ========== This work was partly presented at the 74th American Society for Reproductive Medicine Annual Meeting, Denver, Colorado, USA. Conflicts of Interest ===================== There are no conflicts of interest to declare regarding the publication of this article. Authors\' Contributions ======================= Kotaro Kitaya and Yoko Nagai equally contributed to the study. ![Calculation of α-diversity values for comparison of bacterial communities between EF and VS. (a) Rarefaction analysis of sequences per sample in EF and VS. Comparison of Shannon index (b), mean number of observed species (c), and Chao1 richness (d) between EF and VS. Each graph represents mean (column) and SE (bars). ^∗^p \< 0.05 and ^∗∗^p \< 0.01 by two-tailed t-test.](MI2019-4893437.001){#fig1} ![Principal coordinate analysis plotting of EF and VS microbiota in whole samples (n = 46). The plots were generated using weighted (a) and unweighted (b) UniFrac distance metrics.](MI2019-4893437.002){#fig2} ![Comparison of α-diversity values in the control (n = 18) and RIF (n = 28) groups. Shannon index (a, c) and Chao1 richness (b, d) of EF (a, b) and VS (c, d) microbiota. Each graph represents mean (column) and SE (bars). ^∗^p \< 0.05 by two-tailed t-test.](MI2019-4893437.003){#fig3} ![Principal coordinate analysis plotting of unweighted UniFrac distance metrics comparing EF (a) and VS (b) microbiota between the control (n = 18) and RIF (n = 28) groups.](MI2019-4893437.004){#fig4} ![Heatmap representing dominant bacterial genera found in the EF (a) and VS (b) in the RIF and control groups. The rows show bacterial genera in alphabetical order, and the columns represent subjects. For each subject, the dominant genera are shown in red. Bacteria with a total rate less than 0.1% are not shown.](MI2019-4893437.005){#fig5} ###### Demographics of infertile patients with the RIF and control groups. RIF group (n = 28) Control group (n = 18) ----------------------------------------------------- ---------------------- -------------------------- Age (years) (mean ± SD) 38.7 ± 3.2 37.6 ± 4.2 Body mass index (kg/m^2^) (mean ± SD) 21.8 ± 1.7 22.4 ± 2.1 Gravidity (median (range)) 0 (0-4) 0 (0-3) Parity (median (range)) 0 (0-1) 0 (0-1) Infertility diagnosis^a^ Male factor 8 (28.6%) 6 (33.3%) Polycystic ovarian syndrome 9 (32.1%) 4 (22.2%) Endometriosis 5 (17.9%) 4 (22.2%) Tubal factor 4 (14.3%) 5 (27.8%) Unexplained 9 (32.1%) 3 (16.7 %) Diminished ovarian reserve 1 (3.6%) 0 (0%) Controlled ovarian stimulation protocol^b^ Short GnRH agonist cycle 31 (76.1%) --- Long GnRH agonist cycle 5 (0.7%) --- Ultralong GnRH agonist cycle 1 (0%) --- Flexible GnRH antagonist cycle 41 (27.5%) --- Mild stimulation cycle 8 (0.7%) --- Natural cycle 2 (0%) --- Past embryo transfer history (mean ± SD) Number of cycles 5.5 ± 0.4 --- Number of embryos transferred 8.1 ± 0.6 --- Number of morphologically good embryos transferred 5.6 ± 0.6 --- Number of assisted hatching use 5.7 ± 0.8 --- Number of hyaluronan-rich medium use 3.4 ± 0.5 --- Footnotes: ^a^Totals are not 100 percent due to some patients having more than one diagnosis. ^b^Totals are not 100 percent due to some patients undergoing more than one controlled ovarian stimulation/oocyte-pickup cycle. Abbreviations: RIF: repeated implantation failure; SD: standard deviation; GnRH: gonadotropin-releasing hormone. [^1]: Academic Editor: Emmanuel Economou	{ "pile_set_name": "PubMed Central" }
All relevant data are within the manuscript and its Supporting Information files. Introduction {#sec001} ============ In vivo micro-computed tomography (micro-CT) is an efficient tool for the non-destructive evaluation of laboratory animals and the in vivo tracking of longitudinal changes in bone mass and bone microstructure due to disease and/or bone adaptation processes \[[@pone.0207323.ref001], [@pone.0207323.ref002]\]. Micro-CT has emerged as an advancement from the simple X-ray imaging into an essential technique, which is now used for laboratory research, tissue engineering, and numerical modeling \[[@pone.0207323.ref003]--[@pone.0207323.ref006]\]. Micro-CT can be used to longitudinally monitor bone micro-architecture in growing animals at different developmental stages. It can provide animal specific high-resolution data of time-related changes in desired bone locations. Changes can result from pathological or therapeutic stimuli, assuming minimal or no effects of the micro-CT scanning radiations on the radiated bone structural system \[[@pone.0207323.ref007]--[@pone.0207323.ref009]\]. However, as the micro-CT system might impose relatively high ionizing radiation doses \[[@pone.0207323.ref010], [@pone.0207323.ref011]\], frequent or recurrent exposures to such doses of the scanned bony parts could induce some side effects, including growth hindrance, deformities of the skeleton, bone loss or other hematological abnormalities \[[@pone.0207323.ref007], [@pone.0207323.ref008], [@pone.0207323.ref012]--[@pone.0207323.ref014]\]. High-radiation doses scans provide better image sets, which further facilitate the assessment of trabecular and cortical bone structures with higher accuracy \[[@pone.0207323.ref009], [@pone.0207323.ref015], [@pone.0207323.ref016]\]. However, this dose increment might pose a risk to the normal bone development process. Bone tissue damage can occur with doses as low as 250 mGy \[[@pone.0207323.ref002], [@pone.0207323.ref017]\]. Cell death might occur due to the irreparable DNA damage resulting from excessive doses \[[@pone.0207323.ref009], [@pone.0207323.ref018]\]. Low radiation doses can also trigger the DNA damage checkpoint activation, which results in a decreased cell proliferation \[[@pone.0207323.ref019]\]. Hence, an effective approach must be established to acquire high-quality images while using minimal radiation exposure. This can be achieved by efficaciously optimizing the scanning parameters to produce a low radiation dose which will provide an acceptable image quality without affecting the bone tissue. Different studies use different approaches to investigate bone structure. Some studies need a single micro-CT scan whereas some need repeated CT scans. The impact of single radiation dose on longitudinal bone growth has been extensively investigated. Human long bones can exhibit swelling and fragmentation symptoms for doses ranging from 3--5 Gy \[[@pone.0207323.ref008]\]. Also, it has been reported that a radiation dose in the order of 5 Gy can affect the bone regeneration process while a dose limit of 2.5 Gy showed no such impacts \[[@pone.0207323.ref020]\]. A rabbit femur exposed to 3.5 Gy radiation dose showed a significant reduction in the growth of long bones \[[@pone.0207323.ref021]\], whereas no adverse effects were noticed for 400 mGy and lower radiation doses on the proliferation and differentiation of osteoblasts in adult Sprague-Dawley rats \[[@pone.0207323.ref022]\]. Repeated micro-CT measurements deemed to be necessary especially when tracking changes in bone development. Repeated measurements can provide valuable information on bone quality in post-surgical scenarios or in response to physical exercise or pharmaceutical treatment. However, repeated CT-scans can also cause a threat to the bone if it crosses a safe limit. Numerous animal studies have been performed to assess the impact of repeated micro-CT radiation doses on the whole body or the exposed limb. In a recent study \[[@pone.0207323.ref023]\], repeated (4 scans) doses effects of 1255 mGy and 453 mGy were investigated in adult mice (17 weeks old) femurs and no effects were found. In another study \[[@pone.0207323.ref024]\], adult Wister rats (30 weeks old) underwent 8 weeks in vivo scanning on their right tibia using doses as high as 939 mGy per scan. Bone structural measurements remained unaffected under the applied scanning regime. Another study \[[@pone.0207323.ref025]\] used adult mice aged 12 week old (exposed to 845.9 mGy) and adult rats aged 8 months old (exposed to 596.6 mGy) and found a decrease in the trabecular bone volume fraction in the radiated tibiae compared to the control ones. Both single and repeated radiation studies demonstrate that various animal protocols showed divergent adaptability for the level of radiation doses applied. In addition to the difference in experimental protocols, variances in animal size, shape and anatomy, which put the skeleton under thicker or thinner skin, could be partly responsible for such differences in response to radiations. Hence, radiation results from one animal model and protocol could not be directly extrapolated to another. Moreover, most of the radiation doses related studies were performed on adult animal models, where the bone tissue has already peaked to its skeletal maturity. However, no such studies have been performed to define limit values below which radiation doses can be used safely for a growing animal model, in which bones have not reached their skeletal maturity. Thus, the objective of this study was to evaluate radiation effects on bone morphometry, bone marrow cells, bone growth rate and growth plate histomorphometry in growing tibiae for three radiation doses from repeated in vivo micro-CT scanning in adolescent rats. Results of this study will provide knowledge on weekly radiation doses protocol which can provide high-quality image sets to adequately investigate trabecular and cortical compartments, without causing damage to bone development during the rat adolescent growing period. The present study covered the rat adolescent period, which spans from the beginning of the 4^th^ week of age to the end of 12^th^ week period \[[@pone.0207323.ref026]\], resulting in a 9-week scanning period to investigate the radiation doses effects by comparing the irradiated and non-radiated limbs. The radiation dose of the first group was set at 0.83 Gy/scan, evaluated as the baseline to produce reasonable image quality for bone development investigation purpose. Two-fold (1.65 Gy/scan) and three-fold (2.47 Gy/scan) dose values were tested along with the same protocols for the second and third radiation groups. Materials and methods {#sec002} ===================== Animals {#sec003} ------- 21 days old male Sprague-Dawley rats (n = 33) were obtained from Charles River Laboratories, Montreal, Canada. Rats were randomly divided into three doses groups: 0.83 Gy, 1.65 Gy and 2.47 Gy (n = 11 per group). They were given 1-week of acclimatization before starting the experiment. Rats were housed two and three per cage (dimension 53 × 35.5 cm) at 25°C with a 12:12-hour light-dark cycle and provided with a standard laboratory diet and water ad libitum. Body weight was monitored weekly. The experimental protocol and all animal procedures were carried out in accordance with the guidelines of the Canadian Council on Animal Care (CCAC) and were approved by the Institutional Animal Care Committee at the Research Center of Sainte-Justine University Hospital, Montreal, Canada. Repeated micro-CT scanning {#sec004} -------------------------- A micro-CT scanner was used to perform nine weekly basis repeated CT scans of the proximal right tibia of the rats from their 4^th^ to 12^th^ weeks of age. A final scan was performed at the 14^th^ week. The two-week interval for the last scan was chosen to assess the maximal radiation exposure effect after the end of the exposure protocol \[[@pone.0207323.ref027], [@pone.0207323.ref028]\]. The imaging system was a Skyscan 1176 in-vivo micro-CT (Skyscan, N.V., Belgium) scanner with rotatable X-ray source and detector. Each rat was anesthetized (2% isoflurane, 1.0 L/min O~2~) and maintained on anesthetic gasses for the duration of the scanning. The rat was secured in the carbon fiber half-tube bed of the Skyscan 1176, and the right tibia was positioned into a Styrofoam holder of cylindrical shape. This procedure was performed to place the rat tibia in the scanning midline of the scanner and to eliminate any unwanted movement of the tibia during the radiation period. ([Fig 1](#pone.0207323.g001){ref-type="fig"}) \[[@pone.0207323.ref029]\]. The left tibia together with the tail were folded towards the animal's head and placed alongside the animal on the carbon fiber half-tube bed using masking tape. An ophthalmic gel was applied to the eyes of the rat during the entire scanning period to prevent dryness. The radiated tibia was subjected to X-rays solely, without irradiating the contralateral limb (left tibia). ![Rat positioning on the Skyscan 1176 scanner for in vivo scanning.\ The rat was placed sideways on the scanning bed while kept anesthetized (anesthesia mask not shown). This configuration was adapted to facilitate the positioning of the irradiated leg (right) into the iso-center of the scanning chamber. The right tibia was secured into a Styrofoam holder (1 cm thick) of cylindrical shape and firmly held with a medical adhesive tape. The non-radiated leg (left) was folded towards the animal's head and placed alongside the animal with its tail.](pone.0207323.g001){#pone.0207323.g001} For the first radiation group (0.83 Gy), all scans were performed on the anesthetized rats with an isotropic voxel size of 18μm. The choice of 18μm was made based on the previous findings \[[@pone.0207323.ref030], [@pone.0207323.ref031]\], which enables a reasonable high-quality image for the trabecular and cortical bone tissue investigation. An overview of image acquisition and reconstruction parameters for three radiation groups have been given in a tabular format ([Table 1](#pone.0207323.t001){ref-type="table"}). For 0.83 Gy group, the isotropic voxel size generated 1,336×1,680 CCD detector array. Total irradiation time was 5 min 34 sec and the scanning consisted of a stack of 304 images. For the second radiation group (1.65 Gy), image acquisition parameters were similar to the first group except for an improvement in the additional frame averaging (2 frame averaging versus 1 frame averaging) ([Table 1](#pone.0207323.t001){ref-type="table"}). This improvement resulted in a finer detector array compared to the previous one (1,336×2,000 CCD detector array versus 1,336×1,680 CCD detector array). Total irradiation time was 11 min 9 sec and the scanning consisted of a stack of 395 images. For the third radiation group (2.47 Gy), an isotropic voxel size of 9μm was chosen for acquiring high-quality image sets for assessing trabecular and cortical bone microarchitecture. Image acquisition parameters were similar to the first group ([Table 1](#pone.0207323.t001){ref-type="table"}). However, due to the improvement in isotropic voxel size (9μm versus 18μm), a finer detector array was generated compared to the first group (2,672×3,560 CCD detector array versus 1,336×1,680 CCD detector array). Total irradiation time was 16 min 39 sec and the scanning consisted of a stack of 304 images. 10.1371/journal.pone.0207323.t001 ###### Image acquisition and reconstruction parameters of the rat proximal tibiae for the three doses groups. ![](pone.0207323.t001){#pone.0207323.t001g} 0.83 Gy 1.65 Gy 2.47 Gy ------------------------------------ -------------- -------------- -------------- *Scanning parameters* Voxel size (μm) 18 18 9 Voltage (kV) 65 65 65 Current (μA) 385 385 385 Rotation step (over 180°) 0.65° 0.50° 0.65° Exposure time (ms) 350 350 1140 Frames averaged per projection 1 2 1 Filter AL 1mm AL 1mm AL 1mm Approximate scan time (min) 6 11 17 *Reconstruction Parameters* Filter Gaussian Gaussian Gaussian Smoothing kernel 1 1 1 Ring artifact reduction 4 4 4 Beam hardening correction (%) 10 10 10 Attenuation coefficient 0.000--0.049 0.000--0.049 0.000--0.049 *Analysis Parameters* Thresholding Global, 65 Global, 65 Global, 65 All micro-CT scans were obtained using the SkyScan 1176 model, Bruker-microCT. For all groups, the left tibia was used as a control and scanned only on the last (14^th^ week) scanning time point. Euthanasia of the rats was performed after the last scan (14^th^ weeks of age) using a CO~2~ chamber. For all rats, weight monitoring was conducted on a weekly basis to assess the impact of anesthesia and irradiation on rat development. The acquisition covered the proximal tibial section of the rat tibia. The delivered doses of 0.83 Gy, 1.65 Gy and 2.47 Gy computed tomography dose index (CTDI) were calculated based on the manufacturer specifications (Bruker micro-CT). The provided specifications (Bruker micro-CT) followed the dose measurements using a UNFORS PS-2 patient skin dosimeter. Shielding was provided with acrylic plastic (PMA) tubes of various wall thicknesses to simulate soft biological tissue. Local absorbed radiation dose rate (mGy/min) for tibia, femur, etc. have been provided by the manufacturer for different scan settings scenarios \[[@pone.0207323.ref032]\]. The data that accurately matched with our scanning parameters (65 kV, 385 μA, full x-ray, and 1-mm Al filter) have been extracted and local absorbed dose rate have been multiplied by the scanning time to get the resulted doses for our study \[[@pone.0207323.ref033]\] ([S3 File](#pone.0207323.s003){ref-type="supplementary-material"}). For an approximation, the tissue at all depths was assumed to be a cylinder and the dose rate of all tissue cylinder diameters averaged between the dose in the air (zero depth) and the dose at the cylinder center (half diameter) \[[@pone.0207323.ref032]\]. Scanned image sets were reconstructed by applying filtered back-projection algorithm (software NRecon, v.1.6.10, Skyscan, Kontich, Belgium) \[[@pone.0207323.ref029]\]. A total height of 10 mm cross-sectional images was reconstructed for every scanned set. The reconstruction started from the beginning of the knee joint and extended distally into the tibial diaphysis. The resolution of the processed images for first and second radiation groups was 1500 × 1500 pixels each, 17.48 μm isotropic voxel size, and the images were 8-bit in size (256 gray levels). The third radiation group produced images with 2700 × 2700 pixels each, 8.74 μm isotropic voxel size, and the images were 8-bit in size (256 gray levels). Calcein injections {#sec005} ------------------ For measurement of longitudinal bone growth rate, calcein was used to label the bone line on the surface of the tibia. Injections of calcein (Sigma-Aldrich, St. Louis, MO, USA), a fluorescent marker, were made intraperitoneally at a dosage of 15 mg/Kg \[[@pone.0207323.ref034]\]. Injections were done 5 and 2 days prior to euthanasia. Bone marrow cell assessment {#sec006} --------------------------- After CO~2~ asphyxiation, followed by decapitation, both tibiae were collected. Left (control) and right (irradiated) tibiae were sawed off to keep 10 mm on both proximal and distal sides using an ISOMET 1000 Precision Saw (Buehler, An ITW Company, Illinois, USA). To determine cell radiation damage, bone marrow cells were collected from both control and radiated tibiae. Bone marrow cells were flushed by applying pressure with a needle filled with HBSS (Hank\'s balanced saline solution) in the sawed part of the tibiae. A cell count was performed on the collected cell suspensions on HBSS with trypan blue (0.4% solution, Sigma-Aldrich, Oakville, ON, Canada). Using the trypan blue test, the number of living and dead cells and their corresponding percentages were determined for both control and radiated tibiae. From these values, percentage of unaffected bone marrow cells was calculated by dividing the total number of live cells by the number of total cells (live + dead). Tissue processing {#sec007} ----------------- Formalin solution (Anachemia, Montreal, QC, Canada) was used to fix the proximal sections (\~10 mm) from each tibia for a duration of 48h. Thereafter, graded alcohol solutions were used for dehydration, xylene was used for clarification and methylmethacrylate (MMA) (Fisher Scientific Canada, Nepean, ON, Canada) was used for embedding process \[[@pone.0207323.ref035]\]. When the polymerization was completed, a microtome (Leica SM2500) setup was used to cut the blocks of the tibiae into 6 μm sections. Only the proximal sections were used in this study. To cover the 40--50% of the growth plate depth, the tibiae were cut along the longitudinal bone axis for 36 slides, six series of six slides, which contain two sections per slide. To facilitate the growth rate measurements, the first slide of each series (6 slides, 12 sections total) per proximal tibia were set aside from light. A microscope (Leica DMR with Retina Qimaging Camera) was used for slice observation while using 5x magnification for growth rate measurements. Bone growth rate {#sec008} ---------------- The distance between two calcein labels was divided by the time interval (3 days) between the two applied injections to calculate the bone growth rate \[[@pone.0207323.ref036]\]. An in-house built Matlab program was used for this purpose. The distance was automatically calculated as the mean value of 100 segments parallel to the longitudinal growth direction with both calcein lines modeled as splines \[[@pone.0207323.ref034], [@pone.0207323.ref035]\] ([Fig 2A](#pone.0207323.g002){ref-type="fig"}). ![Bone growth rates (μm/day) measurements.\ (A) 5x magnified microscopic images of the tibial metaphysis labeled twice with calcein for representative irradiated and control tibiae from three doses groups (Ⅰ-ⅤⅠ). Bone growth (ΔX, μm) measured as the mean distance between the two calcein lines, which were modeled as splines and divided by the time interval (3 days) between the two applied injections. (B) Growth rates (μm/day) of rat proximal tibiae for 0.83, 1.65 and 2.47 Gy radiation groups (mean value ± SD). \: a significant difference (p* \< 0.05) between the control (left) and irradiated (right) tibiae for each radiation dose.](pone.0207323.g002){#pone.0207323.g002} Growth plate histomorphometry {#sec009} ----------------------------- Heights of the proliferative and hypertrophic zones, the hypertrophic cell height as well as the number of proliferative cells per column were measured for the histomorphometric analysis, similarly to previous work \[[@pone.0207323.ref034], [@pone.0207323.ref035]\] ([Fig 3A and 3B](#pone.0207323.g003){ref-type="fig"}). Hypertrophic cell height and the number of proliferative chondrocytes per column were measured as they are considered to be the indirect markers of bone growth \[[@pone.0207323.ref034], [@pone.0207323.ref035]\]. To measure heights, a similar approach to the bone growth rate measurements was implemented with 10x magnified image sets. Values from 100 segmental measurements were averaged for the assessment of zonal heights ([Fig 3A](#pone.0207323.g003){ref-type="fig"}). A 20x magnified image set was used to measure the hypertrophic cell height along the longitudinal growth direction ([Fig 3B](#pone.0207323.g003){ref-type="fig"}). The number of proliferative chondrocytes per column was measured from 20x magnified image sets for six random columns per growth plate ([Fig 3B](#pone.0207323.g003){ref-type="fig"}). For a single proximal tibial segment, histomorphometric parameters were measured by averaging 72 values, 6 values per section, 12 values per microscope slide with a six series repetition. ![Histomorphometry measurement.\ (A) Growth plate section embedded in MMA and stained with toluidine blue (10x). Evaluation of the hypertrophic and proliferative zonal thicknesses for three doses groups. (B) Growth plate section embedded in MMA and stained with toluidine blue (20x). Evaluation of the hypertrophic cell height and number of proliferative cells per column for three doses groups.](pone.0207323.g003){#pone.0207323.g003} Trabecular and cortical bone morphometry {#sec010} ---------------------------------------- A volume of interest (VOI) was defined for morphometric analysis from the reconstructed image sets. The VOI included the proximal metaphysis, covering both trabecular and cortical bony segments (software CT Analyzer v.1.13, Skyscan, Kontich, Belgium). The proximal metaphysis of the tibia contains the growth plate and is responsible for blood supply and vascular stasis in growing bone. This part is also very sensitive to radiation exposure compared to the other regions of the bone \[[@pone.0207323.ref037]\]. So, absence of radiation effects on this bony region could presumably be considered as to have no effects on the epiphysis and the metaphysis parts as well \[[@pone.0207323.ref024]\]. The VOI was selected as a percentage of the entire tibial length (L) to keep consistency with the growing tibial length from 4^th^ to 14^th^ week of age. To exclude the primary spongiosa, the VOI started at \~1mm distal to the growth plate and extended for 10% of the total tibial length (L) \[[@pone.0207323.ref038]\] ([Fig 4](#pone.0207323.g004){ref-type="fig"}). ![In vivo scanning of proximal tibia and bone segmentation process.\ (a) A representative 3D reconstructed tibia showing the total tibial length (L). (b) Scanned proximal tibial cross-section (10 mm in height) of the rat tibia. This representative image was acquired from a 17.48-μm pixel size scanning at 0.83 Gy radiation dose. VOI consisting trabecular and cortical bone, for morphometric parameters evaluation, beginning at \~1mm distal to the growth plate and extending for 10% of the total tibial length (L). Proximal (f) and distal (c) tibial sections are illustrated. The cortical (d, g) and trabecular (e, h) bone regions were segmented using a semi-automatic bone segmentation algorithm.](pone.0207323.g004){#pone.0207323.g004} An in-house algorithm was developed for semi-automatically segmenting the trabecular and cortical bone. The segmentation was done by delineating the periosteum and endosteum surface in a semi-automatic algorithm based approach \[[@pone.0207323.ref039], [@pone.0207323.ref040]\]. A global gray threshold value of 65 corresponding to an equivalent density of 0.413 g/cm^3^ of calcium hydroxyapatite (CaHA), was set for all the analysis \[[@pone.0207323.ref038], [@pone.0207323.ref039]\]. Morphometric analysis was performed using CTAn software v.1.13 for the selected VOI of trabecular bone to evaluate the following bone structural parameters: bone mineral density (BMD), bone volume fraction (BV/TV), connectivity density (Conn.Dn), trabecular number (Tb.N), trabecular thickness (Tb.Th), and trabecular spacing (Tb.Sp) \[[@pone.0207323.ref041]\]. Excluding the VOI of trabecular bone from the selected dataset, cortical bone VOI was also extracted. Cortical microarchitectural measurements, including tissue mineral density (TMD), cross-sectional area inside the periosteal envelope (Tt.Ar), cortical bone area (Ct.Ar), cortical thickness (Ct.Th), periosteum perimeter (Ps.Pm), endocortical perimeter (Ec.Pm), medullary area (Ma.Ar), and mean eccentricity (Ecc) were evaluated using the cortical bone VOI \[[@pone.0207323.ref041]\]. The morphometric measurement process was appraised for reproducibility test. To do so, five scans of the right tibia were acquired from a dead rat in different orientations. After the completion of each scan, the rat was completely removed from the scanner bed and repositioned again in a different orientation. The same micro-CT scanning, image reconstruction, VOI selection and morphometric analysis protocols as the ones used for the radiation effects experiment were used in this reproducibility evaluation. The coefficient of variation (CV) was then determined by the five scans. The resulting reproducibility was high, with CV found to be less than 2% for BV/TV, Ct. Th., Ec. Pm., and Ma. Ar., less than 3% for BMD, TMD, Tb.Th, Tb.N, Tt. Ar., Ps. Pm., and Ecc., and less than 4% for Ct. Ar., Tb.Sp, and Conn.Dn. Statistical analysis {#sec011} -------------------- Statistical analyses were performed using SPSS Statistics (v. 23, IBM). Comparisons were made at the 14^th^ week between the irradiated and non-radiated tibiae for impacts on bone marrow cells, bone growth rate, growth plate histomorphometry, and bone morphometry for each dose group \[[@pone.0207323.ref042]\]. ANOVA test (general linear model) was performed to determine time effects, radiation dose, and their interaction on body weight. A paired Student's t-test was performed for determining any significant differences in absolute and percentage numbers of viable cells, in average bone growth rates and in histomorphometric and bone structural parameters measured at the 14^th^ week for both irradiated and control tibiae. Moreover, structural properties of trabecular and cortical bone microstructure of the irradiated tibiae from three doses groups were statistically analyzed on 14^th^ week scanning data. A one-way ANOVA with Tukey's multiple comparisons was performed to assess the significant group difference and pairwise comparisons. For each group, the series mean value was used to replace any values which were missing due to the movement of rats during a scanning procedure or due to the reconstruction error. For all the groups, this missing value incident occurred a total of five times (once in the 0.83 Gy group at 8^th^ week of age, twice in the 1.65 Gy group at 6^th^ and 9^th^ week of age, and twice in the 2.47 Gy group at 7^th^ and 11^th^ week of age). Results were considered statistically significant for p \< 0.05. Results {#sec012} ======= Bone growth rate {#sec013} ---------------- The average bone growth rate measured at the 14^th^ week in irradiated tibiae for 1.65 and 2.47 Gy group resulted in growth rate reductions of 13.1% and 21.8% respectively with respect to the control tibiae. These reductions were statistically significant (p \< 0.05) ([Fig 2B](#pone.0207323.g002){ref-type="fig"}). No significant difference was observed for the bone growth rate in the 0.83 Gy group ([Fig 2B](#pone.0207323.g002){ref-type="fig"}). Growth plate histomorphometry {#sec014} ----------------------------- Significant differences were found in the zone thickness for both HZ and PZ in 1.65 and 2.47 Gy groups, whereas no significant difference was found for the 0.83 Gy group between the irradiated and control tibiae ([Fig 5A and 5B](#pone.0207323.g005){ref-type="fig"}). Hypertrophic cell heights and numbers of proliferative chondrocytes per column were found to be similar for irradiated and non-radiated tibiae for all three groups ([Fig 5C and 5D](#pone.0207323.g005){ref-type="fig"}). ![Histomorphometry measurements comparison for control and irradiated tibiae.\ (a-d) Growth plate histomorphometry measurements of rat proximal tibiae for 0.83, 1.65 and 2.47 Gy radiation groups (mean value ± SD). \: a significant difference (p* \< 0.05) between the control (left) and irradiated (right) tibiae for each radiation dose.](pone.0207323.g005){#pone.0207323.g005} ### Trabecular and cortical bone morphometry: Comparative analysis at the 14^th^ week {#sec015} The effect of repeated in vivo irradiation was assessed by comparing the repeatedly irradiated right tibiae to the singly irradiated left tibiae at the 14^th^ week of age. Morphometric parameters of both trabecular and cortical bones were compared within each group to assess the radiation effect. For the trabecular bone morphometry, 0.83 Gy group showed no significant difference between the irradiated tibiae and their contralateral controls ([Fig 6](#pone.0207323.g006){ref-type="fig"}). For both the 1.65 and 2.47 Gy group, a significant decrease in BMD, Tb.Th, Tb.N, and a significant increase for Tb.Sp was observed between the irradiated and control tibiae ([Fig 6](#pone.0207323.g006){ref-type="fig"}). Moreover, a significant decrease in BV/TV was also observed for the 2.47 Gy group ([Fig 6](#pone.0207323.g006){ref-type="fig"}). ![Mean values and standard deviations of the trabecular bone parameters for the left (hatched columns), and right tibia (black columns) at 14^th^ week of age (n = 11/group). \: a significant difference (p* \< 0.05) between the control (left) and irradiated (right) tibiae for each radiation dose.](pone.0207323.g006){#pone.0207323.g006} For the cortical bone morphometry, no differences were found between the irradiated and control tibiae at the 14th week of age for both 0.83 and 1.65 Gy group ([Fig 7](#pone.0207323.g007){ref-type="fig"}). However, irradiated tibiae resulted in lower Ct.Th compared to the controlled ones for the 2.47 Gy group ([Fig 7](#pone.0207323.g007){ref-type="fig"}). ![Mean values and standard deviations of the cortical bone parameters for the left (hatched columns), and right tibiae (black columns) at 14^th^ week of age (n = 11/group). \: a significant difference (p* \< 0.05) between the control (left) and irradiated (right) tibiae for each radiation dose.](pone.0207323.g007){#pone.0207323.g007} Trabecular and cortical bone morphometry: 9-week longitudinal comparative analysis {#sec016} ---------------------------------------------------------------------------------- Bone morphometric changes were assessed in the right proximal tibia during the entire adolescent period (from 4^th^ to 14^th^ week of age) for each rat. Trabecular bone parameters showed changes with rat development in the different groups. For the 0.83 Gy group, a significant increase (p \< 0.05) was observed for BMD, BV/TV, Tb.Th, and Tb.N from the 4^th^ to the 14^th^ week old period ([Table 2](#pone.0207323.t002){ref-type="table"}). However, a decrease was observed for Conn.Dn values within the same study period ([Table 2](#pone.0207323.t002){ref-type="table"}). A significant increase for Tb.Sp and a decrease for Tb.Th were observed for both 1.65 Gy and 2.47 Gy group ([Table 2](#pone.0207323.t002){ref-type="table"}). However, an increase in Tb.N for 1.65 Gy and a decrease in Conn.Dn and BMD were observed for 1.65 Gy and 2.47 Gy group respectively ([Table 2](#pone.0207323.t002){ref-type="table"}). A significant increase (p \< 0.05) was observed for Tt.Ar, Ct.Ar, Ps.Pm, Ec.Pm, Ma.Ar, and Ecc for all three groups ([Table 3](#pone.0207323.t003){ref-type="table"}). However, for TMD and Ct.Th values, a significant increase was only observed for 0.83 Gy and 1.65 Gy groups ([Table 3](#pone.0207323.t003){ref-type="table"}). 10.1371/journal.pone.0207323.t002 ###### Longitudinal assessment of trabecular microarchitecture of the right proximal tibial metaphysis in three doses groups of rats. ![](pone.0207323.t002){#pone.0207323.t002g} Trabecular structural properties Dose (Gy) Age (week) ------------------------------------ --------------- --------------- --------------- --------------- --------------- --------------- --------------- --------------- --------------- ----------------------- *BMD (g.cm^-3^)* 0.83 0.156 ± 0.035 0.163 ± 0.027 0.150 ± 0.024 0.171 ± 0.029 0.183 ± 0.04 0.172 ± 0.051 0.179 ± 0.044 0.185 ± 0.026 0.200 ± 0.031 0.205 ± 0.037 ^α^ 1.65 0.162 ± 0.028 0.173 ± 0.046 0.163 ± 0.038 0.179 ± 0.041 0.121 ± 0.031 0.131 ± 0.016 0.131 ± 0.02 0.111 ± 0.036 0.125 ± 0.035 0.139 ± 0.045 2.47 0.173 ± 0.015 0.192 ± 0.045 0.184 ± 0.029 0.186 ± 0.035 0.141 ± 0.051 0.14 ± 0.018 0.121 ± 0.051 0.121 ± 0.036 0.132 ± 0.034 0.128 ± 0.042 ^γ^ *BV/TV (%)* * * 0.83 18.78 ± 5.23 20.05 ± 4.12 21.89 ± 6.42 22.27 ± 7.13 20.72 ± 4.13 21.07 ± 3.12 19.94 ± 6.14 21.51 ± 7.12 22.78 ± 5.12 27.56 ± 7.21 ^α^ 1.65 20.55 ± 8.23 21.52 ± 7.4 22.72 ± 8.24 23.77 ± 5.13 19.74 ± 8.23 19.77 ± 5.13 19.34 ± 8.13 22.3 ± 6.12 21.46 ± 7.24 23.76 ± 11.23 2.47 22.47 ± 10.12 23.63 ± 8.23 20.12 ± 7.40 21.23 ± 6.40 18.75 ± 8.24 19.34 ± 7.23 18.84 ± 5.34 19.44 ± 6.14 17.76 ± 8.42 19.14 ± 9.23 *Tb.Th (mm)* * * 0.83 0.078 ± 0.012 0.083 ± 0.012 0.081 ± 0.012 0.086 ± 0.013 0.087 ± 0.012 0.079 ± 0.011 0.084 ± 0.012 0.085 ± 0.015 0.087 ± 0.015 0.107 ± 0.015 ^α^ 1.65 0.093 ± 0.014 0.094 ± 0.011 0.086 ± 0.010 0.078 ± 0.010 0.075 ± 0.010 0.074 ± 0.012 0.076 ± 0.010 0.072 ± 0.012 0.062 ± 0.013 0.055 ± 0.006 ^β^ 2.47 0.103 ± 0.011 0.105 ± 0.011 0.097 ± 0.012 0.089 ± 0.015 0.093 ± 0.011 0.094 ± 0.011 0.088 ± 0.012 0.078 ± 0.011 0.076 ± 0.011 0.067 ± 0.007 ^γ^ *Tb.N (mm^-1^)* * * 0.83 2.208 ± 0.457 2.104 ± 0.435 2.453 ± 0.508 2.578 ± 0.533 2.473 ± 0.482 2.447 ± 0.477 2.642 ± 0.515 2.577 ± 0.483 2.82 ± 0.529 2.946 ± 0.552 ^α^ 1.65 1.972 ± 0.379 1.863 ± 0.358 2.15 ± 0.414 2.12 ± 0.408 2.234 ± 0.414 2.163 ± 0.401 2.22 ± 0.411 2.41 ± 0.402 2.33 ± 0.388 2.601 ± 0.434 ^β^ 2.47 2.651 ± 0.496 2.55 ± 0.477 2.678 ± 0.501 2.489 ± 0.466 2.102 ± 0.359 1.961 ± 0.335 2.16 ± 0.369 2.1 ± 0.326 1.83 ± 0.284 2.072 ± 0.321 ^γ^ *Tb.Sp (mm)* * * 0.83 0.839 ± 0.187 0.817 ± 0.182 0.784 ± 0.174 0.763 ± 0.159 0.819 ± 0.17 0.71 ± 0.151 0.766 ± 0.163 0.751 ± 0.153 0.85 ± 0.173 0.794 ± 0.162 1.65 0.646 ± 0.133 0.612 ± 0.126 0.671 ± 0.138 0.726 ± 0.14 0.752 ± 0.146 0.781 ± 0.154 0.751 ± 0.148 0.795 ± 0.145 0.831 ± 0.152 0.855 ± 0.156 ^β^ 2.47 0.752 ± 0.131 0.723 ± 0.126 0.723 ± 0.126 0.711 ± 0.124 0.777 ± 0.135 0.795 ± 0.134 0.743 ± 0.125 0.773 ± 0.126 0.892 ± 0.145 0.929 ± 0.151 ^γ^ *Conn.Dn (mm^-3^)* * * 0.83 189.8 ± 32.5 195.5 ± 39.1 234.7 ± 46.9 188.4 ± 32.6 178.5 ± 30.9 189.3 ± 32.8 234.1 ± 34.5 173.2 ± 25.5 164.3 ± 24.2 157.9 ± 29.2 ^α^ 1.65 263.4 ± 40.9 258.2 ± 40.1 273.8 ± 42.5 295 ± 54.2 219.5 ± 40.3 232.6 ± 42.7 266.4 ± 45.2 145.3 ± 24.6 132.1 ± 22.4 144.7 ± 26.1 ^β^ 2.47 208.8 ± 30.3 220.2 ± 31.9 255.2 ± 37 210.2 ± 30.5 233.2 ± 43.2 189.2 ± 35.1 225.8 ± 41.8 205.2 ± 27.7 178.3 ± 24.1 186.6 ± 27.4 Values are expressed as Mean ± SD, n = 11/group. Within each dose (0.83, 1.65 and 2.47 Gy), different symbols (α, β, γ) denote statistical significance (**p \< 0.05*) from 4^th^ to 14^th^ week of age. Abbreviations: BMD, bone mineral density; BV/TV, bone volume fraction; Tb.Th, trabecular thickness; Tb.N, trabecular number; Tb.Sp, trabecular spacing; Conn.Dn, connectivity density. 10.1371/journal.pone.0207323.t003 ###### Longitudinal assessment of cortical microarchitecture of the right proximal tibial metaphysis in three doses groups of rats. ![](pone.0207323.t003){#pone.0207323.t003g} Cortical structural properties* Dose (Gy) Age (week) ---------------------------------- --------------- --------------- --------------- --------------- --------------- --------------- --------------- --------------- --------------- ----------------------- **TMD (gm.cm^-3^)** 0.83 0.773 ± 0.063 0.803 ± 0.065 0.828 ± 0.08 0.873 ± 0.085 0.824 ± 0.08 0.858 ± 0.084 0.831 ± 0.082 0.852 ± 0.084 0.884 ± 0.087 0.961 ± 0.078 ^α^ 1.65 0.743 ± 0.08 0.783 ± 0.085 0.782 ± 0.085 0.823 ± 0.078 0.852 ± 0.08 0.802 ± 0.076 0.783 ± 0.074 0.883 ± 0.102 0.902 ± 0.104 1.043 ± 0.101 ^β^ 2.47 0.718 ± 0.069 0.744 ± 0.072 0.752 ± 0.069 0.783 ± 0.071 0.802 ± 0.073 0.783 ± 0.09 0.802 ± 0.092 0.773 ± 0.089 0.794 ± 0.092 0.83 ± 0.08 **Tt.Ar (mm^2^)** 0.83 10.54 ± 1.04 12.05 ± 1.19 15.42 ± 1.52 17.73 ± 1.81 18.6 ± 1.9 19.41 ± 1.99 19.84 ± 2.03 20.07 ± 0.17 20.45 ± 0.18 18.08 ± 1.46 ^α^ 1.65 11.75 ± 1.11 14.23 ± 1.35 18.23 ± 1.73 18.07 ± 1.45 19.69 ± 1.58 20.38 ± 1.63 20.52 ± 2.15 19.19 ± 2.01 18.89 ± 1.98 20.05 ± 1.78 ^β^ 2.47 10.1 ± 0.99 11.04 ± 1.09 13.37 ± 1.32 14.88 ± 1.31 16.34 ± 1.44 17.34 ± 1.52 17.58 ± 1.88 16.23 ± 1.73 16.34 ± 1.74 17.03 ± 1.34 ^γ^ **Ct.Ar (mm^2^)** 0.83 3.289 ± 0.356 4.586 ± 0.496 5.53 ± 0.599 5.132 ± 0.556 5.808 ± 0.592 5.884 ± 0.6 6.542 ± 0.667 7.261 ± 0.74 7.469 ± 0.762 7.013 ± 0.685 ^α^ 1.65 2.96 ± 0.302 4.62 ± 0.472 6.002 ± 0.614 5.994 ± 0.711 6.424 ± 0.762 7.32 ± 0.869 6.784 ± 0.867 7.728 ± 0.988 7.685 ± 0.982 7.55 ± 0.817 ^β^ 2.47 2.835 ± 0.321 4.05 ± 0.459 4.735 ± 0.536 4.23 ± 0.479 5.323 ± 0.649 6.23 ± 0.759 5.954 ± 0.726 6.34 ± 0.773 6.67 ± 0.952 6.823 ± 0.732 ^γ^ **Ct.Th (mm)** 0.83 0.196 ± 0.02 0.251 ± 0.026 0.235 ± 0.024 0.281 ± 0.029 0.288 ± 0.03 0.335 ± 0.035 0.364 ± 0.036 0.383 ± 0.037 0.373 ± 0.036 0.344 ± 0.033 ^α^ 1.65 0.232 ± 0.024 0.244 ± 0.025 0.215 ± 0.022 0.242 ± 0.028 0.263 ± 0.031 0.272 ± 0.032 0.312 ± 0.037 0.353 ± 0.042 0.325 ± 0.039 0.326 ± 0.035 ^β^ 2.47 0.22 ± 0.031 0.223 ± 0.031 0.18 ± 0.025 0.212 ± 0.034 0.249 ± 0.04 0.224 ± 0.036 0.258 ± 0.041 0.263 ± 0.051 0.281 ± 0.055 0.309 ± 0.042 **Ps.Pm (mm)** 0.83 13.48 ± 1.29 13.5 ± 1.29 15.45 ± 1.48 18.68 ± 2.24 19.23 ± 2.3 19.19 ± 2.3 17.23 ± 1.86 19.71 ± 2.12 19.78 ± 2.13 20.48 ± 1.64 ^α^ 1.65 11.96 ± 1.22 12.67 ± 1.29 16.18 ± 1.65 15.43 ± 1.76 18.08 ± 2.06 18.8 ± 2.14 18.57 ± 2.56 16.4 ± 2.26 18 ± 2.48 17.72 ± 1.51 ^β^ 2.47 12.67 ± 1.52 14.22 ± 1.71 16.34 ± 1.96 16.34 ± 1.77 18.75 ± 2.03 19.34 ± 2.09 18.84 ± 1.73 17.89 ± 1.65 18.63 ± 1.72 19.49 ± 1.73 ^γ^ **Ec.Pm (mm)** 0.83 9.51 ± 2.96 9.68 ± 3.01 11.15 ± 3.47 12.58 ± 4.62 12.68 ± 4.66 13.03 ± 4.79 12.92 ± 3.82 12.68 ± 3.75 12.77 ± 3.77 11.88 ± 3.84 ^α^ 1.65 10.45 ± 3.35 10.98 ± 3.52 12.39 ± 3.97 12.32 ± 4.27 12.91 ± 4.48 12.81 ± 4.44 13.13 ± 3.85 12 ± 3.51 11.86 ± 3.47 12.66 ± 4.32 ^β^ 2.47 9.54 ± 2.68 9.37 ± 2.64 10.42 ± 2.93 11.57 ± 4 11.76 ± 4.07 11.81 ± 4.08 12.08 ± 3.81 11.14 ± 3.51 11.02 ± 3.47 12.18 ± 3.23 ^γ^ **Ma.Ar (mm^2^)** 0.83 7.25 ± 1.37 7.46 ± 1.41 9.89 ± 1.87 12.59 ± 2.14 12.79 ± 2.17 13.52 ± 2.3 13.3 ± 2.18 12.81 ± 2.1 12.98 ± 2.13 11.34 ± 1.67 ^α^ 1.65 8.79 ± 1.86 9.6 ± 2.03 12.23 ± 2.59 12.08 ± 2.36 13.26 ± 2.59 13.06 ± 2.55 13.73 ± 2.58 11.46 ± 2.15 11.2 ± 2.11 12.81 ± 2.26 ^β^ 2.47 7.27 ± 1.39 6.99 ± 1.33 8.64 ± 1.65 10.65 ± 1.86 11.01 ± 1.92 11.11 ± 1.94 11.62 ± 1.93 9.89 ± 1.64 9.67 ± 1.61 11.91 ± 1.88 ^γ^ *Ecc* 0.83 0.383 ± 0.07 0.372 ± 0.068 0.437 ± 0.08 0.482 ± 0.082 0.473 ± 0.081 0.525 ± 0.09 0.539 ± 0.081 0.569 ± 0.086 0.582 ± 0.087 0.486 ± 0.069 ^α^ 1.65 0.421 ± 0.098 0.442 ± 0.103 0.483 ± 0.113 0.534 ± 0.109 0.505 ± 0.103 0.539 ± 0.11 0.528 ± 0.09 0.613 ± 0.105 0.627 ± 0.107 0.523 ± 0.087 ^β^ 2.47 0.341 ± 0.065 0.362 ± 0.069 0.382 ± 0.073 0.44 ± 0.075 0.442 ± 0.075 0.473 ± 0.081 0.55 ± 0.083 0.512 ± 0.078 0.522 ± 0.079 0.492 ± 0.07 ^γ^ Values are expressed as Mean ± SD, n = 11/group. Within each dose (0.83, 1.65 and 2.47 Gy), different symbols (α, β, γ) denote statistical significance (**p \< 0.05) from 4^th^ to 14^th^ week of age. Abbreviations: TMD, tissue mineral density; Tt.Ar, cross-sectional area inside the periosteal envelope; Ct.Ar, cortical bone area; Ct.Th, cortical thickness; Ps.Pm, periosteum perimeter; Ec.Pm, endocortical perimeter; Ma.Ar, medullary area; Ecc, mean eccentricity. Tukey's post hoc multiple comparison tests revealed differences among different groups for the 14^th^ week scanning data ([Table 4](#pone.0207323.t004){ref-type="table"}). 0.83 Gy group showed significant difference with the 1.65 Gy group for BMD, Tb.Th, Conn.Dn, and Tt.Ar parameters ([Table 4](#pone.0207323.t004){ref-type="table"}). Comparing 0.83 Gy and 2.47 Gy groups, significant differences were found for BMD, Tb.Th, Tb.N, Tb.Sp, Conn.Dn, Ct.Th, and Ps.Pm parameters ([Table 4](#pone.0207323.t004){ref-type="table"}). The 1.65 Gy group showed significant differences with the 2.47 Gy group for Tt.Ar parameter only ([Table 4](#pone.0207323.t004){ref-type="table"}). 10.1371/journal.pone.0207323.t004 ###### ANOVA test with Tukey's multiple comparisons for the trabecular and cortical bone structural properties of the irradiated rat tibiae for three radiation groups on the 14^th^ week. ![](pone.0207323.t004){#pone.0207323.t004g} Statistical Comparison --------------------------------------------- ------------------------ ----- ----- ----- Trabecular bone structural properties BMD (gm/cm^3^) \< 0.001 Yes Yes No BV/TV (%) 0.067 \- \- \- Tb.Th (mm) \< 0.001 Yes Yes No Tb.N (mm^-1^) 0.003 No Yes No Tb.Sp (mm) 0.039 No Yes No Conn.Dn (mm^-3^) 0.025 Yes Yes No Cortical bone structural properties TMD (gm.cm^-3^) 0.056 \- \- \- Tt.Ar (mm^2^) 0.048 Yes No Yes Ct.Ar (mm^2^) 0.704 \- \- \- Ct.Th (mm) 0.044 No Yes No Ps.Pm (mm) 0.017** No Yes No Ec.Pm (mm) 0.421 \- \- \- Ma.Ar (mm^2^) 0.572 \- \- \- Ecc 0.302 \- \- \- The given p-values are the results of a one-way ANOVA comparing the bone structural properties of the irradiated tibiae on the 14^th^ week among three doses groups. A bold value indicates a significant difference at **p \< 0.05*. The "Statistical Comparison" columns indicate whether the radiation groups were significantly different using Tukey's post-hoc pairwise comparisons. Body weight {#sec017} ----------- Body weights were similar for rats of all groups at the beginning of the experiment (4^th^ week of age) ([Fig 8](#pone.0207323.g008){ref-type="fig"}). A time effect (weight gain) was observed in rats as they were in their growing phase. However, no effects of dose or dose/time interaction were found and no loss of hair was observed during the study period ([Fig 8](#pone.0207323.g008){ref-type="fig"}). At the end of the experiment, average body weights of 0.83, 1.65 and 2.47 Gy groups were 572.2 ± 40.8, 534.1 ± 25.5, and 519.2 ± 23.1 g, respectively. ![Body weight of male Sprague Dawley rats for three doses groups over the adolescent period.\ ANOVA test (general linear model) was performed to determine time effects, radiation dose, and their interaction on body weight. N = 11 rats per group (mean value ± SD).](pone.0207323.g008){#pone.0207323.g008} Bone marrow cells {#sec018} ----------------- Results showed no significant difference (p* = 0.93) between percentage of unaffected bone marrow cells for control (93.2%) and irradiated tibiae (91.1%) at the 14^th^ week for 0.83 Gy group ([Table 5](#pone.0207323.t005){ref-type="table"}). However, for 1.65 Gy group, a significant difference was observed (p = 0.04) between percentage of unaffected bone marrow cells for control (87.3%) and irradiated tibiae (71.6%) ([Table 5](#pone.0207323.t005){ref-type="table"}). A significant difference was also observed (p = 0.02) between control (88.7%) and irradiated tibiae (70.8%) for the 2.47 Gy group ([Table 5](#pone.0207323.t005){ref-type="table"}). 10.1371/journal.pone.0207323.t005 ###### Percentage of unaffected bone marrow cells for 0.83, 1.65 and 2.47 Gy radiation groups extracted from trypan blue test (mean value ± SD). ![](pone.0207323.t005){#pone.0207323.t005g} Radiation group Tibiae Unaffected bone marrow cells (%) p-values ----------------- -------------- ---------------------------------- ------------ 0.83 Gy Control 93.20 ± 3.45 0.926 Irradiated 91.11 ± 4.23 1.65 Gy Control 87.34 ± 7.37 0.037 Irradiated 71.56 ± 9.27 2.47 Gy Control 88.67 ± 6.62 0.021 Irradiated 70.84 ± 8.51 Values are expressed as Mean ± SD. Both tibiae (n = 11 rats/group) were used for the analysis. A bold value indicates a significant difference (**p \< 0.05*) between the control (left) and irradiated (right) tibiae for each radiation dose. Discussion {#sec019} ========== In this study, we investigated the effects of nine weeks in vivo* scanning regime on the rat proximal tibiae under three different radiation doses. We used growing rats (n = 33), for which the right proximal tibia was irradiated while the left tibia was used as a non-radiated contralateral control. Bone growth, histomorphometry, morphology, and bone architecture during the growing period were assessed to identify the effects of repeated in vivo irradiation in the adolescent period. This study would optimally provide an effective radiation doses protocol, which would be "safe" to use for the growing rats. An effective radiation dose can be marked as "safe" if high-quality image sets can be acquired during the bone growing period without influencing the bone tissue health. We induced radiation doses with a higher frequency than generally used in bone investigation studies \[[@pone.0207323.ref043], [@pone.0207323.ref044]\], but similar to recent radiation effect investigation studies \[[@pone.0207323.ref024], [@pone.0207323.ref025]\]. Our adapted highest dose of radiation (2.47 Gy/scan) for 9 weeks is also within the limit of single dose of irradiation (2.5 Gy) for the tibial metaphysis of adult (10 and 14 months old) rabbits \[[@pone.0207323.ref020]\], where no significant alteration in bone formation was found. However, we investigated the growing animals (4^th^ to 14^th^ week of age) and our adapted radiation groups (0.83, 1.65 and 2.47 Gy) demonstrated mixed impacts on the bone microstructure during the study period. Radiation doses of 1.65 and 2.47 Gy adversely impacted tibial bone development during the adolescent growth period {#sec020} ------------------------------------------------------------------------------------------------------------------ Indeed, our results showed that these radiation doses reduced the hypertrophic and proliferative zone heights, which eventually inhibited bone growth rate of proximal tibiae. Both hypertrophic and proliferative cellular activities have an important impact on endochondral bone formation \[[@pone.0207323.ref045]\]. The main functions of the proliferative zone consist of matrix production and cellular proliferation \[[@pone.0207323.ref046]\]. Active cell replication takes place in this zone and chondrocytes are oriented in column formation along longitudinal bone growth \[[@pone.0207323.ref047]\]. The main functions of the hypertrophic zone include generating hypertrophic chondrocytes by terminal differentiation of the proliferative zone chondrocytes farthest from the epiphysis, preparing the matrix for calcification and to calcify the matrix \[[@pone.0207323.ref048]\]. Proliferative chondrocytes eventually increase in volume to generate the hypertrophic chondrocytes \[[@pone.0207323.ref049]\]. In the proliferative zone, cells undergo rapid replication \[[@pone.0207323.ref046]\]. In this region, chondrocyte divides, assume a flattened appearance, and become organized into columns parallel to the long axis of the bone \[[@pone.0207323.ref047]\]. Eventually, column elongation occurs through spatially coordinated cell division and rotational movements \[[@pone.0207323.ref047]\]. Hence, it is expected that any significant changes in these two zones will influence bone growth \[[@pone.0207323.ref045]\]. For the 0.83 Gy group, growth plate histomorphometry remained unaffected for the irradiated tibiae ([Fig 5](#pone.0207323.g005){ref-type="fig"}). However, for 1.65 and 2.47 Gy groups, a significant reduction in zone heights was observed for irradiated tibiae in the hypertrophic and proliferative area ([Fig 5A and 5B](#pone.0207323.g005){ref-type="fig"}). As a result, a significant reduction in overall bone growth rate for the irradiated tibiae was observed for both groups ([Fig 2B](#pone.0207323.g002){ref-type="fig"}). This decline in bone growth rate can be correlated with the reduction in proliferative and hypertrophic zone heights, which has also been observed in other studies \[[@pone.0207323.ref046], [@pone.0207323.ref050]\]. The average bone growth rates measured for both tibia in 0.83 Gy group are moreover similar to normal longitudinal bone growth rates observed in the rat tibia \[[@pone.0207323.ref035]\]. This indicates that the longitudinal bone growth was not affected by the 0.83 Gy radiation doses, which agree with other studies \[[@pone.0207323.ref024], [@pone.0207323.ref036]\], where also no effects of irradiation on the longitudinal bone growth were reported when using a similar radiation exposure level. The significantly reduced bone growth measured in 1.65 and 2.47 Gy groups also agree with the findings from other studies \[[@pone.0207323.ref027], [@pone.0207323.ref051], [@pone.0207323.ref052]\], where inhibition of bone growth was reported due to the effects of in vivo irradiation. Trabecular bone, together with bone marrow cells, were negatively affected when undergoing repeated radiation doses of 1.65 and 2.47 Gy {#sec021} --------------------------------------------------------------------------------------------------------------------------------------- Our results showed that trabecular bone quantity and microstructure were adversely impacted for 1.65 and 2.47 Gy groups (Figs [6](#pone.0207323.g006){ref-type="fig"} and [9](#pone.0207323.g009){ref-type="fig"}). ![Trabecular and cortical bone representation after the 9-weekly in vivo micro-CT scans.\ (a---f) Representative 3D micro-CT images of metaphyseal bone structure of the irradiated (right) and non-irradiated control (left) tibiae at 14^th^ week of age after 0.83, 1.65 and 2.47 Gy radiation doses during the rat adolescent period. 3D micro-CT images within each radiation dose portray tibiae from the same rat, randomly selected to be representative of its respective dose group.](pone.0207323.g009){#pone.0207323.g009} However, our findings showed no significant differences between the irradiated and contralateral tibiae for the trabecular bone microarchitectures for 0.83 Gy group of rats ([Fig 6](#pone.0207323.g006){ref-type="fig"}). Our findings are supportive of a study using adult rats \[[@pone.0207323.ref030]\] (12 weeks old), where no radiation effects (0.60 Gy) were found on the proximal tibiae after a 3-month study period with monthly scanning regime. In another study \[[@pone.0207323.ref024]\], adult Wister rats (30 weeks old) underwent 8 weeks in vivo tibial scanning under doses of 939 mGy per scan, but the bone structural measurements remained unaffected. However, our findings showed significant effects on the irradiated tibiae for 1.65 and 2.47 Gy groups. Our repeated weekly in vivo irradiation resulted in a lower BMD in the irradiated tibia only for 1.65 and 2.47 Gy groups ([Fig 6](#pone.0207323.g006){ref-type="fig"}). From the longitudinal data, it can be observed that the 0.83 and 2.47 Gy group showed respectively a significant increase and decrease for the irradiated tibia in the BMD value from 4^th^ to 14^th^ week period ([Table 2](#pone.0207323.t002){ref-type="table"}). In general, bone mineral content tends to increase at the young age for healthy bone \[[@pone.0207323.ref053]\]. Also, in the adolescent period, soft tissue thickness of the proximal tibia increases due to the bone growth in this period \[[@pone.0207323.ref054]\]. It might be possible that for the 1.65 and 2.47 Gy group, the radiation doses affected the proximal tibial thickness by increasing the osteoclastic activity during the irradiation process \[[@pone.0207323.ref055]\]. This phenomenon might have triggered the significant decrease in BMD in the irradiated tibiae for these groups (Figs [6](#pone.0207323.g006){ref-type="fig"} and [9](#pone.0207323.g009){ref-type="fig"}). This reasoning is supported by another study, where the effect of radiation was assessed in the spine and the hip on 49 radiology and 40 non-exposed workers over ten years period \[[@pone.0207323.ref056]\]. A significant decrease in BMD was found among the workers who were exposed to the radiation. Also, other studies irradiating mice with 1--2 Gy doses reported a lower BMD after 12 weeks of post-irradiation \[[@pone.0207323.ref052], [@pone.0207323.ref057], [@pone.0207323.ref058]\]. An increase in BV/TV is often correlated with a rise of BMD for normal bone growth \[[@pone.0207323.ref059]\], which indicates a higher bone quality. This normal bone development phenomena can be observed for the 0.83 Gy group as the BV/TV values increased significantly during the adolescent period ([Table 2](#pone.0207323.t002){ref-type="table"}). For the 0.83 and 1.65 Gy group, no significant difference was found between the contralateral tibiae on the 14^th^ week ([Fig 6](#pone.0207323.g006){ref-type="fig"}). However, BV/TV values decreased significantly compared to the control ones for the 2.47 Gy group ([Fig 6](#pone.0207323.g006){ref-type="fig"}), which could be associated with the diminishing trend observed earlier in the longitudinal BMD values for the same group. Our findings are supported by a study where a weekly radiation dose of 0.846 Gy over 5 weeks resulted in a decreased BV/TV in adult mice (12-week old) \[[@pone.0207323.ref025]\]. Also, another study reported a 30% loss in 10-week old mice BV/TV after performing three 0.776 Gy dose scans separated by 2-week intervals \[[@pone.0207323.ref009]\]. As for Tb.Th, Tb.N, and Tb.Sp, no difference was found between the irradiated and control tibiae for the 0.83 Gy group ([Fig 6](#pone.0207323.g006){ref-type="fig"}), while Tb.Th and Tb.N were significantly increased in the growing period for this group ([Table 2](#pone.0207323.t002){ref-type="table"}). The increment of Tb.Th during the growing period indicates normal bone growth process \[[@pone.0207323.ref024], [@pone.0207323.ref060], [@pone.0207323.ref061]\]. Moreover, the observed increase in Tb.N is associated with the concomitant increase in BV/TV for the young age period \[[@pone.0207323.ref062]\]. These findings agree with results from other radiation effects investigation studies using 30 weeks old rats (0.60 Gy) \[[@pone.0207323.ref024]\], and 17 weeks old ovariectomized mice (1.30 Gy) \[[@pone.0207323.ref023]\]. In both of these studies, Tb.N, Tb.Th and Tb.Sp remained unaffected. However, for the 1.65 and 2.47 Gy groups, Tb.Th and Tb.N values were significantly lower and the Tb.Sp values were significantly higher in the irradiated tibiae ([Fig 6](#pone.0207323.g006){ref-type="fig"}). From the longitudinal data, a significant decrease for Tb.Th and a significant increase in Tb.N and Tb.Sp were observed for 1.65 Gy group ([Table 2](#pone.0207323.t002){ref-type="table"}), whereas the 2.47 Gy group showed a significant decrease for Tb.Th and an increase for Tb.Sp only in the scanning period ([Table 2](#pone.0207323.t002){ref-type="table"}). These phenomena indicate the occurrence of a radiation-induced bone loss through a decreased connectivity and a gradual thinning of the trabecular structure ([Fig 9](#pone.0207323.g009){ref-type="fig"}). Our data are supportive of previous findings from a mice study where a 5--6 Gy radiation exposure for 3 days and 14 days resulted in decreased Tb.Th, Tb.N, and an increased Tb.Sp \[[@pone.0207323.ref063]\]. Another mice study using 0.846 Gy radiation dose for 5 weekly scans at 2 weeks interval reported a lower Tb.Th, Tb.N, and a higher Tb.Sp \[[@pone.0207323.ref025]\]. As for Conn.Dn, no significant difference was observed between the irradiated and control tibiae at the 14^th^ week for all three groups ([Fig 6](#pone.0207323.g006){ref-type="fig"}). However, a significant decrease was observed from the longitudinal data for both 0.83 and 1.65 Gy groups ([Table 2](#pone.0207323.t002){ref-type="table"}). Connectivity density is vital in the maintenance of bone strength and trabecular connectivity is a fundamental property of 3D bone networks. As Conn.Dn provides a measure of unconnected trabeculae, this decrement could occur because the bone was still in the growing phase while the trabecular structure was changing with time. This observation also agrees well with the findings from other rat studies \[[@pone.0207323.ref024], [@pone.0207323.ref039]\], where Conn.Dn was also decreased with the age of the rats. Also from ANOVA test with Tukey's multiple comparisons, it has been observed that for trabecular bone, 0.83 Gy group showed significant differences with the 1.65 and 2.47 Gy groups for BMD, Tb.Th, Tb.N, Tb.Sp, and Conn.Dn, whereas, no significant differences for the trabecular bone microstructure were found between 1.65 and 2.47 Gy groups ([Table 4](#pone.0207323.t004){ref-type="table"}). This indicates higher similarities of trabecular morphometric data between 1.65 and 2.47 Gy group compared with the 0.83 Gy group. This observation agrees with our morphometric findings as both of this group demonstrated similar adverse effects on the bone microarchitecture compared to the 0.83 Gy group ([Fig 9](#pone.0207323.g009){ref-type="fig"}). It has been reported that a radiation dose, if too high, can cause cell death, and the effects can be apparent within hours, days, or weeks after the exposure period \[[@pone.0207323.ref064]\]. Therefore, a two-week interval for the last scan was implemented in this study considering the possibility that the maximal radiation exposure effect could occur after the end of the exposure protocol \[[@pone.0207323.ref027], [@pone.0207323.ref028]\]. Bone marrow cells remained unaffected for 0.83 Gy group, which agrees with the conjecture based on CTDI \[[@pone.0207323.ref012], [@pone.0207323.ref022]\], and with a recent study \[[@pone.0207323.ref024]\] (30 weeks old rats, 600 mGy), where no cell damage due to radiation was reported. However, for the 1.65 and 2.47 Gy groups, significant differences were observed between control and irradiated tibiae. This phenomena confirm the negative impacts of 1.65 and 2.47 Gy doses on bone tissue health and can also be correlated with the detrimental effects observed by these doses on the trabecular structure found in our study \[[@pone.0207323.ref065]\]. Cortical bone quantity and microstructure were slightly deteriorated under repeated radiation dose of 2.47 Gy {#sec022} ------------------------------------------------------------------------------------------------------------- As opposed to trabecular bone, tested radiation doses had no profound effects on cortical bone microarchitecture. Indeed, for all three groups, there was no significant difference observed between the irradiated and the control tibia for TMD, Tt.Ar, Ct.Ar, Ps.Pm, Ec.Pm, Ma.Ar, and Ecc ([Fig 7](#pone.0207323.g007){ref-type="fig"}). From the longitudinal data, a significant increase was observed for Tt.Ar, Ct.Ar, Ps.Pm, Ec.Pm, Ma.Ar, and Ecc in all groups, whereas TMD and Ct.Th increased for 0.83 and 1.65 groups indicating the normal bone growing phenomenon ([Table 3](#pone.0207323.t003){ref-type="table"}). However, the cortical thickness (Ct.Th) showed a significant difference at the 14^th^ week scanning time point and decreased in the irradiated tibia compared to the control ones for the 2.47 Gy group ([Fig 6](#pone.0207323.g006){ref-type="fig"}). It could be possible that the radiation dose might have affected more intensively the vascularization of the proximal tibia. Vascularization is essential for bone formation and bone remodeling, transporting nutrients and the oxygen supply and allowing endothelial cells to communicate with osteoprogenitors and osteoclasts \[[@pone.0207323.ref066]\]. Moreover, if this process gets affected, a potential bone tissue destruction can occur \[[@pone.0207323.ref067]\]. Another possible reason could be the redistribution of bone mass from the endosteal region to the sub-periosteal region of the tibia. If this redistribution happens, it generally results in reduced cortical thickness of the bone diaphysis \[[@pone.0207323.ref068]\]. Our findings can be confirmed from another study \[[@pone.0207323.ref069]\] where a single 80-Gy radiation exposure for the 8-week-old rat hind limbs substantially decreased the cortical thickness and created wide bone gaps in the bone microstructure. Also from ANOVA test with Tukey's multiple comparisons, it has been observed that for cortical bone, 0.83 Gy group showed significant differences with the 1.65 and 2.47 Gy groups for Tt.Ar, Ct.Th, and Ps.Pm, whereas, significant difference for only Tt.Ar was found between 1.65 and 2.47 Gy groups ([Table 4](#pone.0207323.t004){ref-type="table"}). This findings indicate the vulnerability of cortical bone microarchitecture under the radiation doses of 1.65 and 2.47 Gy group which is in agreement to our morphometric findings. Comparisons among protocols used in similar radiation studies and strengths of the current study {#sec023} ------------------------------------------------------------------------------------------------ Differences found between our results and published studies might result from different factors. The positioning of the animal's limb in the scanner bed can be considered a possible factor for discrepancies among the studies. Since the right proximal tibia (irradiated) was exposed to radiation for 9 weeks (from 4^th^ to 14^th^ week of age), the frequent stretching of the right tibiae might have induced an effect on the bone tissue microstructure. As the radiation chamber rotates around the object for scanning, the right tibia was always pulled away from the body and fixed on the Styrofoam holder with the masking tape during the scanning period. The contralateral tibia was folded along with the tail outside the Styrofoam holder. This stretching could make the rat put a reduced pressure on right tibia for a short period of time right after the scanning period, which could lead to bone loss \[[@pone.0207323.ref024]\]. However, as we have followed the same approach throughout the whole study, this effect (if any) should be similar for all the animals and hence the relative comparison allows to draw conclusions. Also, it was presumed that the left tibia remained unaffected during the scanning of the right tibia. Nevertheless, it is possible that systemic radiation effects have occurred and affected the left tibia \[[@pone.0207323.ref024]\]. However, as we only irradiated the proximal tibial portion, which covered a small segment compared to the whole body, these systemic effects are expected to be non-significant. Also, changes in the body weight were compared with the literature \[[@pone.0207323.ref070], [@pone.0207323.ref071]\] to check for any sudden weight loss and no anomalies were found. Using various animal models might also be another contributing factor for discrepancies. It has been observed from the literature that, for mice, the scanning time interval might be more critical than the radiation doses \[[@pone.0207323.ref072]\]. Also, in some cases, a similar amount of radiation exposure for both mice and rats have produced divergent results. Rat bone structure seems to be more resilient to the same amount of radiation exposure compared to the mice \[[@pone.0207323.ref023]--[@pone.0207323.ref025], [@pone.0207323.ref052], [@pone.0207323.ref057]\]. One possible reason could be the presence of larger and thicker skeletons in rats compared to mice, which might provide an additional absorbing capacity of the induced radiation for rats. Another reason for the discrepancies might be the age of the animals used in different studies. In most studies, an adult animal model has been used compared to our adolescent model \[[@pone.0207323.ref023], [@pone.0207323.ref024], [@pone.0207323.ref039]\]. Bone remodeling gets slower with aging \[[@pone.0207323.ref073]\], and bone turnover rate shifts towards bone resorption \[[@pone.0207323.ref074], [@pone.0207323.ref075]\]. As a result, the bone microstructure behavior is expected to be different in these studies compared to ours. Nonetheless, it remains difficult to make a comparison of our findings with other studies as none of them investigated the effects of in vivo micro-CT irradiation in a rat model during its growing period (4^th^ to 14^th^ week of age) \[[@pone.0207323.ref026]\]. Also, the scanning protocol, radiation doses, types of scanner, and animal positioning during scanning can contribute to differences in results between different animal studies. Despite some limitations, our current study possesses a number of strengths. First, to authors' best knowledge, this study is the first of its kind to investigate the effects of repeated in vivo micro-CT irradiation using an animal model during its entire growing period. Second, three different levels of radiation doses have been investigated using the same image reconstruction parameters to facilitate the comparison among results from different groups. Third, we scanned the non-radiated control legs (left tibiae) only at the end point (14^th^ week of age). This method of using an internal control decreases the use of extra animals for control and reduces the variability in the extracted data sets. Furthermore, results from the current study possess insightful information regarding bone microarchitecture during the bone development period, which would be useful to the bone and orthopedic research community. Conclusion {#sec024} ========== In conclusion, using 1.65 and 2.47 Gy doses might yield better image quality for bone tissue investigation but possess a high risk of altering the bone growing process in the rat adolescent period. Our results showed that, under radiation doses of 1.65 and 2.47 Gy, trabecular bone, together with bone marrow cells, as well as tibial bone development were adversely impacted. Also, cortical bone quantity and microstructure was slightly deteriorated under repeated radiation doses of 2.47 Gy. Hence, it appears from our results that 1.65 and 2.47 Gy doses affected significantly the bone marrow cells, histomorphometric and morphological parameters, and longitudinal bone growth of the immature rats. However, the 0.83 Gy radiation exposure did not affect the bone tissue structure for the growing rats. These findings can be used as a proof of concept for using the reasonable high-quality image acquisition under 0.83 Gy radiation doses during the entire growing period of rats without interfering with the bone development process. Our study also advances the knowledge on the evaluation of the radiation effects during the adolescent period of animal models in order to provide functional information for the design of future in vivo studies, in which the repeated radiation exposure is necessary and can induce additional impacts on the outcomes. Considering that the radiation damage also depends on other factors (scanning protocol, systemic effects, site-specificity), which are not micro-CT system specific, careful consideration should be adapted for future studies. Supporting information {#sec025} ====================== ###### Certification of animal care. (PDF) ###### Click here for additional data file. ###### The ARRIVE guidelines checklist. (PDF) ###### Click here for additional data file. ###### Calculation of radiation doses per scan. (PDF) ###### Click here for additional data file. The authors acknowledge helpful contributions and technical skills of laboratory team members as well as Sainte-Justine University Hospital's animal care technicians. Funding for this study was provided by NSERC (IV), the CRC Program (IV) and the NSERC/CREATE program (TM). [^1]: Competing Interests:The authors have declared that no competing interests exist.	{ "pile_set_name": "PubMed Central" }
Introduction ============ Radish (Raphanus sativus L., 2n = 18) is a member of Brassicaceae family, and has a relatively small genome size. The size is similar to those of other Brassicaceae species such as Brassica rapa ([@b10-bs-61-413]). Radish is thought to have originated in Mediterranean areas, and is now widely cultivated in East Asian countries as an important vegetable root crop ([@b13-bs-61-413]). The edible part of this crop is called root, but in histology, it consists of the true root and the hypocotyl. A wide morphological diversity of this part is known in this species, and many cultivars having various shapes of thickened root have been developed in East Asia ([@b15-bs-61-413]). In contrast, a rat-tail radish produces no thickened root, and is cultivated in tropical Asia for edible puffed pods ([@b3-bs-61-413]). Though radish has an extreme importance from the economic and agricultural views, molecular research of radish such as genome mapping and genetic diversity is behind compared to that of Brassica species. Molecular markers have been extensively used to study genetic diversity, genetic relationships and mapping studies in various crop species. Of these, simple sequence repeats (SSRs) or microsatellites are frequently utilized as DNA markers. The SSR is a DNA repeat consisting of 1--6 nucleotide repeat units, abundantly distributed in most eukaryotic genomes. The SSR marker has the advantages of high variability, ease of detection, codominant inheritance nature, and good transferability between populations and in different research groups (see [@b11-bs-61-413] for a review). For these reasons, SSRs have become an important marker system in cultivar fingerprinting, diversity studies, molecular mapping and marker-assisted selection. On the other hand, because isolation of SSRs and establishment of the specific primers generally require high cost and a long development time, the number of available SSR markers differs in species. Actually, despite a number of SSR markers in Brassica species, there have been only a few reports on development of SSR markers in radish ([@b12-bs-61-413], [@b21-bs-61-413], [@b31-bs-61-413], [@b32-bs-61-413]). In the previous studies, we have made linkage maps of radish to detect loci for morphological characters and clubroot resistance ([@b12-bs-61-413], [@b28-bs-61-413], [@b29-bs-61-413]). Even including our efforts, reports on the linkage map are still limited in radish compared to other Brassicaceae species ([@b4-bs-61-413], [@b6-bs-61-413], [@b12-bs-61-413], [@b28-bs-61-413], [@b29-bs-61-413]). Since limited numbers of SSR markers were used in these studies, correspondence of the published linkage groups (LGs) is mostly unclear. Under such circumstances, increasing the number of available SSR markers would be of importance to conduct further genetic studies in radish. Here, we designed a total of 417 SSR primer pairs in radish from SSR-enriched genomic libraries and cDNA data in the database. Utilities of the radish SSR markers were tested in the two following ways. To assess the usefulness of these markers among radish and other related species in Brassicaceae, the polymorphism information contents (PICs) were calculated. Of the 73 SSR markers tested for mapping, 21 markers were located on a radish linkage map. These results showed effectiveness of the SSRs for mapping study in addition to potential utility for the genetic diversity analysis. Materials and Methods ===================== Plant materials and DNA extraction ---------------------------------- Three following radish lines were used for preparing genomic DNA libraries; a rat-tail radish, a Japanese cultivar, 'Harufuku', and an F~1~ plant derived from a cross between them. The homozygosity of the two parental lines was increased by five or six successive selfings. DNA was extracted from plant leaves using DNeasy Plant Mini Kit (Qiagen, Valencia, CA, USA). For allele analysis, 16 radish cultivars and four lines of other Brassicaceae species were used ([Table 1](#t1-bs-61-413){ref-type="table"}). Development of SSR markers -------------------------- The SSR-enriched genomic DNA libraries were constructed according to [@b20-bs-61-413]. The SSR-enriched fragments were ligated into a pCR-TOPO vector (Invitrogen, Carlsbad, CA, USA). Resultant plasmids were used to transform TOP10 competent cells (Invitrogen). Clones with a 0.5--1.5 kb insert were sequenced using a CEQ8800XL sequencer (Beckman Coulter, Fullerton, CA, USA). Nucleotide sequences of 47 redundant clones were removed manually. Nucleotide sequence data of the genomic SSRs developed here are deposited in the DDBJ/EMBL/GenBank databases under accession nos. AB425071, AB425072, AB425076-AB425078 and AB608338-AB608637 ([Supplemental Table 1](#SD1){ref-type="supplementary-material"}). EST-SSR markers were designed from cDNA contig data in the RadishDB (<http://radish.plantbiology.msu.edu>). Sequences containing SSR with \>10 mononucleotide, \>6 dinucleotide and \>5 trinucleotide repeats were used for primer design as described previously ([@b9-bs-61-413], [@b17-bs-61-413]). Primers were designed using the Primer3 software ([@b22-bs-61-413]). Amplification of primer pairs was tested in the two radish lines above. PCR amplification was carried out as described previously ([@b23-bs-61-413]). The annealing temperature was initially fixed at 50°C and then slight modifications were made to achieve optimal amplification ([Supplemental Table 1](#SD1){ref-type="supplementary-material"}). The amplified products were electrophoresed in a 3% agarose gel or with the CEQ8800XL sequencer (Beckman Coulter). Primer pairs showing clear polymorphic bands in these lines were then used for further analysis. Fragment analysis and allele detection -------------------------------------- Out of the developed SSRs, 16 primer pairs were used for the analysis of cultivars listed in [Table 1](#t1-bs-61-413){ref-type="table"}. The 5′-end of forward primer was labeled with a fluorescent dye (Sigma-Aldrich, St. Louis, MO, USA), and PCR amplification was carried out as described above. The fragment sizes of SSR loci were analyzed with the CEQ8800XL sequencer (Beckman Coulter). The value of the polymorphic information content (PIC) at each locus was calculated for the 16 radish cultivars as described ([@b2-bs-61-413]). Linkage mapping --------------- An improved linkage map of radish has been developed in this study based on our previous data ([@b29-bs-61-413]) plus the SSR marker data as described below. For the mapping, 73 radish SSR markers developed here, seven already-reported radish SSRs ([@b12-bs-61-413], [@b21-bs-61-413], [@b31-bs-61-413]) and six B. rapa SSRs ([@b25-bs-61-413], [@b26-bs-61-413]) were newly tested in this study. The polymorphic markers were then used for scoring the segregation in the F~2~ population (n = 106) ([@b29-bs-61-413]). Linkage analysis was performed using the JoinMap ver. 3.0 software ([@b30-bs-61-413]), but markers deviating significantly (P \< 0.001) from the expected segregation ratio were excluded from the analysis. The Kosambi map function ([@b16-bs-61-413]) was used to calculate the genetic distance between markers. Results ======= Development of SSR markers in radish ------------------------------------ Three SSR-enriched libraries were developed from the two radish lines (rat-tail radish and 'Harufuku') and their F~1~ plant. From the rat-tailed radish library, 326 clones were sequenced. Similarly, 92 and 299 clones were sequenced from 'Harufuku' and the F~1~ plant libraries, respectively ([Table 2](#t2-bs-61-413){ref-type="table"}). Of these, 245 (75.2%), 63 (68.5%) and 191 (63.9%) contained SSR motifs in the sequenced clones, respectively. Out of the 293 primer pairs initially designed from the non-redundant SSR-containing sequences, 156 primer pairs amplified clear bands. Of the 156 pairs succeeded in PCR, 90 primer pairs (57.7%) showed polymorphisms between the two radish lines ([Table 2](#t2-bs-61-413){ref-type="table"}). We also designed 124 primer pairs of EST-SSRs from cDNA contig data in the RadishDB. Of these EST-based primer pairs, 100 produced clear bands within the predicted size range. Of the 100 successful EST-SSRs, 40 primer pairs (40.0%) showed polymorphisms between the two radish lines ([Table 2](#t2-bs-61-413){ref-type="table"}). Finally, 256 SSR markers (156 genomic and 100 EST-SSRs) resulted in clear bands, of which 130 markers were polymorphic between the two radish lines ([Table 2](#t2-bs-61-413){ref-type="table"} and bold types in [Supplemental Table 1](#SD1){ref-type="supplementary-material"}). Assessment of SSR polymorphisms by a fragment analysis ------------------------------------------------------ Out of the primers that amplified clear polymorphic bands, 16 SSR loci were tested for the ability to detect alleles using radish cultivars ([Table 3](#t3-bs-61-413){ref-type="table"}). Of the markers tested, 13 primer pairs amplified clear polymorphic bands for all the 16 radish cultivars ([Table 4](#t4-bs-61-413){ref-type="table"}). These markers detected 3--15 alleles with an average of 9.6. PIC values ranged from 0.54 to 0.92 with an average of 0.78. We also tested nine primer pairs for amplification in other Brassicaceae species. PCR gave detectable amplicons in more than half of the SSR markers tested in the Brassica species though their sizes were often different from those in radish ([Table 4](#t4-bs-61-413){ref-type="table"}). Linkage mapping of the radish SSRs ---------------------------------- The present SSRs were tested for a mapping study. Of the 80 radish SSRs tested, 23 markers were successfully localized on the linkage map ([Fig. 1](#f1-bs-61-413){ref-type="fig"}, arrows), of which 21 were ones developed here ([Fig. 1](#f1-bs-61-413){ref-type="fig"}, bold types). Finally, the present map was constructed with the 336 loci (278 AFLPs, 23 radish SSRs, 34 B. rapa SSRs and 1 radish CAPS, see legend of [Fig. 1](#f1-bs-61-413){ref-type="fig"} for marker nomenclatures), whereas only seven loci (4 AFLPs, 1 radish SSR and 2 B. rapa SSRs) were unmapped. This map spanned 672.7 cM with nine LGs, which are expected from the chromosome number of radish. The length of the LGs ranged from 35.6 to 122.4 cM, which contained from 22 to 56 markers, respectively, with an average map interval of 2.0 cM. The mapped SSR markers were distributed throughout the LGs. Discussion ========== The present study seems to be successful in the SSR-enrichment from the radish genomes, judging from the efficiencies of SSR-enrichment as compared with those in the previous studies using cucumber ([@b8-bs-61-413]), Vaccinium ([@b9-bs-61-413]) and water lotus ([@b17-bs-61-413]). We have used three DNA materials (two radish lines and their F~1~ hybrid) for the SSR development. There was no obvious difference among the three with regard to the SSR-enrichment efficiency and the polymorphisms of markers ([Table 2](#t2-bs-61-413){ref-type="table"}). Out of the 293 primer pairs initially designed for genomic SSR markers, 156 primer pairs succeeded in PCR amplifying clear bands (53.2%), whereas 81 pairs produced faint, smear or multiple bands (27.6%) ([Supplemental Table 1](#SD1){ref-type="supplementary-material"}). The rest of 56 pairs resulted in no amplification (19.1%). One of the reasons for failure in the PCR amplification could be due to the presence of repetitive DNA sequences or similar sequences in other genomic regions, as observed in marker development from species with large genomes ([@b24-bs-61-413]). The unsuccessful markers might be converted into useful markers by re-designing the primer pairs based on the nucleotide sequences flanked by the SSR repeats ([Supplemental Table 1](#SD1){ref-type="supplementary-material"}, accession nos.). The success rate of the PCR amplification in the EST-SSRs (80.6%) was higher than that in the genomic SSRs (53.2%). Similar observation has been reported in different crops such as coffee and common bean (e.g. [@b1-bs-61-413], [@b5-bs-61-413]). This could be because exon-derived sequences are more conservative than the intergenic regions. In spite of such the conservative nature of ESTs, some of the EST-derived SSRs were failed in stable amplification. This could be because some primer sites may have been designed across splicing sites or because of chimerical origins of the cDNA clones in the database, as discussed by [@b27-bs-61-413]. Although polymorphism of EST-SSRs is generally lower than that of genomic SSRs ([@b5-bs-61-413]) because of the conservative nature of exon sequences, no obvious difference was found in polymorphism rate between the genomic and EST-SSRs as long as comparison of the present two radish lines. We previously applied B. rapa SSRs ([@b25-bs-61-413], [@b26-bs-61-413]) to the radish mapping studies ([@b28-bs-61-413], [@b29-bs-61-413]). In the present study, we showed that the inverse case was also effective in part. More than half of the tested radish SSRs clearly detected alleles in four lines of the Brassicaceae genera in addition to the radish cultivars ([Table 4](#t4-bs-61-413){ref-type="table"}), suggesting the effectiveness of the radish SSRs across the Brassicaceae genera. The PIC values obtained in this study (average = 0.78) were comparable to those of Brassica species ([@b25-bs-61-413]). Although limited numbers of SSRs were tested in this allele analysis as a test case, we confirmed that the radish SSRs developed in the previous and the present studies were applicable to the genetic diversity study in a wild radish ([@b21-bs-61-413]) and Japanese radish landraces (data not shown). Therefore, the SSRs developed here would be informative at the level of within-species variation, and may be useful even in other Brassicaceae species. Another aspect of usage of SSRs is for mapping study. We were able to map approximately one third of the applied SSRs on the radish map ([Fig. 1](#f1-bs-61-413){ref-type="fig"}, arrows). This ratio was similar to those of the mapping studies in Brassica species ([@b7-bs-61-413], [@b14-bs-61-413], [@b18-bs-61-413]). Because the present map was integrated to the radish chromosome number (n = 9), the map location of radish-specific SSRs would be useful for comparison of the radish linkage maps in the previous and future studies. Further mapping of the radish SSRs will enable us to construct a denser, SSR-based radish map and to detect agronomically important loci. However, this is beyond scope of the present study that focused on the development of hundreds of radish SSRs and their evaluation. In conclusion, the radish SSRs developed here would provide the useful tools for genetic analysis in radish and its related species. Supplementary Data ================== We thank Dr. H. Tsukazaki, the National Institute of Vegetable and Tea Science, Tsu, Japan, for providing the rat-tail radish, Dr. T. Nunome for helpful advice on the SSR-enriched library construction, Dr. S. Matsumoto for providing B. rapa SSR information, Dr. H. Budahn and Dr. T. Ohsako for valuable comments, and Ms. H. Kasaoka for technical assistance. This work was partly supported by the Program for Promotion of Basic and Applied Research for Innovations in the Bio-oriented Industry (BRAIN) and an ACTR grant from Kyoto Prefectural University to M.H. ![Positions of radish SSRs on a radish linkage map. Arrows indicate the radish SSRs, of which bold types are ones developed in this study. Asterisks indicate nine of the 16 SSR markers used in allele detection among 16 radish cultivars and four other Brassicaceae species. The loci denoted with BRMS- and SLG\_CAPS are B. rapa SSRs ([@b25-bs-61-413], [@b26-bs-61-413]) and a CAPS marker for S-locus specific glycoprotein gene ([@b19-bs-61-413]), respectively. Others are AFLP markers.](bs-61-413f1){#f1-bs-61-413} ###### Cultivars, lines and an ecotype used in this study Species name Cultivars, lines and an ecotype Origin ------------------------ --------------------------------- ------------------------------------------------- Raphanus sativus Akasuji Takii & Co., Ltd Everest Takii & Co., Ltd Harufuku National Institute of Vegetable and Tea Science Horyo Takii & Co., Ltd Koga-benimaru Sakata Seed Co. Kunitomi Matsunaga Seed Co., Ltd Kurobakei-Minowase Takii & Co., Ltd Miyashige ohnaga Takii & Co., Ltd Moriguchi hosonaga Matsunaga Seed Co., Ltd Rat-tail radish National Institute of Vegetable and Tea Science Red Globe Tohoku & Co., Ltd Sakurajima ohmaru Takii & Co., Ltd Shogoin daikon Sakata Seed Co. Tokinashi Marutane Co., Ltd Wakayama Takii & Co., Ltd Yakumidaikon Noguchi Seed Co. Arabidopsis thaliana Columbia Brassica juncea Akaohba takana Takii & Co., Ltd Brassica oleracea Grand Duke Takii & Co., Ltd Brassica rapa Shogoin ohmarukabu Takii & Co., Ltd ###### Development of radish SSR markers Genomic SSRs EST-SSRs Total ------------------------------------------------------------------------------------ ----------------- ---------- ------ ---------- ---------- ------- Origin[a](#tfn1-bs-61-413){ref-type="table-fn"} Rat-tail radish Harufuku F~1~ Subtotal RadishDB No. of clones sequenced 326 92 299 717 -- -- No. of clones containing SSR 245 63 191 499 -- -- SSR enrichment (%) 75.2 68.5 63.9 69.6 -- -- No. of primer pairs desinged 183 31 79 293 124 417 No. of primer pairs amplified clear bands[b](#tfn2-bs-61-413){ref-type="table-fn"} 81 26 49 156 100 256 No. of polymorphic primer pairs[b](#tfn2-bs-61-413){ref-type="table-fn"} 48 17 25 90 40 130 Polymorphism (%)[b](#tfn2-bs-61-413){ref-type="table-fn"} 59.3 65.4 51.0 57.7 40.0 50.8 F~1~: an F~1~ hybrid of rat-tail radish and Harufuku. RadishDB: cDNA contig data from the RadishDB. Data based on the experiments in rat-tail radish and Harufuku. See footnotes in Supplemental Table 1 for details. ###### Radish SSR loci used for allele detection Marker name[a](#tfn3-bs-61-413){ref-type="table-fn"} Primer sequence (5′-3′) Repeat motif Product size (bp)[b](#tfn4-bs-61-413){ref-type="table-fn"} No. of alleles detected Allele size range (nt)[c](#tfn5-bs-61-413){ref-type="table-fn"} PIC Accession No. ------------------------------------------------------ ------------------------------- ------------------------ ------------------------------------------------------------ ------------------------- ----------------------------------------------------------------- ------ --------------- RsSA012 F: GGATCGTTCCTTTTTAGGGTAAT (GA)~23~ 187 15 152--237 0.90 AB608423 R: GCTAAAAATCCGTGAGAAAGAG RsSA014 F: AATAAGCATGTGGTGGGAAGTTA (GA)~11~ 183 3 171--183 0.54 AB608424 R: GGGTTTATGAAAGGGATTTTGTC RsSA020 F: TCAGGGGTAAAACCGTCAATTA (CT)~17~ 227 8 193--227 0.76 AB608427 R: AGGATCGGAGATACGATTCAAA RsSA027 F: CTAGCCGTTTCCAAATTTGTTC (GA)~42~ 190 15 154--198 0.89 AB608430 R: AGTACTTTAACCACTGCCCAACA RsSA033 F: ACAATTTCACGACAGTAAACATGAA (TC)~26~ 228 14 181--285 0.87 AB608432 R: CCGAGTTGATTAAAACACACATACA RsSA120 F: TCTTACCATTGGTGTAAGTCAATCC (GA)~27~ 253 15 209--256 0.92 AB608477 R: GAAAGGTGGAGAAAATGAAGTAACA RsSH001 F: AACTCAGGTCCCTTGTGCTAGA (TC)~6~(CA)~7~ 237 4 201--243 0.65 AB608483 R: GGAACTATGTTGTTGTCGGAAA RsSH016 F: GTTTGTTGTTGTTTGTGTCACCT (CT)~4~(GT)~3~(CT)~10~ 136 6 132--140 0.76 AB608488 R: CAGAAGCAAGCACTATTTGAGAA RsSH048 F: TCGTCCGTTATGTATGTTACTCTCA (GT)~11~ 200 7 188--206 0.60 AB608489 R: TATGCGTACTCCGTAAGACAATGTA RsSH093 F: CAATTCTTTGTATGCTTTTGTCTGAT (GA)~17~ 233 7 231--239 0.76 AB608516 R: TGGCAAGATATATATAACCCTCGTTT RsSR025 F: ACACTTTCAGTCACCGACACATA (GA)~20~ 239 14 213--252 0.89 AB608556 R: ACTTTCTTTAGGTAACCCCACCA RsSR040 F: CGTCTCTTTCTTTTTCAGACCAA (TC)~14~ 221 10 202--228 0.73 AB608564 R: GCTTGAGATGAGATGAGGAGAAA RsSR042 F: ATAAAGCAGCAGAAGATGGTGAG (AC)~14~ 171 9 156--206 0.80 AB608565 R: GAATGAAACTCCTTTAAGAAGAAGC RsHH016 F: CTGATCGAACTGGAACCACAATT (AG)~24~ 189 10 179--209 0.82 AB608620 R: GAGGGTTTTAGGGCACCTGA RsHH023 F: CTGGTCTCACAATCAAACATCT (TA)~10~(TG)~13~ 169 11 162--206 0.83 AB608624 R: CTTATCTGTCACTTATTAATAGGCT RsHR026 F: AAGCGTGTCATCAGATCCCAGA (GA)~13~ 131 6 119--135 0.68 AB608635 R: CATTCTCTCAATGCATAAGATTGAGC Average 9.6 0.78 A complete list of radish SSRs developed in this study is shown as Supplemental Table 1. Estimated from the nucleotide sequences used for primer design. Sizes determined by fragment analyses. ###### Alleles in 16 radish cultivars and 4 other Brassicaceae species Marker name RsSA012 RsSA014 RsSA020 RsSA027 RsSA033 RsSA120 RsSH001 RsSH016 RsSH048 RsSH093 RsSR025 RsSR040 RsSR042 RsHH016 RsHH023 RsHR026 ------------------------------------------------------------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- ---------- Radish cultivars Akasuji 192/192 174/174 213/225 162/197 278/278 210/253 243/243 132/139 200/200 233/234 222/241 221/221 161/171 181/189 198/198 131/133 Everest 152/152 171/171 203/203 186/186 226/242 233/253 239/239 134/134 198/200 231/231 213/251 221/221 159/171 189/189 169/186 135/135 Harufuku 224/237 171/171 203/203 186/186 236/236 250/250 237/237 163/163 200/200 233/233 241/241 223/223 161/161 189/189 169/169 131/131 Horyo 187/235 171/171 225/225 161/198 208/234 244/244 239/243 132/132 198/198 234/234 235/235 228/228 206/206 181/189 186/186 131/135 Koga-benimaru 187/187 171/171 217/217 154/158 181/190 215/228 243/243 134/139 188/190 234/234 240/252 223/223 171/171 179/179 164/164 133/135 Kunitomi 188/188 174/174 225/225 162/162 228/234 242/252 201/239 134/134 198/206 234/234 234/236 221/221 159/161 -- -- 135/135 Kurobakei-Minowase 190/228 174/174 213/225 165/165 279/279 257/257 239/239 132/132 198/200 234/234 222/222 221/221 169/171 181/181 211/211 119/135 Miyashige ohnaga 182/182 174/174 209/209 159/163 228/234 210/212 239/239 132/134 200/200 233/233 235/241 223/223 156/161 181/189 186/186 135/135 Moriguchi hosonaga 188/188 171/174 225/225 161/161 234/234 244/248 239/239 139/139 200/200 232/232 235/235 211/221 156/171 193/193 197/205 134/135 Rat-tail radish 187/187 183/183 227/227 190/190 228/228 253/253 237/237 136/136 200/200 233/233 239/239 221/221 171/171 181/181 186/186 133/133 Red Globe 232/232 174/174 193/193 165/165 285/285 215/228 243/243 136/140 198/200 237/237 217/217 204/221 156/157 179/179 162/162 135/135 Sakurajima ohmaru 187/231 174/174 216/217 161/161 234/234 210/212 237/239 132/132 198/198 233/234 250/250 213/221 165/167 181/189 186/186 131/131 Shogoin daikon 216/232 171/171 217/225 177/177 228/228 242/255 239/239 139/139 198/200 233/237 235/235 223/223 171/171 181/189 186/186 131/131 Tokinashi 228/228 171/174 225/227 177/192 222/234 212/256 237/237 132/132 197/198 232/239 213/236 220/220 161/161 185/185 174/174 131/135 Wakayama 222/222 174/174 225/225 161/161 232/232 210/250 237/243 139/139 199/200 234/235 242/241 221/221 159/161 191/195 169/169 130/131 Yakumidaikon 187/220 174/174 225/225 179/188 224/224 209/209 239/239 134/134 200/200 237/237 238/238 202/210 159/159 201/209 204/206 -- Expected size (bp)[a](#tfn6-bs-61-413){ref-type="table-fn"} 187 183 227 190 228 253 237 136 200 233 239 221 171 189 169 131 Size range (nt)[a](#tfn6-bs-61-413){ref-type="table-fn"} 152--237 171--183 193--227 154--198 181--285 209--256 201--243 132--140 188--206 231--239 213--252 202--228 156--206 179--209 162--206 119--135 PIC 0.90 0.54 0.76 0.89 0.87 0.92 0.65 0.76 0.60 0.76 0.89 0.73 0.80 0.82 0.83 0.68 Other Brassicaceae species Arabidopsis thaliana (Columbia) -- -- 142/142 161/161 N.E. N.E. 193/193 204/204 N.E. N.E. -- 210/210 -- N.E. N.E. N.E. Brassica juncea (Akaohba takana) 195/195 116/197 169/199 203/297 N.E. N.E. 231/242 113/113 N.E. N.E. 191/217 230/230 154/154 N.E. N.E. N.E. B. oleacea (Grand Duke) 170/170 191/191 176/206 225/278 N.E. N.E. 246/246 113/119 N.E. N.E. 219/220 236/290 140/140 N.E. N.E. N.E. B. rapa (Shogoin ohmarukabu) -- 167/167 199/199 152/152 N.E. N.E. 179/179 113/113 N.E. N.E. 189/189 290/290 163/163 N.E. N.E. N.E. See footnotes in [Table 3](#t3-bs-61-413){ref-type="table"}. --: no specific amplification detectable. N.E.: not examined. [^1]: Communicated by T. Terachi	{ "pile_set_name": "PubMed Central" }
Session: 236. Healthcare Epidemiology: Epidemiologic Methods Saturday, October 6, 2018: 12:30 PM	{ "pile_set_name": "PubMed Central" }
Scientific Reports 6: Article number: 21102; 10.1038/srep21102published online: 02162016; updated: 04222016 In the Supplementary Information file originally published with this Article, the lower part of Fig. S1 'sGnRH-II' was incorrectly labelled 'GnRH-II'. This error has been corrected in the Supplementary Information that now accompanies the Article.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#S1} ============ Cyclic AMP is the prototypical second messenger. It has been implicated in many physiological processes, including cell proliferation, apoptosis, and differentiation. In recent years it has become clear how this single second messenger can mediate so many diverse, sometimes seemingly contradictory, processes. It is now appreciated that cAMP acts locally within independently-regulated, spatially-restricted microdomains (reviewed in^[@R1]^). Cyclic AMP is produced from ATP by adenylyl cyclases (AC). Traditionally, cAMP had been thought to be produced exclusively via G protein-regulated, transmembrane adenylyl cyclases (tmACs) in response to hormones signaling via G protein coupled receptors (GPCRs). It is now known that mammalian cells contain a second type of AC, soluble adenylyl cyclase (sAC; ADCY10)^[@R2]^. sAC is a widely distributed source of cAMP^[@R3]^, which unlike tmACs, is localized throughout the cell and targeted to intracellular compartments and cellular organelles, including inside the nucleus and the mitochondrial matrix^[@R4]^. Consistent with its localization, sAC defines multiple, independently-regulated, intracellular cAMP signaling microdomains. Because signaling occurs in spatially restricted microdomains, understanding cAMP biology requires methods to distinguish between sAC-dependent and tmAC-dependent cascades. sAC is also distinct from tmACs in its regulation. Unlike tmACs, sAC is not regulated by heterotrimeric G proteins^[@R5]^; instead, it is directly regulated by bicarbonate (HCO~3~^−^)^[@R6],[@R7]^, calcium (Ca^2+^)^[@R8],[@R9]^, and due to its millimolar Km for its substrate, ATP^[@R9]^, sAC is a physiological ATP sensor^[@R10]^. Inside cells, HCO~3~^−^ levels are dynamically regulated by ubiquitously expressed carbonic anhydrases (CA), which equilibrate carbon dioxide (CO~2~) and intracellular pH (pH~i~) with HCO~3~^−^. Due to this nearly instantaneous equilibrium, mammalian sAC, and its HCO~3~^−^-regulated orthologs throughout the kingdoms of life, serve as Nature's physiological CO~2~/HCO~3~^−^/pH~i~ sensors (Reviewed in^[@R3],[@R11]^). In response to changes in the CO~2~/HCO~3~^−^/pH~i~ equilibrium, sAC regulates sperm activation and motility^[@R12],[@R13]^; ciliary beat frequency in airway^[@R14],[@R15]^; luminal pH in the epididymis^[@R16]^; the mitochondrial electron transport chain^[@R17]^; apoptosis^[@R18]--[@R20]^; activity dependent feeding of neurons in the brain^[@R21]^; glucose-stimulated insulin release from β cells of the pancreas^[@R10]^; and intraocular pressure in the eye^[@R22]^. The physiological roles of sAC have been identified genetically, using two different sAC knockout (KO) mouse strains^[@R12],[@R13],[@R23]^ and sAC-specific siRNA^[@R24]--[@R26]^, and pharmacologically. Two sAC inhibitors useable in cellular systems have thus far been identified: catechol derivatives of estrogen (CEs) and KH7^[@R2],[@R27]^. CEs are natural products which were the first known inhibitors of sAC^[@R28]^; they inhibit sAC by binding to a hydrophobic cleft distinct from the active site and chelating essential metal cofactors^[@R29]^. While CEs have been useful for demonstrating sAC's role in cAMP dependent processes, CEs can also inhibit tmACs^[@R29]^, which limits their utility. The second pharmacological sAC inhibitor, KH7, was identified in a small molecule screen against purified sAC protein^[@R13]^. In contrast to CEs, KH7 is specific for sAC relative to tmACs^[@R27]^, and it has grown into the most widely used reagent for identifying cAMP signaling functions mediated by sAC. KH7's mechanism of action remains unknown, and it exhibits non-specific cellular effects^[@R30],[@R31]^. To identify additional sAC specific inhibitors to use as chemical probes and if sufficiently safe, for potential therapeutic applications, we developed a novel mass-spectrometry-based high throughput screening method. Using this screen, we identified an improved sAC-specific inhibitor LRE1. LRE1 inhibits sAC via a unique mechanism; it occupies the binding site of the physiological activator HCO~3~^−^ and allosterically inhibits sAC. LRE1 effectively inhibits sAC-mediated functions in sperm and mitochondria, and it is non-toxic to cells. Identification of LRE1 as a second sAC-specific inhibitor facilitates studies of the physiological roles played by sAC-defined intracellular and intraorganellar cAMP microdomains and fosters a greater understanding of CO~2~/HCO~3~^−^/pH~i~ chemosensing. Furthermore, because it has more favorable properties, LRE1 may enable development of first-of-their-kind sAC inhibition therapeutic strategies. RESULTS {#S2} ======= Development of a novel adenylyl cyclase screening assay {#S3} ------------------------------------------------------- We developed a mass spectrometry (MS) based assay that accurately quantifies levels of both product cAMP and substrate ATP in a single sample over several orders of magnitude. Thus, by correlating cAMP production ([Fig.1A](#F1){ref-type="fig"}) with reduction of input ATP ([Fig. 1B](#F1){ref-type="fig"}), this MS-based assay provides assurance that the measured product derives from substrate similar to traditional, radioactivity-based assays where \[α^32^P\] ATP is converted into \[^32^P\] cAMP^[@R32]^. The RapidFire 365 High-throughput MS System (Agilent Technologies; RF-MSS) can process samples every 15 seconds allowing analysis of a 384 well plate in under two hours; thus, RF-MSS provides a platform for high throughput MS screening which can simultaneously measure both cAMP produced and ATP consumed in an individual sAC assay. Using the RF-MSS we observed no appreciable sample carryover between assay samples, and we were able to detect both a sAC-dependent increase in cAMP signal ([Fig. 1A](#F1){ref-type="fig"}) and decrease in ATP signal ([Fig. 1B](#F1){ref-type="fig"}) over time. Activity of sAC measured using this RF-MSS based adenylyl cyclase assay was consistent with sAC's known kinetics^[@R9]^. When measured in the presence of increasing concentrations of substrate ATP in the presence of Mg^2+^ as the sole divalent cation, HCO~3~^−^ stimulated sAC activity by increasing the V~max~ with little effect on the apparent K~m~ for substrate ATP-Mg^2+^ ([Fig. 1C](#F1){ref-type="fig"}). Addition of Ca^2+^ stimulated sAC activity by decreasing its apparent K~m~ for ATP, and addition of both Ca^2+^ and HCO~3~^−^ synergistically activated sAC. These data confirm that using RF-MSS as a tool for measuring in vitro adenylyl cyclase activity is comparable to the classic, radioactivity based, "two-column" method^[@R32]^ for measuring adenylyl cyclase activity. For screening, we chose in vitro assay conditions which reflect physiologically stimulated enzyme (i.e., in the presence of Mg^2+^, Ca^2+^, and HCO~3~^−^) to maximize our chances for identifying therapeutically useful sAC inhibitors. Under these conditions, the cAMP generated in the presence of active human sAC protein compared to the RF-MSS signal for cAMP in the presence of inactive sAC yields a Z' score of \~0.7 ([Fig. 1D](#F1){ref-type="fig"}). Thus, to screen for inhibitors, we used denatured sAC protein as "positive control." Under these conditions, a pilot screen of the LOPAC reference library (Sigma) containing small organic compounds with well-documented pharmacological activities also yielded a Z' factor of \~0.7 ([Fig. 1E](#F1){ref-type="fig"}). Repeated screening of individual plates of the LOPAC library confirmed reproducibility with an R^2^ value of 0.81 ([Fig. 1F](#F1){ref-type="fig"}). A total of 33,135 compounds, selected (as described in Methods) from a larger set of 7.87 million commercially available screening compounds to maximize coverage of chemical space and "drug-likeness" properties, were screened at 10 µM for their ability to inhibit in vitro human sAC activity (details summarized in [Supplementary Results, Supplementary Table 1](#SD1){ref-type="supplementary-material"}). From the 33,135 compounds screened, 46 compounds which inhibited sAC activity by ≥ 35% (0.14% hit rate) were tested for reproducibility and concentration-dependency. To avoid assay dependent artifacts, reproducibility and concentration-dependency were simultaneously assessed in the RF-MSS cyclase assay as well as in an independent colorimetric competitive enzyme immunoassay (i.e., ELISA-based Correlate-EIA Direct cAMP Assay). Four (4) of the 46 compounds proved to be concentration-dependent sAC inhibitors in both types of assay ([Supplementary Table 2](#SD1){ref-type="supplementary-material"}). Two compounds, RU-0204277 (1) and RU-0207148 (2), reproducibly had IC~50~s in the low µM range, which is comparable to KH7^[@R13],[@R27]^. We decided to focus on RU-0204277 due to its lower molecular weight, better Q.E.D. score, and unique structural features relative to KH7. The systematic name of RU-0204277 is 6-chloro-N4-cyclopropyl-N4-\[(thiophen-3-yl)methyl\]pyrimidine-2,4-diamine; hereafter, we refer to it as LRE1. The purity and identity of the "cherry picked" compounds were confirmed by LC-MS, and both commercially obtained LRE1 and independently synthesized LRE1 inhibited sAC in vitro activity with similar potency as the "cherry picked" compound. The concentration-response of independently synthesized LRE1 on sAC protein is shown in [Fig. 2A](#F2){ref-type="fig"}. LRE1 is specific for sAC {#S4} ------------------------ LRE1 is not a "frequent hitter;" it was negative in 17 other in vitro and cell-based screens against various targets including other nucleotidyl cyclases, proteases, ion channels, and receptors. In mammals, the enzymes most related to sAC are the G protein-regulated transmembrane adenylyl cyclases (tmACs)^[@R5]^. Nine genes encode tmAC isozymes in mammals, but among these, tmACs I, III, and VIII define a highly homologous and biochemically related subclass; tmACs II, IV, and VII define a second subclass; and tmACs V and VI define a third^[@R33]^. Therefore, to assess LRE1's selectivity for sAC relative to other mammalian adenylyl cyclases, we tested whether LRE1 affected the in vitro G protein stimulated activities of heterologously expressed tmACs I, II, V, VIII, and IX. At 50 µM, which completely inhibited sAC, LRE1 did not inhibit any tmAC ([Fig 2B](#F2){ref-type="fig"}). LRE1 is efficacious in cellular systems {#S5} --------------------------------------- We next tested whether LRE1 could inhibit sAC in intact cells. Cellular levels of cAMP reflect a balance between its synthesis by adenylyl cyclases and its catabolism by phosphodiesterases (PDEs). When cells are grown in the presence of a cocktail of PDE inhibitors, they accumulate cAMP which reflects the activity of endogenous adenylyl cyclases. To assess adenylyl cyclase inhibitor efficacy in cellular systems, we used 4-4 cells, which stably overexpress sAC in a HEK293 cell background. The cAMP accumulation in 4-4 cells in the presence of PDE inhibitors is almost exclusively due to sAC^[@R10],[@R27]^. LRE1 inhibited cAMP accumulation in 4-4 cells with an IC~50~ of 11 µM ([Fig. 2C](#F2){ref-type="fig"}), which is comparable to KH7 and which is in good agreement with LRE1's ≤ 10 µM IC~50~ on sAC protein ([Fig. 2A](#F2){ref-type="fig"}). To confirm LRE1's specificity for sAC relative to tmACs in a cellular context, we used a second cell line, immortalized mouse embryo fibroblasts derived from sAC knockout mice (sAC KO MEFs). sAC KO MEFs are devoid of sAC, and they express a mixture of tmAC isoforms, specifically tmAC Types I, III, IV, VI, VII, VIII, and IX^[@R27]^. LRE1 was inert towards forskolin-stimulated cAMP accumulation in sAC KO MEFs ([Fig. 2D](#F2){ref-type="fig"}). Together, these assays identify LRE1 as a sAC-specific inhibitor suitable for use in cell-based assays. We demonstrated LRE1's ability to inhibit a physiological sAC-dependent cAMP response. Pancreatic β cells sense serum glucose and respond to elevated glucose levels by releasing insulin. We previously demonstrated that sAC is essential for glucose sensing^[@R24]^ and for glucose dependent insulin secretion^[@R10]^. INS-1E cells are an insulinoma cell line which recapitulate β cell glucose sensing^[@R34]^, and we previously showed that while tmACs are largely responsible for the basal cAMP production in INS-1E cells growing in low glucose, sAC is responsible for the increased cAMP production in cells grown in high glucose^[@R24]^. Consistent with it being a sAC specific inhibitor, LRE1 effectively blocks the glucose induced cAMP response in INS-1E cells, while having little effect on the basal, tmAC-dependent cAMP synthesized in low glucose ([Fig. 2E,F](#F2){ref-type="fig"}). LRE1 inhibits sAC allosterically {#S6} -------------------------------- To obtain molecular insights into the LRE1 binding site and inhibition mechanism, we solved the crystal structure of a sAC/LRE1 complex. LRE1 was soaked into sAC-cat apo crystals, and excellent additional electron density for the LRE1 ligand was visible after molecular replacement phasing of the 1.79 Å resolution diffraction data ([Supplementary Table 3](#SD1){ref-type="supplementary-material"}). The final sAC/LRE1 complex showed good refinement statistics (R/R~free~ = 16.1/20.3%) and a ligand geometry well-defined by electron density ([Fig. 3A,B](#F3){ref-type="fig"}; [Supplementary Fig. 1A](#SD1){ref-type="supplementary-material"}). LRE1 occupies the HCO~3~^−^ binding site (BBS) between Lys95 and Arg176, both of which are essential for HCO~3~^−^ regulation^[@R7]^. The inhibitor extends into a channel connecting the BBS to the active site (see below). The overall B-value of 20.7 Å^2^ for LRE1 corresponds to the B-values of its protein environment and indicates a tightly bound ligand. The substituted pyrimidine ring of LRE1 occupies a rather hydrophobic pocket formed by Phe165, Leu166, Leu102, Val167, Met337 and Lys95 ([Fig. 3B,C](#F3){ref-type="fig"}). Hydrogen bonds are formed by the backbone oxygens of Val167 (2.7 Å) and Met337 (2.9 Å) to the amino group of the LRE1 pyrimidine, and by the Val167 α-amino group to a ring nitrogen. The chlorine substituent points into a highly hydrophobic pocket formed by Leu102, Val167, Phe165, Leu166, and the side chain carbons of Lys95, and it might for a weak dipole-charge interaction with the Lys95 amino group. The LRE1 cyclopropene ring is accommodated in a hydrophobic pocket lined by Phe336 and Phe45/Ala97. Phe45 also interacts, through a T-shaped π-stacking interaction, with the LRE1 thiophene ring, which is further surrounded by Val175/172, Phe338, and Arg176. This Arg is a "trigger arm," which links BBS and substrate binding site^[@R7]^. In the LRE1 complex, the Arg176 side chain is in a unique orientation; it points away from active site and BBS and is fixed in this orientation through polar interactions with Asp339 and Asn180. Consistent with LRE1 occupying the allosteric regulatory region mediating HCO~3~^−^-dependent sAC activation, kinetic experiments revealed that LRE1 inhibition is competitive with HCO~3~^−^ ([Fig. 4A](#F4){ref-type="fig"}). Overlaying the sAC/LRE1 complex with a sAC complex in the presence of HCO~3~^−^ (PDB code 4cll)^[@R7]^ shows that part of the LRE1 pyrimidine ring assumes the positions of the HCO~3~^−^ atoms ([Fig. 4B](#F4){ref-type="fig"}) and appears to simulate their polarity pattern. However, the bulky LRE1 is not compatible with the Arg176 conformation assumed to be relevant for sAC activation^[@R7]^. For its interaction with the LRE1 thiophene, Phe338 from the β2--β3 loop is pulled \~3 Å toward the thiophene pocket, away from its normal position next to the active site. The LRE1 thiophene itself and the rearranged Phe338 block the region accommodating Arg176 during HCO~3~^−^ binding. Due to the size of LRE1, Lys95, which forms the BBS border opposite from Arg176, is also shifted \~2.1 Å away from its normal position. These distortions explain why occupying the BBS with LRE1 does not stimulate sAC activity. To investigate how LRE1 inhibits basal sAC activity (i.e., in the absence of HCO~3~^−^), we compared the inhibitor complex to a sAC-cat complex with the substrate analog α,β-methylene-ATP (ApCpp; PDB code 4clk)^[@R7]^. The closest distance between inhibitor and substrate analog would be a 6.4 Å gap between the LRE1 cyclopropene and the ApCpp ribose ([Supplementary Fig. 1B](#SD1){ref-type="supplementary-material"}). Thus, the inhibitor does not appear to overlap with the substrate binding site. Consistently, kinetic experiments revealed LRE1 inhibition to be non-competitive with the substrate ATP ([Fig. 4C](#F4){ref-type="fig"}). We thus solved a sAC structure after soaking with both substrate analog ApCpp and the inhibitor LRE1. In this structure, solved at 1.86 Å resolution ([Supplementary Table 3](#SD1){ref-type="supplementary-material"}), both ligands are well defined by electron density ([Fig. 4D](#F4){ref-type="fig"}). The inhibitor occupies an identical position as it occupies in the binary sAC/LRE1 complex, while part of ApCpp binds differently in the trimeric complex relative to how it binds in the sAC/ApCpp complex ([Fig. 4E](#F4){ref-type="fig"}). The adenosine moiety of the substrate analog is shifted toward the area vacated by the Arg176 reorientation, and it directly interacts with the inhibitor. Phe336 and Phe338, which are shifted due to direct interactions with the inhibitor, appear to be major factors contributing to the adenosine relocation upon LRE1 binding. Phe336 is shifted into the original adenosine binding site which prevents regular ATP binding, and Phe338 is moved away from the active site and tears the directly interacting substrate base away from its original position. Thus, while LRE1 is not competitive with substrate binding ([Fig. 4C](#F4){ref-type="fig"}), its presence dramatically affects ATP's position in the active site forcing it into an orientation inconsistent with catalysis. Comparison of sAC/LRE1 with a tmAC/forskolin structure ([Supplementary Figure 1C](#SD1){ref-type="supplementary-material"}; PDB code 1cjt) illustrates why LRE1 is specific for sAC. The bulky tmAC activator forskolin fits into the much deeper and wider regulator binding site in tmACs, but it does not fit in the tighter sAC activation site^[@R7]^. The position of LRE1 in sAC overlaps only with the small portion of the forskolin site in tmACs which directly corresponds to the BBS. LRE1 extends into the access channel toward the BBS, and the relatively small LRE1 is well suited for this tight channel. In contrast, LRE1 would be unable to fill most of the large activator site in tmACs, thus lacking sufficient interactions for tight binding to tmACs. LRE1 inhibits sperm and mitochondrial functions of sAC {#S7} ------------------------------------------------------ To assess its efficacy as a physiologically relevant inhibitor, we tested whether LRE1 could inhibit known sAC-dependent functions. The first identified role of sAC was in sperm^[@R35]^. Mammalian sperm acquire fertilization capacity as they transit through the reproductive tract in a process known as capacitation. At the molecular level, an early event in capacitation is HCO~3~^−^-induced activation of cAMP synthesis by sAC followed by the consequent activation of the main cAMP effector, Protein Kinase A (PKA). Activation of sAC and PKA precedes stimulation of a prototypical pattern of tyrosine (tyr) phosphorylation, induction of a "hyperactivated" form of motility, and finally competence to fertilize^[@R35]^. sAC activity has been genetically^[@R13],[@R36]^ and pharmacologically^[@R13]^ demonstrated to be essential for these physiological changes sperm undergo during capacitation. In sperm, LRE1 concentration-dependently blocks PKA dependent phosphorylation of its substrates ([Fig. 5A](#F5){ref-type="fig"}; [Supplementary Figure 2A](#SD1){ref-type="supplementary-material"}), as well as the downstream increase in tyr phosphorylation ([Fig. 5B](#F5){ref-type="fig"}; [Supplementary Figure 2B](#SD1){ref-type="supplementary-material"}), hyperactivation ([Fig. 5C](#F5){ref-type="fig"}), and fertilization ([Fig. 5D](#F5){ref-type="fig"}). Consistent with its effective concentration at inhibiting sAC in isolated cells ([Fig. 2B,D](#F2){ref-type="fig"}), LRE1 inhibits these processes with an IC~50~ of \~10 µM ([Supplementary Figure 2](#SD1){ref-type="supplementary-material"}). KH7 also blocked each of these sAC-dependent functions^[@R13]^, but because the inhibitory activity of KH7 is quenched by the presence of BSA, which is an important component of capacitation-media^[@R37],[@R38]^, KH7's effect on capacitation had to be studied using other cholesterol-binding compounds (e.g., beta cyclodextrin)^[@R13]^. In contrast to KH7, LRE1 inhibition was unaffected by the presence of BSA in the incubation media. Finally, confirming that LRE1 mediates its effects exclusively via inhibition of sAC, these effects of LRE1 can be rescued by addition of exogenous membrane permeable cAMP ([Fig. 5A,B,D](#F5){ref-type="fig"}). A second well established function of sAC is regulation of mitochondrial respiration. sAC resides in the mitochondrial matrix where it senses Krebs cycle generated CO~2~ and hormonally stimulated calcium signals to regulate the activity of the electron transport chain^[@R39]^. Among other potential sites of regulation^[@R40]^, sAC generated cAMP stimulates the activity of cytochrome c oxidase (COX)^[@R41]^. LRE1 inhibits COX activity in WT MEFs, and consistent with LRE1's effects being solely mediated via inhibition of sAC, COX activity in sAC KO MEFs was unaffected by LRE1 ([Fig. 5E](#F5){ref-type="fig"}). LRE1 is non-toxic and does not uncouple mitochondria {#S8} ---------------------------------------------------- Reliable chemical probes and effective therapeutic compounds should have limited toxicity to cells with as few off-target effects as possible. Thus far, pharmacologic experiments exploring sAC function with the widely-used sAC specific inhibitor KH7 have focused on short-term assays or single doses in vivo^[@R22]^. In our own experience, we have observed non-sAC mediated toxicity in cells treated with KH7 for prolonged periods of time. We used the Cell-Titer Glo high throughput viability assay (Promega Inc.) to directly compare cellular toxicity of LRE1 and KH7. While neither compound exhibited significant toxicity after a single day in culture ([Supplementary Figure 3A](#SD1){ref-type="supplementary-material"}), after 2 days, KH7 was toxic to MEFs derived from WT C57Bl/6 mice at low µM concentrations ([Supplementary Figure 3B](#SD1){ref-type="supplementary-material"}). This toxicity was independent of sAC because KH7 exhibited a similar concentration-dependent toxicity in MEFs derived from sAC KO mice ([Supplementary Figure 3C](#SD1){ref-type="supplementary-material"}). In contrast, LRE1 exhibited no significant toxicity after two days in culture ([Supplementary Figure 3B,C](#SD1){ref-type="supplementary-material"}) at concentrations (i.e., up to 50 µM) which effectively inhibited sAC in cellular systems ([Fig. 2C](#F2){ref-type="fig"}, and [Fig. 5](#F5){ref-type="fig"}). These data identify LRE1 as the first pharmacological tool suitable for probing sAC functions in long-term assays. KH7 has been reported to lead to mitochondrial uncoupling in a sAC-independent manner^[@R30]^. Under conditions which reproduced KH7's uncoupling effect, LRE1 exhibited no uncoupling of isolated mouse brain mitochondria ([Supplementary Figure 3D](#SD1){ref-type="supplementary-material"}) and did not appreciably perturb bilayer properties (O.S. Andersen, personal communication). These studies suggest that LRE1 is a chemically inert nontoxic compound that does not interfere with essential cellular pathways. DISCUSSION {#S9} ========== Historically, adenylyl cyclase activity has been studied using two different types of in vitro assays. The classical method for measuring adenylyl cyclase activity involves supplying \[α^32^P\] ATP and measuring the amount of \[^32^P\] cAMP produced after separating formed product from input substrate using cyclic nucleotide separation chromatography methods^[@R32]^. This assay has exquisite specificity (the only way to generate radiolabelled product cAMP is from the input substrate ATP) and a broad dynamic range. It has been the 'standard' for measuring in vitro cyclase activity for the past \~40 years. Unfortunately, this method is not suitable for high throughput screening. For screening, numerous methods, which are dependent upon antibodies or other binding proteins which discriminate cAMP from other nucleotides, including input ATP, are available to quantitate cAMP formed after a cellular or in vitro reaction^[@R42]^. These methods are well-suited for high throughput screening, and while they can be extremely sensitive, they suffer from narrow dynamic ranges limiting their utility, especially for kinetic studies. Additionally, because sAC has a \~ 20 fold higher K~m~ for its substrate ATP relative to other mammalian adenylyl cyclases (its K~m~ for ATP is 1 mM under physiological conditions)^[@R9],[@R10]^, in vitro sAC assays are performed with mM concentrations of ATP. When input ATP is this high, the assays we have tested suffer from high backgrounds due to cross reactivity with input ATP which masked the concentrations of cAMP produced. The RF-MSS cyclase assay we have developed, based upon MS determination of both cAMP produced and ATP consumed, combines the specificity and broad dynamic range of the radioactive substrate assay with the high throughput adaptability of more modern methods. By measuring both product formation and substrate consumption, the RF-MSS assay substantiates that product is derived from input substrate. While the sensitivity and dynamic range of the assay will depend upon the particular Mass Spectrometer used, the 15 second per sample cycle time of the RapidFire delivery system provides high throughput capability. In addition, the RF-MSS is non-radioactive which provides for a safer assay with less environmental impact, and it is not dependent upon consumable and often non-renewable reagents such as antisera. Thus, the RF-MSS assay has the potential to replace previously existing adenylyl cyclase assays. Using the RF-MSS assay, we identified a new sAC-specific inhibitor, LRE1. Among the different adenylyl cyclase inhibitors known, those structurally studied and mechanistically understood target the substrate binding region (e.g., P site inhibitors) or its immediate environment (e.g., catechol estrogens, DIDS)^[@R2],[@R7]^. Crystallographic and kinetic data reveal that LRE1 specifically inhibits sAC via a novel allosteric mechanism induced through binding to its unique bicarbonate binding site (BBS). The corresponding site in tmACs is where the tmAC-specific activator forskolin binds^[@R43]^. Forskolin is considerably larger than bicarbonate ([Supplementary Fig. 1C](#SD1){ref-type="supplementary-material"}), and it does not fit in sAC's BBS^[@R7]^. This difference in size between the forskolin binding site of tmACs and the BBS of sAC allows highly specific ligand interactions explaining the selective inhibition of sAC by LRE1. LRE1 not only occupies the BBS, it also extends into the channel connecting the BBS to the active site which provides additional favorable interactions for specific and tight ligand binding. This channel is also the binding site of the structure-based designed inhibitor, ASI-8^[@R44]^ ([Supplementary Fig. 1D](#SD1){ref-type="supplementary-material"}), and it is postulated to be the site of KH7 inhibition of sAC^[@R2]^. However, both KH7 and ASI-8 are long compounds which would extend into the active site and inhibit, at least partially, via competition with the substrate. In fact, bithionol, a cytotoxic organochlorine, was recently found to inhibit sAC by binding to the BBS and this connecting channel^[@R45]^, similar to ASI-8. And bithionol indeed inhibits sAC by at least partially competing with substrate ATP. In contrast, LRE1 only occupies the connecting channel with its short cyclopropane moiety and does not reach the active site. Consistently, LRE1 is non-competitive with ATP ([Fig. 4C](#F4){ref-type="fig"}); thus, LRE1 is the only known, exclusively allosteric inhibitor of sAC. Despite this shorter cyclopropane extension (as compared to ASI-8 and KH7), LRE1 still exhibits high affinity. Presumably, its affinity is augmented by additional interactions from the branching thiophene moiety, which exploits an additional sAC pocket ([Supplementary Fig. 1D](#SD1){ref-type="supplementary-material"}). These data identify LRE1 as the first potent and selective pharmacological modulator which acts allosterically via BBS binding. Thus, our study reveals that drug-like ligands can bind tightly to the BBS and allosterically inhibit sAC, and it raises the possibility that pharmacological ligands could mimic the stimulating effect of the small physiological sAC activator bicarbonate. Understanding cAMP signaling in mammalian cells requires methods to distinguish between the two different sources of cAMP, tmACs and sAC. We previously identified pharmacological tools and conditions for selectively inhibiting tmAC-dependent cAMP generation using the P site inhibitor 2'5' dideoxyadenosine and for specifically inhibiting sAC using KH7^[@R27]^. Unfortunately, because KH7 exhibits an intrinsic fluorescence, it has limited utility for studying intracellular cAMP microdomains using fluorescence-based live cell cAMP sensors. In contrast, LRE1 does not appear to have intrinsic fluorescence. Thus, our identification of LRE1 as an improved sAC-specific inhibitor provides an essential chemical probe for the study of cAMP microdomains. Finally, the many identified adenylyl cyclase inhibitors, in particular those affecting sAC, lack high potency and/or selectivity or possess non specific cellular toxicity^[@R2]^. In contrast, LRE1 combines high potency and selectivity with stability, solubility and, most importantly, lack of cytotoxicity. In particular, LRE1 exhibits less non-specific toxicity than KH7, the widely used sAC specific inhibitor^[@R2],[@R27]^, and bithionol, a recently characterized sAC selective inhibitor^[@R45]^. Therefore, LRE1 is an ideal candidate to explore the therapeutic possibilities of sAC inhibition. In summary, we have exploited the RapidFire mass spectrometry system (RF-MSS) to develop a safe and robust adenylyl cyclase assay suitable for a wide variety of applications. Using this RF-MSS cyclase assay, we identified an improved sAC specific inhibitor, LRE1, which occupies the regulatory "activator" site to allosterically inhibit the enzyme, and which exhibits pharmacologically favorable characteristics. METHODS {#S10} ======= Chemicals and Cell Lines {#S11} ------------------------ All chemicals were purchased from Sigma Aldrich unless otherwise noted. KH7 was synthesized by The Milstein Synthetic Chemistry Core Facility at Weill Cornell Medical College. LRE1 was purchased from Nanosyn and verified by The Milstein Synthetic Chemistry Core Facility by LC/MS ([Supplementary Figure 8](#SD1){ref-type="supplementary-material"}) and NMR ([Supplementary Figure 9](#SD1){ref-type="supplementary-material"}). Small molecules were purchased as powder and suspended in DMSO. Anti-phosphotyrosine (anti-PY) monoclonal antibody (clone 4G10) was obtained from Millipore, and rabbit monoclonal anti-phosphoPKA substrates (clone 100G7E) was purchased from Cell Signaling. 4-4 cells, sAC KO MEFs, and WT MEFs were generated and functionally authenticated in our laboratory as previously described^[@R27]^ and grown in DMEM + 10% FBS. INS-IE cells^[@R34]^ were obtained directly from P. Antinozzi and C. Wollheim and functionally authenticated in our laboratory^[@R10],[@R24]^. They were cultured in RPMI media containing 10% heat-inactivated FBS, 10 mM HEPES, and 1 mM sodium pyruvate, 50 µM β-mercaptoethanol, and passaged every 2--3 days. All cells were maintained at 37°C in 5% CO~2~ and are periodically checked for mycoplasma contamination. RapidFire Mass Spectroscopy {#S12} --------------------------- Analyses of cyclase assays were performed on an Agilent Rapid Fire 365 high throughput Mass Spectroscopy System (RF-MSS), equipped with an Agilent 6520 TOF accurate mass spectrometer as a detection system (Agilent Technologies). The Agilent 6520 TOF MS has a theoretical limit of high femtogram sensitivity and up to 5 orders of magnitude dynamic range. This instrument also includes a Zymark Twister robotic arm that handles microtiter plates and a purification SPE system. Sample (35 µl) was aspirated from each well of a 384-well reaction microtiter plate and injected onto a graphitized carbon SPE column extraction cartridge. Columns were washed with aqueous alkaline buffer (5 mM ammonium acetate, pH 10) and eluted using an alkaline/organic solvent (25% acetonitrile, 25% acetone in 5 mM ammonium acetate, pH 10) onto the electrospray-MS, where the mass spectra were collected. The RapidFire sipper was washed between sample injections using alkaline/organic and aqueous alkaline solvents. RapidFire mass spectrometry screening data were processed and analyzed using Agilent MassHunter Software. For in vitro cyclase assays, human sAC~t~ protein^[@R9]^ was assayed in the presence of 50 mM Tris pH 7.5, 2 mM DTT, 5 mM MgCl~2~, and the indicated concentration of ATP in the presence or absence of 5 mM CaCl~2~ and/or 40 mM NaHCO~3~. cAMP (MW 328.0412) produced and ATP (MW 505.9816) consumed were determined by RF-MSS in comparison with standards of known concentrations, and for all quantitative cAMP measurements, we confirmed there was a corresponding decrease in input ATP. Mammalian tmAC isozymes tmAC I (ADCY1; bovine), tmAC II (ADCY2; rat), tmAC V (ADCY5; rat), tmAC VIII (ADCY8; rat), and tmAC IX (ADCY9; mouse) were transfected and expressed in HEK293 cells using the CMV promoter and whole cell extracts were assayed in the presence of 100 µM GTPγS via the classical two column radioassay measuring conversion of \[α^32^P\] ATP into \[^32^P\] cAMP as described previously^[@R46]^. To determine tmAC-specific activities, the activities of empty vector transfected lysates were subtracted. The activity in vector transfected HEK293 lysates in the presence of 100 µM GTPγS alone was 5.5 ± 0.8 nmol cAMP/min; its activity in the presence of 100 µM GTPγS + 50 µM LRE1 was 7.7 ± 1.7 nmol cAMP/min. Small Molecule Library {#S13} ---------------------- 33,135 compounds were selected from a larger set of 7.87 million commercially available screening compounds. The strategy for purchasing these compounds was to maximize coverage of chemical space and "drug-likeness" properties based on quantitative scores, while simultaneously minimizing cost. Pipeline Pilot software (Dassault Systemes Biovia Corp.) was used to execute a published algorithm for scoring compounds based on drug-likeness, called the quantitative exponent of drug-likeness (Q.E.D.) score^[@R47]^. From the 7.87 million commercially available compounds, 3.05 million compounds had a weighted Q.E.D. score greater than 0.7. The "choose diverse" component of the Pipeline Pilot software utilizes ECFP6 fingerprint descriptors to cluster the compounds into structurally related groups and chooses a single representative of each cluster, based on Tanimoto distances, called "cluster centers". The "choose diverse" component identified 100,000 relatively diverse compounds from among these 3.05 million structures. From these 100,000 "cluster centers," 33,135 could be purchased from Enamine Ltd. in 1 mg quantities at a price which was within our budgetary limits. These compounds were formatted in 10 copies of 5 mM DMSO stocks in 384-well deep well plates for screening. High-throughput Screening {#S14} ------------------------- The primary screen (summarized in [Supplementary Table 1](#SD1){ref-type="supplementary-material"}) was performed in polystyrene 384-well microplates (Greiner Bio One International AG). 33,135 compounds were screened at 10 µM. The reagents were dispensed using a liquid handling instrument Multidrop Combi. Assay buffer (50 mM Tris pH 7.5, 5 mM CaCl~2~, 5 mM MgCl~2~, 40 mM NaHCO~3~, 2 mM DTT) was added first on the 384-well plate (Greiner-Bio One) followed by the compounds. The small molecule compounds were dispensed with a nanohead containing a 384-well array of Norgen Kloehn nanosyringes, and each plate was bar-coded. The human sAC~t~ protein^[@R9]^ was added next followed by the remainder of the assay buffer. The stop reaction buffer (1% formic acid) was added to the last column of each plate (denatured protein; T0). Reactions were started by addition of ATP at a final concentration of 1 mM. Reactions were incubated for 3 hours at 37°C, and stopped with the addition of 1% formic acid. Cellular cAMP accumulation assays {#S15} --------------------------------- sAC KO MEFs or 4-4 cells (2.0 × 10^6^ cells/ml) in suspension were transferred to 1.5 ml tubes and incubated at 37°C, 5% CO~2~ for one hour. A time zero value for each condition was determined by adding 100 µl of cells directly into 100 l stop solution (0.2 M HCl). To measure cAMP accumulation, cells in suspension were incubated for the indicated period of time in the presence of 500 M IBMX (± LRE1 or ± KH7) at 37°C after which 100 l of cells were transferred to a fresh tube containing stop solution. Intracellular cAMP content was determined using Correlate-EIA Direct Assay (Assay Designs, Inc). INS-1E insulinoma cells were incubated in 2.5 mM glucose Krebs-Ringer buffer (pH 7.5) supplemented with 2 mM sodium bicarbonate, 10 mM HEPES, and 0.1% BSA for 2 h before start of the experiment. At time zero for each experiment, media was switched to Krebs-Ringer buffer containing 2.5 mM glucose or 16 mM glucose in the presence of 500 µM IBMX and inhibitor at the shown concentrations. After 10 min cells were lysed in 200 µl 0.1 M HCl. Intracellular cAMP contents were determined using Correlate-EIA Direct Assay (Assay Designs, Inc). Accumulated cAMP quantitated after 10 minutes are presented as pmol cAMP accumulated per 2.5 × 10^5^ cells. Crystal structure determination of sAC complexes {#S16} ------------------------------------------------ Protein for crystallization comprised the catalytic domains of human sAC (residues 1--469; sAC-cat), corresponding to the native sAC~t~ isoform, and a C-terminal his-tag. sAC-cat was expressed in Hi5 insect cells using the BIIC-method and purified by nickel affinity chromatography, anion exchange and size exclusion chromatography as described previously^[@R48]^. Apo sAC-cat was crystallized using the hanging drop method at 4 °C as described^[@R48]^. Crystals were soaked with LRE1 by transferring them into cryo solution (100 mM sodium acetate pH 4.8, 200 mM tri-sodium citrate, 18% (w/v) PEG 4000, 20% (v/v) glycerol) supplemented with 20 mM LRE1 alone or together with 150 mM ApCpp, 75 mM MgCl~2~/CaCl~2~ and incubation for 24 h at 4 °C. Crystals were flash-frozen in liquid nitrogen, and a complete dataset (space group P6~3~) was collected from one crystal at 100 K at the Berlin Electron Storage Ring society for Synchrotron Radiation beamline 14.1 (BESSY BL14.1) operated by Helmholtz-Zentrum Berlin^[@R49]^. All diffraction data were processed with XDSAPP^[@R50]^, with the resolution limit set automatically using CC^\\ [@R51]^. Molecular replacement phasing with the apo sAC structure (pdb code 4cll) as a search model, manual model building and refinement were done as described for previous sAC/ligand complexes^[@R7]^. Structure figures were generated with PyMol ([www.pymol.org](http://www.pymol.org)). Cheminformatics and Data Handling {#S17} --------------------------------- Data from all screening studies are archived and analyzed using the CDD Vault from Collaborative Drug Discovery ([www.collaborativedrug.com](http://www.collaborativedrug.com)). MarvinSketch (ChemAxon, version 14.9.8.0) was used for drawing and naming chemical structures. Instant Jchem for Excel (ChemAxon, Budapest Hungary, version 15.8.24.0) was also used for creating structure spreadsheets and tables. Compound similarity search was performed using Pipeline Pilot version 9.2 (Dassault Systemes Biovia Corp.). Graphpad PRISM ([www.graphpad.com](http://www.graphpad.com), version 6.05) was used for curve fitting. Sperm Assays {#S18} ------------ Mouse sperm were isolated as described^[@R52]^. Sperm were incubated in Toyoda--Yokoyama--Hosi (standard TYH) medium buffered with 25 mM bicarbonate in 5 % CO~2~ for sperm capacitation and fertilization assays. Western blot and motility assays were performed in Hepes-buffered TYH (H-TYH). For capacitation, 15 mM NaHCO~3~ and 5 mg/ml BSA was added, and sperm incubated at 37 °C for at least 1 hour. For LRE1 treatment, sperm were pretreated for 10 min with LRE1 before addition of capacitation medium and/or dibutyryl cAMP. Anti-PKA substrates and anti-pY western blots, in vitro* fertilization (IVF), and sperm motility analysis using a CEROS computer-assisted semen analysis (CASA) system (Hamilton Thorne Research) were performed as described^[@R52]^. At least 20 microscopy fields corresponding to a minimum of 200 sperm were analyzed in each motility experiment. For fertilization, wells containing 25--40 eggs were inseminated with sperm that had been incubated for 1 h and 20 min in capacitation supporting Whitten's medium with the shown amounts of LRE1. Sperm in the presence of 100 µM LRE1 were also incubated in the presence of 1 mM dibutyryl cAMP. After 4 h of insemination, eggs were washed and put in fresh Whitten's media. The eggs were evaluated 24 h post-insemination. To assess fertilization, the following criteria were considered: (1) the formation of the male and female pronuclei, (2) the emission of the second polar body, and (3) two-cell stage events. Mitochondrial COX activity {#S19} -------------------------- Wild type and sAC KO MEFs were plated in P10 dishes. At 70% of confluence cells were incubated at 37°C with DMSO or LRE1 50µM for 30 min in serum free medium. Subsequently the cells were scraped in PBS and centrifuged for 5 min at 1000 rpm. The pellet was snap frozen in liquid nitrogen and stored at −80° C. Before the experiment the pellets were resuspended in PBS, COX activity was measured as described previously^[@R53]^ and normalized to protein concentration. Cytotoxicity Assays {#S20} ------------------- Cells were plated in 384- or 96-well plates. The next day, compounds were added (ten concentrations), and the plates were incubated for 24 or 48 hrs at 37°C/ 5% CO~2~, and viability was quantified by measuring cellular ATP using luminescence (CellTiter-Glo, Promega). Measurement of the membrane potential in isolated mouse brain mitochondria {#S21} -------------------------------------------------------------------------- Mouse brain mitochondria were isolated by the Percoll gradient method as described^[@R54]^ and re-suspended in 90 µl of isolation medium containing 225 mM mannitol, 75 mM sucrose, 20 mM HEPES-KOH (pH 7.4), 1 mM EGTA, and 0.2 mg/ml fatty acid-free bovine serum albumin (BSA). The membrane potential changes were assessed from the changes in fluorescence of a permeating cation Safranin O^[@R55]--[@R57]^ at 495 nm excitation and 586 nm emission wavelength with Hitachi F-7000 spectrofluorimeter (Hitachi). The incubation medium was composed of 125 mM KCl, 20 mM HEPES (pH 7.4), 4 mM KH~2~PO~4~, 0.5 mM EGTA, 0.2 mg/ml BSA, 2 µM Safranin O (membrane potential probe), 5 mM Pyruvate, 2.5 mM malate, and 0.1 mg/ml mitochondria protein. Supplementary Material {#SM} ====================== Accession Codes Crystal structures and diffraction data have been deposited with the worldwide protein data bank (wwPDB) under accession codes 5IV4 (sAC/LRE1) and 5IV3 (sAC/LRE1/ApCpp). Author contributions L.R.E., S.K., C.S., F.G., L.R.L., and J.B. designed the research project and analyzed data; L.R.E. and C.A. performed the HTS; S.K. and C.S. solved the crystal structure; J.v.H. expressed protein; L.R.E. performed the kinetic and cellular inhibition studies; H.B. performed the assays of individual tmACs; F.A.N., A.A., and P.E.V. performed the sperm studies; F.V., A.S., and G.M. performed the mitochondrial studies; and L.R.E. and J.H.Z. performed the cytotoxicity studies. All authors contributed specific parts of the manuscript, with L.R.E., L.R.L. and J.B. assuming responsibility for the manuscript in its entirety. Conflict of interest Drs. Buck, Levin and Zippin own equity interest in CEP Biotech which has licensed commercialization of a panel of monoclonal antibodies directed against sAC. All other authors declare that they have no conflicts of interest with the contents of this article. We thank Dr. David Warren and The Milstein Synthetic Chemistry Core Facility at Weill Cornell Medical College and Drs. Steven S. Gross and Qiuying Chen. CS and SK thank the BESSY staff and the Helmholtz Protein Sample Production Facility (PSPF) for technical assistance. L.R.E. and F.G. thank the HTSRC staff, the Helmsley Trust for funding of the Rapid-Fire Automated SPE TOF Mass-Spectrometry Instrument, and Agilent staff, especially to Peter Rye and Lauren Frick, for technical assistance. This work was supported, in whole or in part, by National Institutes of Health grants GM107442 and EY025810 and by the Weill Cornell Medicine CTSC, NIH/NCATS Grant UL1TR00457 (to L.R.L. and J.B.); NIH grants HD38082 and HD44044 (to P. E. V.), R01GM088999 (to G.M.), and K08 CA160657 (to J.H.Z.); Melanoma Research Alliance Team Science and Clinique Clinical Scholars Awards (to J.H.Z.); and Deutsche Forschungsgemeinschaft grant STE1701/11 (to C.S.). ![Validation of RapidFire Mass Spectrometry System (RF-MSS) cyclase assay and high throughput screening conditions\ An example of a RF-MS chromatogram showing (a) the product cAMP as extracted ion intensity (EIC) × 10^3^, and (b) substrate ATP as EIC × 10^4^. Peaks derive from increasing input ATP (in mM) recorded at time "0" (black) and after 180 minutes incubation with sAC (red). (c) Human sAC activity measured using RF-MSS as a function of substrate of ATP-Mg^2+^ for 120 min in the presence of excess MgCl~2~ (20 mM). Shown are curves for Mg^2+^ alone (red dots), Mg^2+^/40 mM HCO~3~^−^ (green squares), Mg^2+^/10 mM Ca^2+^ (blue triangles), and Mg^2+^/40 mM HCO~3~^−^/10 mM Ca^2+^ (purple triangles). Determinations are representative of at least two independent experiments and curves are nonlinear fits generated by Prism. (d--f) RF-MSS screen assaying sAC in the presence of 1 mM ATP/5 mM MgCl~2~/5 mM CaCl~2~/40 mM NaHCO~3~. (d) Comparison of active (black dots) and denatured (open squares) sAC enzyme: Z' score = 0.7. (e) LOPAC library (1280 compounds) pilot screen (black); DMSO control (red); denatured sAC protein (blue). (f) Results of screening LOPAC library twice. LOPAC compounds (gray); DMSO control (green square); denatured sAC protein (red). (R^2^ = 0.81).](nihms789743f1){#F1} ![LRE1 is a potent and selective inhibitor of sAC in vitro and in cells\ (a) Concentration-response curve of LRE1 on sAC protein in the presence of 1 mM ATP/5 mM MgCl~2~/5 mM CaCl~2~/40 mM NaHCO~3~. Values are averages of triplicate determinations with SEM. (b) AC activities of 293 cells overexpressing each of the indicated tmACs in the absence (black bars) or presence of 50 µM LRE1 (red bars). Shown are representative assays repeated at least three times; values represent averages of quintuple determinations of tmAC specific activity with SEM. (c) Concentration-response of LRE1 (red squares) and KH7 (black circles) on cellular accumulation of cAMP in sAC overexpressing 4-4 cells. (d) Concentration-response of LRE1 on cellular accumulation of cAMP in forskolin-stimulated sAC knockout fibroblasts. (c,d) Representative assays repeated at least three times; values are means ± SEM of triplicate determinations normalized to activity in the absence of any compound. (e) Concentration-response of LRE1 on glucose-induced cAMP production in INS-1E cells in 2.5 mM (low) or 16 mM (high) glucose with 0 (black), 10 (blue), 30 (cyan), 50 (red), or 100 (brown) µM LRE1. Shown are means ± SEM of triplicate determinations of at least three independent experiments. (f) The glucose-induced cAMP, calculated from panel (e) by subtracting the cAMP accumulated in low glucose from the cAMP accumulated in high glucose, in the presence of the indicated concentration of LRE1.](nihms789743f2){#F2} ![Crystal structure of sAC/LRE1 complex\ (a) Overall structure of the sAC/LRE1 complex, with C~1~ in grey and C~2~ in cyan. LRE1 is shown in stick representation, and active site and BBS are indicated. (b) Close-up of the BBS with ligand and interacting residues as sticks colored according to atom type. LRE1 is overlaid with 2F~o~-F~c~ electron density (blue) contoured at 1σ. (c) Interaction scheme for sAC and LRE1. Interactions of side chains are indicated by black dots and by arcs for aromatic residues, and backbone interactions by dashed lines.](nihms789743f3){#F3} ![Mechanistic characterization of sAC inhibition by LRE1\ (a) LRE1 is competitive with HCO~3~^−^. sAC activity, measured by RF-MSS in the presence of the indicated concentration of LRE1 and 1 mM ATP/5 mM MgCl~2~/5 mM CaCl~2~ with 0 (red); 10 (purple squares); 20 (blue triangles); or 80 (orange diamonds) mM NaHCO~3~. Data represent means of triplicate determinations of an experiment repeated twice. (b) Overlay of sAC/LRE1 (grey) with a sAC/bicarbonate complex (cyan, RMSD 0.3 Å^5^ for 352 C~α~ atoms). The ligands and the two key residues for bicarbonate binding are shown as sticks colored according to atom type. (c) LRE1 does not compete with ATP. sAC activity, measured by RF-MSS in the presence of the indicated concentration of LRE1 and either 0.6 (pink diamonds), 0.9 (cyan triangles), 1.3 (blue squares), or 2.0 (red dots) mM ATP and 5 mM MgCl~2~/5 mM CaCl~2~/40 mM NaHCO~3~. Data represent individual points of a serial dilution of an experiment repeated three times. (d) Crystal structure of a sAC/ApCpp/LRE1 complex. The substrate analog ApCpp in the active site and the inhibitor in the BBS are shown as sticks colored according to atom type and overlaid with 2F~o~-F~c~ electron density contoured at 1σ (blue). Ca^2+^ is shown as yellow sphere. (e) Overlay of sAC/ApCpp/LRE1 (green) with a sAC/ApCpp complex (blue, RMSD 0.31 Å^2^ for C~α~ atoms). Ligands and key interacting residues are shown as sticks, and Ca^2+^ as a yellow sphere.](nihms789743f4){#F4} ![LRE1 inhibits sAC dependent processes in sperm and mitochondria\ (a--d) Sperm experiments. (a) Western blot using anti-PKA substrates antibodies of mouse cauda sperm activated by incubation in capacitation (Cap) media in the presence of the shown amounts of LRE1 in presence (+) or absence (−) of dibutyryl cAMP (1 mM) and IBMX (100 µM). N.Cap = non capacitated negative control. (b) Western blot using anti-phospho tyrosine antibodies of the same blot as in (a). For a,b, shown are representative Western blots of experiments repeated three times using independent sperm preparations from different mice. The complete gels used for these images are shown in [Supplementary Figures 4 and 5](#SD1){ref-type="supplementary-material"}, respectively. (c) The percentage of hyperactivated motility in sperm capacitated in the presence of the indicated concentration of LRE1 (black). Non capacitated sperm (red). Values are averages (± S.E.M.) of three independent sperm preparations from three different mice collected and analyzed on separate days. (d) Percentage of fertilized eggs from sperm capacitated in the presence of the indicated concentration of LRE1 (black bars) or sperm capacitated in the presence of 100 µM LRE1 + 1 mM dibutyryl cAMP (red bar). Values are averages (± S.E.M.) of four independent sperm preparations from four different mice collected and analyzed on separate days. (e) Cytochrome c oxidase (COX) activities in cells from WT and sAC KO MEFs treated with DMSO or 50 µM LRE1 for 30 min. Values are averages (± S.D.); N=5. COX activity was statistically different in WT cells ± LRE1 (P=0.037) by t-test.](nihms789743f5){#F5}	{ "pile_set_name": "PubMed Central" }
###### Strengths and limitations of this study - The Safety Attitudes Questionnaire (SAQ) was easily answered and consistently well explained. Second, the group administration fully supported the survey, and parallel studies showed an ongoing commitment to patient safety and maintaining a safety culture. Third, all of the respondents were anonymous. Although anonymity may have allowed the participants to feel more comfortable about completing the questionnaire, we cannot rule out the possibility that some potential participants hesitated to participate for fear of persecution or job harm. - We are aware of several limitations of our study. The SAQ tool relies strongly on self-reported behaviour. The resulting information may be biased and not correctly reflect the actual situation. A positive-answering tendency cannot be excluded. Staff perceptions of communication can vary over time and can be influenced by day-to-day events within the hospital setting. The external validity of the study findings was limited by the study design and participants. Finally, we did not examine variations in perceptions among departments or across hospital settings. Background {#s1} ========== Patient safety is viewed as a crucial component of quality in healthcare service.[@R1] Over the last decade, numerous definitions of patient safety have emerged in the literature. The Institute of Medicine[@R2] described patient safety as the prevention of harm. However, the European agency Safety Improvement for Patients in Europe asserted that patient safety focuses on identifying, analysing and minimising patient risk.[@R3] Several studies have noted patient safety issues in different contexts. For example, study results from the USA revealed that one-fifth of the people in a community in New York reported that either they or someone in their household had experienced a medical error[@R4] (an adverse event is defined as an injury resulting from a medical intervention and not caused by an underlying medical condition).[@R5] European data, mostly from European Union (EU) Member States, show that medical errors and healthcare-related adverse events occur in 8--12% of hospitalisations. Infections associated with healthcare affect an estimated 1 in 20 hospital patients on average every year (an estimated 4.1 million patients). The UK National Audit Office estimates the cost of such infections at £1 billion/year.[@R6] A recently released European Commission report titled Patient Safety in EU: 2014 elucidated an array of occurrences related to healthcare-associated infections that are directly responsible for 37 000 deaths/year, contribute to a further 110 000 deaths/year and cost hospitals more than €5.4 billion/year.[@R7] In healthcare, a significant percentage of errors are attributed to communication breakdowns and a lack of effective teamwork.[@R8] Furthermore, poor communication and ineffective teamwork are factors that contribute to the occurrence of patient safety incidents.[@R8] Effective teamwork and communication are considered critical for ensuring high reliability and the safe delivery of care. Teamwork and communication techniques can improve quality and safety, decrease patient harm, promote cross-professional collaboration and the development of common goals, decrease workload issues, and improve staff and patient satisfaction.[@R8] To this end, hospitals need to assess patient safety and promote teamwork principles to create safe hospital systems.[@R5] [@R12] [@R13] The transitional Albanian health system {#s2} ======================================= The Albanian health system {#s2a} -------------------------- Following various reforms that began in 1995 and have gained pace in recent years, the Albanian Health Care System moved from a typical Semashko model to a Bismarck model.[@R14] The decentralisation of primary care management, the complete privatisation of the pharmaceutical sector and dentistry and the founding of the Health Insurance Institute (HII) were the main milestones of these reforms. The health system is funded through a mix of general tax revenues, payroll tax revenues for the compulsory HII, voluntary prepayment for Voluntary Health Insurance (offered by HII), out-of-pocket payments made at the time of service use and various international donors.[@R14] Healthcare in Albania remains mainly public/state provided and is only partly privately provided. It is divided into three levels: primary, secondary and tertiary healthcare services. Healthcare services cover the entire country and are directed by the Ministry of Health.[@R15] The Ministry of Health has been rapidly changing from its traditional role as a 'health directorate' to a leadership role in health policy development and health strategy implementation. However, the Ministry of Health remains the major healthcare financing body, providing two-thirds of the total healthcare budget. The Ministry of Health is also a policymaker, decision maker and manager, and it leads human resources and training.[@R14] There are 4577 physicians in Albania and 709 inhabitants per physician.[@R17] The Albanian hospital decentralisation process {#s2b} ---------------------------------------------- Albania is engaged in health reform initiatives that aim to introduce primary healthcare centred on family medicine to enhance the performance of the health system and to cope with a broader political agenda.[@R18] There is also a focus on hospital decentralisation reforms as part of an overall institutional decentralisation process.[@R19] Since the beginning of 2009, HII has had contracts with 39 hospitals in Albania: 1 tertiary hospital (in Tirana), 3 university hospitals (in Tirana), 11 regional hospitals and 24 district hospitals. The hospitals are financed according to a historical budget. In 2010, the contract between the HII and the hospitals also included elements related to quality and performance indicators; however, monitoring these indicators seems very difficult, if not impossible, because clinical protocols and medical hospital standards are not yet available. The hospitals in Albania are not yet entirely accredited.[@R14] The healthcare sector remains substantially underfunded and understaffed, and it lacks adequate healthcare management.[@R19] It is important to provide patients with much-needed safety because patient safety is an identifiable problem in hospitals. This approach would allow the provision of patient safety while avoiding blame games that divert attention away from the quality of care. Healthcare organisations often learn from errors in the use of local and national reporting systems.[@R20] Patient safety programmes exist primarily in the EU. Outside the EU, they are in a latent position (as in Albania at present). There are no established programmes on error records. Formal hospital registers show that patient safety is an issue in this country. According to some findings, 'medical breaches' are directly connected to presumed medical corruption.[@R21] There is a lack of research on patient safety in Albanian hospitals and primary care settings. Therefore, an instrument to measure healthcare professionals' attitudes about the safety climate in their hospitals would be helpful for understanding and identifying areas that need improvement and for evaluating improvements in interventions. The purpose of the present study was to establish the reliability and validity of the translated version of the Safety Attitudes Questionnaire (SAQ; Hospital Version) by evaluating its psychometric properties. Moreover, we aimed to determine whether there are differences among nurses and physicians regarding safety attitudes. Previous research has assessed the psychometric properties of the SAQ across countries[@R24] [@R25] and in different contexts and settings. The internal consistency and Cronbach\'s α values are acceptable, and the construct validity measured by confirmatory factor analysis (CFA) generally exhibits satisfactory model fit.[@R24] However, no psychometric instruments have been developed to measure patient safety in Albanian hospital settings. Our research hypotheses related to the study aim were as follows: H1. The data from this study confirm the proposed six-factor model of the original SAQ.H2. The SAQ shows good internal consistency.H3. The individual items of the SAQ show high correlations within their respective factors.H4. There are no differences among nurses and physicians in terms of perceived patient safety attitudes in selected hospital settings. Methods {#s3} ======= Setting {#s3a} ------- This was a quantitative, cross-sectional study. The data were collected between May and June 2012 at four regional hospitals in Albania. Nurses and doctors answered the SAQ-A voluntarily and anonymously. Surveys that were blank or had unvarying responses (eg, all of the responses were 'neutral' or 'agree strongly') were excluded from this analysis because they did not provide any diagnostic information. Participants {#s3b} ------------ One-stage cluster sampling was used in this study. Four hospitals were randomly selected from a list of 11 regional hospitals to ensure geographic representativeness. In each of the hospitals selected, both nurses and doctors were included in the survey. The rationale behind choosing only two occupational groups (nurses vs doctors) was to explicitly compare these two groups with the distinctions regarding safety behaviour reported in the academic healthcare literature.[@R27] Nurses and doctors in Albanian hospitals function in a manner similar to that of a team; hence, their attitudes affect patient service and quality. The surveys were administered during predetermined departmental and staff meetings. Measurements {#s4} ============ Safety Attitudes Questionnaire {#s4a} ------------------------------ The instrument used for data collection was the SAQ short adapted 30-item version.[@R31] The instrument comprises two parts. The first part contains questions that address perceptions of patient safety ([table 1](#BMJOPEN2014006528TB1){ref-type="table"}). The second part collects data about the professional respondent, including position held, sex, main job and years of experience. ###### Comparison of the Safety Attitudes Questionnaire (SAQ) scores among nurses and physicians -------------------------------------------------------------------------- Subscale Physicians\ Nurses\ t p Value M (SD) M (SD) -------------------------- ------------- ------------- ------ ------------ Teamwork 52.3 (10.7) 45.7 (11.3) −5.4 0.01 Safety climate 38.7 (11.1) 36.8 (10.3) −1.5 0.1 Stress recognition 39.5 (9.2) 46.7 (11.4) 6.4 \<0.01 Job satisfaction 49.7 (9.2) 40.6 (12.2) −7.8 \<0.01 Perception of management 46.8 (9.6) 44.8 (13.1) −1.6 0.1 Working conditions 42.4 (11.4) 29.2 (13.4 −9.7 \<0.01 -------------------------------------------------------------------------- The instrument measures healthcare professionals' perceptions in six areas, the teamwork climate, the safety climate, job satisfaction, perceptions of management, stress recognition and working conditions, with a response scale ranging from 1 (disagree strongly) to 5 (agree strongly). We chose the SAQ as an evaluation tool because of the strong correlation shown by previous studies between favourable SAQ scores and positive patient outcomes.[@R31] Another reason was the proven validity and reliability of the SAQ in the countries in which it has been tested, including the USA, the UK, Turkey and Norway.[@R24] There are other tools for evaluating the patient safety culture or the patient safety climate, such as Patient Safety Culture in Health care Organizations,[@R32] the Manchester Patient Safety Assessment Framework[@R33] and the Patient Safety Behavioural Intent (PSBI).[@R34] These tools have been used in practice to highlight their potential strengths and weaknesses, as outlined in published research reports. However, the SAQ is one of the most commonly used and rigorously validated tools for measuring the safety climate in healthcare. A distinguishing feature of the SAQ is that higher scores on this survey have been associated with positive patient and staff outcome data. This feature contrasts with other tools that are less likely to have a direct association with patient outcomes.[@R35] Safety culture versus safety climate: definition of terms {#s4b} --------------------------------------------------------- Safety culture has a broad definition. It has been defined as a worldwide issue that incorporates the values, assumptions and drivers that guide an organisation. Alternately, the safety climate may involve a narrower scope by addressing the way employees perceive certain aspects of the organisational culture. Measuring the safety climate is important because the culture of an organisation and the attitudes of teams have been found to influence patient safety outcomes, and measures of the safety climate can be used to monitor change over time.[@R30] [@R35] Some studies suggest that it is easier to measure the safety climate because culture is very broad, whereas the climate focuses on staff members' current perceptions of safety in relation to management support, supervision, risk taking, safety policies and practices, trust and openness. The safety climate is also thought to be more likely than culture to show change following interventions.[@R36] [@R37] Consequently, we chose to examine the safety climate in our study as an easily researched and manageable concept. Statistical analysis {#s5} ==================== The data analyses were conducted using SPSS V.16 (SPSS Inc, Chicago, Illinois, USA). CFA was conducted by structural equation modelling using AMOS V.5.0 software to test the extent to which each SAQ-A dimension was explained by the items and the extent to which safety attitude was explained by the six dimensions. Relative χ^2^\<5.0, a comparative fit index (CFI) value of ≥0.95, a root mean square error of approximation (RMSEA) value of \<0.08 and a standardised root mean square residual (SRMR) value of \<0.09 were considered to demonstrate acceptable model fit. The internal consistency of the Albanian version of the SAQ was assessed using Cronbach\'s α coefficient. Two items were negatively worded and were reversed for the statistical analysis. An independent sample t test was used to compare the mean score of the SAQ subscales between physicians and nurses. The χ^2^ test was used to compare the proportions of the responses between physicians and nurses. Pearson\'s correlation coefficient was used to examine the relationship between the SAQ subscales. A p value ≤0.05 was considered statistically significant. Ethical issues {#s6} ============== Approval by an ethical committee was not necessary because the study had no experimental design and did not involve patients. Moreover, we were informed that it was not mandatory to receive ethical approval from the hospital ethics committee. The study was conducted in line with the Helsinki Declaration using a survey. The participants were informed about the aims of the study. Responding to the questionnaire was voluntary and anonymous, and the participants were given the opportunity to opt-out without giving any explanation if they felt uncomfortable with the content of the questions. Results {#s7} ======= Albanian version of the SAQ {#s7a} --------------------------- The SAQ was translated from English to Albanian and back again by native speakers. To ensure that the version that was translated from English to Albanian was clear and used the correct words, a focus group comprising physicians, nurse experts and faculty members conducted a review. All of the components were reviewed for their conformity with Albanian culture. Linguistic validation of the translation was performed using the back-translation technique. The analytical results demonstrated that all six dimensions had good reliability (see [tables 2](#BMJOPEN2014006528TB2){ref-type="table"} and [3](#BMJOPEN2014006528TB3){ref-type="table"}). The experts' view determined the authenticity of the content on the basis of its relevance, appropriateness and importance to Albanian culture. A content validity ratio (CVR) and a content validity index (CVI) were calculated for each item. The mean CVR for the total scale was 0.96, and the mean CVI was 0.82, indicating satisfactory content validity. (The adapted version of the questionnaire is available from the authors on request). ###### Goodness-of-fit indices for the confirmatory factor analysis of the Safety Attitudes Questionnaire factors Sample size 341 ------------------------------------------------- ------- Standardized root mean square residual (SRMR) 0.075 Root mean square error of approximation (RMSEA) 0.044 Comparative fit index (CFI) 0.97 ###### Internal consistency for the six Safety Attitudes Questionnaire (SAQ) factors: Cronbach\'s α SAQ factors Cronbach\'s α ------------------------------------- --------------- Safety climate (7 items) 0.82 Teamwork climate (6 items) 0.79 Job satisfaction (5 items) 0.78 Stress recognition (4 items) 0.62 Perceptions of management (4 items) 0.64 Working conditions (4 items) 0.76 Survey responses {#s7b} ---------------- A total of 341 valid questionnaires were returned from four hospitals for a response rate of 70%. An analysis of the missing data showed that no item had more than 2% missing values (range 0--1.8%). The majority of the participants (63.6%) were females (36.4% were males); 61.4% were physicians, and 38.6% were nurses ([table 4](#BMJOPEN2014006528TB4){ref-type="table"}). ###### Characteristics of the survey respondents Variables Physicians (n=209) Nurses (n=132) ------------------------- -------------------- ---------------- Gender, n (%) Female 136 (65.0) 114 (86.0) Male 73 (35.0) 18 (14.0) Age group (years) n (%) 21--30 53 (58.2) 38 (41.8) 31--40 42 (66.7) 21 (33.3) 41--50 55 (59.1) 38 (40.9) 51--60 55 (69.6) 24 (30.4) \>60 5 (33.3) 10 (66.7) Internal construct validity {#s7c} --------------------------- The goodness-of-fit values used to evaluate the internal construct validity are displayed in [table 3](#BMJOPEN2014006528TB3){ref-type="table"}. The SRMR value was 0.075, the RMSEA value was 0.044 and the CFI value was 0.97. These values indicate a good model fit approximation of the translated version of the SAQ. The CFA indicated a good model fit for each dimension and for the entire safety construct: the GFI, TLI and CFI were \>0.90, and the RMSEA was \<0.10. Internal consistency {#s7d} -------------------- The internal consistency of the six factors and the 30 items of the translated version of the SAQ had Cronbach\'s α values of 0.62 to 0.82. Safety climate had the highest Cronbach\'s α values, and stress recognition had the lowest value ([table 3](#BMJOPEN2014006528TB3){ref-type="table"}). The test of the hypothesised relationships among the factors and items showed that the correlation ranged from 0.02 to 0.89 and that five of the six factor correlations were significant. The teamwork climate was positively correlated with the safety climate (r=0.55, p\<0.01), job satisfaction (r=0.54, p\<0.01), perceptions of management (r=0.68, p\<0.01) and working conditions (r=0.68, p\<0.01), whereas the stress recognition subscale was not significantly related to any subscales. The intercorrelations among the factors are presented in [table 5](#BMJOPEN2014006528TB5){ref-type="table"}. ###### Correlation matrix for the Safety Attitudes Questionnaire (SAQ) subscales Subscale Safety climate Teamwork climate Job satisfaction Stress recognition Perception of management Working conditions --------------------------- ---------------- ------------------ ------------------ -------------------- -------------------------- -------------------- Safety climate Teamwork climate 0.55\* Job satisfaction 0.46\* 0.54\* Stress recognition 0.25 0.08 0.02 Perceptions of management 0.54\* 0.68\* 0.47\* 0.15 Working conditions 0.68\* 0.71\* 0.61\* 0.25 0.79\* \Significant at the 5% level. SAQ factors and item descriptions and overall positive responsiveness per scale {#s7e} ------------------------------------------------------------------------------- The SAQ factor definitions and items, the missing answers, the mean (SD), and agreement (agree strongly) and disagreement (disagree strongly) responses are described in [table 6](#BMJOPEN2014006528TB6){ref-type="table"}. ###### Safety Attitudes Questionnaire (SAQ) item descriptions Item is reverse scored[@R38] Per cent of items missing data Mean (SD) Per cent agree (minimum agree--agree) Per cent disagree (minimum disagree-- maximum disagree) --------------------------------------------------------------------------------------------------------------------------------- ------------------------------ -------------------------------- ------------- --------------------------------------- --------------------------------------------------------- Teamwork climate (1--6)* 1\. It is easy for personnel in this hospital to ask questions when there is something that they do not understand No 1.3 4.13 (0.96) 80.45 (41.67--100.00) 7.41 (0.00--35.00) 2\. I have the support I need from other personnel to care for patients No 2.1 3.95 (0.99) 74.28 (33.30--98.05) 9.18 (0.00--42.86) 3\. Nurse input is well received in this hospital No 1.9 3.99 (1.05) 73.42 (23.94--100.00) 10.16 (0.00--54.94) 4\. In this hospital, it is difficult to speak up if I perceive a problem with patient care Yes 2.2 2.45 (1.26) 21.69 (0.00--51.00) 59.87 (9.09--100.00) 5\. Disagreements in this hospital are resolved appropriately (ie, not by who is right but by what is best for the patient) No 1.8 3.57 (1.10) 56.97 (22.75--85.24) 18.24 (0.00--54.09) 6\. The physicians and nurses here work together as a well-coordinated team No 1.6 3.75 (1.07) 68.42 (25.72--98.2) 14.24 (0.00--53.12) Safety climate (7--13) 7\. The culture in this hospital makes it easy to learn from the errors of others No 1.5 3.96 (1.01) 71.45 (33.33--100.00) 9.55 (0.00--33.33) 8\. Medical errors are handled appropriately in this hospital No 2.3 3.49 (1.06) 51.08 (14.3--92.7) 17.3 (0.00--57.14) 9\. I know the proper channels to direct questions regarding patient safety in this hospital No 1.6 3.84 (1.01) 64.5 (24.00--100.00) 9.45 (0.00--39.10) 10\. I am encouraged by my colleagues to report any patient safety concerns I may have No 1.5 4.05 (0.94) 77.3 (48.60--100.00) 7.15 (0.00--26.32) 11\. I receive appropriate feedback about my performance No 0.7 3.22 (1.23) 46.48 (4.58--76.5) 31.5 (0.00--75.00) 12\. I would feel safe being treated here as a patient No 1.5 4.02 (1.04) 74.99 (36.35--100.00) 9.48 (0.00--42.67) 13\. In this hospital, it is difficult to discuss errors Yes 1.7 2.57 (1.13) 20.15 (0.00--48.15) 53.87 (20.84--92.7) Job satisfaction (14--18) 14\. This hospital is a good place to work No 0.7 3.75 (1.08) 63.5 (4.55--100.00) 13.54 (0.00--58.09) 15\. I am proud to work at this hospital No 0.9 3.72 (1.07) 62.5 (15.00--100.00) 10.8 (0.00--50.00) 16\. Working in this hospital is like being part of a large family No 0.7 3.12 (1.30) 42.1 (0.00--93.55) 33.4 (0.00--80.00) 17\. Morale in this hospital area is high No 1.5 2.89 (1.25) 38.72 (4.18--84.33) 36.75 (0.00--76.25) 18\. I like my job No 0.2 4.38 (0.88) 85.6 (62.1--100.00) 4.63 (0.00--18.32) Stress recognition (19--22) 19\. When my workload becomes excessive, my performance is impaired No 1.5 3.79 (1.13) 32.16 (27.57--100.00) 34.9 (0.00--52.3) 20\. I am more likely to make errors in tense or hostile situations No 1.3 3.77 (1.16) 46.5 (30.00--87.00) 21.84 (0.00--50.00) 21\. Fatigue impairs my performance during emergency situations (eg, emergency resuscitation, seizure) No 3.4 3.00 (1.28) 29.45 (5.78--79.15) 46.84 (12.30--76.54) 22\. I am less effective at work when fatigued No 1.4 3.75 (1.03) 25.69 (38.5--96.4) 41.21 (0.00--30.00) Perceptions of management (23--26) 23\. The hospital management does not knowingly compromise the safety of patients No 1.7 2.54 (1.27) 49.2 (9.11--87.5) 26.5 (4.89--90.45) 24\. The hospital administration supports my daily efforts No 0.6 2.48 (1.75) 64.5 (0.00--94.2) 29.62 (0.00--100.00) 25\. I receive adequate, timely information about events in the hospital that might affect my work No 1.7 3.21 (1.09) 51.5 (12.00--74.8) 22.1 (0.00--66.65) 26\. The levels of staffing in this clinical area are sufficient to handle the number of patients No 1.8 2.67 (1.33) 53.54 (0.00--84.62) 39.47 (4.21--96.45) Working conditions (27--30) 27\. All of the necessary information for diagnostic and therapeutic decisions is routinely available to me No 2.2 3.5 (1.07) 58.64 (16.8--89.5) 18.4 (0.00--68.45) 28\. This hospital constructively addresses problem physicians and employees No 1.9 2.89 (1.14) 24.85 (0.00--82.33) 35.45 (0.00--81.00) 29\. Trainees in my discipline are adequately supervised No 2.4 3.55 (1.19) 58.64 (10.00--100.00) 21.48 (0.00--64.50) 30\. This hospital does a good job of training new personnel No 1.2 3.50 (1.19) 57.46 (16.2--96.45) 20.42 (0.00--61.4) The percentage of hospital healthcare providers who reported having a positive attitude was 60.3% for the teamwork climate, 57.2% for the safety climate, 58.4% for job satisfaction, 37.4% for stress recognition, 59.3% for the perception of management and 49.5% for working conditions. Group comparisons: nurses' behaviours versus physicians' behaviours {#s7f} ------------------------------------------------------------------- The results of the t test for independent samples did not indicate any significant difference between the scores of physicians and the scores of nurses with regard to the safety climate (t=−1.5, p=0.1) and the perception of management (t=−1.6, p=0.1) subscales, showing the same perceptions of attitude. A significant difference was found between the scores of the physicians and nurses for the teamwork (t=−5.4, p\<0.01), stress recognition (t=6.4, p\<0.01), job satisfaction (t=−7.8, p\<0.01) and working conditions subscales (t=−9.7, p\<0.01). For the statement, "When my workload becomes excessive, my performance is impaired", a lower proportion of physicians and nurses endorsed the 'agree' option; this result achieved statistical significance (39% physicians vs 51% nurses; χ^2^=5.0, p=0.02). For the statement "I am provided with adequate, timely information about events in the hospital that might affect my work", 59% of the physicians responded positively compared with 34% of the nurses (χ^2^=18.8, p\<0.01). Regarding the item "The culture in this hospital makes it easy to learn from the errors of others", a high percentage of the nurses and physicians agreed with the statement (80.2% of the nurses and 83% of the physicians; χ^2^=0.1, p=0.6). The nurses and doctors also agreed with the statement, "In this hospital, it is difficult to discuss errors"; the nurses felt more intimidated by discussing errors (55% of the physicians vs 44% of the nurses; χ^2^=4.9, p=0.02). Discussion {#s9} ========== Previous studies in the area of patient safety have investigated the safety climate in primary care in Albania.[@R39] The aim of the present study was to investigate perceptions of the safety climate among nurses and doctors working in hospital settings to confirm a validated tool (SAQ). According to the results of this study, construct validity based on the CFA and goodness-of-fit indices including CFI, SRMR and RMSEA demonstrated a good model that fit very well. It is imperative that a model should be made more specific and tested a second time in case a hypothetical model fails to fit appropriately.[@R40] [@R41] According to good model fit indices, the Albanian version of the SAQ is a valid measure of the safety attitude in hospitals. This finding is also an indication of the internal construct validity of the SAQ. The data from this study confirm our first hypothesis regarding the proposed six-factor model of the original SAQ (H1: The data from this study confirm the proposed six-factor model of the original SAQ). The internal consistency and internal structure of the Albanian translation of the SAQ were assessed, and the translation showed satisfactory psychometric properties. The mean CVR for the total scale was 0.96 and the mean CVI was 0.82, indicating satisfactory content validity. A good model fit was indicated by the construct validity, and its goodness-of-fit was used to determine the CFA. Based on the Cronbach\'s α values, the internal consistency of the six factors of the Albanian version of the SAQ showed values between 0.62 and 0.82. Job satisfaction and the perception of management had α values of 0.62 and 0.64, respectively, which are slightly below the recommended acceptable α value limit of 0.70 because of missing data that may have influenced the result.[@R41] The study results showed good internal consistency, confirming our second hypothesis (H2: The SAQ shows good internal consistency). To apply the SAQ in Albania and ensure its reliability, the outcomes of this research should be considered in future evaluations. Cultural aspects may exist regarding the perception of management in the Albanian setting. For example, hospital directors are quasi-political appointees, and physicians may have little training in healthcare management[@R16] and thus may lack the competency to properly evaluate overall management issues. The study confirmed the third hypothesis based on the conclusion that all of the factors, with the exception of stress recognition, were interconnected (H3: Individual items of the SAQ show high correlations within their respective factors). According to the study results (see [table 5](#BMJOPEN2014006528TB5){ref-type="table"}), the teamwork climate was more positively correlated with the perceptions reported for the management and working conditions subscales. Stress recognition was not correlated with any of the subscales. These findings complement those of the psychometric testing for the original SAQ. In accordance with previous studies, stress recognition did not show a relationship with the perceptions of management, the teamwork climate, and job satisfaction, factors with moderate-to-high correlations.[@R24] A significant difference in the perceived patient safety attitudes for the subscales of teamwork, safety climate, job satisfaction and working conditions was found between the nurses and physicians, with the nurses scoring lower mean values (see [table 1](#BMJOPEN2014006528TB1){ref-type="table"}). Overall, the nurses perceived lower job satisfaction, worse working conditions, a lower level of teamwork and poorer perceptions of management compared with doctors. Thus, our fourth hypothesis was not confirmed (H4: There is no difference among nurses and physicians in terms of perceived patient safety attitudes in selected hospital settings). Other clinical areas have categorically shown broad distinctions between nurses and physicians.[@R27] [@R28] Two studies in the USA that used the SAQ showed that nurses and doctors differed in their perceptions of safety culture,[@R29] [@R30] possibly because of the personal characteristics of the caregivers, such as their level of education, socioeconomic status and gender. The traditional hierarchy of physicians has often discouraged nurses from speaking up to doctors. Nurses may be hesitant to confront physicians on issues of patient care because they might have less training in or experience with dealing with patients' medical conditions.[@R30] In our study, the nurses consistently showed low mean scores for working conditions, safety climate and job satisfaction. Another study that examined nurses' job satisfaction showed that 41% of nurses were dissatisfied with their work in the USA; in England, 38.9% of nurses intended to abandon the profession. Generally, salary, professional growth and autonomy are some of the factors that influence the nursing professional\'s job satisfaction.[@R42] National researchers in Albania have noted that nurses were less satisfied with their promotion opportunities and coworker satisfaction.[@R43] WHO studies stress the limited formal opportunities to upgrade knowledge and skills through continuing professional development.[@R44] From the study results, it is clear that the international benchmark standards were not met,[@R24] as evidenced by the failure of the mean values in relation to the five safety dimensions. Managerial implications {#s8a} ----------------------- Measuring safety climate dimensions such as perceived teamwork climate, job satisfaction and the perception of management in hospitals can help to diagnose the underlying safety culture of an entire organisation or work unit. This study created a thorough image of nurses' and doctors' behaviour regarding issues such as teamwork, the safety climate, stress at work, job satisfaction and management support in the selected regional hospitals. European integration could increase the mobilisation of human resources for health to other countries (eg, immediately after acceding to the EU in 2007, Romania reported more than 6000 requests for certificates recognising the Romanian diplomas of doctors, dentists, pharmacists and nurses for use elsewhere in the Union[@R44]). In light of hospital decentralisation reforms and EU adherence,[@R16] [@R19] this study serves as a starting point for initiating policy changes to address the issues identified above, such as improving job satisfaction, working conditions and the perceived inadequate information flow among nurses, and to implement interventions targeted to reduce the impact of these factors on the quality of hospital care. The prevailing culture influences safety behaviours and outcomes for both healthcare workers and patients. This study has shown that the SAQ-A is a valid and easily administered instrument. As a first step, hospitals can use this tool to measure their employees' safety attitudes on a regular basis. Moreover, healthcare managers can use the resulting data to design effective safety management systems and possible interventions, such as promoting teamwork or stress recognition among doctors and nurses. The results of this study can serve as baseline information for researchers with a variety of research interests, especially those related to patient safety and human resources for health. From a cultural perspective, the results serve as a basis for comparison with other countries or systems that have their own particularities but share important common features with Albania (in terms of health services organisation, financing and regulation and similar evolutions of these systems). Promoting patient empowerment and a positive safety culture {#s8b} ----------------------------------------------------------- Over the past two decades, many policy documents, national priorities and guidelines across Europe, North America, Australia and sections of the Asian continent have been developed with messages about developing a positive safety culture with regard to healthcare.[@R45] [@R46] These types of initiatives are quite scarce and latent in Albania. Below, we present a contextualised framework ([figure 1](#BMJOPEN2014006528F1){ref-type="fig"}) that takes into account the Albanian Health Care System when considering how to promote a safety culture in an underfinanced healthcare system. The authors hope to broaden this perspective to similar health systems in Eastern European countries. Our main goal was to develop and test a framework for making the concept of safety culture meaningful and accessible to policymakers, healthcare managers and front-line staff, thus helping to facilitate a debate about ways to improve the safety culture in Albanian healthcare settings. We used a comprehensive review of the literature and national policy documents to identify the key dimensions of safety culture in healthcare settings. ![Gabrani and Petela, promoting safety culture frame in Albanian healthcare settings.](bmjopen2014006528f01){#BMJOPEN2014006528F1} The voices of patients and professionals are largely absent in the field of patient safety in Albania. For example, patients are considered to be passive and to maintain a troublesome apathy towards involvement in the healthcare decision-making process [@R47] with regard to the country\'s informal payment phenomenon, which has deleterious effects on the Albanian health system. More active involvement of patients will help to diffuse and scale up patient safety programmes that have been successfully implemented in Western Europe and outside EU-developed countries. Further research needed {#s8c} ----------------------- The topic of patient safety is becoming increasingly prominent in political agendas.[@R48] Reduced revenues and increasing expenditures in times of financial crisis are likely to increase pressure on health systems to further contain costs, thus affecting service quality and patient safety.[@R49] According to one recent cost-effectiveness study on patient safety, interventions show that specific actions related to patient safety can be cost-effective.[@R50] To cope with the EU policy measures and reduce the costs of unsafe care and to develop cost-effective patient safety programmes in Albania, further research is needed. Furthermore, work is needed to better identify and design solutions that fit into existing institutional and organisational frameworks. At the hospital level, there is a need to understand the differences in attitudes among nurses and doctors and to better understand their low mean values on the scales compared with international benchmarks,[@R24] especially in terms of the stress and fatigue recognition scale. Conclusions {#s10} =========== It is apparent that the psychometric aspects and the translation of the SAQ indicate good construct validity. Nonetheless, the reliability analysis suggested that some items need further refinement to establish sound internal consistency. The SAQ is clearly a useful tool for evaluating safety attitudes in Albanian hospital settings, and it confirmed our main hypothesis, H1. The data from this study also confirmed the proposed six-factor model of the original SAQ, H2. The SAQ showed good internal consistency, H3. Individual items of the SAQ showed high correlations within their respective factors. As previous research suggests, the SAQ has potential as a useful tool for evaluating safety attitudes. Regarding our study, there is room for improvement, especially in terms of generalising the findings to larger samples and conducting additional exploratory analyses to identify a better factor model. Moreover, the researchers are committed to adapting a new version of the tool by combining it with in-depth interviews to learn more about the differences in nurses' and doctors' safety attitudes. The authors would like to thank the students of the Faculty of Public Health for their outstanding assistance with data collection. The authors also thank all of the hospital healthcare workers who were part of this research study. Special thanks for your very helpful comments about the safety culture, stress and error reporting in the healthcare environment. Special thanks also for the auspicious guidance of Wendy Knibb, professor at the Schools of Health and Social Care, Faculty of Health and Medical Sciences, University of Surrey, UK, and to Gilles Dussault, professor at the Institute of Hygiene and Tropical Medicine (IHMT), Portugal. They thank American Journal Experts (AJE) for the English language editing. Contributors: AG designed the study. AH and JCG refined the writing strategy and the analysis elaboration. AS and JCG performed the data analysis, and JCG wrote the manuscript. JCG, AH and AS designed (adapted) the questionnaire and led the data collection process. All of the authors assisted with the data collection and the manuscript preparation and modified and approved the final manuscript. AG and AH share the primary contribution to the article. Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: None declared. Provenance and peer review: Not commissioned; externally peer reviewed. Data sharing statement: No additional data are available.	{ "pile_set_name": "PubMed Central" }
Shortly after the discovery of vitamin A (retinol) in 1913, scientists began to realise the extreme importance of this simple fat-soluble molecule in a large number of complex and diverse physiological processes in vertebrates. All-trans-retinoic acid (ATRA) was identified as a key physiologically active metabolite of vitamin A. The understanding of how ATRA exerts its highly pleiotropic effects in vertebrates, as well as its antitumour activities, remained unclear until the discovery of nuclear retinoic acid receptors. Retinoic acid receptors were shown to be members of the superfamily of steroid/thyroid hormone nuclear receptors functioning primarily as ligand-inducible transcription factors. Retinoic acid receptors ======================= Similarly to other key developmental regulators, RARs possess the capacity to function as a molecular switch; when not bound by ligand, they form a complex with co-repressors such as N-CoR (negative co-regulator) or SMRT (silencing mediator for retinoid and thyroid hormone receptors), and histone deacetylases (HDACs) to actively inhibit gene expression ([Figure 1](#fig1){ref-type="fig"}). Upon ATRA binding, co-repressors are released and co-activators, including histone acetyltransferases (HATs), DRIP/TRAP/ARC, or mediator-containing complexes are recruited to decompress chromatin and activate transcription of target genes ([@bib16]). To date, three different RAR and retinoid X (or rexinoid) receptor (RXR) genes have been characterised (RARA, RARB, and RARG), each encoding multiple N-terminal protein isoforms. Retinoid X (or rexinoid) receptors, which bind 9-cis-retinoic acid (9-cis-RA) with high affinity, serve as obligatory heterodimerisation partners for RARs. RARA, in common with other RAR genes, encodes two major isoforms ([Figure 1](#fig1){ref-type="fig"}) that differ in their A region sequence that contribute to transcriptional regulation in a ligand-independent and promoter-specific manner ([@bib22]). These isoforms are identical in their B to F region sequence, which contain DNA (DBD) and ligand binding (LBD) domains as well as structural motifs responsible for dimerisation, co-repressor interaction, and ligand-dependent trans-activation. Expression of the RARα2, RARβ2 and possibly also RARγ2 isoforms is under control of promoters that are inducible by ATRA. Activation of transcription by RARs is intrinsically linked to their proteasome-mediated degradation ([@bib13]) and upregulation of RARα2, RARβ2, and RARγ2 isoform expression by ATRA may therefore have been evolutionarily selected and conserved to renew expression of a given receptor to sustain gene activation and the physiological effects of ATRA over an extended period of time. Non-genomic effects of ATRA =========================== Similar to other nuclear receptor ligands ATRA can also function, via so-called non-genomic pathways, independent of the classical mechanism of nuclear receptor action. All-trans-retinoic acid can effectively and rapidly modulate the activities of proteins involved in signal transduction in a manner that is highly cell type specific. The activation of multiple kinase signalling pathways leads to transcriptional effects independent of nucleolar RARs via transcription factors that lie at the end of these signalling cascades ([@bib1]; see [Figure 2](#fig2){ref-type="fig"}). For example, in neuronal cells, ATRA has been shown to activate Erks via PI3K and Src kinases leading to repression of the antidifferentiative octamer-binding transcription factor 4 (Oct4). In epithelial cells and fibroblasts, RARα can form complexes and activate G protein alpha Q (Gαq) in response to ATRA. Activation of Gαq in turn leads to the activation of RhoGTPase, p38MAPK, and MSK1. Phospho-MSK1 then contributes to transcriptional activation of RARα target genes by phosphorylation of histones and RARα itself. Another non-genomic effect of ATRA is that it can rapidly lead to the accumulation of cyclic adenosine monophosphate (AMP) in acute promyelocytic leukaemia (APL), most likely by adenyl cyclase stimulation and protein kinase A (PKA) activation. Activated PKA can phosphorylate and activate RARα leading to increased expression of RAR target genes and differentiation. In addition, ATRA has been shown to upregulate the granulocyte colony-stimulating factor (G-CSF) and granulocyte macrophage colony-stimulating factor (GM-CSF) receptor resulting in the amplification of the MAP kinase pathway via a positive feedback-loop mechanism ([@bib15]). Therefore, in normal myelomonocytic differentiation ATRA and growth factor (GF) signalling converge via common signalling cascades to affect activities of RARs as well as other transcription factors. The activity of ATRA in leukaemia cells is dependent on complex interactions between genomic and non-genomic pathways. Although much remains to be understood regarding the exact mechanisms underlying the non-genomic effects of ATRA, it is clear that these effects have important consequences and have to be taken into consideration when evaluating the overall response of a particular cell type to retinoids. ATRA in APL =========== Defective or aberrant differentiation is a hallmark of many cancers. It is caused by various mutations that disrupt signalling pathways, epigenetic modifiers, and transcription factors leading to the deregulation of the expression of proteins required for differentiation. Terminal differentiation results in the loss of long-term proliferative capacity and eventually induction of apoptosis. The potential of ATRA to force the completion of disrupted maturation programs underlies the concept of differentiation therapy. Differentiation therapy offers the prospect of a less aggressive treatment coupled to limited damage to normal cells. [@bib4] showed that the HL-60 cell line, which was allegedly derived from a patient with APL, could be induced to undergo granulocytic differentiation by micromolar concentrations of ATRA. One year later, the same group of investigators showed that ATRA causes primary APL cells to differentiate in vitro. Although an obvious step forward, the clinical trial with ATRA did not immediately follow and it was only in 1987 that Huang and colleagues reported the successful use of ATRA to induce complete remission (CR) in patients with APL, without an apparent increase in the severity of the haemorrhagic diathesis observed at presentation ([@bib20]). Incorporation of chemotherapy and up-front use of ATRA in combination with arsenic trioxide (ATO) lead to CR rates in excess of 93% with patients achieving 5-year overall survival rates approaching 100%. A recent phase 3 trial showed that combination treatment of ATRA and ATO without chemotherapy leads to CR and an extended overall survival compared with the standard ATRA and chemotherapy treatment (P=0.02) in patients with low-risk APL ([@bib24]). This success of ATRA in APL shows that the concept of differentiation therapy can be an effective strategy in leukaemia and provided a rationale for its use in the treatment of other cancers. The discovery that the t(15;17) translocation, the cytogenetic marker for APL involves the RARA gene on chromosome 17q21, reaffirmed the importance of ATRA and RARα in myelomonocytic differentiation and paved the way for the understanding of the molecular mechanisms underlying the disease pathogenesis as well as this remarkable response to ATRA ([@bib34]). The most common RARA translocation partner is the PML gene. Fusion of PML sequences to the B-F regions of RARA increases the affinity of the fusion receptor for co-repressors so that higher (pharmacological) levels of ATRA are required to induce its dissociation. Consistent with these findings, ability of PML/RARα to block differentiation of transfected myeloid cells depends on an intact N-CoR-binding region in the RAR moiety of the fusion protein ([@bib31]). ATRA in non-APL AML and other malignancies ========================================== The seemingly selective effectiveness of ATRA in APL poses an important question of whether the presence of the PML/RARα rearrangements in this malignancy renders it uniquely susceptible to ATRA therapy. One argument against such a view is that ATRA effectiveness in AML has been observed in cells that lack PML/RARα, as HL-60 cells, that can be induced to differentiate into mature granulocytes ([@bib4]). Furthermore, clinical studies with ATRA in previously untreated older patients with AML yielded some encouraging results and various clinical trials throughout the World have shown effectiveness of ATRA when used in combination with other agents such as conventional chemotherapy. [@bib11] found that treatment with low-dose cytosine arabinoside (LDAC) and ATRA lead to an improvement of the time to treatment failure rate and longer survival in patients with AML older than 60 years in comparison with LDAC alone ([@bib11]). Similarly, [@bib33] reported that in elderly patients with AML the genotype 'mutant NPM1 without FLT3-ITD\' appears to be a predictive marker for response to ATRA given in addition to intensive chemotherapy. [@bib35] also observed that the combined use of ATRA and LDAC was effective for inducing CR in patients with 'poor prognosis\' AML, especially in patients with \<50% bone marrow infiltration ([@bib35]). However, other phase 2 studies showed no or non-significant effects ([@bib29]; [@bib6]). Overall, these studies show that the efficacy achieved so far with ATRA in patients with non-APL AML is limited and there is much room for improvement. Use of ATRA in anticancer therapy --------------------------------- Given the specific role of ATRA in cell proliferation and differentiation, its potential effectiveness to treat human cancers has captivated researchers for decades. Effects of ATRA on tumour development are extensive and include differentiation and apoptosis induction, as well as tumour prevention ([@bib2]). For example, the ATRA derivative 13-cis-retinoic acid (13-cis-RA) is successfully used in the treatment of children with high-risk neuroblastoma, reducing the risk of recurrence after high-dose chemotherapy and stem cell transplantation and increasing long-term survival ([@bib27]). Nevertheless, so far ATRA is only routinely and effectively used to treat APL and precancerous lesions such as leukoplakia, actinic keratosis, and cervical dysplasia. Consistent with the important role of ATRA signalling in cellular homeostasis and potential tumorigenic consequences of deregulation in this signalling pathway are reports by Losson and colleagues showing that loss of Tif1 alpha renders RARα oncogenic in the liver ([@bib21]). In recent years, ATRA has been studied in combinations with other drugs in a variety of cancers and precancerous conditions. Currently, several clinical trials using ATRA as part of combination therapies are ongoing. Tamoxifen, an antioestrogen with antiproliferative effects, is more effective against breast carcinoma and hepatoma cells when combined with retinoids. The combination of dexamethasone and ATRA has also been shown to inhibit proliferation and induce differentiation in human osteosarcoma cells. It has also been observed that Vitamin D3 analogues combined with ATRA can reduce the breast tumour mass in nude mice and inhibit cell growth and induce apoptosis in lung, prostate, breast, and ovarian cancer cells in vitro. In addition, the use of retinoids against secondary liver cancer, aerodigestive malignancies and secondary breast cancer has shown promising results ([@bib9]). All-trans-retinoic acid has also been evaluated in combination with G-CSF and GM-CSF. All-trans-retinoic acid and G/GM-CSF co-operate to promote myelomonocytic differentiation in murine pluripotent myeloid progenitor cells, different AML cell lines, as well as primary cells from patients with myeloid leukaemia ([@bib26]). The ligand-dependent activities of endogenous and transiently expressed RARα isoforms are enhanced by G/GM-CSF and correlate with an increased expression of ATRA inducible genes as well as leukaemic cell differentiation ([@bib15]). These data suggest that the manipulation with small molecules of the signalling pathways downstream from the G/GM-CSF receptors in AML cells may be an effective strategy to maximise the effects of ATRA in AML therapy. Another way to enhance ATRA-stimulated differentiation is the use of inhibitors of inosine 5′-monophosphate dehydrogenase (IMPDH). Inosine 5′-monophosphate dehydrogenase has an important role in the maintenance of homeostasis of guanosine nucleotides in the cell. Higher levels of guanosine nucleotides as well as IMPDH can be found in many cell types, including myeloid cells, and depletion of guanosine nucleotides are known to accompany their differentiation. The use of telomerase inhibitors has also been proposed to improve ATRA responsiveness. Recent studies suggest that one of the effects of ATRA and ATO in APL is due to a synergy at the level of inhibition of telomerase activity. Stem cells exhibit higher telomerase activity compared with their normal counterparts and downregulation of the hTERT gene and a concomitant reduction in telomerase activity is an early event following induction of differentiation by ATRA. Finally, combination of ATRA with different interferons (IFNs) have been shown to enhance effects of both drugs and lead to growth suppression and cell death in many tumour cell lines via a set of genes associated with retinoid-IFN-induced mortality\' (GRIMs) ([@bib19]). Many mutations involving genes that mediate signalling pathways have a crucial role in tumorigenesis. These genes contribute to cell proliferation (e.g., FLT3, c-KIT, RAS, and PTPN11), are involved in differentiation (e.g., AML1 and CEBPA), or are implicated in cell-cycle regulation or apoptosis (e.g., P53 and NPM1). Despite them being ideal partners for ATRA, only a few treatments combining ATRA with inhibitors of signal transduction pathways have been described to date. In one study, it has been observed that the inhibition of the mitogen-activated protein kinase (MAPK) pathway sensitises AML cells with constitutive MAPK activation to retinoid- and chemotherapy-induced apoptosis ([@bib28]). It has been observed in another study that RAD001 (everolimus), an inhibitor of mammalian target of rapamycin complex 1 (mTORC1), is able to reduce the stem cell potential and enhance the ATRA-induced growth arrest and differentiation of AML cells. Furthermore, RAD001 improves the ability of ATRA to inhibit the proliferation of HL-60 cells growing as tumour xenografts in immune-deficient nude mice ([@bib30]). Inhibitors of the glycogen synthase kinase 3 (GSK3) have also been shown to inhibit growth and drive differentiation of a variety of leukaemic cells in combination with ATRA. GSK3b phosphorylates RARα on multiple residues reducing its transcriptional activity. The inhibition of GSK3b prevents RARα phosphorylation and its degradation by the proteasome leading to increased levels of RARα in the cell. So far only the FMS-like tyrosine kinase 3 (FLT3) inhibitor midostaurin- and cladribine-based induction chemotherapy (CLAG) have been studied in a clinical trial in combination with ATRA in relapsed or refractory AML. Another way to improve responsiveness to ATRA is to increase its intracellular levels by inhibiting the activity of the aldoketoreductase AKR1C3. AKR1C3 is an NADPD-dependent oxidoreductase that has a role in the metabolism of retinoids and can therefore control the level of ATRA in the cell. Acute myeloid leukaemia cells have been shown to express AKR1C at high levels and inhibition of AKR1C leads to increased responsiveness of AML cells to ATRA ([@bib5]). Epigenetic regulation of RARA and combinatorial use of Epi-drugs and ATRA ------------------------------------------------------------------------- The promoter of the RARA isoform RARA2, but not RARA1, is hypermethylated in AML cell lines, and induction of differentiation and expression of RARA2 can be stimulated with DNA demethylating agents and ATRA treatment. Consistent with what has been observed in cell lines, levels of RARα2 expression are also markedly reduced in AML patient samples compared with normal myeloid cells expressing CD33 or CD34. However, in contrast to AML cell lines and normal myeloid cells, expression of the RARα1 isoform is also significantly reduced in primary AML cells. Moreover, both RARA1 and RARA2 promoters ([@bib14]) are unmethylated in AML patient samples, while histones associated with the RARA2 promoter have decreased levels of overall acetylation of histone H3 and decreased levels of di-methylation of histone H3 lysine 4. Both are heterochromatin markers associated with reduced gene expression indicating that epigenetically driven decrease in RARA expression may be a factor in the pathogenesis of AML and contribute to the differentiation blockade. Many mutations contribute to tumorigenesis by causing transcriptional deregulation through epigenetic modifications resulting in aberrant proliferation, impaired differentiation, and impaired apoptosis. These epigenetic modifications include changes in DNA methylation and histone modifications such as acetylation and methylation. Some studies have indicated that the combination of ATRA and epigenetic (epi)-drugs such as DNA methyltransferase inhibitors (DNMTi) or histone deacetylase inhibitors (HDACi) has higher therapeutic value in AML disease and potentially in other malignancies ([Figure 3](#fig3){ref-type="fig"}). Aberrant CpG island methylation in promoter regions in genes coding for tumour suppressors is a well-described phenomenon. CpG island methylation is associated with transcriptional silencing, involving the recruitment of methyl-binding proteins and HDACs to regions close to transcriptional initiation sites ([@bib12]). Over the last years, several epigenetic drugs have been approved or are in clinical development ([@bib10]). Two of the first epigenetic drugs used in AML were the nucleoside analogues 5-azacytidine and 5-aza-2′deoxycytidine. These DNMTi are incorporated into the DNA, where they form a covalent complex with the DNMT enzyme. This results in the degradation of the enzyme and a progressive loss of DNMT activity and the aberrant CpG island hypermethylation in cancer cells. Histone deacetylase inhibitors have been demonstrated to suppress tumours with t(15;17) translocations in nude mice, and induce remission in transgenic mouse models of ATRA-resistant APL ([@bib18]). Histone deacetylase inhibitors also induce differentiation and apoptosis of t(8;21) leukaemia cell lines and primary leukaemia blasts ([@bib36]). Although HDACi have been used both alone and in combination with ATRA in patients with AML (including APL) the results were of limited value. One reason could be that these agents fail to clear leukaemia-initiating cells (LICs) ([@bib23]). An interesting new approach to specifically target the repressive PML-RARα-HDAC complex in APL has been recently published by [@bib3]. Using a hybrid molecule of ATRA with a 2-aminoanilide tail of the HDAC inhibitor MS-275 (MC2392), they were able to selectively induce apoptosis in cells carrying the PML-RARα fusion gene. We have recently demonstrated a critical role for the histone H3 lysine 4 mono/di (H3K4me1/me2) demethylase LSD1 (KDM1A, AOF2) as a negative regulator of the ATRA-mediated myeloid differentiation pathway ([@bib32]). LSD1 is highly expressed in AML patients ([www.proteinatlas.org](http://www.proteinatlas.org)) and its overexpression has been correlated with impaired survival in various other tumours ([@bib17]). Overexpression of LSD1 in AML is associated with alterations in the histone methylation status and repression of ATRA-responsive genes. We showed that combination treatments with inhibitors of LSD1 and ATRA can dramatically enhance the pro-differentiative effects of ATRA in vitro and in vivo. Co-treatment with ATRA and tranylcypromine (TCP) greatly increased the fraction of CD11b^+^ cells by 21-fold and 16-fold in HL-60 and TEX cells, respectively. Mirroring the results in cell lines, co-treatment with ATRA and TCP increased the fraction of CD11b^+^ cells in primary AML patient samples by a factor up to 11-fold and induced morphological changes, including formation of cytoplasmic neutrophil granules, consistent with differentiation of patient blasts into polymorphonuclear leukocytes. The treatment also lead to diminished clonogenic capacity of AML cells in vitro and engraftment of primary AML cells from patients in vivo in NOD.SCID gamma (NSG) mice, suggesting that LSD1i/ATRA combinations may also target LICs. We showed that these effects are at least in part due to gene-specific increases in H3K4me2 methylation leading to expression of a subset of genes associated with the myelomonocytic differentiation pathway. This work shows that LSD1 is an important therapeutic target and highlights the potential for drugs targeting aberrant epigenetics in combination with retinoids ([Figure 3](#fig3){ref-type="fig"}). Several major pharmaceutical companies are currently developing selective LSD1 inhibitors with micromolar potency and activity against a panel of cancer cell lines ([@bib7]). Phase I trials evaluating treatment with LDS1 inhibitors in several hematologic malignancies are expected to begin this year. Future of ATRA in cancer therapy? ================================= Despite important advances in cancer therapeutics over the past few decades, treatment failure, relapse. and mortality rates in the majority of malignant diseases, including non-APL AML, remain unacceptably high. It is reasonable to predict that cure rates will not improve significantly unless alternative treatment modalities to conventional chemotherapy regimens and bone marrow transplantation are developed. Treatment of AML with targeted therapies is particularly promising due to the fact that AMLs have far fewer mutations than most other adult cancers, with an average of only 13 mutations ([@bib25]; [@bib8]). The availability of relatively inexpensive new techniques with which cancer-causing mutations in individual patients can be analysed, open the possibility for personalised anticancer therapies. Therefore, the goal has to be the development of a toolbox of drugs that can be used in a personalised manner. All-trans-retinoic acid is a promising compound for treatment of a variety of cancers because of its specific effects on cell proliferation, differentiation, and apoptosis, as well as its low toxicity. Retinoic acid receptors are not mutated in cancer cells, so ATRA could potentially exert its anticancer effects in many malignancies. Nevertheless, to unleash the therapeutic potentials of ATRA, many studies emphasise the need for a better understanding of the mechanisms that block ATRA signalling and ATRA regulated gene expression in cancer, including AML. It is evident that combinatorial therapies targeting multiple gene silencing mechanisms may be the most effective strategy in reactivating ATRA-sensitive gene expression and differentiation of AML cells, as well as mediating anticancer activities of ATRA in general. The identification of classes of proteins that control gene expression via histone and DNA modifications is driving the development of new therapeutic agents, the so-called 'epi-drugs\', that alter chromatin structure. These epigenetic modifying drugs have been shown to be only partly effective against different cancers when used alone. In the context of non-APL AML, epi-drugs such as LSD1 inhibitors are most effective when combined with ATRA and vice versa. Further explorations of such combinatorial strategies may very well unravel multiple novel pathways/targets to unlock and rejuvenate the potential of ATRA in anticancer therapy. ![Structure and function of the retinoic acid receptor alpha gene and protein upon binding of RA. (A) RARA, in common with other RAR genes, encodes two major isoforms that differ in their promoters (P1 and P2) and A region sequences, but are identical in their B--F region sequences. The B--F regions contain DNA (DBD) and ligand binding (LBD) domains as well as structural motifs responsible for dimerisation, co-repressor interaction and ligand-dependent trans-activation. (B) Schematic representation of the RARα LBD, which is composed of 12 alpha helices (colour coded and labelled as H1--H12). The ribbon diagrams of the crystal structures of un-liganded RXRα LBD (3A9E) and the liganded RARα LBD (3KMR) are illustrating the principle of RAR activation. In the un-liganded state, helices H3 and H4 provide a binding site for co-repressors. Upon ligand binding H3 and H4 undergo conformational changes leading to a destruction of the co-repressor binding site. Helix H12 moves towards a ligand binding pocket and generates a defined binding surface for co-activators such as p160. (C) RARs possess the capacity to function as a molecular switch; when not bound by ligand, they complex with co-repressors, such as N-CoR or SMRT, and HDACs to actively inhibit gene expression. Upon ATRA binding, co-repressors are released and co-activators, including DRIP/TRAP/ARC or mediator-containing complexes as well as HATs, are recruited to decompress chromatin and activate transcription of target genes.](bjc2014412f1){#fig1} ![Non-genomic effects of ATRA. (A) A fraction of the cellular RARα pool is present in membrane lipid rafts, where it forms complexes with G protein alpha Q (Gαq) in response to RA. Activation of Gαq leads to activation of RhoGTPase, p38MAPK and MSK1 in epithelial and fibroblast cells. Phospho-MSK1 contributes to transcriptional activation of RARα target genes by phosphorylation of histones and RARα itself. (B) In neuronal cells, RA was shown to activate Erks via PI3K/RARα and Src/RARγ. Downstream effectors of Erks are still unknown, but in RA-treated P19 cells, the activated Erks phosphorylate testicular nuclear receptor 2 (TR2) that then becomes a repressor of the octamer-binding transcription factor 4 (Oct4) gene, thus facilitating the differentiation program that is expected to take place in response to RA. (C) RA leads to a rapid response to cyclic adenosine monophosphate (AMP) accumulation in APL cells, most likely by stimulation of adenyl cyclase, leading to protein kinase A (PKA) activation and enhanced expression of RAR target genes and differentiation.](bjc2014412f2){#fig2} ![Mechanism of action of ATRA/epi-drug combinations. (A) Retinoic acid receptors, not bound by ligand, complex with co-repressors and HDACs to actively inhibit gene expression. Cancerous cells such as leukaemia initiating cells (LICs) are frequently characterised by activating mutations or an overexpression of epigenetic regulators (DNMTs, HMTs, HDMs, and HDACs) causing a false repression of tumour suppressor genes or genes involved in differentiation signalling pathways. (B) ATRA treatment alone is not sufficient to unblock RAR target genes do to epigenetic dysregulation. (C) Treatment of cancer cells by epigenetic (epi)-drugs in combination with ATRA unblocks the transcriptional repressive state of RAR target genes and allows the activation gene such as those involved in differentiation signalling pathways.](bjc2014412f3){#fig3} [^1]: These authors contributed equally to this work.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== The blood cockleAnadara granosa is a bivalve mollusc in the family Arcidae and subfamily Anadarinae and locally known as "kerang" in Malaysia \[[@B1]\]. The bivalves in this family are renowned for a source of cheap protein in tropical areas, especially in the Indo-Pacific region \[[@B2]\]. Therefore, marine bivalve particularlyA. granosa is of considerable economic importance in Malaysia \[[@B3]\]. SinceA. granosa is a filter feeding organism, contamination of the highly productive mudflats with heavy metals tends to be accumulated in their whole body tissue. This could serve as an important environmental sink of heavy metals \[[@B4]\] and provide an indication of river pollution. Sarawak like West Malaysia is presently undergoing rapid industrial development and there have been incidences of toxic pollution from industry \[[@B5]--[@B7]\]. Asajaya is a small Sarawakian town located in the Samarahan Division and adjacent to industrial areas which were reclaimed from mangrove. The types of industry in this area presently in operation include food processing and canning, processing of agricultural products, feed mills, timber based wood products, and transport equipment. Sabang is a main river flowing into mentioned area where some fishing villages are situated and connected with Sarawak River and South China Sea. Seafood such as cockles is supplied from Asajaya to most of the seafood markets in this division. Solid and liquid wastes emanating from the industrial activities are the inevitable byproducts of manufacturing process. These wastes contain toxic chemicals and other substances including toxic heavy metals \[[@B8]\]. A number of natural and anthropogenic sources produce heavy metals. Heavy metals such as Cd and Pb are toxic even at relatively low concentration and not essential for metabolic activities \[[@B9], [@B10]\]. The abundance of heavy metal may jeopardize human health due to the consumption of contaminated bivalves \[[@B11]\]. For examples, Cd may cause human carcinogen; Pb can damage blood circulation \[[@B12]--[@B14]\]. People are becoming more aware of the complexity of the nature and the delicate balance that exist within the global ecosystem \[[@B15]\]. The discharge of effluents and associated toxic compounds into aquatic systems represents an ongoing environmental problem due to their possible impact on communities in the receiving aquatic water and a potential effect on human health \[[@B16]\]. In particular, in highly polluted and industrial areas, point and nonpoint sources of anthropogenic chemicals and metals have polluted rivers with highly complex mixtures of chemicals and other anthropogenic perturbations to degree where life in rivers is severely impacted \[[@B17]\]. Therefore, the objective of this study has been to determine the concentrations of two toxic metals (Cd and Pb) in blood cockles sampled from four different sites near Sabang River of Asajaya. 2. Method and Materials {#sec2} ======================= 2.1. Description of Study Sites {#sec2.1} ------------------------------- The blood cocklesAnadara granosa were collected during May 14 from four sites of Sabang River, namely, Kampung Sambir, Kampung Tambirat, Beliong Temple, and Kampung Tanjung Apong, which are located at Asajaya, Sarawak, Malaysia, as shown in [Figure 1](#fig1){ref-type="fig"}. Sabang is a main river flowing into mentioned area where some fishing villages are situated. This river is connected with Sarawak River and with South China Sea. The sites are located adjacent to industrial areas which were reclaimed from mangrove. Currently, the types of industries are food processing and canning, processing of agricultural products, feed mills, timber based wood products, and transport equipment. 2.2. Sampling and Metal Analysis {#sec2.2} -------------------------------- Cockles were transported, refrigerated at 4°C, to the laboratory within 4 h of collection. After a preliminary shell clean-up, the specimens were frozen and maintained at −18°C pending processing. For analysis, all individuals from one site were defrosted; the shells were carefully removed using Teflon-covered forceps and a stainless steel surgical blade; the soft tissues were then freeze-dried at −80°C for about 24 h. These dried samples were then ground to obtain a homogenous powder and stored separately in acid-cleaned glass bottles in desiccators at room temperature until analysis. Approximately 0.5 g of the samples was digested with 10 mL of 65% analytical grade HNO~3~ into a microwave digestion vessel and it was then capped tightly. The digestion vessel (bomb) was then placed in a normal, conventional microwave oven and cooked for two minutes at a 40% power level. The vessel was then cooled for five minutes and then recooked for four minutes at 30% power level. The completely digested sample was then allowed to cool until the fumes dissipated. The inner side of the cap was then cleaned with distilled water and the washed liquid was added to the original sample mixture in the vessel and then it was subjected to filtration using number \#44 Whatman filter paper. The sample solution was diluted to a 50 mL graduated volumetric flask and then subjected to metal determination. Flame atomic absorption spectrometer (FAAS; Perkin-Elmer HGA-600) was employed for the analysis of cadmium and lead. Standard mixtures of 0.05, 0.1, 0.2, 0.5, and 1.0 ppm of Cd and Pb were prepared in nitric acid solution for calibration. The accuracy of the methods was assessed using certified reference material DORM-2 in triplicate, and mean recovery was 95.5 ± 1.3% for both metals. All data were expressed on a dry weight basis (μg/g dry wt.). 3. Results and Discussion {#sec3} ========================= It was observed that the concentrations ranges were 1.35 ± 0.16 to 2.22 ± 0.34 μg/g (dry weight) and 2.65 ± 0.34 to 4.36 ± 0.53 μg/g (dry weight) for Cd and Pb, respectively, in cockles ([Table 1](#tab1){ref-type="table"}). The highest values (2.22 ± 0.34 μg/g for Cd and 4.36 ± 0.53 μg/g for Pb) were observed in specimens from Kampung Tanjung Apong which is located near the connecting point of Sabang River and Sarawak River and the second highest values (1.85 ± 0.28 μg/g for Cd and 3.55 ± 0.48 μg/g for Pb) were observed from Kampung Sambir which is located near the connecting point of Sabang River and China Sea. The levels of Cd and Pb from studied sites exceeded the maximum permissible limits set by Malaysian Food Regulations (1985). The highest value of Cd may be due to the influence of external discrete sources like industrial activities, agriculture runoff and other anthropogenic inputs \[[@B11], [@B18]\] and sites are within the vicinity of agricultural areas mostly large oil palm plantations heavy in pesticides and herbicides used \[[@B5], [@B19]\]. After long term consumption, this may be the cause of possible toxicological risks and heavy metal related diseases, such as Parkinson\'s and Wilson\'s diseases \[[@B20]\]. On the other hand, the highest value of Pb may result from burning of fossil fuels from boats used for fishing and also leisure activities \[[@B6], [@B21]\]. This may be the cause of neurological deficits such as mental retardation in children and kidney disease such as interstitial nephritis to adults and also contribute to hypertension and cardiovascular disease \[[@B22]\] to the consumers after long term consumption. Previous study from the northwest coast of Peninsular Malaysia, in the state of Penang, showed that the levels of heavy metals analyzed for cocklesA. granosa were ranged from 0.87 to 0.89 and 0.11 to 0.12 μg/g dry weights for Cd and Pb, respectively \[[@B23]\]. The results indicate that the levels of Cd and Pb inA. granosa from Penang were lower than safety limits set by Malaysian Food Regulations (1985) as well as lower than those obtained in this research findings. Another research revealed that the levels of heavy metals analyzed for green-lipped musselsPerna viridis collected along the west coast of Peninsular Malaysia were ranged from 0.68 to 1.25 and 2.51 to 8.76 μg/g dry weights for Cd and Pb, respectively \[[@B24]\]. The concentration of Cd was below the reference values for human consumption set by Malaysian Food Regulations (1985) whereas the concentration of Pb was above the reference values. The results also indicate that the level of Pb in the present investigation was within the value from the west coast of Peninsular Malaysia. The use of coastal waters as a convenient receptacle for domestic and industrial wastes threatens the quality of seafood rivers and coastal waters are presently exposed to not only the increasing quantities of metals and nutrients, but also the cocktails of industrial derived contaminants, many of which exhibit significant persistence and capabilities for bioaccumulation \[[@B25]\]. However, it revealed that the cockles in this research area have been contaminated through Cd and Pb caused by agriculture runoff and industrial and other anthropogenic activities which can be the cause of serious human health problems for long term consumption. 4. Conclusion {#sec4} ============= It is concluded that the concentrations of Cd and Pb in blood cockles from four sites of Asajaya exceeded the maximum permissible limits set by Malaysian Food Regulation (1985). It showed that cockles from the mentioned area have been contaminated through Cd and Pb caused by agriculture runoff and industrial and other anthropogenic activities. It may be the cause of serious human health problems such as Parkinson\'s and Wilson\'s diseases and also mental retardation in children and kidney disease such as interstitial nephritis to adults and also contribute to hypertension and cardiovascular disease after long term consumption. Thus, consumer should have consciousness about such type of seafood from mentioned sites and need further investigation. The authors wish to acknowledge the financial assistance provided through the Malaysian Government Research Grant no. ERGS/02(08)860/2012(12) and by the University Malaysia Sarawak (UNIMAS). Conflict of Interests ===================== The authors declare that there is no conflict of interests. ![Map of Asajaya showing sampling sites.](TSWJ2014-924360.001){#fig1} ###### The concentrations (µg/g dry wt) of Cd and Pb in blood cockle Anadara granosa from four sites of Sabang River at Asajaya, Sarawak, Malaysia. ----------------------------------------------------------------------------------------------------- Sampling sites Cd\ Pb\ (mean ± SD) (mean ± SD) ------------------------------------------------------------------------- ------------- ------------- Kampung Sambir 1.85 ± 0.28 3.55 ± 0.48 Kampung Tambirat 1.35 ± 0.16 2.65 ± 0.34 Beliong Temple 1.58 ± 0.23 2.94 ± 0.42 Kampung Tanjung Apong 2.22 ± 0.34 4.36 ± 0.53 The maximum permissible limits set by Malaysian Food Regulations (1985) 1.00 2.00 ----------------------------------------------------------------------------------------------------- SD: standard deviation. [^1]: Academic Editor: Richard J. C. Brown	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Missing data resulting from participant attrition, or incomplete study follow-up, threaten the integrity of a study, with consequences ranging from a reduction in statistical power to significantly biased results and diminished validity. While numerous techniques have been developed to address the problem, none are simple, most require manipulation of the data, and the precision of results can be difficult to determine [@pone.0043138-Andersen1]--[@pone.0043138-Vestbo1]. Missing data could be reduced or preemptively addressed by predicting which participants are most likely to miss visits or be lost to follow-up. Eligibility criteria can be designed to limit enrollment to those with more adherent profiles and supplemental support procedures can be implemented to improve contact with those most at risk of follow-up interruptions. Predicting attrition is challenging and has been the subject of a number of studies in the U.S., Western Europe and sub-Saharan Africa. Reviews of published studies have listed up to 18 different definitions for patients who were considered lost to follow-up [@pone.0043138-Geng1], [@pone.0043138-Fox1], ranging from missed visits to complete disappearance from a study. Despite varying operational definitions of study attrition, certain study designs appear to be associated with higher attrition rates, specifically those that have a longer duration of follow-up [@pone.0043138-Ioannidis1] or frequent, complex or time-consuming visits [@pone.0043138-Andersen1], [@pone.0043138-Davis1], [@pone.0043138-Getz1]. The use of unskilled or inconsistently trained staff and the provision of limited incentives also lead to higher attrition [@pone.0043138-Davis1], [@pone.0043138-DeVita1]. Participant characteristics also appear to play an important role in the likelihood of follow-up interruption. In the US, characteristics of study participants at risk for attrition include lower income [@pone.0043138-DeVita1], less education [@pone.0043138-DeVita1], black race [@pone.0043138-Andersen1], [@pone.0043138-DeVita1], lower CD4 count [@pone.0043138-Andersen1], [@pone.0043138-Ioannidis1], younger age [@pone.0043138-Andersen1], [@pone.0043138-Ioannidis1], [@pone.0043138-deBruyn1], [@pone.0043138-Snow1], hospitalization or incarceration [@pone.0043138-Maru1], injection drug use [@pone.0043138-Andersen1], [@pone.0043138-Ioannidis1], [@pone.0043138-Maru1], unwillingness to travel [@pone.0043138-Maru1], and increased life stressors, multiple health problems and inconsistent use of health care resources [@pone.0043138-Andersen1], [@pone.0043138-Davis1]. Gender differences in rates of missed visits have been inconclusive [@pone.0043138-Andersen1], [@pone.0043138-Davis1], [@pone.0043138-DeVita1], [@pone.0043138-Snow1]. In sub-Saharan Africa and other resource-limited settings, reasons for not completing a study are not fully understood. Despite the paucity of data on the subject, some factors have been associated with attrition in antiretroviral treatment (ART) programs in sub-Saharan Africa, including younger age [@pone.0043138-OchiengOoko1]--[@pone.0043138-Kempf1], baseline CD4 cell count [@pone.0043138-Ekouevi1], [@pone.0043138-Brinkhof1]--[@pone.0043138-Lowrance1] and male gender [@pone.0043138-OchiengOoko1], [@pone.0043138-Ekouevi1], [@pone.0043138-Brinkhof1], [@pone.0043138-Lowrance1], [@pone.0043138-WoolsKaloustian1]. Many of the retention difficulties described by patients are similar to those listed by study participants in the United States, including health problems [@pone.0043138-OchiengOoko1], [@pone.0043138-Roura1], concurrent family and/or work commitments [@pone.0043138-Miller1], [@pone.0043138-Hardon1], limited clinic hours [@pone.0043138-Miller1], long appointments [@pone.0043138-Miller1]--[@pone.0043138-Otieno1], lack of child care [@pone.0043138-Geng1], and lost wages or employment [@pone.0043138-Miller1], [@pone.0043138-Hardon1], [@pone.0043138-Tuller1]. Cultural or socio-economic factors may pose considerable difficulty for study participants with HIV in Africa. These include fear of disclosure of HIV status [@pone.0043138-Otieno1], [@pone.0043138-Ezechi1], bias of family and friends against study participation [@pone.0043138-Roura2], spousal abuse [@pone.0043138-Ezechi1], stigma [@pone.0043138-Hardon1], [@pone.0043138-Otieno1], and the financial costs associated with the program [@pone.0043138-Hardon1]--[@pone.0043138-Tuller1], [@pone.0043138-Rosen1], [@pone.0043138-Maskew1], particularly those relating to transportation [@pone.0043138-Roura1], [@pone.0043138-Otieno1], [@pone.0043138-Ezechi1], [@pone.0043138-Maskew1], [@pone.0043138-Painter1]. Distance-decay is used to explain how health care use is negatively correlated with the distance patients must travel to access care [@pone.0043138-Bigogo1]--[@pone.0043138-Rahaman1]. Specifically, distance-decay implies that patients who must travel greater distances may delay seeking health care and have worse outcomes [@pone.0043138-Moisi1]. Research studies often require multiple study visits at a higher frequency than routine care, and similar effects of travel distance on adherence to visit schedules may also occur in research settings. To examine this possibility and assess other correlates of poor adherence to study visits, we used follow-up records from a prospective cohort study of HIV-1-discordant couples conducted in Nairobi, Kenya to determine whether the distance between a participant's home and the research clinic was an important factor in explaining missed study visits. Methods {#s2} ======= Ethics Statement {#s2a} ---------------- Written informed consent was obtained from all participants. This study received ethical approval from the institutional review board of the University of Washington and the Ethics and Research Committee of the University of Nairobi and was conducted according to the guidelines set forth by the United States Department of Health and Human Services. Recruitment and Enrollment {#s2b} -------------------------- HIV-1-serodiscordant heterosexual couples were enrolled in a prospective cohort study between May 2007 and October 2009. Couples were eligible if one member of the couple was HIV-1-infected and the other HIV-1 susceptible, both partners reported having sexual intercourse at least 3 times in the previous 3 months, and planned to stay together for at least two years. Participants who were pregnant or using antiretroviral therapy at enrollment were excluded. Couples were referred to the study from HIV voluntary counseling and testing (VCT) sites throughout the greater Nairobi region and from a clinical trial site at Kenyatta National Hospital that had identified discordant couples. Couples provided written informed consent at the screening and enrollment visits as well as contact and locator information for retention-tracking purposes. All couples received at least one home visit from a peer counselor to verify the location of their residence. Study clinicians collected baseline sociodemographic, sexual/reproductive and medical history data from both members of the couple, and blood and genital samples were collected at screening, enrollment and subsequent visits. Study Follow-up Visits {#s2c} ---------------------- Couples were asked to return to the study clinic quarterly to provide blood and genital specimens, complete questionnaires on recent sexual behavior and current health status, and be tested for HIV-1 seroconversion. HIV-1-infected partners were supplied with trimethoprim-sulfamethoxazole to be taken daily as prophylaxis for opportunistic infections. Couples continued to be followed in the study even if they separated, the female partner became pregnant, or one partner left the study. Support groups for those with and without HIV infection were available throughout the study. Drinks were supplied free of charge during clinic visits, and participants were reimbursed 500 Kenyan Shillings (KSh), (approximately 3.50 US dollars \[USD\]) per visit for transportation expenses to and from the clinic. Laboratory Procedures {#s2d} --------------------- At enrollment and quarterly visits, HIV status of the uninfected partner was determined by two rapid tests conducted in parallel using the Determine HIV-1/2 rapid test (Abbott Laboratories, Tokyo, Japan, now marketed by Inverness Medical as Alere Determine) and the Bioline HIV 1/2 rapid test (Standard Diagnostics Inc., Suwon, Korea), with confirmation by the Vironostika HIV Uni-form II Ag/Ab ELISA kit (bioMérieux SA, Marcy l'Etoile, France) for positives or indeterminates. Urine pregnancy tests were also performed every three months (Quick Vue One Step hCG Urine Pregnancy kit, Quidel Corp, San Diego, CA). CD4 cell counts (FACSCaliber, BS Bioscience, Franklin Lakes, USA) and HIV-1 RNA viral load (Gen-Probe Transcription Mediated Amplification, San Diego, USA) were measured in HIV-1-infected participants at enrollment and every six months. Geographic Mapping {#s2e} ------------------ The locations of participants' residences were mapped using locator information provided by each participant. Based on this information, a member of the study retention team mapped each residence using a web-based mapping application built on the Google Maps engine (Google, Inc, Mountain View, CA). The geographic coordinates for all VCT sites and the study clinic were determined using a handheld global positioning system (GPS) unit (Trimbel Juno, Sunnyvale, USA). The resulting coordinates were used to estimate linear (as the crow flies) distances between each residence and the referring VCT and between each residence and the study clinic using a geospatial algorithm (Google, Inc). 10.1371/journal.pone.0043138.t001 ###### Baseline characteristics of study participants. ![](pone.0043138.t001){#pone-0043138-t001-1} Couples with HIV-1-Infected Male Partner Couples with HIV-1-Infected Female Partner -------------------------------------------------------------- ------------------------------------------ -------------------------------------------- ----------------- ------------------ Characteristic median (IQR) or n (%) Age (Years) 36 (32, 42.5) 29 (25, 35) 34 (29, 39) 28 (24, 34) Relationship duration (years) 6.3 (2.4, 10.9) 5.8 (2.4, 11.2) 4.77 (2.4, 9.5) 4.8 (2.3, 8.7) CD4 Count (cells/µl) 340.5 (231, 503) -- -- 476 (307.5, 673) HIV-1 Viral Load (Log~10~ copies/mL) 4.8 (4.1, 5.4) -- -- 4.6 (3.7, 5.2) Home-to-clinic distance (km)[a](#nt102){ref-type="table-fn"} 0--4.9 36 (21.8) 36 (21.8) 58 (19.3) 58 (19.3) 5--9.9 55 (33.3) 55 (33.3) 114 (37.9) 114 (37.9) 10--14.9 46 (27.9) 46 (27.9) 77 (25.6) 77 (25.6) ≥15 28 (17.0) 28 (17.0) 52 (17.3) 52 (17.3) Married to Study Partner 165 (98.2) 162 (97.0) 289 (96.0) 284 (94.7) Number of Living Children 0 13 (7.7) 18 (10.7) 42 (14.0) 37 (12.3) 1--2 81 (48.2) 86 (51.2) 164 (54.5) 180 (59.8) ≥3 74 (44.1) 64 (38.1) 95 (31.6) 84 (27.9) Less than Primary school education 25 (14.9) 45 (26.8) 45 (15.0) 71 (23.6) Informal Residence 85 (50.6) 83 (49.4) 153 (50.8) 161 (53.7) Home Ownership/Rent Rent \< KSh 2,000[b](#nt103){ref-type="table-fn"} 66 (39.3) 66 (39.3) 134 (44.5) 138 (45.9) Rent ≥ KSh 2,000 87 (51.8) 87 (51.8) 140 (46.5) 140 (46.5) Own home 15 (8.9) 15 (8.9) 27 (9.0) 23 (7.6) Monthly household income (KSh) Couple does not earn income 20 (11.9) 20 (11.9) 35 (11.6) 35 (11.6) Monthly income \<8,000[c](#nt104){ref-type="table-fn"} 57 (33.9) 57 (33.9) 111 (36.9) 111 (36.9) Monthly income ≥8,000 91 (54.2) 91 (54.2) 155 (51.5) 155 (51.5) NOTE. IQR = interquartile range; KSh = Kenyan Shillings. Monthly household income is the sum of the income for both study partners. Linear distance, based on location of residence at enrollment. Data missing for 6 participants. Based on the median rent (2,000 KSh ≈ 26 USD). Based on the median monthly income (8,000 KSh ≈ 104 USD). Statistical Analysis {#s2f} -------------------- The outcome in this analysis was time until study follow-up interruption (FUI), defined as missing two consecutive quarterly study visits. We used Cox proportional hazards regression to evaluate the time from enrollment until the first FUI, calculated as the length of time between the date of the enrollment visit and the scheduled date of the first missed visit (three months after the last completed clinic visit). Participants were censored at the date for their last scheduled follow-up visit, or at the date of death for participants who died during follow-up. If the exact date of death was unknown, the censoring date was defined as three months after the last completed visit. The crude associations between FUI and factors of interest were estimated by hazard ratios (HR) and 95% confidence intervals (CI). We hypothesized a priori that HIV status would modify the associations between FUI and distance variables, and therefore all analyses were conducted separately for HIV-1-infected and HIV-1-uninfected individuals. For regression models, distance was considered both continuously and categorically. We assumed that distances beyond 5 kilometers (km) were likely to necessitate transport other than walking, and therefore categorized linear distance into 5 km intervals (0--5, 5--10, 10--15, and \>15). Potential confounders were included stepwise in a multivariable model and remained in the model if the adjusted association differed from the unadjusted hazard ratio by more than 10% in either the HIV-1-infected or -uninfected model. Robust standard errors were used to relax the assumption of normally distributed standard errors. All statistical analyses were conducted in Stata 10.1 (StataCorp, College Station, TX) with statistical significance criteria set at α\<0.05. ![Cumulative incidence of follow-up interruptions.\ HIV-1-discordant couples were enrolled and followed quarterly for 2 years. A follow-up interruption (FUI) was defined as missing ≥2 consecutive study visits. The Kaplan Meier curves show the cumulative incidence of FUI separately for (---) HIV-1-uninfected females, (---------), -infected females, (---) -uninfected males, and (-- -- --) -infected males.](pone.0043138.g001){#pone-0043138-g001} Results {#s3} ======= Study Population and Cohort Characteristics {#s3a} ------------------------------------------- Four hundred sixty nine HIV-1-discordant couples (938 individuals) were enrolled into the study. The male member of the couple was HIV-1-infected in 168 (36%) of enrolled couples, and the female was the infected partner in 301 (64%) of the couples. Median CD4 cell count at baseline in the HIV-1-infected was 415 cells/µl (Interquartile Range \[IQR\]: 281, 618) and median HIV-1 viral load was 4.6 log~10~ copies/ml (IQR: 3.9, 5.3). Among all enrolled, the median age at enrollment was 35 years (IQR: 30, 41) for men and 29 years (IQR 24, 34) for women ([Table 1](#pone-0043138-t001){ref-type="table"}). A total of 186 (20%) participants had less than a primary school education: 70 (15%) men and 116 (25%) women. Fifty-five (12%) of the couples enrolled reported no household income at study entry, 36% reported earning less than the median of 8,000 KSh per month, and 197 (52%) earned more than 8,000 KSh per month, equivalent to approximately 130 USD at the time of the study. Eighty (9%) participants reported owning a home at study entry with the remaining 858 (91%) renting their current residence. ![Map of distances from the study clinic.\ The map of Nairobi shows the location of the study clinic, with linear distances to the study clinic indicated by concentric white circles at 5 km increments. The percentage of participants living within each section is shown in yellow.](pone.0043138.g002){#pone-0043138-g002} Incidence of Follow-up Interruptions {#s3b} ------------------------------------ Of those enrolled, 197 individuals missed two consecutive study visits and were categorized as having an FUI during 1471 person-years (PY) of follow-up, for a rate of 13.4 FUI per 100 PY. Among HIV-1-infected partners 46 women and 35 men experienced FUI and among HIV-1-uninfected partners, 46 women and 70 men experienced FUI ([Figure 1](#pone-0043138-g001){ref-type="fig"}). The rate of FUI was 10.8 per 100 PY in the HIV-1-infected and was significantly lower than the rate of 16.1 per 100 PY in the uninfected participants (HR = 0.66, p = 0.004). Among the HIV-1-infected, women had a slightly lower FUI rate than men (9.4 versus 13.2 per 100 PY; HR = 0.72, p = 0.138), while among the HIV-1-uninfected, women had a slightly higher FUI rate than men (18.2 versus 15.0 per 100 PY; HR = 1.22, p = 0.229). Among women, the HIV-1-infected had a significantly lower FUI rate than the uninfected (HR = 0.52, p\<0.01). However, among the men, there was no difference in the FUI rate between HIV-1-infected and -uninfected (HR = 0.87, p = 0.485). Relationship between Home-to-Clinic Distance and FUI {#s3c} ---------------------------------------------------- For each couple we ascertained a linear distance between the residence and the study clinic ([Figure 2](#pone-0043138-g002){ref-type="fig"}). Patients lived a median of 9.14 km (IQR: 6.15, 12.79) from the clinic, with three couples living more than 100 km from the clinic. Eighty-three percent of couples lived within 15 km of the study clinic, and 20% of couples lived within 5 km. Among HIV-1-infected participants, living between 5 and 10 km from the study clinic was significantly associated with a greater than two-fold increased likelihood of FUI when compared to those living \<5 km from the clinic (Hazard Ratio \[HR\] 2.17; 95% CI: 1.09, 4.34; p = 0.03) ([Table 2](#pone-0043138-t002){ref-type="table"}). Similar associations were seen in the 10--15 km (HR: 1.55, 95% CI: 0.74, 3.26; p = 0.25) and \>15 km categories (HR: 1.48, 95% CI: 0.66, 3.33; p = 0.35), when compared to the less than 5 km distance category; however, these did not reach statistical significance. No significant associations between FUI and home-to-clinic distance were found among HIV-1-uninfected individuals. Similarly, no associations between FUI and home-to-clinic distance were found when distance was considered as a linear variable in the model. 10.1371/journal.pone.0043138.t002 ###### Correlates of study follow-up interruption. ![](pone.0043138.t002){#pone-0043138-t002-2} HIV-1-Infected HIV-1-Uninfected -------------------------------------------------------------- ---------------------------------------- ------------------ ------ -------------- Female gender 0.72 (0.47, 1.11) 1.26 (0.88, 1.81) Age 1.00 (0.87, 1.15) 0.95 (0.84, 1.07) Relationship duration 0.98 (0.94, 1.02) 0.98 (0.95, 1.01) CD4 count 0.95 (0.88, 1.04) -- -- HIV-1 viral load (Log~10~ copies/mL) 1.23 [\](#nt109){ref-type="table-fn"} (1.01, 1.51) -- -- Home-to-clinic distance (km)[a](#nt106){ref-type="table-fn"} 0--4.9 1 ref 1 ref 5--9.9 2.17 [\](#nt109){ref-type="table-fn"} (1.09, 4.34) 0.81 (0.50, 1.30) 10--14.9 1.55 (0.74, 3.26) 0.67 (0.39, 1.13) ≥15 1.48 (0.66, 3.33) 0.88 (0.50, 1.53) Married to study partner 0.75 (0.27, 2.04) 0.53 (0.25, 1.10) Number of living children 0 1 ref 1 ref 1--2 0.61 (0.33, 1.15) 0.73 (0.44, 1.22) ≥3 0.62 (0.32, 1.20) 0.87 (0.51, 1.47) Less than primary education 0.51 [\](#nt109){ref-type="table-fn"} (0.27, 0.98) 1.34 (0.88, 2.04) Informal residence 0.99 (0.65, 1.52) 1.22 (0.86, 1.75) Own home/monthly rent Rent \< KSh 2,000[b](#nt107){ref-type="table-fn"} 1 ref 1 ref Rent ≥ KSh 2,000 1.67 [\](#nt109){ref-type="table-fn"} (1.05, 2.65) 1.14 (0.78, 1.67) Own home 1.23 (0.51, 2.97) 1.15 (0.61, 2.18) Monthly household income (KSh) Couple does not earn income 1 ref 1 ref Monthly income \<8,000[c](#nt108){ref-type="table-fn"} 0.69 (0.36, 1.35) 0.68 (0.39, 1.19) Monthly income ≥8,000 0.78 (0.42, 1.46) 0.64 (0.37, 1.09) NOTE. HR = hazard ratio; CI = confidence interval; KSh = Kenyan Shillings. The HR for age is per 5 years, the HR for relationship duration is per 1 year, and the HR for CD4 count is per 100 cells/µl. Monthly household income is the sum of the income for both study partners. Confidence intervals were calculated using robust standard errors. Linear distance, based on location of residence at enrollment. Data missing for 6 participants. Based on the median rent. 2,000 KSh ≈ 26 USD. 8,000 KSh ≈ 104 USD. p\<0.05. To further investigate the finding that a residence 5--10 km from the study clinic was the highest risk group among HIV-1-infected participants, we conducted a series of post-hoc analyses. We found no significant difference in risk of FUI between HIV-1-infected participants living ≥10 km from the clinic compared to those living 5--10 km (HR = 0.70, 95% CI 0.44 to 1.11, p = 0.13), and when we collapsed these categories, there was an attenuated association between distance from study clinic and FUI (HR = 1.80 comparing ≥5 km to \<5 km, 95% CI 0.94, 3.47, p = 0.08). Sociodemographic Correlates of Study Follow-up Interruption {#s3d} ----------------------------------------------------------- We evaluated other sociodemographic factors as potential correlates of FUI using Cox regression with separate models for HIV-1-infected and -uninfected participants. Among HIV-1-infected participants, higher baseline log~10~ HIV-1 viral load was associated with increased FUI rates (HR = 1.23, 95% CI: 1.01, 1.51), and having less than a primary school education was associated with lower FUI rates (HR = 0.51, 95% CI: 0.27, 0.98). Also among HIV- infected individuals, paying greater than or equal to the median monthly rent (≥2000 KSh, approximately 26 USD) compared to rent less than the median rent was associated with a 67% increase in the rate of FUI (95% CI: 1.05, 2.65). We found no significant associations between baseline characteristics and FUI among HIV-1-uninfected participants. A multivariable model fit separately for HIV-1-infected and -uninfected patients including primary school education, monthly rent and log~10~ HIV-1 viral load (for the HIV-1 infected group only) found no independent associations with FUI (data not shown). To determine whether gender modified the association between linear distance and FUI, we stratified the analysis by both gender and HIV-1 status. Among the HIV-1-positive women, living between 5 and 10 km from the study clinic was associated with a nearly three-fold increase in the FUI rate compared to those living \<5 km from the clinic (95% CI: 1.01, 8.61; p = 0.05) ([Table 3](#pone-0043138-t003){ref-type="table"}). Distance categories of 10--15 km and \>15 km had increased FUI rates, but these did not achieve statistical significance (HR = 2.04, p = 0.22 and HR = 2.45, p = 0.14, respectively). Conversely, among HIV-1-uninfected males, there was a pattern of decreased FUI with home-to-clinic distances greater than 5 km, but these associations were not statistically significant. Overall we did not find that gender modified the association between distance and FUI; the likelihood ratio test for a distance/gender interaction was not significant in either HIV-1-infected (p = 0.420) or -uninfected groups (p = 0.999). Rent/home ownership was the only confounder in the association between distance and FUI. After adjustment for rent/home ownership, the 5--10 km distance was no longer statistically significant among the HIV-positive females. 10.1371/journal.pone.0043138.t003 ###### Association between home-to-clinic distance and follow-up interruption, by gender and HIV status. ![](pone.0043138.t003){#pone-0043138-t003-3} Couples with HIV-1-Infected Male Partner Couples with HIV-1-Infected Female Partner ---------------- ------------------------------------------ -------------------------------------------- ------ -------------- ------ -------------- ------ -------------- ------ -------------- ------ -------------- ---------------------------------------- -------------- ------ -------------- 0 to \<5 1 ref 1 ref 1 ref 1 ref 1 ref 1 ref 1 ref 1 ref 5 to \<10 1.74 (0.70, 4.30) 1.63 (0.67, 4.00) 1.01 (0.47, 2.18) 0.68 (0.37, 1.25) 0.72 (0.39, 1.31) 0.94 (0.41, 2.20) 2.95 [\](#nt112){ref-type="table-fn"} (1.01, 8.61) 2.65 (0.88, 8.00) 10 to \<15* 1.21 (0.46, 3.20) 1.19 (0.46, 3.08) 0.56 (0.23, 1.39) 0.69 (0.35, 1.35) 0.74 (0.38, 1.41) 0.55 (0.22, 1.40) 2.04 (0.65, 6.38) 1.78 (0.54, 5.86) ≥15 0.83 (0.24, 2.82) 0.84 (0.63, 2.66) 1.16 (0.48, 2.78) 0.66 (0.29, 1.51) 0.74 (0.36, 1.52) 1.20 (0.50, 2.92) 2.45 (0.76, 7.89) 2.05 (0.62, 6.75) NOTE. HR = hazard ratio; CI = confidence interval. Adjusted for housing status (own home, renting for less than the median, or renting for greater than or equal to the median). p\<0.05. Discussion {#s4} ========== In a longitudinal cohort of HIV-1-discordant couples followed quarterly, we found that rates of FUI were high (21% overall) and that distance between home and the study clinic was significantly associated with FUI, particularly among HIV-1-infected women. While HIV-1-infected female partners were the least likely to miss two consecutive quarterly visits, they were approximately three times more likely to experience FUI if they resided greater than walking distance (\>5 km) from the clinic. We found a complicated relationship between distance from participants' residences and the study clinic and their likelihood of an FUI. The association between home-to-clinic distance and FUI was restricted to HIV-1-infected participants, and the strongest association was among those living 5--10 km from the study clinic with a weaker relationship among those living \>10 km from the clinic. This pattern was stronger among HIV-1-infected women compared to infected men. Transportation challenges related to traveling more than 5 km to reach the clinic may pose a barrier to attending study visits. For example, those living within 5--10 km from the centrally located study clinic resided within city limits, potentially resulting in higher transportation costs (approximately 200 KSh per study visit) if they could not walk the distance to clinic. Participants living \<5 km from the study clinic may have walked to study visits, or if not, were unlikely to have spent a large proportion of the compensation on direct transportation costs (likely less than 60 Ksh). Therefore, the 500 KSh per visit incentive, an amount equal to a quarter of the median monthly rent (2,000 KSh), may pose a greater incentive for those residing closer to the clinic compared to those living at or near the city limits. We did not find a significant association between longer home-to-clinic distances (i.e., \>10 km) and the risk of FUI. However, the point estimates for this category were slightly lower than in the 5--10 km category, and the lack of statistical significance may simply represent a loss of statistical power. We hypothesize that those who lived a greater distance from the study clinic and chose to enroll may have a greater level of motivation to continue participation, possibly due to a desire to obtain specialized health care not available near their home or to avoid stigma related to HIV infection. Additionally, living farther from the center of town may be indicative of lower socioeconomic status and increased reliance on the financial reimbursement provided by the study, and thus decreased rates of FUI. In this study, we found that women experienced fewer FUI than men, which is consistent with other studies that have found greater adherence to clinic visits and less program attrition among HIV-1-infected women compared to -infected men [@pone.0043138-Geng1], [@pone.0043138-Ekouevi1], [@pone.0043138-WoolsKaloustian1], [@pone.0043138-Toure1]. It has been hypothesized that this is due to women's motivation to remain healthy for her family [@pone.0043138-OchiengOoko1]. While our study confirmed that HIV-1-infected women were less likely to miss study visits, we also found that within this group, home-to-clinic distance may play a role in FUI. HIV-1-infected women who are otherwise motivated to adhere to a study visit schedule may find the \>5 km distance a challenge if they begin to experience HIV-related symptoms. A link between ill health and missed visits is supported by the fact that a higher viral load at enrollment was a significant predictor of increased risk of FUI. Higher viral load, a marker for HIV disease progression, could result in difficulty adhering to scheduled visits since participants who are ill may be unable to travel or may need to spend time seeking care and treatment rather than attending study visits. If antiretroviral treatment is commenced, participants with poor adherence to their medications may also demonstrate poor adherence to study visits as well. Additionally, previous studies have found that women spend more time traveling to obtain basic needs (e.g., food and water) for their families. Therefore, a longer distance to reach the study clinic may strain the physical and time resources of HIV-1-infected women compared to their male counterparts, resulting in the greater observed impact of distance on FUI among this group [@pone.0043138-Republic1], [@pone.0043138-Wasike1]. We also found that paying higher rent was associated with a 67% increase in the FUI rate compared to paying lower rent. If a higher rent is an indicator of greater employment obligations, then the need to take time off from work to attend study visits may explain this association. However, since we saw no statistically significant association between household income and FUI, the association with rent may be influenced by other factors, such as the distance between the place of employment and study clinic. Having less than a primary education was also significantly associated with nearly half the rate of FUI compared to those with at least a primary education. Lower levels of education could limit a person's ability to find stable employment, which could increase the need for the financial benefits of participating in a research study. This study had several limitations. We were able to obtain measurements of the distance between participants' residences and the clinic; however, we were unable to determine the time or financial cost required, or methods of transportation available. In one study, travel cost was found to be significantly associated with FUI, while the distance was not [@pone.0043138-Zachariah1]; in another, travel time and distance were significantly associated, while cost was not [@pone.0043138-Buor1]. Also, because this secondary study was conducted retrospectively, we were unable to assess participant perceptions of the impact of travel and distance on their ability to adhere to visits, or any other variables that could have affected study follow-up. In addition, we were only able to access locator information obtained at the first clinic visit. As a result, our data did not account for subsequent changes in residence. Despite its limitations, our study provides important information for further investigation. We found that living further from the study clinic was not significantly associated with an increased risk of FUI, with the exception of HIV-1-infected women between 5 and 10 km, indicating that enrollment of participants living far from the clinic may not be inherently problematic. We also found that HIV-1-uninfected men and women had higher rates of interruption of study follow-up and are an important group on which to concentrate retention efforts in clinical trials and prospective studies that depend on retaining both HIV-1-infected and -uninfected participants. In this study, higher rates of FUI among HIV-1-uninfected partners may have been due to the lack of direct benefits from attending study visits. Unlike their infected partners who received CD4 cell count testing and drugs to prevent opportunistic infections at the study clinic, the uninfected partners may have experienced limited direct benefits from study participation, and consequently may have been less motivated to attend visits. There has been limited research on retention of HIV-1-discordant couples in clinical research studies in sub-Saharan Africa. Yet these couples bear one of the greatest burdens of HIV transmission, and are already the focus of several large HIV prevention studies. Understanding and addressing transportation challenges in settings without easily accessible and inexpensive modes transportation could improve retention, both in research studies as well as in routine care and disease management. Studies combining quantitative and qualitative methods to address the barriers and incentives to study retention are needed to improve study retention and more importantly, to improve utilization and access to health care in sub-Saharan Africa settings. We thank the couples who participated in the study and the clinical, laboratory, and data management personnel who made this research possible. We would also like to thank Dr. Matthew Dunbar for his assistance in mapping the location of participants' residences. [^1]: Competing Interests:The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: NJC BLG RB CF. Performed the experiments: BLG RDM AM RYC RB AG CF. Analyzed the data: NJC PBP BLG RDM. Wrote the paper: NJC PBP BLG CF.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#SEC1} ============ In recent decades, non-coding RNA (ncRNA) has gradually become a research hotspot because of its important roles in a wide range of biological processes. ncRNA can interact with various macromolecules, including DNA, RNA and proteins, and regulate gene expression at transcriptional, post-transcriptional and epigenetic levels ([@B1],[@B2]). The number of identified ncRNAs has also sharply increased with the widespread usage of high-throughput technology in different cells and tissues. The latest GENCODE version 29 annotated more than 40 000 ncRNAs in the human genome. Many ncRNAs are important oncogenes or tumour suppressor genes, such as HULC ([@B3]) and HOXB-AS3 ([@B4]). However, compared with protein-coding genes, the function of most ncRNAs remains to be deciphered. Single-nucleotide polymorphisms (SNPs), the most common type of human genetic variation, play important roles in human complex traits and diseases ([@B5]). Genome-wide association studies (GWAS) have found extensive SNPs associated with various traits and diseases. However, most GWAS-detected risk SNPs are located in the genomic non-coding regions ([@B8]), which indicates that ncRNAs may be possible causal targets of some GWAS loci ([@B9]). For example, rs6983267 on human chromosome 8q24.21 is a potential genetic biomarker of colorectal cancer predisposition and is located far from protein-coding genes. Recent studies demonstrated that rs6983267 may exert its role through influencing the expression of lncRNA CCAT2 ([@B10],[@B11]). Therefore, investigation of the effects of SNPs on ncRNA expression will help to understand how risk alleles contribute towards tumorigenesis and cancer development. Expression quantitative trait locus (eQTL) analysis links variations in gene expression to genotypes and has been considered a powerful tool to understand the effects and molecular mechanism of functional SNPs ([@B12]). However, most eQTL analyses are focused on the associations between genotypes and protein-coding genes; only a few ncRNA-related eQTL (ncRNA-eQTL) analyses have been performed at the genome-wide level ([@B16]). In recent years, although some ncRNA-related SNP databases such as LincSNP 2.0 ([@B17]), MSDD ([@B18]) and lncRNASNP2 ([@B19]) have been developed to explore the relationship of ncRNAs and SNPs, no database has been developed to specifically and comprehensively quantify the association between SNP and ncRNA expression. The Cancer Genome Atlas (TCGA) consortium ([@B20]) has generated DNA germline genotype datasets, transcriptome profiling and patient survival data for over 10 000 primary tumours in 33 cancer types. Using these valuable datasets, we previously performed eQTL analyses for all cancer types and developed the pancanQTL online database ([@B21]), which provides eQTLs of 20 531 genes. However, most of the genes in pancanQTL are protein-coding genes. Recently, the TCGA database has updated its gene expression profiles and provided expression profiles of more than 40 000 lncRNAs. In addition, the TCGA has also provided microRNA (miRNA) expression profiles. These data enable us to systematically analyse associations between SNPs and the gene expression of ncRNAs (including lncRNAs and miRNAs). Thus, in this study, using TCGA genotype data and the latest expression profiles, we developed a computational pipeline to systematically identify ncRNA-eQTLs across 33 cancer types. In addition, we linked ncRNA-eQTLs to known GWAS loci and patient survival times and identified thousands of GWAS-related eQTLs and survival-related eQTLs. Finally, we constructed a user-friendly database, ncRNA-eQTL (<http://ibi.hzau.edu.cn/ncRNA-eQTL>), for users to browse and download data. DATA COLLECTION AND PROCESSING {#SEC2} ============================== Collection and processing of genotype data {#SEC2-1} ------------------------------------------ Genotype data of 33 cancer types (full names of cancer types are shown in [Supplementary Table S1](#sup1){ref-type="supplementary-material"}) of each individual were obtained from the TCGA data portal (<https://tcga-data.nci.nih.gov/tcga/>), which were called by the Affymetrix SNP 6.0 array platform and included 898 620 SNPs (Figure [1A](#F1){ref-type="fig"}). To increase the power for eQTL discovery, we imputed autosomal variants for all samples using IMPUTE2 ([@B22]) in each cancer type, with 1000 Genomes Phase 3 as the reference panel ([@B23]). A two-step imputation procedure was used to improve the computation efficiency. TCGA genotype data were first pre-phased to certain haplotypes and then imputed from the reference panel into the estimated study haplotypes. After imputation, the following criteria ([@B24]) were employed to remove SNPs: (i) imputation confidence score, INFO \< 0.4, (ii) minor allele frequency (MAF) \< 5%, (iii) SNP missing rate \> 5% for best-guessed genotypes at posterior probability ≥ 0.9 and (iv) Hardy--Weinberg equilibrium P-value \< 1 × 10^−6^ estimated by Hardy--Weinberg R package ([@B25]) (Figure [1B](#F1){ref-type="fig"}). ![Flowchart of the ncRNA-eQTL database. (A) Data collection of sequencing, genotype and GWAS data. (B) Processing of sequencing and gene expression data. (C) Database content of four types of eQTLs (cis-eQTLs, trans-eQTLs, survival-eQTLs, GWAS-eQTLs).](gkz711fig1){#F1} Expression data processing and covariates {#SEC2-2} ----------------------------------------- Gene expression profiles generated from RNA sequencing (RNA-seq) and small RNA sequencing (smRNA-seq) for each sample were obtained from the TCGA data portal (<https://gdc-portal.nci.nih.gov/>). The gene annotation, which was used for TCGA RNA-seq and miRNA-seq annotation, was downloaded from the GENCODE (version 22) website (<https://www.gencodegenes.org/>) and miRBase (<http://www.mirbase.org/>). According to the annotation, we removed all the protein-coding genes from the profiles. In each cancer type, lncRNAs with an average expression of ≥ 0.01 FPKM and miRNAs with an average expression of ≥0.01 TPM were retained. To minimize the effect of outliers on the regression scores, the expression values for each gene across all samples were transformed into a standard normal based on rank ([@B24]) (Figure [1B](#F1){ref-type="fig"}). Global expression data may be influenced by several factors, such as batch effects ([@B26]) and genetic and non-genetic biases ([@B27]). Thus, covariates are often included to correct known and unknown confounders and increase the sensitivity of eQTL analyses ([@B24]). To adjust the global effect of population structure, we first used smartpca in the EIGENSOFT program ([@B28]) to perform principal component analysis (PCA) for each cancer type. The top five principal components in the genotype data were selected as covariates. To eliminate the possible batch effects and other confusions hidden in the expression data, we used PEER software ([@B29]) to select the first 15 PEER factors from the expression data as covariates. Other factors, such as tumour stage ([@B30]), gender ([@B24]) and age ([@B12]), were also counted as additional covariates (Figure [1A](#F1){ref-type="fig"}). Identification of cis- and Trans-eQTL using Matrix eQTL {#SEC2-3} ------------------------------------------------------- Cis-eQTL and trans-eQTL analyses were performed in our study. The cis-eQTLs were defined if the SNP was within 1 Mb from the gene transcriptional start site (TSS) and regulating the corresponding gene expression ([@B24]), and trans-eQTLs were defined if the eQTL was beyond that region or on another chromosome. To perform cis-eQTL analyses and trans-eQTL analyses, we first downloaded SNP annotations (GRCh38) (<https://www.ncbi.nlm.nih.gov/projects/SNP>) from the dbSNP database. We analysed the associations between each ncRNA and autosomal SNP through linear regression by employing a computationally efficient eQTL analysis called Matrix eQTL ([@B31]), controlling for population bias, sex, age, tumour stage and unobserved factors in the expression data for each cancer type ([@B29],[@B31]). We defined eQTLs as SNPs with false discovery rates (FDRs) calculated by MatrixEQTL \< 0.05 (Figure [1C](#F1){ref-type="fig"}). Identification of GWAS-associated eQTLs {#SEC2-4} --------------------------------------- Overlaps between ncRNA-eQTLs and SNPs in GWAS regions were identified to explore the possible target genes of existing GWAS loci. To achieve that, we first downloaded all the known risk tag SNPs identified by GWAS from the National Human Genome Research Institute (NHGRI) GWAS Catalog ([@B32]) (<http://www.ebi.ac.uk/gwas>, accessed September 2018), a collection of data from GWAS for various human diseases and traits. Then, we obtained the SNPs in linkage disequilibrium (LD) with these tag SNPs from SNAP ([@B33]) (<https://personal.broadinstitute.org/plin/snap/ldsearch.php>). The parameters were set as follows: SNP dataset: 1000 Genomes; r^2^ (the square of the Pearson correlation coefficient of linkage disequilibrium) threshold: 0.5; population panel: CEU (Utah residents with northern and western European ancestry); distance limit: 500 kb. Finally, these GWAS tag SNPs and LD SNPs were mapped to ncRNA-eQTL results. Identification of survival-associated eQTLs {#SEC2-5} ------------------------------------------- To prioritize promising ncRNA-eQTLs, we identified ncRNA-eQTLs that may be associated with patient survival times. The clinical data, including patient overall survival times, were downloaded from the TCGA data portal. For each ncRNA-eQTL, we divided the samples into three different groups: AA (homozygous genotype), Aa (heterozygous genotype) and aa (homozygous genotype), and removed the group with samples fewer than three for reliable survival analysis and Kaplan--Meier (KM) P-value calculation. Then, the differences in survival times between the groups were detected by log-rank test, and KM curves were introduced to visualize the survival time difference of each group. In addition, survival-associated eQTLs were confirmed for ncRNA-eQTLs that fulfilled the condition of FDR \< 0.05. Database construction {#SEC2-6} --------------------- We organized all results in MySQL (version 5.6) relation tables and constructed a web interface using HTML, CSS and PHP (version 5.4) running on an Apache web server (version 2.4.6). A highly flexible and editable plug-in for the jQuery JavaScript library called DataTables (<https://datatables.net/>) was integrated to display the data content in a dynamic way. DATABASE CONTENT AND THE WEB INTERFACE {#SEC3} ====================================== Database content and statistics of lncRNA-related eQTLs {#SEC3-1} ------------------------------------------------------- Using genotype data and RNA-seq data, we first performed lncRNA-related cis-eQTL and trans-eQTL association analyses in each of the 33 cancer types independently. The sample sizes of cancer types ranged from 36 in cholangiocarcinoma (CHOL) to 1067 in breast invasive carcinoma (BRCA), with a median of 177 (Table [1](#tbl1){ref-type="table"}). After genotype imputation, the incorporated SNPs ranged from 2 745 615 in BRCA to 5 078 753 in acute myeloid leukaemia (LAML), with a median of 4 525 414. There were 40 670 ncRNAs in the expression profiles of each cancer type. We filtered the ncRNAs with expression \>0.01 FPKM, resulting in a median of 12 554 ncRNAs included in the analyses. All SNPs within the $\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$ \pm$\end{document}$1 Mb region of the TSS of each gene were tested in cis-eQTL analyses, while others were used for trans-eQTL analyses. ###### The summary of samples and lncRNA related eQTLs in this study Cis Trans ------- ------ -------- ----------- ----------- -------- ----------- --------- -------- --------- ACC 77 10 673 3 567 953 6906 229 6547 1030 49 934 BLCA 403 12 090 4 191 159 205 824 4077 156 228 29 840 935 23 714 BRCA 1067 13 170 2 745 615 498 969 8124 308 016 62 764 2 328 46 137 CESC 242 12 410 4 276 554 101 968 2626 82 242 20 779 735 17 555 CHOL 36 12 217 4 012 151 0 0 0 0 0 0 COAD 282 11 063 4 505 758 169 256 3558 129 931 25 618 859 20 423 DLBC 47 11 447 4 819 767 122 8 121 82 4 82 ESCA 148 19 921 4 431 385 44 450 1238 37 484 6177 218 4723 GBM 139 15 247 4 525 414 74 593 1743 61 211 6109 202 5518 HNSC 492 11 768 4 249 925 294 234 4698 217 470 38 283 1 075 31 312 KICH 65 11 736 3 755 519 9 289 265 8 075 382 26 320 KIRC 520 14 537 4 578 071 558 380 7216 380 619 57 962 1 492 47 020 KIRP 287 12 554 4 881 400 228 243 4350 175 762 28 242 905 22 640 LAML 96 19 856 5 078 753 44 537 1069 34 233 5570 148 4802 LGG 498 14 213 4 626 469 723 868 7844 465 249 71 517 1 595 54 899 LIHC 367 9691 4 157 271 171 255 3333 128 444 22 425 764 17 725 LUAD 506 13 624 4 384 017 347 537 5714 249 589 39 040 1 130 31 147 LUSC 495 14 319 3 745 439 321 036 5556 226 779 44 235 1 236 38 111 MESO 81 12 393 4 759 523 15 045 372 14 140 2152 97 1779 OV 251 16 660 2 966 217 102 136 3423 79 467 11 534 515 9317 PAAD 177 13 298 4 991 769 139 332 2623 112 129 16 561 494 13 926 PCPG 174 11 971 4 709 166 119 083 2571 93 599 16 630 552 14 066 PRAD 478 12 945 4 822 300 604 359 7181 412 073 69 412 1 686 56 307 READ 91 11 298 4 540 674 18 312 539 16 750 4645 153 4138 SARC 257 11 454 4 087 361 105 751 2607 85 278 17 121 600 14 261 SKCM 103 11 315 4 854 570 14 150 427 13 014 4856 159 4501 STAD 371 19 117 4 300 207 175 519 3637 128 258 21 276 635 15 847 TGCT 148 14 304 4 811 363 95 579 1989 79 971 11 477 325 9839 THCA 495 12 874 4 876 701 702 674 7426 464 482 57 645 1 353 44 077 THYM 119 13 223 4 930 920 92 773 1907 74 540 9355 330 7028 UCEC 173 11 548 4 957 767 44 594 1327 38 605 9933 411 8385 UCS 55 13 439 3 871 537 51 5 51 0 0 0 UVM 77 9182 4 692 767 15 620 405 14 530 3300 128 2993 Total 8817 26 839 7 169 904 6 045 445 98 087 4 294 887 715 952 21 139 573 526 We identified a total of 6 045 445 eQTL-lncRNA pairs in 33 cancer types of cis-eQTL analyses at a per-tissue FDR of \< 0.05, which corresponded to a median P-value = 5.28 × 10^−6^. No cis-eQTL or trans-eQTL were identified in CHOL because of the low sample size of 36. Among other cancer types, the number of cis-eQTLs ranged from 51 in uterine carcinosarcoma (UCS) to 465 249 in lower grade glioma (LGG). GTEx multi-tissue eQTL studies have reported a directly proportional relationship of egenes (genes associated with eQTLs) with sample size, and no plateauing has been reported at a maximum 300 sample size ([@B24],[@B34]). In our study, we also observed that the number of cis-eQTLs was significantly positively correlated with the sample size, even after adjusted numbers of genotype and ncRNAs (P-value = 6.12 × 10^−8^, R~s~ = 0.79, [Supplementary Figure S1A](#sup1){ref-type="supplementary-material"}), and a similar trend was exhibited for egenes (P-value = 8.66 × 10^−10^, R~s~ = 0.85, [Supplementary Figure S1B](#sup1){ref-type="supplementary-material"}). The distribution of cis-eQTLs relative to the transcription start site shows that the majority of eQTLs are approximately symmetrically centred around the TSS ([Supplementary Figure S1C](#sup1){ref-type="supplementary-material"}), with 90% within ±300 kb of the TSS. Overall, 80.2% of lncRNA-eQTLs only regulate the expression of one lncRNA, and 19.8% of lncRNA-eQTLs can regulate the expression of more than two genes ([Supplementary Figure S1D](#sup1){ref-type="supplementary-material"}). Across all cancer types, we identified a total of 715 952 eQTL--lncRNA pairs in trans-eQTL analyses at per-tissue FDR of \<0.05, which corresponded to a median P-value = 3.45 × 10^−13^. The number of trans-eQTLs/egenes was also significantly positively correlated with the sample size (P-value = 5.52 × 10^−9^, R = 0.83 for eQTLs, P-value = 2.38 × 10^−11^, R = 0.88 for egenes, [Supplementary Figure S1E and F](#sup1){ref-type="supplementary-material"}). Data summary of miRNA-related eQTLs {#SEC3-2} ----------------------------------- Using genotype data and miRNA-seq data, we further performed analyses of miRNA-related eQTLs (miRNA-eQTLs) across 33 cancer types. In the cis-eQTL analysis, we identified a total of 87 833 eQTL--miRNA pairs at a per-tissue FDR \< 0.05, and the number of pairs ranged from six in lymphoid neoplasm diffuse large B-cell lymphoma (DLBC) to 11 779 in THCA ([Supplementary Table S2](#sup1){ref-type="supplementary-material"}). There was only one cis-regulated miRNA in UCS and DLBC, while 301 egenes were identified in THCA. For trans-eQTL analysis, a total of 5170 eQTL--miRNA pairs were identified, and the number of trans-eQTLs ranged from two in adrenocortical carcinoma (ACC) to 658 in lung squamous cell carcinoma (LUSC). The number of miRNA-related cis-eQTLs and trans-eQTLs was also significantly correlated with the number of samples (P-value = 1.41 × 10^−7^, R = 0.79 for cis and P-value = 7.75 × 10^−5^, R = 0.69 for trans, respectively). Case study of eQTLs associated with patient survival times and GWAS loci {#SEC3-3} ------------------------------------------------------------------------ To prioritize promising ncRNA-eQTLs, we linked our eQTLs with TCGA patients' clinical data and identified eQTLs that may be associated with overall survival times. We identified a total of 8235 lncRNA-eQTLs and 116 miRNA-eQTLs associated with patient overall survival times across 33 cancer types at FDR \< 0.05. For example, rs10965295 and rs3131997 were significantly associated with patient overall survival times in LGG (Figure [2A](#F2){ref-type="fig"}), and these two SNPs could regulate CDKN2B-AS1 expression and hsa-miR-5699-5p in LGG, respectively (Figure [2B](#F2){ref-type="fig"}). These survival-eQTLs and related ncRNAs may play important roles in cancer development and could serve as predictive and prognostic biomarkers. ![Case study of eQTLs in survival-eQTLs and GWAS-eQTLs. (A) nc-eQTL rs10965295 and rs3131997 affect patient overall survival times in LGG. (B) egene CDKN2B-AS1 of rs10965295 and egene hsa-miR-5699--5p of rs3131997 were significantly differentially expressed among genotypes in LGG. (C) nc-eQTL rs17191861 located in BRCA GWAS locus. (D) egene LINC01235 of rs17191861 was significantly differentially expressed in paired tumour and normal samples.](gkz711fig2){#F2} To explore ncRNA-eQTLs and possible causal genes in known GWAS loci, we identified overlaps between ncRNA-eQTLs and SNPs in GWAS regions. A total of 45 826 tag SNPs were downloaded from the GWAS Catalog; 920 379 SNPs in LD with these tag SNPs were obtained, and all these SNPs were defined as GWAS SNPs. By mapping ncRNA-eQTLs to GWAS SNPs, we identified 253 080 lncRNA-eQTLs and 9252 miRNA-eQTLs overlapping with known disease/traits associated loci. To provide an example of a GWAS-eQTL application, we further mapped BRCA ncRNA-eQTL results to GWAS SNPs of corresponding breast cancer. Among breast GWAS SNPs, we found a total of 1989 eQTLs, which could regulate the expression of 161 ncRNAs. We further analysed the expression of these ncRNAs in tumours and their matched normal samples and found that several ncRNAs were significantly differentially expressed between paired tumour and normal samples at FDR \< 0.05. These ncRNAs could be possible causal targets of BRCA GWAS loci. For example, Michailidou et al. found that rs77457752 on chromosome 9p23 is significantly associated with breast cancer risk (P-value = 8 × 10^−7^), but they did not report its possible target genes ([@B35]). rs77457752 is located in the intron of the lncRNA LINC00583. We retrieved the expression level of LINC00583 and found that it is not expressed in 91% of breast cancer samples. In the upstream and downstream regions of rs77457752, there are only a few genes (Figure [2C](#F2){ref-type="fig"}), and we did not find any correlated protein-coding genes of rs77457752 and its LD SNPs. In our results, we found that rs17191861, which is in LD with rs77457752 (LD r^2^ = 0.73), is significantly correlated with LINC01235. Differential expression analysis shows that the expression of LINC01235 in tumours is significantly lower than that of adjacent normal samples (P-value \< 2 × 10^−16^, fold change = -2.44, Figure [2D](#F2){ref-type="fig"}), suggesting that LINC01235 may be a potential causal gene in this risk locus. Web interface {#SEC3-4} ------------- To facilitate broad access to these ncRNA-eQTLs and associated data, we developed a user-friendly data portal, ncRNA-eQTL (<http://ibi.hzau.edu.cn/ncRNA-eQTL/index.php>) (Figure [3A](#F3){ref-type="fig"}), which includes two sub-databases, lncRNA-eQTL home and miRNA-eQTL home. The two sub-databases can be switched from the button on the top right of our data portal. Each sub-database provides four major datasets: cis-eQTLs, trans-eQTLs, survival-eQTLs and GWAS-eQTLs (Figure [3B](#F3){ref-type="fig"}). On the homepage, we designed a quick search option in which users can input their interested SNPs or genes. Then, four dynamic tables displaying cis-eQTLs, trans-eQTLs, survival-eQTLs and GWAS-eQTLs with related information will be presented (Figure [3C](#F3){ref-type="fig"}). The cis-eQTL and trans-eQTL table display SNP ID, SNP genomic position, SNP alleles, ncRNA, ncRNA position, β-value (effect size of SNP on gene expression), r-value (correlation coefficient between SNP and expression) and P-value of eQTLs. For each record, a vector diagram of a boxplot is provided to display the association between SNP genotypes and gene expression. The survival-eQTL table displays SNP ID, SNP genomic position, SNP alleles, log-rank test P-value and median survival times of different genotypes. For each record, a vector diagram of the Kaplan--Meier plot is embedded to display the association between SNP genotypes and overall survival times. The GWAS-eQTLs table will return the SNP information, gene information and related GWAS traits. ![Web interface of the ncRNA-eQTL database. (A) ncRNA-eQTL database header with navigation bar. (B) Four different searching modules we offer. (C) The results of a quick search. (D) Heat map of cis-eQTLs across 33 cancer types. The colour of each box indicates the values of the correlation coefficient (r). (E) Heat map of trans-eQTLs across 33 cancer types.](gkz711fig3){#F3} We also designed a 'Pancan-eQTL' page, where users can submit a batch of SNPs and/or gene symbols. Then, two interactive heat maps of cis-eQTL (Figure [3D](#F3){ref-type="fig"}) and trans-eQTL (Figure [3E](#F3){ref-type="fig"}) will display all the values of the correlation coefficient (r). Users can download all four datasets for each cancer type from the 'Download' section. The 'Help' page provides information for data collection, processing and result summary. ncRNA-eQTL welcomes any feedback by email to the address provided in the 'Contact' page. We have tested the database on various web browsers, including Chrome (recommended), Firefox, Opera, Windows Edge and Safari of macOS. DISCUSSION {#SEC4} ========== In summary, ncRNA-eQTL is a comprehensive ncRNA (lncRNA and miRNA)-related eQTL resource that uses large cancer samples to evaluate the effects of genetic variants on ncRNA expression. It comprises cis/trans-eQTLs, survival-eQTLs and GWAS-eQTLs and provides a user-friendly interface for users to query, browse and download data of interest. To the best of our knowledge, this is the first database specifically identifying ncRNA-eQTLs in multiple cancer types. In addition, we observed that the number of eQTLs increased with the sample size. The sample sizes of most of the other eQTL studies were below 300 ([@B30],[@B36]), but in our analyses, 12 cancer types had more than 300 samples, indicating that the ncRNA-eQTL database would be the most comprehensive resource for ncRNA-eQTLs. Two important features of our resource are linking ncRNA-eQTLs to patient survival times and known GWAS loci, which will help users narrow down their candidate eQTLs and egenes. Many researchers have been looking for genetic biomarkers related to disease susceptibility, development and prognosis and have committed to analysing the biological mechanisms behind genetic determinations. In our study, we found that thousands of genetic variants could influence cancer prognosis and provided their related non-coding egenes. By integrating ncRNA-eQTL data with known GWAS data, we identified thousands of ncRNA-eQTLs in known GWAS regions. ncRNAs and related genetic determinations are still poorly functionally characterized. These survival and GWAS-related ncRNA-eQTLs and egenes could be important candidates for further experimental validation. Our database will facilitate the fine mapping of post-GWAS and identification of therapeutic biomarkers for cancer. We believe that this valuable resource will be of significant interest to the research community, especially in the field of ncRNA and cancer-related studies. Supplementary Material ====================== ###### Click here for additional data file. We thank all members in Zhang lab for helping in paper writing and website construction. Server support from the Hubei Key Laboratory of Agricultural Bioinformatics is also appreciated. SUPPLEMENTARY DATA {#SEC5} ================== [Supplementary Data](https://academic.oup.com/nar/article-lookup/doi/10.1093/nar/gkz711#supplementary-data) are available at NAR Online. FUNDING {#SEC6} ======= Huazhong Agricultural University Scientific & Technological Self-innovation Foundation \[11041810351 to J.G., W.Z\]. Funding for open access charge: Huazhong Agricultural University Scientific & Technological Self-innovation Foundation \[11041810351\]. Conflict of interest statement. None declared.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION ============ Periosteal osteosarcoma (PEROSA) is an intermediate-grade malignancy arising on the surface of bones, frequently in femur and tibia, followed by ulna and humerus in extremities ([@B1], [@B2]). PEROSA of the chest wall, unlike that of the extremities, presents a diagnostic challenge because of the extreme rarity of the lesion, variable clinical symptoms, and suboptimal visualization on radiography of the characteristic features, resulting in misinterpretation of the lesion as a variety of other diseases ([@B3]). We present with radiologic findings and review of literature, of the two cases of PEROSA arising from the rib and the scapula. CASE REPORTS ============ Case 1 ------ A 17-year-old male presented with a painless swelling on the right lateral chest wall with duration of 3 months. The past medical history was unremarkable. The patient did not recall any traumatic episodes that preceded his discomfort. On physical examination, a hard, non-movable, and non-tender mass (5 × 6 cm in size) was palpated on the area. The hematologic and biochemical studies were normal. The chest radiography showed a local soft tissue bulging with 4-cm in size in the vicinity of the right 7th rib. A poorly defined triangular calcific opacity, like a Codman\'s triangle, was detected on the outer surface of the lateral arc of the right 7th rib. There was no evidence of the rib destruction ([Fig. 1A](#F1){ref-type="fig"}). On the computed tomography (CT) with bone setting, a small juxtacortical hypodense mass was seen on the surface of the rib. The underlying rib showed the periosteal reaction with fine perpendicular spicules extending from the thickened cortex ([Fig. 1B](#F1){ref-type="fig"}). The CT with contrast enhancement showed the fine septal and wall enhancement of the hypodense soft tissue mass with periosteal reaction ([Fig. 1C](#F1){ref-type="fig"}). No definite evidence of cortical destruction or medullary invasion of the rib was seen on either chest radiograph or CT scan. Also, there were no abnormalities in the lung parenchyma. The pre-operative differential diagnoses include the primary bone surface tumor (such as periosteal osteosarcoma or chondrosarcoma), soft tissue tumor with secondary periosteal reaction, and tuberculosis. The patient underwent segmental wide resection of the 7th rib, including the 6th through 8th ribs. The tumor of gross specimen measured 4.5 cm in longest diameter, and it was firmly attached to the outer surface of the 7th rib. Gross specimen in cut-section showed surface tumor of the rib with medullary infiltration ([Fig. 1D](#F1){ref-type="fig"}). The CT examination of the specimen demonstrated a juxtacostal mass with an elevated periosteum, a Codman\'s triangle, and suspicious medullary involvement. Finally, the tumor was proved to be PEROSA microscopically with medullary extension, predominantly consisting of the chondrosarcomatous components and definite areas of the calcified tumor osteoid ([Fig. 1E](#F1){ref-type="fig"}). Over the 5 years of follow-ups, the patient remained free of recurrent or metastatic disease with no adjuvant therapy. Case 2 ------ A 17-year-old female presented with a painful mass in the right chest wall with the duration of a month. The patient did not have any previous trauma, and she accidentally found the mass while taking a shower. The past medical history and laboratory findings were unremarkable. The physical examination revealed a firm, tender mass (2 × 2 cm in size) on the posterosuperior aspect of the right shoulder. There was no limitation in the range of motion or sensory change of the right shoulder. A posteroanterior view on a radiograph of the shoulder showed ill-defined and increased soft tissue density in the superior aspect to the right scapula. On MR imaging, the mass showed intermediate signal intensity on T1-weighted image (WI), compared to that of the chest wall muscle ([Fig. 2A](#F2){ref-type="fig"}), and heterogeneous high signal intensity on T2WI ([Fig. 2B](#F2){ref-type="fig"}). The rim enhancement of the mass was seen on the axial, fat-saturated, contrast-enhanced T1WI ([Fig. 2C](#F2){ref-type="fig"}). On sagittal scan of the contrast-enhanced T1WI, the mass was seen to be located in the supraspinatus muscle of the suprascapular fossa, and it looked to be attached to the superior margin of the underlying scapular spine with some cortical destruction. The mass was also observed to partially extend posteriorly to the subcutaneous fat layer ([Fig. 2D](#F2){ref-type="fig"}). Fluoroscopy in craniocaudal projection, with the patient in the supine position, well-depicted the soft tissue opacity containing calcified matrix ([Fig. 2E](#F2){ref-type="fig"}). The pre-operative differential diagnoses included the primary surface bone tumors, such as periosteal osteosarcoma or conventional chondroblastic osteosarcoma, and the soft tissue tumors with secondary marrow invasion. Incisional biopsy was performed for the mass, and the biopsy specimen demonstrated a malignant tumor with chondrosarcomatous features. One week later, a wide excision of the tumor with partial scapulectomy was performed. At the operation, the mass was mainly found in the supraspinatus muscle, attached to the scapular spine, with some portion extending posteriorly to the subcutaneous layer, which correlated to the preoperative imaging findings. Gross specimen showed a 6 × 4 cm-sized solid mass, with gray to yellowish appearance, and cartilage formation in the surface of the cut. The sagittal CT scan of the specimen also revealed a soft tissue mass with calcifications. The tumor was pathologically confirmed as PEROSA, which is predominantly composed of moderately differentiated chondrosarcomatous components and tumor osteoid. The focal intramedullary involvement was noted ([Fig. 2F](#F2){ref-type="fig"}). The patient received adjuvant chemotherapy for 6 months, and there were no evidence of recurrence or metastasis for the subsequent 5 years. DISCUSSION ========== To our knowledge, PEROSA is extremely rare in flat bones ([@B1]). In the study of Burt et al. ([@B4]) with 38 cases of chest wall osteosarcoma, 13 cases (34%) including the post-radiation osteosarcoma arose in the rib and 12 cases (32%) arose in the scapula, but no PEROSA was found. We searched through medical literature published in English before August 2013, in PubMed website with keywords such as \'periosteal osteosarcoma\', \'rib\', \'clavicle\', \'scapula\', and \'sternum\'. There was only four reported cases of PEROSA involving the chest wall: two in the clavicle ([@B5]); one in the rib ([@B2]); and one in the scapula ([@B6]). Classical radiologic features of PEROSA of the long bones have been well-known. The tumor appears as a soft tissue mass that is broadly attached to the bony cortex, which shows cortical thickening, scalloping, and perpendicular spiculation of the mineralization. It elevates the periosteum to produce Codman\'s triangles and fusiform surface enlargement ([@B7]). However, with simple radiograph, the characteristic features of the PEROSA arising from the rib or scapula may be visualized suboptimally, and may be misinterpreted as a lung cancer, pleural lesion, and other variety of disease ([@B2], [@B3], [@B6], [@B7]). The radiologically differential diagnoses of PEROSA include fracture with callus, periostitis, heterotopic ossification, calcific tendinopathy, other primary bone surface tumors (parosteal osteosarcoma, periosteal chondrosarcoma, high-grade surface osteosarcoma), metastatic carcinoma, and tuberculous infection of the chest wall ([@B2], [@B3]). The post-fracture callus tends to have a more conspicuous and orderly deposition of the osteoid and immature bone ([@B2]). The parosteal osteosarcoma shows a well-defined, lobulated osseous mass attached to the bone by a narrow stalk and the characteristic radiolucent line interposed between the mass and subjacent cortex ([@B7]). Periosteal chondrosarcoma tends to be a round mass with extensive granular opacities and the underlying cortex showing saucer-like depression with thickening and sclerosis ([@B8]). The high-grade surface osteosarcoma usually arises in the diaphysis and they show perpendicular periosteal reaction, which makes the differentiation with PEROSA difficult. However, unlike PEROSA, the high-grade surface osteosarcoma tends to surround the bony circumference more extensively ([@B7]). The tuberculosis of the chest wall has a wide spectrum of radiologic findings of commonly encountered juxtacortical lesion, with or without the bone destruction, which makes it difficult to distinguish from PEROSA, especially in endemic areas. The rarity of the incidences of PEROSA on the flat bone also leads to differential diagnosis. CT may provide valuable information such as assessment of tumor origin, tumor matrix, calcification within soft tissue mass, and localization of the masses. This indicates of the differentiation from more common lesions that may be confused on radiographs ([@B3]). Although our case on the rib did not show characteristics of calcification, the CT is useful for differentiation of chondroid tumor, in which the hypodense chondroid matrix and punctate calcifications are some of the specific findings. The MR imagings may reveal more precisely the margin and extent of the mass, and also may provide information for differential diagnosis. PEROSA shows similar signal intensity to the muscle on T1WI and heterogeneous high signal intensity on T2WI, which reflects the predominancy of the chondroid components within the mass. The peripheral or septal patterns of contrast enhancement also suggests the cartilaginous lesion. The adjacent bone may show focal areas of the signal change due to reactive marrow edema. Less commonly, intramedullary invasion of the tumor may be seen as continuity from the soft tissue mass ([@B7]). These MR findings are helpful for differentiating PEROSA from the high-grade surface osteosarcoma, which does not show high signal intensity on T2WI or peripheral enhancement, due to lack of high-water content and cartilaginous component ([@B7]). However, MR imaging is less sensitive than CT for detection of mineralization in the mass. Medullary involvement is rarely found in the cases of PEROSA, showing direct continuity between the overlying surface tumor and the area of bone marrow infiltration. It is distinguished from the areas of simple marrow signal change, which shows intervening cortex without definite continuity with the overlying surface tumor. Also, cortical bone adjacent to PEROSA usually appears to be thickened and scalloped, in which the bone permeation or destruction is not a common finding of PEROSA ([@B7]). In the study of Hall et al. ([@B1]) with 61 cases of PEROSA, the gross medullary extension was observed in 2 cases, microscopic medullary extension in 1 case, and cortical bone permeation in another case. Murphey et al. ([@B7]) also reported 1 case of medullary invasion, accounting for 2% of their series. Moreover, Sonobe et al. ([@B9]) and Suehara et al. ([@B10]) each reported 1 case of PEROSA with medullary invasion. Therefore, with the unusual and distinctive radiologic and histologic findings of PEROSA, the medullary involvement or cortical bone permeation should not be precluded as the diagnosis ([@B1], [@B7]). In our case of the rib, the medullary cavity did not seemed to be involved according to the pre-operative imaging studies including the CT, but the intramedullary involvement was found on the specimen CT and was pathologically proved. In our case with scapula involvement, the intramedullary involvement was suggested on sagittal scan of fat-saturated and contrast-enhanced T1WI, and it was confirmed by the pathologic examination. Periosteal osteosarcoma is locally aggressive malignant tumor, so that inadequate surgical resection can result in recurrence. Therefore, currently recommended treatment for PEROSA consists of wide segmental resection, with or without neoadjuvant chemotherapy ([@B5]). The metastatic potential of PEROSA is much lower than that of conventional osteosarcoma, with a rate of distant metastasis of 10-20%, and mostly common to the lung ([@B1]). The prognosis of PEROSA is better than that of the conventional intramedullary osteosarcoma and high-grade surface osteosarcoma, but it is worse than that of parosteal osteosarcoma ([@B5]). In summary, we report two rare cases of PEROSA arising from the flat bones of chest wall, the rib, and the scapula. Although diagnosis of PEROSA arising from the rib and scapula is challenging, due to their rarity in location and suboptimal visualization on radiographs, the cross-sectional imaging modalities such as CT and MR are helpful for detection, diagnosis, and differentiation of PEROSA from other chest wall diseases. ![17-year-old male with periosteal osteosarcoma in right 7th rib.\ A. Simple chest radiograph shows poorly defined triangular calcific opacity (Codman\'s triangle, black arrow), with periosteal reactions (white arrows) on outer surface of lateral arc of right 7th rib, without evidence of rib destruction. B. Non-enhanced CT scan with bone setting shows thin-walled, juxtacortical, hypodense mass (arrows) and periosteal reaction with fine, perpendicular spicules extending from thickened, adjacent cortex of rib (arrowheads). C. CT with contrast enhancement demonstrated fine septal and wall enhancement of hypodense mass (black arrowheads) and periosteal reaction with cortical thickening (white arrows). D. Gross specimen in cut-section shows surface tumor of rib with medullary infiltration. E. Microscopic examination shows definite areas of calcified tumor osteoid (arrows) within chondroid matrix (hematoxylin and eosin stain, × 100).](kjr-15-370-g001){#F1} ![17-year-old girl with periosteal osteosarcoma in right scapula.\ A, B. MR imaging shows intermediate signal intensity in mass on T1-weighted image (A, arrows) and heterogeneous high signal intensity on T2-weighted image (B, arrows). C. Axial scan of fat-saturated and contrast-enhanced T1-weighted image shows rim enhancement of mass (arrows). D. On sagittal scan of contrast-enhanced T1-weighted image, mass is mainly seen located in supraspinatus muscle in suprascapular fossa (white arrows), with cortical destruction of underlying scapular spine (black arrowhead), and partially extending posteriorly to subcutaneous fat layer (black arrows). E. Post-MR fluoroscopy on craniocaudal projection, with patient in supine position, well-depicts soft tissue opacity (white arrows) with inner calcified matrix (black arrowhead) in suprascapular fossa. F. Microscopically, cortical bony erosion (arrowheads) and focal intramedullary involvement (arrows) are present (hematoxylin and eosin stain, × 10).](kjr-15-370-g002){#F2}	{ "pile_set_name": "PubMed Central" }
Introduction ============ Deoxynivalenol (DON, vomitoxin) is an economically important mycotoxin affecting cereals. It is produced by multiple fungal pathogens, such as Fusarium graminearum, during infection and disease development, and can act as a potent virulence factor ([@B24]; [@B18]). F. graminearum costs the cereal industry millions in losses yearly, which can be amplified during years of high disease severity ([@B19]), not only due to yield losses but also to grain contamination with mycotoxins, mainly DON. DON contamination is widespread throughout the globe ([@B26]) and can cause health issues for both humans and animals. When ingested by animals, DON leads to emesis and a loss of appetite ([@B7]; [@B25]), while long term effects include reduced immunity, increased sensitivity to disease and a lack of weight gain ([@B2]; [@B21]). When grains contaminated with high levels of DON are processed, the resulting animal feed can have higher than acceptable limits; this results in discounts being applied to contaminated grains. Reducing the amount of DON accumulation in grains is of high importance to industry and many strategies to mitigate the problem have been explored ([@B29]). There are currently no effective chemical processes for effectively and economically detoxifying or eliminating DON during grain processing. DON has been shown to be detoxified by several bacterial species via specific modifications ([@B16]). The de-epoxidation of DON has been demonstrated by several species and occurs under anaerobic conditions ([@B4]; [@B8]; [@B27]). The C3 carbon is a common point of modification. Plants often glycosylate DON at the C3 position, but the resulting masked mycotoxin can be hydrolyzed back to DON via multiple mechanisms ([@B3]). The toxicity of DON can also be reduced by the acetylation of the C3 carbon ([@B17]), but again the attached acetyl group can be hydrolyzed during digestion to reform DON. The oxidation and subsequent reduction of DON to produce 3-keto-DON and 3-epi-DON, respectively, has also shown to be a promising detoxification method. Both of these compounds have reduced toxicity compared to DON, especially 3-epi-DON ([@B10]; [@B23]). Devosia mutans 17-2-E-8 (Devosia sp. 17-2-E-8) has been shown to transform DON to 3-epi-DON through a two-step process ([@B11]; [@B9]). Recently, the first enzyme in [D]{.ul}ON [ep]{.ul}imerization (Dep) pathway, DepA, was identified as a pyrroloquinoline quinone (PQQ)-dependent dehydrogenase ([@B5]). This enzyme readily converts DON to 3-keto-DON and can be expressed heterologously in E. coli. Here we have identified DepB from D. mutans 17-2-E-8, the second enzyme in the DON detoxification pathway. The ability to convert DON to 3-epi-DON (Figure [1](#F1){ref-type="fig"}), a compound with at least 50-fold less toxicity, would be a large step forward in producing high quality feed even during large scale fusarium disease epidemics. Regular use of a treatment to detoxify DON would also help alleviate some of the long term issues/decreased productivity faced by livestock exposed even to low levels of DON. These enzymes could be used directly as a feed additive, incorporated into industrial feed production (either as pure enzymes or incorporated into microorganisms), or inserted into plants in an attempt to detoxify DON before harvesting. ![The DON epimerization pathway. DON is oxidized to 3-keto-DON by DepA followed by a reduction to 3-epi-DON by DepB.](fmicb-09-01573-g001){#F1} Experimental Procedures ======================= Chemicals --------- All chemicals were from Fisher Scientific (Nepean, ON, Canada) unless otherwise stated. DON was from Sigma--Aldrich (Oakville, ON, Canada) or TripleBond (Guelph, ON, Canada) and 3-keto-DON was from TripleBond. 3-epi-DON was previously purified and its identity confirmed ([@B12]). Partial Purification of Potential DON Transforming Enzymes ---------------------------------------------------------- A culture of D. mutans 17-2-E-8 was grown for 2 weeks at 28°C in LB medium supplemented with 34 μg ml^-1^ of kanamycin while shaking at 200 RPM. This starter culture was used to inoculate 4 L of LB medium which was incubated for 6 days at 28°C while shaking at 150 RPM. The culture was then centrifuged at 8,000 × g for 30 min to pellet the cells. The cell pellets were then frozen at -20°C. Frozen pellets were thawed and cells resuspended in 50 mM Tris, pH 8.0 with 150 mM NaCl. The cells were then lysed by sonication using a QSONICA Q500 sonicator (Qsonica LLC., Newtown, CT, United States) while on ice. Sonication parameters were as follows: 30% amplitude, 30 s on and 50 s off for 20 cycles. The cells were allowed to cool for 10 min on ice before being subjected to another 20 cycles of sonication. The lysed cells were then centrifuged at 7,142 × g for 30 min to remove the insoluble fraction. The supernatant was filtered through a 0.45 μm syringe filter and Halt Protease Inhibitor Cocktail (Thermo Fisher, Mississauga, ON, Canada) without EDTA, was added. The active fraction of the lysate was then salted out using ammonium sulfate precipitation ([@B6]). Briefly, the ammonium sulfate concentration was brought to 30% of saturation and insoluble proteins were removed by centrifugation at 7,142 × g for 20 min. The ammonium sulfate concentration was brought to 50% of saturation and the insoluble fraction was collected by centrifugation at 7,142 × g for 20 min. This was repeated for 70% saturation. Each of the above fractions was subjected to dialysis using Slide-A-Lyzer mini dialysis device 3.5K MWCO (Thermo Fisher), to remove the excess salt, and tested for the ability to convert 3-keto-DON to 3-epi-DON (DepB activity). The active fraction was then subjected to a heat treatment at 60°C for 15 min. The insoluble protein was removed by centrifugation at 7,142 × g. During this initial purification, enzymatic activity was assayed by adding 20 μL of protein solution to a solution with a final concentration of 100 μg ml^-1^ 3-keto-DON in Tris pH 8.0. The amount of 3-epi-DON produced was measured after approximately an hour and/or overnight. The dialyzed active fraction obtained after ammonium sulfate precipitation and heat treatment, was subjected to anion exchange column chromatography using a HiTrap Q FF column (GE Healthcare) attached to AKTAPrime plus and monitored at 280 nm. The sample was loaded onto the column which had been equilibrated with 20 mM Tris (Buffer A). The column was then rinsed with 8 ml of Buffer A and proteins were eluted by a linear gradient from 0% Buffer B (20 mM Tris, 1 M NaCl) to 100% over 40 ml. Fractions containing the highest activity were pooled, concentrated by ultrafiltration using an Amicon Stirred Cell with an Ultracel 10 kDa ultrafiltration Disk (Millipore). The sample was then loaded onto a HiPrep 16/60 Sephacryl S-200 HR size exclusion column and eluted over 140 min at 1 ml per minute with 20 mM Tris and 300 ml NaCl. Fractions were tested for DepB activity and the most active fractions were pooled and concentrated by ultrafiltration as above. Protein Sequencing ------------------ The concentrated pooled sample of the most active fractions was sent for protein sequencing to the Advanced Analysis Centre at the University of Guelph (Guelph, ON, Canada). The results can be found in Supplementary Table [S1](#SM5){ref-type="supplementary-material"}. Gene Synthesis and DNA Manipulation ----------------------------------- Six potential candidates responsible for the transformation of 3-keto DON to 3-epi-DON activity were selected from Supplementary Table [S1](#SM5){ref-type="supplementary-material"}, their subcellular localization was predicted using PSORTb 3.0 ([@B28]) and SignalP 4.0 ([@B22]); all proteins were predicted to be cytosolic. The genes were then codon optimized for synthesis in E. coli, synthesized and cloned into pET28A by Genscript (Piscataway, NJ, United States). The genes were inserted to pET28a using NdeI and BamHI restriction sites to produce proteins with N-terminal polyhistidine tags. Their protein sequences can be found in Supplementary Table [S2](#SM6){ref-type="supplementary-material"}. Each construct was transformed separately into E. coli BL21; a single colony was selected for protein expression. Expression and Purification of Recombinant Proteins --------------------------------------------------- E. coli BL21 containing each synthesized construct (see above) was propagated at 37°C in 500 ml of LB medium supplemented with 34 mg l^-1^ kanamycin while shaking at 175 rpm. When the cultures reached an optical density of approximately 0.7, 150 μM of isopropyl β-d-1-thiogalactopyranoside (IPTG) was added to initiate protein expression. The cultures were incubated overnight at 18°C and harvested by centrifugation at 8,000 × g for 10 min and frozen at -20°C. Each pellet was thawed and resuspended in 5 ml of 50 mM Tris, pH 8.0, 150 mM NaCl. The cells were then lysed by sonication using the following parameters: 30% amplitude with a cycle of 30 s on, 40 s off for 8 cycles. Lysates were centrifuged at 7,142 × g for 30 min to remove the insoluble pellet and filtered through a 0.45 μm filter. The clarified lysates (approximately 5 mL) were then mixed with an equal volume of 50 mM sodium phosphate with 300 mM NaCl and 20 mM imidazole, pH 8.0 (wash buffer) and incubated with 300 μL of HisPure Ni-NTA resin (Thermo Fisher) for 1 h. Each mixture was then briefly centrifuged and the supernatant was removed from the tubes. The beads were resuspended in 1 ml of wash buffer and transferred to 1.5 ml tubes. The tubes were briefly spun, supernatant removed and resuspended with 1 ml of wash buffer. This was repeated for a total of 5 washes. The proteins which bound to the beads were eluted with 1 ml of 50 mM sodium phosphate with 300 mM NaCl and 250 mM imidazole (elution buffer), pH 8.0 and the beads removed by centrifugation. Each of the six samples was then tested for DepB activity. Testing the Candidates for Activity ----------------------------------- Twenty microliter of purified protein from each construct was added to a reaction containing 400μM NADPH and 100μg ml^-1^ 3-keto-DON in 50 mM Tris, pH 8.0. The reduction in absorbance at 340 nm was monitored using an Ultrospec 3100 pro UV/Vis Spectrophotometer (GE Healthcare/Amersham Biosciences). Activity was later confirmed by HPLC analysis as described below. HPLC Analysis ------------- Samples and standards (20μL) were separated by a Proteo Phenomenex column (Jupiter, 4 μm, 250 mm × 4.6 mm) attached to a Shimadzu uHPLC system (Mandel Scientific Company, Guelph, ON, Canada). The mobile phases used were: solvent A (100% water) and solvent B (100% acetonitrile). To separate and detect the level of DON and 3-epi-DON in a sample the following elution was used: An 11 min isocratic elution consisting of 12% solvent B was used to detect DON and 3-epi-DON followed by a column wash of 80% solvent B for 3 min and a re-equilibration of the column at 12% solvent B for 6 min. 3-keto-DON was eluted using the following settings: A 15 min isocratic elution consisting of 23% solvent B was used to detect 3-keto-DON followed by a column wash of 90% solvent B for 3 min and a re-equilibration of the column at 23% solvent B for 6 min. DON, 3-keto-DON and 3-epi-DON were detected by monitoring the absorbance at 218 nm and were quantified by comparison to a known standard. Purification of DepB -------------------- Once a specific protein was identified as DepB a different method for purification was used. The expression was the same except two, 1 L cultures were grown at a time and the pellets pooled. Cells were lysed under the same conditions except 15 sonication cycles were used. The clarified lysate was mixed with an equal volume of wash buffer and incubated with one ml of Ni-NTA agarose for an hour at 4°C. The mixture was then poured into a Flex-Colum (Kimble Chase Life Science, Vineland, NJ, United States) and washed with 40 ml of wash buffer. DepB was eluted with eight ml of elution buffer. The buffer was changed by repeated dilutions in the stirred cell to 50 mM Tris, 150 mM NaCl, pH 8.0. The purified enzyme was then aliquoted stored at -80°C. Buffer, pH, and Cofactor Specificity ------------------------------------ Reactions assessing the buffer specificity of DepB contained 50 mM of various buffers, 100μg ml^-1^ 3-keto-DON, 400μM NADPH, and 7μg DepB. Each reaction was allowed to proceed for 10 min before it was stopped with acidified methanol. The activity of DepB at various pH values was examined. Reactions contained 50 mM of three component buffer (0.1 M Tris, 0.05 M acetic acid, and 0.05 M 2-(N-morpholino)ethanesulfonic acid) at various pH's, 100μg ml^-1^ 3-keto-DON, 400μM NADPH, 4.7μg DepB and the reaction was allowed to proceed for 15 min before it was stopped with acidified methanol. Cofactor specificity was tested under the following conditions: reactions contain 50 mM of Tris pH 7.5, 100μg ml^-1^ 3-keto-DON or DON, 400μM NADPH or NADH, NADP or NAD, 4.7μg DepB and the reaction was allowed to proceed for 15 min or overnight before it was stopped with acidified methanol. Reactions were carried out in triplicate and the amount of 3-epi-DON produced was assessed by HPLC. Thermostability and the Effect of Lyophilization on DepB -------------------------------------------------------- The stability of DepB (purified from E. coli) and the ability of the enzyme to catalyze the reaction was tested at multiple temperatures. Aliquots of DepB were incubated at a specific temperature between 30 and 70°C for 1 h, and cooled on ice. Heat treated DepB and untreated DepB (kept at 4C) were tested for activity at room temperature (23 ± 2°C). The activity of DepB was monitored from 20 to 60°C in 5°C intervals. Each reaction contained 100μg ml^-1^ 3-keto-DON, 400μM NADPH, 5μg DepB in 50 mM of Tris pH 7.5. The reactions were carried out in triplicate and were allowed to proceed for 7.5 and 15 min before they were stopped by addition of acidified acetonitrile. The amount of 3-epi-DON produced per minute was determined by HPLC. Separately, frozen aliquots of DepB were lyophilized, stored at room temperature for 1 week before resuspension in water. Samples which were lyophilized were assessed for activity at 25°C and were compared to those which had been freshly removed from storage at -80°C and thawed on ice. Each reaction contained 100μg ml^-1^ 3-keto-DON, 400μM NADPH, 4.7μg DepB in 50 mM of Tris pH 7.5. The reactions were allowed to proceed for 12 min before they were stopped by addition of acidified methanol. Reactions were carried out in triplicate and the amount of 3-epi-DON produced was assessed by HPLC. Comparative Genomic Analysis ---------------------------- The amino acid sequences of all predicted CDSs from 19 strains of Devosia (Devosia chinhatensis GCA_000969445.1; Devosia epidermidihirudinis GCA_000971295.1; Devosia geojensis GCA_000969415.1; Devosia insulae DS-56 GCA_000970465.2; Devosia limi DSM 17137 GCA_000970435.1; Devosia psychrophila GCA_000971275.1; Devosia riboflavina GCA_000743575.1; Devosia soli GCA_000970455.1; Devosia sp. A16 GCA_001402915.1; Devosia sp. DBB001 GCA_000689495.1; Devosia sp. H5989 GCA_001185205.1; Devosia sp. LC5 GCA_000735585.1; Devosia sp. Leaf420 GCA_001425445.1; Devosia sp. Leaf64 GCA_001421945.1; Devosia sp. Root105 GCA_001424865.1; Devosia sp. Root413D1 GCA_001425235.1; Devosia sp. Root436 GCA_001426345.1; Devosia sp. Root635 GCA_001427605.1; Devosia sp. Root685 GCA_001427875.1) were aligned against the D. mutans (Devosia sp. 17-2-E-8 GCA_000743515.1) genome sequence with tblastn, using an e-value cut-off of 1e-5. The best-scoring alignments from each CDS were visualized with Circos v.0.69. Statistical Analysis -------------------- Statistical Analyses were performed using GraphPad Prism 7.02 ([@B20]). All reactions were run in triplicates and where appropriate Tukey's multiple comparison tests were used to determine if there was a significant difference between treatments. Results ======= Partial Purification of DepB From D. mutans 17-2-E-8 Crude Lysate ------------------------------------------------------------------- After ammonium sulfate precipitation at 30, 50, and 70% of saturation, the fraction containing proteins which salted out between 30 and 50% of saturating ammonium sulfate had the highest level of DepB activity. After dialysis, proteins which did not precipitate during heat treatment were subjected to anion exchange chromatography (Supplementary Figure [S1](#SM1){ref-type="supplementary-material"}). The most active fractions were pooled, concentrated and DepB was further purified by size exclusion chromatography (Supplementary Figure [S2](#SM2){ref-type="supplementary-material"}). The DepB activity of fractions roughly correlated with the intensity of a band at 38 kDa. The most active fractions were pooled, concentrated and used for protein sequencing. Identification of DepB ---------------------- The list of proteins in the sample was quite extensive (Supplementary Table [S1](#SM5){ref-type="supplementary-material"}), with 101 proteins identified based on sequence identity of peptides with the draft D. mutans genome (Genbank JQGB00000000). From this list, six candidates were selected based on their predicted function and size (Supplementary Table [S2](#SM6){ref-type="supplementary-material"}). The codon optimized candidate genes were cloned into pET28a and expressed in E. coli BL21. After an initial Ni-NTA purification each candidate was tested for DepB activity by measuring the reduction of 3-keto-DON. One gene, fig\\| 6666666.163324.CDS.25 (GenBank: KFL28068.1), hereafter termed depB, was shown to encode a functional protein, DepB. The activity of DepB was confirmed by HPLC, whereby after the addition of DepB to solution 3-keto-DON disappeared from solution while 3-epi-DON appeared. The identity of 3-epi-DON was confirmed by LCMS/MS. DepB was then re-purified (Supplementary Figure [S3](#SM3){ref-type="supplementary-material"}) and aliquots were stored at -80°C. Genomic Location of depB ------------------------ The predicted DepB CDS is located on an ∼127 kb contig that is distinct from the ∼128 kb contig on which depA is located (Figure [2A](#F2){ref-type="fig"}). These two proteins are therefore not part of the same operon, and are in fact quite distant from one another on the D. mutans genome. BLAST comparisons of the depB region with 19 other Devosia strains demonstrates that the CDSs in this region are mostly conserved among different strains (Figure [7A](#F7){ref-type="fig"}), although synteny is not (data not shown). In D. mutans, depB is flanked by a predicted alpha-glucosidase/alpha-galactosidase and a SDR family NAD(P)-dependent oxidoreductase; however, both flanking genes are in the opposite orientation of depB, indicating that they are not co-transcribed with depB (Figure [2B](#F2){ref-type="fig"}). ![Genomic location of D. mutans depB region. (A) D. mutans genome atlas showing protein sequence similarities with 19 other Devosia strains. The inner three rings represent the forward CDS (green), reverse CDS (red) and contigs (gray). The remaining rings (red) represent the tblastn hits of the predicted proteins from 19 Devosia strains against D. mutans (inner to outer): D. mutans, Devosia sp. DBB001, Devosia sp. A16, D. chinhatensis, D. epidermidihirudinis, D. geojensis, Devosia sp. H5989, D. insulae, Devosia sp. LC5, Devosia sp. Leaf420, Devosia sp. Leaf64, D. limi, D. psychrophila, D. riboflavina, Devosia sp. Root105, Devosia sp. Root413D1, Devosia sp. Root436, Devosia sp. Root635, Devosia sp. Root685, D. soli. The region surrounding depB is magnified (outlined in black) and depB highlighted in blue. (B) Gene diagram of depB (red) and surrounding genes (blue, predicted function; black, hypothetical protein).](fmicb-09-01573-g002){#F2} Preliminary Biochemical Characterization ---------------------------------------- The activity of DepB was tested with multiple common buffers. There was no significant difference in the relative activity of DepB between any of the buffers except for ammonium bicarbonate, which was lower than the HEPES buffer (Supplementary Figure [S4](#SM4){ref-type="supplementary-material"}). Tris was ultimately utilized as the buffer in which future experiments were performed. The activity of DepB at pH values from 5.0 to 9.0 was assessed (Figure [3](#F3){ref-type="fig"}). The highest activity of DepB was at a pH of 7.5 but it was not significantly different from activity at near neutral pH. This shows that DepB has a relatively broad pH range in which it can function to transform 3-keto-DON to 3-epi-DON. ![The pH dependency of DepB. Reactions contained 50 mM of three component buffer at various pH-values, 100μg ml^-1^ 3-keto-DON, 400μM NADPH, 4.7μg DepB, the reaction was stopped after 15 min.](fmicb-09-01573-g003){#F3} It was previously shown that the reduction of 3-keto-DON to 3-epi-DON was enhanced by the presence of NADPH ([@B9]), so the ability of NADH to be utilized as a cofactor was also tested. After a 15 min incubation with 3-keto-DON, only reactions containing NADPH produced substantial amounts of 3-epi-DON, although a small amount of 3-epi-DON was also detected after the NADH reaction was left overnight (approximately 16 h) (Figure [4](#F4){ref-type="fig"}). This demonstrates that DepB is indeed NADPH dependent. The ability of DepB to convert DON to 3-keto-DON was tested using NAD^+^ and NADP^+^ as cofactors, but no trace amount of 3-keto-DON could be detected even after the reactions were incubated overnight (data not shown). ![DepB activity using different nicotinamide cofactors. Reactions contain 50 mM of Tris pH 7.5, 100μg ml^-1^ 3-keto-DON, 400μM NADPH or NADH, 4.7μg DepB and the reaction was allowed to proceed for 15 min or overnight before it was stopped with acidified methanol.](fmicb-09-01573-g004){#F4} DepB remained stable when incubated at temperatures from 30 to 55°C (Figure [5](#F5){ref-type="fig"}). The activity was reduced after an incubation at 60°C and abolished after incubations of 65°C and higher. DepB catalyzed the reaction the fastest at 30 and 35°C while had no detectable activity at 60°C (Figure [6](#F6){ref-type="fig"}). As this enzyme may have commercial uses, its activity after being lyophilized and resuspended in water was assessed. There was no significant change in the activity of DepB after lyophilization (Figure [7](#F7){ref-type="fig"}). ![The effect of heat treatment on the activity of DepB. Each reaction contained 100μg ml^-1^ 3-keto-DON, 400μM NADPH, 5μg DepB in 50 mM of Tris pH 7.5. Treated samples were subjected to a 1 h incubation at temperature then cooled on ice. The activity of the enzyme after treatment was plotted relative to the activity to the treatment at 35°C. a, not significantly different from 35°C. b, not significantly different from 60°C. c, not significantly different from 65°C.](fmicb-09-01573-g005){#F5} ![The effect of temperature on the activity of DepB. Each reaction contained 100μg ml^-1^ 3-keto-DON, 400μM NADPH, 5μg DepB in 50 mM of Tris pH 7.5. The activity of the enzyme at each temperature was plotted relative to the activity at 30°C.](fmicb-09-01573-g006){#F6} ![The effect of lyophilization on the activity of DepB. Each reaction contained 100μg ml^-1^ 3-keto-DON, 400μM NADPH, 4.7μg DepB in 50 mM of Tris pH 7.5 and was stopped after 12 min. Activity is relative to DepB which was not lyophilized. Neg Ctl had no enzyme added to the solution.](fmicb-09-01573-g007){#F7} Discussion ========== Here we have identified the second enzyme in the Dep pathway, DepB, responsible for the reduction of 3-keto-DON to 3-epi-DON. DepB is the first enzyme identified that is capable of performing this important detoxification function and its identification may be a key component of developing a system to mitigate the impact of DON in grains. The identification of DepB from a partially purified lysate from its native host shows that it is unnecessary to completely purify enzymes in order to identify an enzyme of interest. A combination of ammonium sulfate precipitation, mild heat treatment, followed by two column chromatography steps did not drastically reduce the number of proteins in the sample but increased the relative abundance of the protein of interest to a degree that made the identification of the enzyme possible. In essence, this work demonstrates that a partial purification combined with biochemical knowledge about the enzyme(s) involved (such as cofactor utilization or molecular size) is sufficient to generate a short list of candidate genes. Having these genes synthesized and codon optimized can aid in rapid screening of the candidate genes and identification of the gene of interest. The gene identified as depB in the presented work is annotated by the Rapid Annotation using Subsystem Technology (RAST) server as a putative oxidoreductase and in the National Center for Biotechnology Information (NCBI) as an aldo/keto reductase or dehydratase (as were all of its closest BLAST hits). DepB is a member of the AKR family, containing the signature TIM barrel motif and the predicted catalytic tetrad (D48, Y53, K81, and H122). Although depB and its surrounding gene sequences are largely conserved among different Devosia strains, in most cases little to no synteny exists in this region. Together with the fact that depB is in the opposite orientation to its surrounding genes in D. mutans, it is difficult to infer its original function based on genomic context. Notably, depB is located distantly to depA in the D. mutans genome, despite the two proteins participating in a common biochemical pathway. This suggests that the two genes were not initially acquired as a single functional unit with DON transforming activity (e.g., via horizontal gene transfer), but rather had different initial functions that eventually co-evolved to metabolize DON. It is also interesting to note that depB was annotated as "putative" by the RAST server and has general annotation by NCBI ([@B1]). Genes with a less specific annotation may be more suitable candidates when investigating novel functions as those with more specific annotations may be quite similar to known proteins, thus lacking novel functions. Although this may not always be true, it may prove helpful when only a limited number of candidates can be selected. As DON is one of the most economically important mycotoxins worldwide, identifying detoxification mechanisms has potentially major implications for the agriculture industry ([@B14]). Although many microorganisms which can detoxify DON have been known for some time, due to the demanding conditions of their growth and more importantly to issues related to their safe usage in feed and food matrixes, they are not currently used in practical applications under industrial settings. Significant efforts have been put into identifying enzymes capable of detoxifying DON, but until recently, have been largely unsuccessful ([@B16]; [@B14]). Now that the enzymes responsible for Dep have been identified, purified enzymes could be used to detoxify DON or, the enzymes could be incorporated into microorganisms (such as yeasts or lactic acid bacteria) which are currently part of production processes. As DepB may be useful in the future as part of a system to detoxify DON, its ability to work in various buffers and at a range of pH was tested. DepB effectively transformed 3-keto-DON to 3-epi-DON in all the buffers tested. This is important as the enzyme may be utilized in very different matrices and may need to be utilized in the presence of other enzymes requiring specific conditions. DepB also catalyzes the reaction reasonably well over a broad pH range (5--9) which is important as certain uses might have a lower pH (cereal milling) while others are closer to neutral pH (alcohol fermentation). The ability of DON detoxifying enzymes to work under multiple processing conditions will be advantageous (and might be necessary in some cases) for the development of an effective detoxification system. DepB is clearly a NADPH dependent enzyme. Although catalysis is possible with NADH, it is approximately 5,000-fold less efficient. Previously, an enzyme from Sphingomonas sp. strain S3-4, AKR18A1 (GenBank: MF314460.1), with moderately high sequence similarity (42%) to DepB was shown to convert DON to 3-keto-DON, using NADP as a cofactor ([@B13]). DepB was unable to convert DON to 3-keto-DON even after prolonged incubation. This enzyme was identified in Sphingomonas sp. strain S3-4, a strain capable of transforming DON to 3-keto-DON and 3-epi-DON in sequence. While unable to complete this reaction, DepB is able to complete the second step of the pathway, transforming 3-keto-DON to 3-epi-DON. It is interesting that in the DON detoxification pathway from D. mutans 17-2-E-8 the first step is completed by a PQQ-dependent dehydrogenase while the second set is performed by a NADPH-dependent aldo-keto reductase ([@B5]) whereas in Sphingomonas sp. strain S3-4, the first step is performed by an aldo-keto reductase that shares the sequence homology of the second step enzymes in the 17-2-E-8 strain ([@B13]). Considering both active sites bind the same structure with the exception of the 3′ group, small changes in the active site might allow for the altered stereospecificity. Comparing the involved active sites and possibly identifying the responsible residues that dictate outcomes of such enzyme/substrate interactions merits further investigation. A 1 h heat treatment at 55°C (or lower) did not significantly affect the activity of DepB, a treatment at 60°C did reduce activity while treatments at 65°C or higher abolished activity. Stability at moderately high temperatures may be critical if DepB is to be used in certain industries (such as corn milling) as several processing steps occur at elevated temperatures. Although DepB is stable at 55°C, it had drastically reduced activity at this temperature compared to room temperature. DepB has maximal activity between 30 and 35°C, while retaining robust activity between 20 and 40°C. DepB catalyzes the conversion of 3-keto-DON to 3-epi-DON slowly at temperature at or above 50°C, however, it could retain activity until the temperature becomes lower. If this enzyme is to be incorporated into a system to detoxify DON in corn milling, stability at 55°C could be a significant advantage over less stable enzymes. A promising property of DepB is that it was able to be lyophilized, and resuspended in water without a loss of activity. This could be quite important and a valued characteristic if a product were to be developed as it would enable the product to be more easily shipped and stored by the end users, especially if those users were to be farmers treating feed. The identification of DepB, the second enzyme in the Dep pathway provides an opportunity for developing a system to detoxify DON in contaminated grains. Currently many DON mitigation strategies focus on pre-harvest prevention of Fusarium sp. infection or prevention of fungal growth post-harvest ([@B15]). The Dep pathway may lead to a viable route to detoxify DON, thus reducing/eliminating the negative aspects of DON toward livestock. There are several possibilities for both the delivery method and applications of the system. Purified enzymes could be utilized during the corn milling process, during ethanol fermentation process or direct to animal feed. Alternatively, microorganisms expressing the system could be incorporated into pre-existing systems. In addition, if the system was expressed in wheat, it may be able to alleviate the toxicity of DON to the plant, thus reducing losses due to Fusarium. Although both enzymes in the Dep pathway have now been identified, significant work is still needed in order to develop these enzymes into a system capable of having a meaningful impact to industry. Author Contributions ==================== TZ, YH, and JC, conceived the project. JC and YH designed and carried out the experiments. DL provided the genetic analysis. JC prepared the manuscript and JC, YH, DL, and TZ contributed to the final version of the paper. All authors read and approved the final manuscript. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Funding. The authors are listed as inventors on a patent application involving DepB as identified in this paper (U.S. Provisional Application No. 62/555,996) but have no financial interest as ownership is held by the Government of Canada. The authors would like to thank Dyanne Brewer and Armen Charchoglyan from the University of Guelph Advanced Analysis Centre and Honghui Zhu and Xiu-Zhen Li from the Guelph Research and Development Center for technical assistance. The financial support from Agriculture and Agri-Food Canada for this research is much appreciated (Project ID J-001498). Supplementary Material ====================== The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fmicb.2018.01573/full#supplementary-material> ###### (A) Elution profile of the protein solution separated by anion exchange. Fractions with the highest activity are marked (^∗^) and were pooled for further purification. (B) SDS--PAGE of fractions 8-13. Lane 1, protein standards, Lanes 2-3, Fractions from DepA purification, Lane 4, Flow Through, Lanes 5-10, Fractions 8-13; respectively. ###### Click here for additional data file. ###### (A) Elution profile of the protein solution separated by size exclusion. Most active fractions indicated with a (^∗^) (B) SDS--PAGE of fractions 10-18. Lane 1, protein standards, Lane 2-9 Fractions 10-18; respectively. ###### Click here for additional data file. ###### Purified DepB from E. coli. Lane 1, protein standards. Lane 2, crude lysate. Lane 3 flow through. Lane 4, final wash. Lane 5, purified DepB. ###### Click here for additional data file. ###### The relative activity of DepB with various buffers. Each reaction contained 50 mM of the respective buffer, 100 μg ml^-1^ 3-keto-DON, 400 μM NADPH, 7 μg DepB and the reaction was allowed to proceed for 10 min before it was stopped with acidified methanol. Activity is relative to the activity in HEPES. ^∗^Activity in ammonium bicarbonate was significantly lower than in HEPES as determined by a Tukey's multiple comparison test. ###### Click here for additional data file. ###### Protein sequencing results. ###### Click here for additional data file. ###### Candidate genes for having DepB activity. ###### Click here for additional data file. [^1]: Edited by: Michael Benedik, Texas A&M University, United States [^2]: Reviewed by: Santosh Kr. Karn, Sardar Bhagwan Singh Post Graduate Institute of Biomedical Science and Research, India; Lígia O. Martins, Instituto de Tecnologia Química e Biológica (ITQB-NOVA), Portugal [^3]: This article was submitted to Microbiotechnology, Ecotoxicology and Bioremediation, a section of the journal Frontiers in Microbiology	{ "pile_set_name": "PubMed Central" }
Key PointsAcute and chronic pain are highly prevalent in children and young people presenting to health care yet are often poorly assessed and managed.Worldwide, it is clearly evident that there is a significant need for improvements in professional education of paediatric pain across disciplines.The provision of online education is one feasible short-term solution to the lack of prelicensure pain education in health care professionals.Further development of paediatric pain education is essential at every level along with research to determine whether such initiatives improve professional knowledge, attitudes, and clinical outcomes for patients. 1. The impact of pain in children {#s1} ================================= Both acute and chronic pain are highly prevalent in children presenting to health care, and are known to have important and detrimental impacts on recovery from illness and injury. It is also well known that poorly managed pain causes unnecessary suffering, adversely impacts quality of life, and carries a significant economic cost to society.^[@R29],[@R43]^ Moderate to severe pain in hospitalised children is nevertheless common, yet is often poorly assessed and managed. Research indicates severe acute pain is prevalent in approximately 62% of hospitalised adolescent patients. Experiences of severe acute pain during hospitalisation are primarily procedure-related and often present intermittently in concurrence with each procedure.^[@R8],[@R38]^ The management of paediatric chronic pain also presents an increasingly recognised problem. The prevalence of paediatric chronic pain varies substantially depending on the location of the pain. For example, prevalence of chronic headache ranges between 8% and 18%, whereas musculoskeletal pain prevalence ranges from 4% to 40%. Pain prevalence increases with age, and for most conditions, it is notably higher in girls than boys.^[@R26]^ Chronic paediatric pain can negatively impact emotional functioning and significantly reduce school ability.^[@R16]^ The cost of treating chronic pain in adolescents in the United Kingdom has been estimated at pound 3.8 billion per annum^[@R37]^ and may be as high as \$19.5 billion per annum across the United States.^[@R17]^ The per-person cost of treating patients with chronic pain in Canada has also been evaluated as 50% higher comparatively to the cost of treating patients without pain.^[@R20]^ Pain in paediatric cancer provides an example of where the boundary between acute and chronic pain is blurred. A systematic review of pain in adolescents with leukaemia or brain tumour found that pain may be procedure-related, treatment-related, or associated with the cancer itself, and is likely to persist chronically after treatment completion.^[@R32]^ Untreated or undertreated pain in oncology patients is common along with reliance on primarily pharmacological management, although holistic management using interdisciplinary care, multimodal therapies, and family-centred treatment is necessary. 2. Pain education of health professionals is limited {#s2} ==================================================== Despite the known impacts of paediatric pain, and that the International Association for the Study of Pain (IASP) offers 8 discipline-specific curricula plus one interprofessional curriculum (<http://www.iasp-pain.org/Education/CurriculaList.aspx?navItemNumber=647>), basic pain education for health care professionals is known to be inadequate.^[@R28]^ Furthermore, a major obstacle to adequate pain relief is the small number of clinicians who are knowledgeable about pain. Although there have been some recent improvements (see below), there are still limited high-quality, validated pain education resources available for health care professionals who regularly work with children and who may be called upon to manage pain.^[@R28]^ Physicians, nurses, psychologists, physiotherapists, pharmacists, child life specialists (play specialists), and other allied health professionals all have important roles in providing good pain management for children. A biopsychosocial management model involving multiple disciplines is widely acknowledged as the most appropriate as physiological, psychological, and social factors contribute to the pain experience, and hence, it is unlikely that unimodal approaches to acute or chronic pain management will be successful.^[@R29]^ Barriers to implementing multimodal pain management are myriad and include inconsistent or inappropriate pain assessment, lack of knowledge of pain physiology and relevant evidence-based treatments, poor access to suitably trained professionals, and low prioritisation of pain education and training by health care providers.^[@R14],[@R18],[@R19],[@R31],[@R33]^ 3. A need for professional education in paediatric pain across disciplines {#s3} ========================================================================== It has been accepted for some time that although there is an expansive body of literature available to aid health care professionals in understanding the treatment of pain in children, this knowledge is not necessarily used in clinical practice.^[@R25]^ This knowledge-practice "gap" phenomenon does not only apply to pain and is a known challenge that needs to be addressed from the earliest stages of training. Personal beliefs on how to treat health conditions such as pain are difficult to change postqualification,^[@R4]^ although interprofessional pain education delivered during undergraduate stages of training can improve interdisciplinary collaboration in practice, including within pain management.^[@R21]^ In medicine, despite gradual advances in undergraduate pain education, the investigation of undergraduate medical programmes in Canada has highlighted that some continue to lack structure in delivering proper integrated, interdisciplinary content.^[@R13],[@R42]^ These findings are corroborated internationally in the APPEAL study, a Europe-wide review of pain education for undergraduate health care professionals evaluating curricula from 242 medical schools.^[@R1],[@R9]^ It was found that 69% of medical schools across Europe have no dedicated pain teaching, with notable inconsistencies in content and a lack of practical teaching methods in pain management---only 26% of courses used placement-based teaching. The APPEAL Taskforce has called for pain education for undergraduate medical students to be made "fit for purpose" so that public health needs can be met, suggesting that a clear European framework should be outlined for pain education, and that it should be a requirement of undergraduate courses to attain a defined minimum level of competency in pain management. In nurse education and practice, the management of pain in children is strongly emphasised, especially where nurses often directly administer pain treatments as recommended by the supervising clinician. Nevertheless, research investigating the inclusion of pain education (adult and paediatric) in preregistration nursing programmes across 71 Higher Education Institutes in the United Kingdom indicated a lack of pain-related content.^[@R30]^ Nurses play a fundamental role in acute pain services that are increasingly stretched due to a blur of the boundaries between acute, chronic, and palliative pain. Similarly, research investigating nurses\' knowledge of paediatric pain throughout universities and hospitals in Mexico found that undergraduate curricula do not dedicate enough time to paediatric pain management, and that this knowledge deficit continues into clinical practice.^[@R33]^ One study investigating paediatric pain management in a UK children\'s hospital noted increased demand for pain services, which was not accompanied by an increase in training. In particular, nurses emphasised disempowerment and a lack of skills to match the change in needs.^[@R6]^ Despite this, even brief education interventions can be effective, with paediatric nurses reporting increased confidence in providing pain management for children in palliative care after a 5-hour training course on pain management.^[@R11]^ Researchers emphasize the need for future interdisciplinary training to improve the effectiveness of palliative pain management in children, focusing on effective therapies and patient--clinician communication. In physiotherapy education, there has been a call for up-to-date pain education curricula for preregistration physiotherapy programmes, which includes all factors that contribute to pain.^[@R24]^ Physiotherapists are becoming increasingly expected to be able to integrate biopsychosocial approaches into treatment; however, more training is required as part of continuing professional development.^[@R18]^ Although the IASP provides subject-specific curricula for physiotherapy, there is no set regulatory standard for physiotherapy pain education; graduates must become better equipped in managing patients\' experience of pain through structured addition of pain education into current physiotherapy courses.^[@R40]^ In psychology education, an integrated approach combining biological dimensions of pain with behavioural, cognitive, and emotional factors from psychology is required to address the multifaceted nature of pain and aid professionals in selecting appropriate pain management strategies.^[@R22]^ A recent review of evidence-based psychological interventions for the management of paediatric chronic pain concluded that psychological research must continue to develop interventions that reduce biomedical biases in the management of paediatric pain. Further to developing interventions, psychologists in the field must continue to educate patients and professionals towards an in-depth understanding of the ways in which psychological interventions and interdisciplinary approaches to pain management can facilitate recovery.^[@R12]^ In addition to efforts by specialist psychologists to improve others\' understanding of paediatric pain, a needs assessment of pain psychology in the United States indicates a need for feasible changes within psychology training and education to include and emphasize pain. Implementation of enhanced pain education in psychology trainees and clinicians is supported across a range of relevant professional and patient groups: psychologists, individuals with chronic pain, pain physicians, primary care physicians, nurse practitioners, and directors of graduate and postgraduate psychology training programs.^[@R14]^ However, research into education for paediatric pain psychologists indicates that currently available training programmes lack organized structure, with a central focus on clinical work that prevents professionals from defining career trajectories; furthermore, postdoctoral accreditation both of training programs and certification of individual postdoctoral trainees in pediatric pain psychology remains poorly defined.^[@R27]^ Clearly, psychology training in paediatric pain would benefit from implementation of integrated training programs at all phases of professional development, from undergraduate through to postdoctoral studies. 4. Online education as a solution to improving professional knowledge of paediatric pain {#s4} ======================================================================================== Online education, as part of continuing professional development, is a possible solution to lack of prelicensure pain education in health care professionals, and has the potential to improve both knowledge and attitudes towards paediatric pain assessment and management.^[@R28]^ Guidelines in the subspecialty of paediatric pain psychology emphasise the use of novel technologies for professional training purposes.^[@R7]^ Supporting this, research investigating online educational videos for paediatric needle pain found that educational videos can offer clinicians current, relevant, and accurate evidence-based techniques for acute pain management.^[@R15]^ A recent systematic review of the effectiveness of online pain resources for health professionals^[@R28]^ identified 6 online resources that specifically addressed paediatric pain, and 2 resources that addressed both adult and paediatric pain education. Outcomes from 2 paediatric-specific programmes indicated increased professional competence in pain management.^[@R2],[@R10]^ Postintervention improvements in beliefs and attitudes towards paediatric pain, and improvements in clinician skills (specifically pain assessment and opioid administration), were found in 2 studies.^[@R2],[@R41]^ Improved adherence to clinical practice guidelines in paediatric palliative care was found in one Dutch study on paediatric nurse specialists.^[@R23]^ Considering pain education interventions holistically, variations in instructional methods and rapid advancement of technology make it difficult to determine which elements facilitate effective online learning for health professionals. Pain education interventions require assessment coherent with developed guidelines^[@R34]^ in future to determine which elements are effective at improving professional knowledge and skills, as well as whether improved professional education results in improved health outcomes for patients. Although online resources are available for professional paediatric pain education, only 2 resources stand out as freely available and addressing factors in paediatric pain management from a biopsychosocial perspective. Modules from the UK programme^[@R36]^ and the Canadian programme,^[@R39]^ both of which aim to improve professional knowledge of paediatric pain, are outlined in [Box 1](#FB1){ref-type="boxed-text"} (additional educational resources are listed in [Box 2](#FB2){ref-type="boxed-text"}). The clinical benefit to patients following completion of these programmes by professionals remains unknown; further assessment and development of online resources targeting paediatric pain education through integration of medical, psychological, and physiological perspectives is necessary. ###### Freely available online professional pain education resources\* ###### Useful organizations and resources regarding paediatric pain education IASP Special Interest Group on Pain in Childhood [www.childpain.org/ispp.shtml](http://www.childpain.org/ispp.shtml);Resources in French <https://www.pediadol.org/-Journees-Pediadol-.html>;Resources in German <http://www.deutsches-kinderschmerzzentrum.de/ueber-uns/fort-und-weiterbildung>;Annual Pediatric Pain Master Class in Minneapolis <http://noneedlesspain.org/ppmc/>;Pain in Child Health research training program [www.sickkids.ca/PICH](http://www.sickkids.ca/PICH). 5. Conclusion {#s5} ============= Acute and chronic pain are widespread issues in children, can have negative impacts on the quality of life of individuals, and present a significant economic cost. Professional knowledge of pain, particularly paediatric pain, is limited. Education regarding the assessment and treatment of pain in children is needed across all relevant disciplines including within medicine, nursing, physiotherapy, and psychology. It is important that professionals in each discipline understand all factors in a biopsychosocial approach to addressing paediatric pain, and moreover, accessible tools need to be developed to address the issue. In addition to recognised time constraints in professional practice, interprofessional education programmes for health care providers lack an evidence base.^[@R35]^ Furthermore, innovative pain education programmes are generally not well implemented; both accessibility to and assessment of these programmes must be improved to facilitate positive changes in current practice.^[@R5]^ One area in which professional education has been shown to improve clinicians\' confidence in managing paediatric pain is in palliative medicine^[@R3]^; nonetheless, longitudinal follow-up is still needed to assess practice patterns. Online education programmes for professionals may present an innovative solution to the lack of interdisciplinary knowledge on paediatric pain; however, there is currently a lack of available resources. Current and future online resources must be assessed to determine whether they can improve professional knowledge, attitudes, and clinical outcomes for patients. Disclosures {#s6} =========== The authors have no conflict of interest to declare. Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.	{ "pile_set_name": "PubMed Central" }
1.. Introduction ================ Paralytic shellfish toxins (PSTs), one of the most potent groups of neurotoxins in marine biotoxins, are produced by toxic marine dinoflagellates and freshwater cyanobacteria. Around 30 analogues have been reported ([Figure 1](#f1-marinedrugs-09-00466){ref-type="fig"}). Some of these analogues are metabolic products of algae-derived toxins in shellfish \[[@b1-marinedrugs-09-00466]\]. Approximately 20 analogues are found to show a wide range of toxicity \[[@b2-marinedrugs-09-00466]\]. Natural and cultivated shellfish are contaminated with PSTs by feeding on toxin-producing algae, and are a potential cause of paralytic shellfish poisoning (PSP) to humans. Although mouse bioassay (MBA), which is the official testing method to determine shellfish toxicity in many countries, is effective in protecting the consumers from PSP, alternative methods are required due to some drawbacks including accuracy, reproducibility and ethical concerns \[[@b3-marinedrugs-09-00466]\]. A number of alternative analytical methods have been developed to date. Some have been evaluated by a single laboratory validation \[[@b4-marinedrugs-09-00466],[@b5-marinedrugs-09-00466]\]. Among different types of methods, high performance liquid chromatography fluorescent detection (HPLC-FD) is one of the reliable methods to identify and quantify individual toxins. It is reported that the toxicity obtained by HPLC-FD is usually consistent with the total toxicity of MBA by measuring the major toxins C1--4, gonyautoxin (GTX) 1--6, decarbamoylgonyautoxin (dcGTX) 2 and dcGTX3, neosaxitoxin (neoSTX), decarbamoylsaxitoxin (dcSTX) and saxitoxin (STX) \[[@b2-marinedrugs-09-00466]\]. However, although the calibration standards are commercially available from the National Research Council Canada (NRC), shortage of the calibration standards has hampered implementation of HPLC-FD in routine monitoring of PSTs in bivalves. In preparation of high amount of the calibration standards, selection of natural source materials is important. Shellfish with a high level of toxicity \[[@b6-marinedrugs-09-00466],[@b7-marinedrugs-09-00466]\] are one of the important natural sources but this source has hardly been obtainable in Japan over the last decade due to very low levels of toxicity (at most 10--50 MU/g). Therefore, some toxic dinoflagellates such as Alexandrium excavatum \[[@b8-marinedrugs-09-00466]\] and A. tamarense \[[@b9-marinedrugs-09-00466]\] are used as appropriate source materials in terms of stable toxin source materials. Some toxin-producing cyanobacteria such as Anabaena circinalis and Aphanizomenon flos-aquae would also be ideal source materials in toxin production and cell growth. Because primary toxins produced by toxic algae are insufficient to account for the complicated PST analogues in shellfish, chemical conversion of the primary toxins to other analogues is used to produce a variety of toxins. Some chemical or enzymatic reactions were summarized for each substituent moiety; C11, C13 and N1 positions \[[@b10-marinedrugs-09-00466]\]. Among the known toxins, N-sulfocarbamoyl gonyautoxins, C1/C2, are useful to produce other N1-H analogues by chemically removing the substituent moieties. Although adequate conditions for conversion of C1/C2 to dcGTX2/dcGTX3, dcSTX, and GTX2/GTX3 were reported in previous papers \[[@b11-marinedrugs-09-00466],[@b12-marinedrugs-09-00466]\], detailed investigation of chemical conversion of other analogues (e.g., GTX5) that are more dominant toxins in shellfish, has not been reported yet. In this paper, we focused on large-scale cultivation of the cyanobacteria A. circinalis to prepare a variety of N1-H analogues from C1/C2 by chemical conversion. 2.. Results and Discussion ========================== 2.1.. Large-Scale Culture of Cyanobacteria A. Circinalis (TA04) ----------------------------------------------------------------- The cultured A. circinalis (strain TA04) has been maintained in modified C medium since 1996 \[[@b13-marinedrugs-09-00466]\]. The culture showed dark-green color in mid-stationary phase, and turned to chrome yellow color in late-stationary phase. The biomass and the toxin concentration also increased in the exponential growth phase, followed by the short induction growth phase (∼7 days, [Figure 2](#f2-marinedrugs-09-00466){ref-type="fig"}). As the toxin concentration declined when the culture reached the late-stationary phase, the cells were harvested at 28 days. The cultured cyanobacteria produced C1/C2 (70--90%) as the dominant toxins. Although GTX2/GTX3 and dcGTX2/dcGTX3 (5--15% each) were produced as the minor toxins, N1-hydroxylated toxins such as GTX1/GTX4 were not found in the harvested culture. The toxin components determined in our present study were different from those obtained in our previous study \[[@b13-marinedrugs-09-00466]\]. This difference is probably caused by physiological natural changes due to the long-term successive culture over 15 years and the change of culture medium. On the other hand, the α/β-epimer ratio of C-toxins was 1.1, which was close to previous data \[[@b14-marinedrugs-09-00466]\]. The toxin production of C1/C2 in large-scale culture was 1.3 ± 0.1 μmol/L culture (7% RSD, n = 3) at 28 days. This value was somewhat lower than that found in our previous study (1.8--2.0 μmol/L culture). Approximately 80% of the total toxins were found inside the cells. Our result was similar to that obtained in another strain, MB06 \[[@b14-marinedrugs-09-00466]\], which was collected in the same location as our material strain, TA04, in the ratio of epimers and intra-/extracellular toxin in 1.2 L small scale culture, whereas the cell growth of strain TA04 was slower than strain MB06. The cyanobacteria were cultured under several modified conditions of Fitzgerald medium toward efficient toxin production. The biomass and toxin production for 28 days were investigated in each condition ([Table 1](#t1-marinedrugs-09-00466){ref-type="table"}). As a result, the concentration of nitrate and phosphate hardly affected the toxin production and biomass of the cyanobacteria. Two different inoculation volumes were determined to shorten the culture period, but the cell growth and the toxin production were the same as those of the control. In PST biosynthesis of cyanobacteria, the precursors such as arginine and S-adenosylmethionine were identified by the cultivation with stable isotopes \[[@b12-marinedrugs-09-00466]\] and a hypothetical pathway was proposed. The in vitro experiments for PST biosynthesis suggested that pyridoxal phosphate (PLP)-dependent enzymes were strongly involved in PST biosynthesis \[[@b15-marinedrugs-09-00466]\]. Therefore, the toxin production in the cell was investigated in vivo by adding a common cofactor, PLP, to the medium. Our study showed PLP hardly affected both cell growth and toxin production in our culture. PLP could be insufficiently incorporated into cells to be available to enzymes. Moreover, the addition of organic phosphates, β-glycerophosphate sodium salt of 10 and 100 μM each, was also investigated and no significant change was observed. Finally, Good's buffer, bicine, was also used to control the pH of the culture medium during culture. The toxin production was slightly improved compared to the control, but the cell harvest was hampered and the intracellular toxins leaked outside in harvest because of the increased viscosity of the culture. [Figure 2](#f2-marinedrugs-09-00466){ref-type="fig"} indicates that approximately 140 μmol of C1/C2 could be produced from 100 L culture. It was demonstrated that over 220 μmol of C1/C2 was obtained from 200 L of the culture which was produced over 6 months cultivation. This amount of toxins would be sufficient to prepare the calibration standards of STX analogues. 2.2.. Preparation of Paralytic Shellfish Toxin Analogues -------------------------------------------------------- C1/C2 are one of the most modified STX analogues among the known toxins. C1/C2 are also known as the lowest toxicity group among STX analogues \[[@b2-marinedrugs-09-00466]\]. [Figure 3](#f3-marinedrugs-09-00466){ref-type="fig"} shows the preparation scheme of PST analogues by chemical conversion from C1/C2. N-sulfocarbamoyl group on C1/C2 is hydrolyzed in neutral pH, and this reaction leads C1/C2 to dcGTX2/dcGTX3. Subsequently, the sulfate ester group of the toxins is reductively eliminated by 2-mercaptoethanol (2-ME) to prepare dcSTX \[[@b11-marinedrugs-09-00466]\]. Thus, our route to produce dcSTX is useful compared to other routes through STX \[[@b16-marinedrugs-09-00466]\] because production of STX is legally restricted in Japan and many countries. GTX5 was directly converted from C1/C2 by reductive elimination of the sulfate. C1/C2 were reacted with 2-ME to give 48% of GTX5, 18% of the remaining C1/C2, and 34% of unknown reactants or the loss. Of the reaction temperatures tested (50--80 °C), a good yield (50%) of GTX5 was obtained at 50 °C. The yield reached 30% within 30 min and slowly increased with time, and then gradually declined over 3 h. The remaining substrate was decreased to 20--40% within 30 min and then slightly declined. HPLC analysis of the reactant mixture revealed that dcGTX2/dcGTX3 and dcSTX, as by-products in the reaction, were not produced, indicating that the reaction specifically proceeded without hydrolyzing the N-sulfocarbamoyl chain of C1/C2. Interestingly, HPLC analysis of the reaction mixture revealed that the relative proportion of C1 to C2 decreased after the reaction, implying that the α-epimer reacted selectively with a thiol compound compared to the β-epimer. Chromatographic behavior of the product was the same as that of the authentic GTX5 ([Figure 4](#f4-marinedrugs-09-00466){ref-type="fig"}). [Figure 5](#f5-marinedrugs-09-00466){ref-type="fig"} shows LC-MS/MS spectra of the authentic GTX5 and the product converted from C1/C2. The MS/MS spectrum of the authentic GTX5 gave ion peaks at m/z 282, and 300 corresponding to \[M − H~2~O − SO~3~ + H\]^+^ and \[M − SO~3~ + H\]^+^, respectively ([Figure 5A](#f5-marinedrugs-09-00466){ref-type="fig"}). Basically the same fragment ions were obtained for the product ([Figure 5B](#f5-marinedrugs-09-00466){ref-type="fig"}), demonstrating that the product was GTX5. GTX5 was prepared in NaHCO~3~ solution containing dithiothreitol (DTT) in a previous study \[[@b20-marinedrugs-09-00466]\]. They also attempted preparation of GTX5 by the reaction of C1/C2 with 2-ME in place of DTT, but GTX5 was not obtained. The fact that our method was successful to produce GTX5 with 2-ME was probably because of acidic condition in which the thioester intermediate obtained in neutral pH \[[@b21-marinedrugs-09-00466]\] was not formed. Conversion efficiency of GTX5 from C1/C2 prepared in our present method was better than that (36%) reported in the previous study \[[@b20-marinedrugs-09-00466]\]. The toxins with an N-sulfate group such as C3/C4, GTX5 and GTX6 are easily converted to the corresponding carbamate toxins by hydrolyzing in diluted mineral acids \[[@b12-marinedrugs-09-00466],[@b19-marinedrugs-09-00466],[@b22-marinedrugs-09-00466]\]. In our present study, this hydrolysis reaction was applied to prepare GTX2/GTX3 from C1/C2. The products showed the same chromatographic behavior with the authentic GTX2/GTX3 in HPLC-FD. The products, GTX2/GTX3, were obtained as an equilibrium mixture (ratio: 3.0 to 3.5), although C1/C2 used for the reaction were not an equilibrium mixture. In LC-MS/MS analysis, epimers such as GTX2/GTX3 show different fragmentation patterns. GTX2 as α-epimer gives a predominant ion peak at m/z 316 corresponding to \[M − SO~3~ + H\]^+^, whereas GTX3 as β-epimer gives several major peaks at m/z 298, 316 and 378 corresponding to \[M − H~2~O + H\]^+^. The products were confirmed to show the same fragmentation pattern in each epimer. In the process of preparation of PST analogues, it was a requirement to avoid generation of STX because production of STX is illegal in Japan and many countries. In our method, STX was not detected in any processes by HPLC-FD. Although STX is not prepared in this study as mentioned above, STX could be prepared from GTX2/GTX3 or GTX5 as described in the previous reports \[[@b12-marinedrugs-09-00466],[@b17-marinedrugs-09-00466]\], if necessary. Thus, C1/C2 as starting materials could be converted to six analogues and STX, giving a relatively high yield in each reaction ([Figure 3](#f3-marinedrugs-09-00466){ref-type="fig"}). The starting materials of C1/C2 can be obtained at a yield of 50% from the extract of the cell by purifying them with several chromatography steps such as activated charcoal, gel filtration, and ion exchange. The purified C1/C2 were converted to dcGTX2/dcGTX3 (step A), GTX5 (step C), and GTX2/GTX3 (step D), and subsequently a portion of the resulting dcGTX2/dcGTX3 were converted to dcSTX (step B). At least 10 μmol of the toxin calibrants could be prepared from 200 L of culture of the toxic cyanobacteria ([Table 2](#t2-marinedrugs-09-00466){ref-type="table"}). 3.. Experimental Section ======================== 3.1.. Large-Scale Culture of Cyanobacteria and the Harvest ---------------------------------------------------------- The cyanobacterium A. circinalis used in this study is a non-axenic strain TA04. The field sample of A. circinalis was collected at Tullaroop reservoir, Victoria, Australia, and strain TA04 was one of single-trichome isolates prepared by Negri et al. \[[@b14-marinedrugs-09-00466]\]. Strain TA04 was cultured in Fitzgerald medium in 2 L culture, and then scaled-up to 10 L volume (Thermo Fisher Scientific, Rochester, NY, USA). The cyanobacteria were cultured on a 16 L/8 D cycle at 17 °C and 25 μmol/m^2^/s, passing air through a disc filter (0.20 μm pore size, Advantec, Japan). The bottle cap was equipped with three pores, where one was used as air-inlet with the disc filter, one was as air-outlet with the disc filter and one was as a sampling tube with a cock. The cyanobacteria were maintained by inoculating 1 L of the culture in exponential growth phase (two weeks later) to 9 L of the sterilized medium. The 10 L batch culture was carried out in quintuplicate, of which one was used for inoculation of seed population, and repeated three times. Four weeks later after inoculation of the seed population, the A. circinalis cells were harvested by a continuous centrifugation or filtration (No. 2, Advantec, Japan). The cell-free culture filtrate and the cell fractions were treated according to our previous procedure \[[@b11-marinedrugs-09-00466]\]. 3.2.. Cell Growth Measurement and Toxin Analysis ------------------------------------------------ Cell growth was estimated from fluorescent intensity derived from chlorophyll a in the cell. The culture was taken in axenic through a sampling tube. Culture samples (200 μL) were collected, and added to wells of a 96-well microplate in triplicate. The cell-free culture filtrate (200 μL) as blank was prepared by filtering the culture with a syringe filter (Dismic-25cs, 0.45 μm, Advantec, Japan) or a glass fiber paper (GA-100, 25 mm, Advantec, Japan). It is subtracted from the culture fluorescence using a fluorescence plate reader (Ex: 460 nm; Em: 645 nm, FL600, BioTek). Toxin content (C1/C2) of the cell and the cell-free filtrate was analyzed with HPLC-FD \[[@b2-marinedrugs-09-00466]\]. The culture of 10 to 30 mL at each growth phase was passed through a glass fiber paper (GA-100, 25 mm, Advantec, Japan) to collect cells. The paper was dissected and suspended into 1 mL of 0.5 M acetic acid. The suspension was centrifuged, following ultrasonication, to obtain the supernatant as analyte. On the other hand, the filtrate was adjusted to become acidic by adding 0.5 M acetic acid. The filtrate was evaporated to dryness and then dissolved in 500 μL of 0.5 M acetic acid. The toxin solution was ultrafiltered (Ultracell YM-10, 10000 MWCO, Millipore, Billerica, MA, USA) to obtain the resulting filtrate as analyte. 3.3.. Serial Preparation of dcGTX2/dcGTX3 and dcSTX from C1/C2 -------------------------------------------------------------- The preparation of dcGTX2/dcGTX3 and dcSTX from C1/C2 was carried out according to the method of Watanabe et al. \[[@b11-marinedrugs-09-00466]\]. C1/C2 as starting materials were purified from the harvested cell fraction in the cyanobacteria A. circinalis (TA04) culture. Typically, dcGTX2/dcGTX3 were obtained by the hydrolysis of C1/C2. C1/C2 were heated in 100 mM potassium phosphate buffer at pH 7.0 for 60 min, stirring at 70 °C, and then cooled on ice. The resulting dcGTX2/dcGTX3 were isolated by gel filtration chromatography (Bio Gel P-2, 10 mm × 400 mm, Bio-Rad) and weak-cation exchange chromatography (Bio-Rex 70, 200--400 mesh, 10 mm × 400 mm, Bio-Rad) at overall yield of 50%. The purified dcGTX2/dcGTX3 was reacted under stirring conditions at 60 °C in 100 mM potassium phosphate buffer at pH 6.0 containing an equal volume of 2-ME for 60 min to quantitatively give the resulting dcSTX. 3.4.. Preparation of GTX5 from C1/C2 ------------------------------------ C1/C2 (1.4 μmol) were evaporated to dryness and dissolved in 2 mL of 100 mM acetate buffer at pH 5.2 containing equal volume of 2-ME in a 4 mL glass vial. The toxin solution was reacted at 50 °C for 2 h. The reactant was evaporated and then diluted with acetonitrile to be a final concentration of 90% (v/v) acetonitrile solution and then the diluted solution was passed through a ZIC-HILIC cartridge (3 mL, 200 mg, Merck SeQuant, Umeå, Sweden) preconditioned with 6 mL of distilled water and 6 mL of acetonitrile in order. The cartridge was washed with 3 mL of acetonitrile and then eluted with 3 mL of distilled water as toxic fraction. The toxic fraction was purified by Bio-Rex 70 (200--400 mesh, 10 mm × 450 mm, Bio-Rad) to give GTX5 at an overall yield of 48%. The product was confirmed and identified by the comparison with the authentic standard that was previously prepared by Oshima and confirmed in structure and purity with NMR \[[@b2-marinedrugs-09-00466]\]. Chromatographic behavior of the toxin was determined with HPLC-FD. The parameters on LC-MS instrument (3200 Q TRAP LC/MS/MS system, AB-SCIEX, Concord, Canada) were optimized with the authentic GTX5, prior to the analysis. The flow injection analysis was performed in mass range of m/z 200--500 and positive ion mode at 500 °C. The eluents were flowed at 0.2 mL/min with 65% B, where eluent A was distilled water and eluent B was acetonitrile, both containing 20 mM formic acid. The MS/MS analysis was carried out in mass range of m/z 50--450 at declustering potential of 11 V and collision energy of 21 V. The fragmentation pattern between the product and the authentic was compared and identified. 3.5.. Preparation of GTX2/GTX3 from C1/C2 by Acidic Hydrolysis -------------------------------------------------------------- Preparation of GTX2/GTX3 was carried out according to a previous method \[[@b12-marinedrugs-09-00466]\]. C1/C2 (1.4 μmol) were evaporated to dryness and dissolved in 1 mL of 0.13 M HCl aq. The solution was heated at 100 °C, stirring for 15 min and then cooled on ice. The products, GTX2/GTX3, were recovered using a ZIC-HILIC cartridge as described above. The reaction quantitatively proceeded. The structure was confirmed as described above. 4.. Conclusions =============== In this study, major N1-H toxins were prepared from C1/C2 as starting material. In particular, this is the first report of the detailed conditions for preparation of GTX5 from C1/C2 analogues. The chemical conversion allows us to prepare the toxins systematically. The toxin yield, of over 220 μmol of C1/C2 extracted from 200 L of culture of the cyanobacteria, was estimated to be capable of preparing at least 10 μmol of each toxin, based on our reaction yields ([Table 2](#t2-marinedrugs-09-00466){ref-type="table"}). We previously reported a method using NMR to quantify concentrations of PST standards \[[@b23-marinedrugs-09-00466]\]. Based on the preparation technique of PSTs and the quantitative NMR, large amount of the N1-H toxin standards could be prepared and supplied constantly for shellfish safety monitoring programs using chemical analyses. The N1-OH toxins such as GTX1/GTX4 are also essential for the chemical analyses. The preparation technique of these toxins is currently being investigated, and will be reported elsewhere. The authors are greatly appreciative to Blackburn, CSIRO, Australia, for providing cyanobacteria A. circinalis (TA04). This study was funded by Ministry of Health, Labour and Welfare in Japan. ![Structures of paralytic shellfish toxins.](marinedrugs-09-00466f1){#f1-marinedrugs-09-00466} ![Growth curve (A) and toxin production (B) in a large-scale culture of A. circinalis. (A) Fluorescent intensity shows the average level of chlorophyll a in cells (n = 3); (B) Only C1/C2 produced by the cells were analyzed by high performance liquid chromatography fluorescent detection (HPLC-FD). Solid line shows total toxin concentration of C1/C2 in culture. Dashed line shows intracellular toxin concentration of C1/C2. The difference corresponds to the extracellular toxin concentration.](marinedrugs-09-00466f2){#f2-marinedrugs-09-00466} ![Preparation scheme of paralytic shellfish toxin analogues. (A) Phosphate buffer (pH 7.0), 70 °C, 60 min; (B) Phosphate buffer (pH 6.0), 2-mercaptoethanol, 60 °C, 60 min; (C) Acetate buffer (pH 5.2), 2-mercaptoethanol, 50 °C, 2 h; (D) 0. 13 M HCl aq., 100 °C, 15 min; (E) 2-mercaptoethanol, 100 °C, 15 min.; (F) 0.05 M HCl, 100 °C, 30 min \[[@b17-marinedrugs-09-00466]\]; (G) 7.5 M HCl, 100 °C, 3 h \[[@b18-marinedrugs-09-00466],[@b19-marinedrugs-09-00466]\]. Solid arrows, this study; dashed arrows, previous studies.](marinedrugs-09-00466f3){#f3-marinedrugs-09-00466} ![Chromatograms of the authentic GTX5 (upper) and the product (bottom). The authentic GTX5 was used as a mixture of GTX1--5.](marinedrugs-09-00466f4){#f4-marinedrugs-09-00466} ![MS/MS spectra of the authentic GTX5 (A) and the product (B). Both toxins were analyzed at declustering potential of 11 V and collision energy of 21 V. The precursor ion was set to 380.](marinedrugs-09-00466f5){#f5-marinedrugs-09-00466} ###### Biomass and toxin production of C1/C2 at 28 days of A. circinalis under modified condition of Fitzgerald medium. Entry Parameters Biomass Toxin concentration of C1/C2 (μmol/L) ----------- --------------------------------------------------------------- --------- ------------- ------------------------------------------- ---- ---- Control 2300 1.39 80 20 1 NaNO~3~ 0 mM 400 0.82 75 25 2 2.9 mM 1200 1.26 86 14 3 11.6 mM 1600 1.03 84 16 4 K~2~HPO~4~ 0.14 mM 1600 1.43 86 14 5 0.54 mM 2000 1.49 84 16 6 Inoculation 0.5 L 2000 1.36 77 23 7 2 L 2500 1.05 84 16 8 Pyridoxal phosphate 10 μM 2500 1.36 58 42 9 Glycerophosphate Na 10 μM 3100 1.89 55 45 10 100 μM 2300 1.33 59 41 11 Bicine 0.6 mM 2700 1.78 3 97 Average [\](#tfn1-marinedrugs-09-00466){ref-type="table-fn"} 1.35 75 25 Average toxin concentration was calculated using the data from entry 1 to 10. ###### Amounts of toxins prepared from large-scale culture (200 L) of cyanobacteria A. circinalis.* The amounts of C1/C2 as substrates are calculated as 110 μmol, because the toxins are isolated at 50% yield from the extract of the cyanobacteria. Reaction steps ^[§](#tfn2-marinedrugs-09-00466){ref-type="table-fn"}^ Substrates Amounts (μmol) Products Amounts (μmol) Yields [^¶^](#tfn4-marinedrugs-09-00466){ref-type="table-fn"} (%) Remaining substrates Residual Amounts (μmol) Residual ratio [^†^](#tfn5-marinedrugs-09-00466){ref-type="table-fn"} (%) --------------------------------------------------------------------------- ---------------- -------------------- -------------- -------------------- ----------------------------------------------------------------------- -------------------------- ----------------------------- ------------------------------------------------------------------------------- A C1/2 70 dcGTX2/3 35 50 C1/2 21 30 *B [\](#tfn3-marinedrugs-09-00466){ref-type="table-fn"} dcGTX2/3 15 dcSTX 14 90 C C1/2 25 GTX5 12 48 C1/2 5 18 D** C1/2 15 GTX2/3 14 90 The steps correspond to [Figure 3](#f3-marinedrugs-09-00466){ref-type="fig"}; dcGTX2/3 were used as a part of the products in step A; (Amount of Product)/(Amount of Substrate) × 100; (Residual Amounts)/(Amount of Substrate) × 100.	{ "pile_set_name": "PubMed Central" }
![](hosplond70099-0012){#sp1 .108}	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-sensors-18-03081} =============== Window functions are widely used to reduce the well-known Gibbs oscillations resulting from the truncation of a Fourier series \[[@B1-sensors-18-03081]\]. This century-old technology is employed in many applications in the sensor signal processing field \[[@B2-sensors-18-03081],[@B3-sensors-18-03081],[@B4-sensors-18-03081]\]. In most of these applications, a window function is expected to behave as a Dirac function in the frequency domain, that is, to concentrate all the spectral power into an extremely narrow band with zero sidelobes. However, this behaviour exists neither in theory nor in practice. A window function always has a mainlobe width and sidelobes. As stated in \[[@B5-sensors-18-03081]\], no window is suitable for all cases, and one should be selected according to the requirements of a particular application. Thus, a method for designing window functions with flexible spectral characteristics is greatly needed. Commonly used spectral characteristics of a window function include the mainlobe width (MW), the peak sidelobe level (PSL), and the sidelobe fall-off rate (SLFOR), which are closely related to the resolution and spectrum leakage. In addition, steerable sidelobes over specified regions are needed in some applications. Due to the importance and significance of window functions, the literature on this topic is unsurprisingly extensive (e.g., see \[[@B1-sensors-18-03081],[@B2-sensors-18-03081],[@B3-sensors-18-03081],[@B4-sensors-18-03081],[@B5-sensors-18-03081],[@B6-sensors-18-03081],[@B7-sensors-18-03081],[@B8-sensors-18-03081],[@B9-sensors-18-03081],[@B10-sensors-18-03081],[@B11-sensors-18-03081],[@B12-sensors-18-03081],[@B13-sensors-18-03081],[@B14-sensors-18-03081],[@B15-sensors-18-03081],[@B16-sensors-18-03081],[@B17-sensors-18-03081],[@B18-sensors-18-03081],[@B19-sensors-18-03081],[@B20-sensors-18-03081],[@B21-sensors-18-03081],[@B22-sensors-18-03081],[@B23-sensors-18-03081],[@B24-sensors-18-03081],[@B25-sensors-18-03081]\] and the references therein). Traditional windows, including triangular, Hamming, Hanning, Blackman, and other well-known windows, aim at smoothing the truncated impulse response in the time domain, resulting in lower sidelobes at the expense of a wider mainlobe \[[@B1-sensors-18-03081]\]. However, the spectral characteristics of these windows cannot be controlled because of the fixed expressions. Scholars later presented many window families with adjustable parameters, and trade-offs can be made to control different properties \[[@B6-sensors-18-03081],[@B7-sensors-18-03081],[@B8-sensors-18-03081],[@B9-sensors-18-03081],[@B10-sensors-18-03081],[@B11-sensors-18-03081],[@B12-sensors-18-03081],[@B13-sensors-18-03081],[@B14-sensors-18-03081],[@B15-sensors-18-03081],[@B16-sensors-18-03081],[@B17-sensors-18-03081]\]. In recent years, a triangular self-convolution window has been obtained by convolving the triangular window as presented in \[[@B18-sensors-18-03081]\], and it exhibits a good deal of lower PSL and higher SLFOR. Its applications in harmonic estimation are discussed in \[[@B19-sensors-18-03081],[@B20-sensors-18-03081]\]. In \[[@B21-sensors-18-03081]\], optimized trapezoid convolution windows are proposed, leading to a narrow main lobe width but retaining a low PSL as well as a fast SLFOR. In particular, Tseng provided a novel window function with steerable sidelobes with which a deep dip can be steered to any frequency, effectively improving the detectability of a small tone without degrading its resolvability \[[@B22-sensors-18-03081],[@B23-sensors-18-03081]\]. Several windows combining steerable sidelobes and other properties were presented in \[[@B24-sensors-18-03081],[@B25-sensors-18-03081]\]. However, the parameters of the expressions proposed above should be carefully chosen and repeatedly tuned to obtain satisfactory performance, which is complicated and unintuitive. Moreover, to the best of our knowledge, none of these methods can control all spectral characteristics of window functions at the same time. Window functions combining steerable sidelobes with controllable spectral characteristics have not been proposed to date. In addition, in a particular case, window functions should be specially designed according to specific applications; however, the methods mentioned above are not very applicable. For example, in the radar signal processing field, the complex linear frequency modulation (LFM) signal has a rectangle-like shape in the frequency domain. The common pulse compression method for echoes of the LFM signal is usually carried out by using the matched filter (MF) method. The final step of the MF method is to conduct an inverse discrete Fourier transform (IDFT) on a rectangle-like spectrum. The output of the MF in the time domain has a high PSL. Thus, window functions in the frequency domain are needed in the final step. Obviously, the output of the MF is related to not only the window functions but also the spectrum of the LFM. Hence, many properties of the window functions cannot be acquired by using the existing methods, which are based on rectangular windows by default, and partial or total effectiveness will be lost in these applications. In other recent research, Stoica P. proposed many algorithms, including cyclic algorithms (CAs), CA-pruned (CAP), CA-new (CAN), and weighted CAN (We CAN), for applications in unimodular phase-coded waveform design \[[@B26-sensors-18-03081]\]. The designed waveforms have ultra-low range sidelobes over specified regions after pulse compression. Then, an FFT-based CA for designing a unimodular sequence train with low central and recurrent autocorrelation sidelobes is proposed in \[[@B27-sensors-18-03081]\]. Some numerical properties of the CA for the estimation of the parameters of a multinomial model is presented in \[[@B28-sensors-18-03081]\]. Performance of the CA is also compared with some of the best available optimization algorithms. It should be noted that CA is not a technical term but simply a name to describe the iteration procedure, and all CAs in \[[@B26-sensors-18-03081],[@B27-sensors-18-03081],[@B28-sensors-18-03081]\] and this paper are different in the details. In addition, the convergence analysis and the recursion formula in these papers are not given. Interestingly, CA is a good iteration idea for solving optimization problems and may be well applied in the window-design field. In this paper, we aim to propose a method for designing general window functions in which all spectral characteristics and steerable sidelobes can be easily and flexibly controlled. In addition, these windows should be adaptable to different cases such as non-rectangular applications. Thus, a novel method, i.e., the inverse of the shaped output using the cyclic algorithm (ISO-CA), for designing general window functions with flexible spectral characteristics is proposed. The ISO-CA is based on only the fast Fourier transform (FFT) and the inverse FFT (IFFT) and thus can accelerate the iterative procedure. The recursion formula is also provided, and the convergence of the iterations is proven. The results of experiments and simulations show that flexible spectral characteristics, including the MW, PSL, SLFOR, and steerable sidelobes, for the designed window function can be easily obtained using the proposed CA method. In addition, it is more flexible and intuitive for designers to use instead of choosing and adjusting several parameters. Moreover, any existing window with good properties can be chosen as an initial window function of the algorithm so that the generated CA window function will have both the properties of the initial window and the designated spectral characteristics. Thus, using the proposed method, windows can be specially designed for specific applications to which existing methods would not apply. This paper is organized as follows. In [Section 2](#sec2-sensors-18-03081){ref-type="sec"}, the model derivation is introduced. Some simulations and experiments are presented in [Section 3](#sec3-sensors-18-03081){ref-type="sec"}. [Section 4](#sec4-sensors-18-03081){ref-type="sec"} concludes this paper. The convergence analysis of the ISO-CA is provided in the [Appendix](#app1-sensors-18-03081){ref-type="app"}. 2. Model Derivation {#sec2-sensors-18-03081} =================== Window functions are selected principally based on their spectral characteristics. The ISO-CA proposed in this paper is a novel method for designing general window functions with flexible spectral characteristics, for which an existing window should be chosen as the initial input. Here, we choose the rectangular window function as an example:$$\mathbf{w} = \left\lbrack {1,1,\cdots,1} \right\rbrack^{T},$$ where $\left\lbrack \cdot \right\rbrack^{T}$ denotes the transpose operator. The length of the rectangular window is set to N. The rectangular window in the frequency domain can be expressed as:$$\mathbf{W} = \mathbf{F}_{M}\begin{bmatrix} \mathbf{w} \\ 0 \\ \end{bmatrix}_{M \times 1},$$ where $\mathbf{F}_{M}$ is the M-point DFT matrix:$$\mathbf{F}_{M} = \begin{bmatrix} 1 & 1 & \cdots & 1 \\ 1 & \mathsf{\Omega}_{M}^{1} & \cdots & \mathsf{\Omega}_{M}^{M - 1} \\ \vdots & \vdots & \ddots & \vdots \\ 1 & \mathsf{\Omega}_{M}^{M - 1} & \cdots & \mathsf{\Omega}_{M}^{{({M - 1})}{({M - 1})}} \\ \end{bmatrix},\mathsf{\Omega}_{M}^{k} = \exp\left( {- j\frac{2\pi}{M}k} \right).$$ We pad $\left( {M - N} \right)$ zeros at the end of the window in ([2](#FD2-sensors-18-03081){ref-type="disp-formula"}) to decrease the well-known picket fence effects of the DFT caused by the truncation and discreteness. In addition, N and M are usually set to even numbers to maintain the "even" property of the window \[[@B5-sensors-18-03081]\]. For convenience, we can rewrite ([2](#FD2-sensors-18-03081){ref-type="disp-formula"}) as:$$\mathbf{W} = \mathbf{F}_{MN}\mathbf{w},$$ where $\mathbf{F}_{MN}$ denotes the first N columns of $\mathbf{F}_{M}$. Then, we define the spectral characteristics of the window functions. For the PSL:$${PSL}\left( {dB} \right){= 10} \cdot {lo}g_{10}\left( \frac{\max\left\{ \left\| {W\left( k \right)} \right\|^{2} \right\}}{\left\| {W\left( 0 \right)} \right\|^{2}} \right),k \in {sidelobe}\;{area},$$ which denotes the maximal level of the sidelobes or the sidelobe attenuation. Note that steerable sidelobes can be obtained by the proposed method. The ISL over a specified interval $\left\lbrack {a,b} \right\rbrack$ is defined as follows:$$\underset{a\rightarrow b}{ISL}\left( {dB} \right) = 10 \cdot \log_{10}\left( \frac{\sum\limits_{k = a}^{b}\left\| {W\left( k \right)} \right\|^{2}}{\left\| {W\left( 0 \right)} \right\|^{2}} \right),a,b \in {sidelobe}\;{area}.$$ The SLFOR (dB/oct) denotes how many dB the sidelobes decrease by per octave, which measures the sidelobe fall-off rate. For the MW, the width of the mainlobe at 3 dB below the mainlobe peak is commonly used \[[@B1-sensors-18-03081]\]. Windows with different spectral characteristics are needed in different situations. Thus, a method for designing window functions that can control all these spectral characteristics is greatly needed. Now, we design a mask vector:$$\begin{matrix} {\Gamma = \left\lbrack {\Gamma\left( 0 \right),\Gamma\left( 1 \right),\cdots,\Gamma\left( {M - 1} \right)} \right\rbrack^{T}} \\ {\Gamma\left( p \right) = 1,p \in {mainlobe}\;{area}} \\ {\Gamma\left( q \right) \in \left\lbrack {0,1} \right\rbrack,q \in {sidelobe}\;{area}} \\ \end{matrix}$$ to shape the output $\mathbf{w}$ and obtain the desired spectral characteristics:$$\mathbf{W}_{\Gamma} = \Gamma \odot \mathbf{W},$$ where ⊙ is the Hadamard (element-wise) product. Note that window functions are usually real, even and nonnegative in the time domain \[[@B5-sensors-18-03081]\]. Thus, $\Gamma$ should satisfy:$$\Gamma\left( i \right) = \Gamma\left( {M - 1 - i} \right),\forall i \in \left\lbrack {0,M - 1} \right\rbrack$$ to maintain these properties according to the DFT theorem. q and $\Gamma\left( q \right)$ can be designed and adjusted easily and flexibly according to different requirements of the PSL, SLFOR, and steerable sidelobes. The advantage of the mask vector $\Gamma$ is that we can change only the spectral characteristics of interest while maintaining the others as much as possible. The problem now is how to design a window function $\mathbf{w}_{\Gamma}$ such that:$$\mathbf{F}_{MN}\mathbf{w}_{\Gamma} = \mathbf{W}_{\Gamma}.$$ Equation ([10](#FD10-sensors-18-03081){ref-type="disp-formula"}) is actually an overdetermined matrix equation with M equations and N unknowns, where $M \gg N$ generally. Only in very special circumstances does an overdetermined equation have an accurate solution, but a least squares solution can always be acquired by calculating the left inverse of the matrix $\mathbf{F}_{MN}$:$${\hat{\mathbf{w}}}_{\Gamma} = \mathbf{F}_{MN}^{\dagger}\mathbf{W}_{\Gamma},$$ where $\mathbf{F}_{MN}^{\dagger} = \left( {\mathbf{F}_{MN}^{H}\mathbf{F}_{MN}} \right)^{- 1}\mathbf{F}_{MN}^{H}$, and $\left\lbrack \cdot \right\rbrack^{H}$ denotes the conjugate transpose or the Hermitian transpose operator. We know that matrix $\mathbf{F}_{MN}$ is part of $\mathbf{F}_{M}$, the columns of which are orthonormal base vectors:$$\mathbf{F}_{MN}^{H}\mathbf{F}_{MN} = M \cdot \mathbf{I}_{NN},$$ where $\mathbf{I}_{NN}$ is an $N \times N$ unit matrix. Thus, we can obtain:$$\mathbf{F}_{MN}^{\dagger} = \frac{1}{M}\mathbf{F}_{MN}^{H}.$$ According to the DFT theorem, the IDFT matrix satisfies:$$\mathbf{F}_{M}^{- 1} = \frac{1}{M}\mathbf{F}_{M}^{H} = \begin{bmatrix} {\frac{1}{M}\mathbf{F}_{MN}^{H}} \\ \vdots \\ \end{bmatrix} = \begin{bmatrix} \mathbf{F}_{MN}^{\dagger} \\ \vdots \\ \end{bmatrix}$$ Equation ([14](#FD14-sensors-18-03081){ref-type="disp-formula"}) is an interesting conclusion. It indicates that we can replace the left inverse calculation with the IFFT and simply take the first N values of the result to obtain ${\hat{\mathbf{w}}}_{\Gamma}$. We now obtain a simple closed-form solution to the expected window function. However, we may not acquire the anticipatory spectral characteristic as expressed in ([8](#FD8-sensors-18-03081){ref-type="disp-formula"}) because of the natural errors of the least squares solution. To compensate for these errors of solution ([11](#FD11-sensors-18-03081){ref-type="disp-formula"}), the ISO-CA is proposed. The solution provided in ([11](#FD11-sensors-18-03081){ref-type="disp-formula"}) has only some of the features of the desired spectral characteristic as expressed in ([8](#FD8-sensors-18-03081){ref-type="disp-formula"}). Thus, the idea is to enhance the features via further iterations. The calculated ${\hat{\mathbf{w}}}_{\Gamma}$ is chosen as the new initial window function $\mathbf{w}$, and then the procedure above is repeated to allow the spectral characteristics to gradually fit the mask vector $\Gamma$. The ISO-CA can be summarized as Algorithm 1 shows. Algorithm 1 ISO-CA Set initial $\mathbf{w}$ and $\Gamma$;\ repeat\ 1. Calculate the M-point FFT of $\mathbf{w}$ to obtain $\mathbf{W}$ (Equation ([4](#FD4-sensors-18-03081){ref-type="disp-formula"}));\ 2. Update the mask vector $\Gamma$ and calculate: $\mathbf{W}_{\Gamma} = \Gamma \odot \mathbf{W}$ (Equation ([8](#FD8-sensors-18-03081){ref-type="disp-formula"}));\ 3. Calculate the M-point IFFT of $\mathbf{W}_{\Gamma}$, and take the first N values to obtain ${\hat{\mathbf{w}}}_{\Gamma}$ (Equation ([14](#FD14-sensors-18-03081){ref-type="disp-formula"}));\ 4. Replace $\mathbf{w}$ with ${\hat{\mathbf{w}}}_{\Gamma}$;\ until the stop criterion is met. It can be seen from the flow that each iteration contains: calculating the FFT of the window in the time domain, shaping the output to obtain the desired spectral characteristics, inverting the shaped output using the IFFT, and replacing the old window with the new one. Thus, we call it the inverse of the shaped output using the ISO-CA method. The iteration of the ISO-CA has linear convergence rates. The recursion formula and derivation are presented in the [Appendix](#app1-sensors-18-03081){ref-type="app"}. This algorithm procedure shows that only the FFT, the Hadamard product, and the IFFT are used in each iteration without complex computations. Thus, by taking advantage of the simplicity of the FFT and IFFT, the efficiency of the iteration is greatly improved. It is possible to create a wide dip centred at a specified frequency without broadening the mainlobe \[[@B22-sensors-18-03081]\]. However, the suppression of the near-sidelobes is obtained at the expense of a broadened mainlobe. Thus, in this case, the data range of p in ([7](#FD7-sensors-18-03081){ref-type="disp-formula"}) should be updated during the iteration. In addition, to control the spectral characteristics of the window function, the criterion for stopping the iteration can be designed as follows:$${stop}\;{criterion}\left\{ \begin{array}{l} {i \geq I_{\max}} \\ {{\parallel {\mathbf{w}_{i} - \mathbf{w}_{i - 1}} \parallel}_{2} \leq \varepsilon_{0}\left( {i \geq 2} \right)} \\ {{PSL}\left( i \right) \leq \varepsilon_{1}} \\ {\underset{a\rightarrow b}{ISL}\left( i \right) \leq \varepsilon_{2}} \\ {{SLFOR}\left( i \right) \geq \varepsilon_{3}} \\ {{MW}\left( i \right) \geq \varepsilon_{4}} \\ \end{array} \right.$$ where $I_{\max}$ denotes the maximum number of iterations; $\varepsilon_{k},k \in \left\lbrack {0,4} \right\rbrack$ is the prescribed tolerance; and ${PSL}\left( i \right)$, $\underset{a\rightarrow b}{ISL}\left( i \right)$, ${SLFOR}\left( i \right)$, and ${MW}\left( i \right)$ denote the respective spectral characteristics of $\mathbf{w}_{i}$ in the i-th iteration. The first two criteria in ([15](#FD15-sensors-18-03081){ref-type="disp-formula"}) are regular iteration stop conditions, where $I_{\max}$ constrains the maximum number of iterations and $\varepsilon_{0}$ denotes the convergence condition. $\varepsilon_{0}$ is set to a very small number that denotes the condition of convergence, and trade-offs can be achieved by setting $\varepsilon_{1}$ to $\varepsilon_{4}$. The remaining four criteria can be designed in any combination to control the spectral characteristics of the generated window function according to the requirement. Typically, $\varepsilon_{0}$ is set as low as possible, and designers must then balance the PSL, SLFOR, MW, and steerable sidelobes by setting $\varepsilon_{1}$ to $\varepsilon_{4}$. 3. Simulations and Experiments {#sec3-sensors-18-03081} ============================== In this section, we present some simulations and experiments, which are divided into four main parts, to show the effectiveness, flexibility, versatility and practicability of the proposed method. 3.1. Controllable Spectral Characteristics {#sec3dot1-sensors-18-03081} ------------------------------------------ In this subsection, we generate several window functions using the proposed ISO-CA based on the rectangular window function to compare the spectral characteristics with those of some frequently used windows. To suppress all the sidelobes, let $\Gamma\left( q \right) = 0$. The interval $q \in \left\lbrack {a,b} \right\rbrack$ of the mask vector $\Gamma$ in ([7](#FD7-sensors-18-03081){ref-type="disp-formula"}) should contain the entire sidelobe region and should be adjusted adaptively in each iteration because of the continually widening MW. The abscissas of the figures in the time domain and frequency domain are discrete samples and the normalized frequency, respectively. Eight CA window functions with different designed spectral characteristics are presented in [Figure 1](#sensors-18-03081-f001){ref-type="fig"} and [Figure 2](#sensors-18-03081-f002){ref-type="fig"}; the length of each window function is 128 discrete samples. [Figure 1](#sensors-18-03081-f001){ref-type="fig"}a,b shows that trade-offs between the MW and PSL can be made. To better compare the properties of different window functions, the spectral characteristics of the generated CA window functions and other commonly used windows are listed in [Table 1](#sensors-18-03081-t001){ref-type="table"}. Comparing $CA_{1}$ to $CA_{4}$ with the traditional window functions, when the MWs are the same, the proposed windows provide smaller PSLs. It is not difficult to infer that if the PSLs are the same, the MWs of the proposed windows must be narrower. Additionally, note that the Chebyshev window has the smallest PSL for a given MW with the SLFOR equal to zero \[[@B1-sensors-18-03081]\]. Thus, the proposed windows have lower sidelobes over distant regions, although the PSL is slightly higher than that of the Chebyshev window. With the fall-off rate of the mask vector $\Gamma$ controlled, four window functions in the frequency domain with a fixed PSL (−20 dB) and different SLFORs are presented in [Figure 2](#sensors-18-03081-f002){ref-type="fig"}a,b. [Figure 1](#sensors-18-03081-f001){ref-type="fig"} and [Figure 2](#sensors-18-03081-f002){ref-type="fig"} show that the spectral characteristics of windows can be controlled flexibly and effectively using the proposed method. In summary, in addition to the flexibility of the proposed method, the window functions generated by the proposed methods yield better results overall, achieving a narrow MW, a small PSL, and a controllable SLFOR. 3.2. Steerable Sidelobes {#sec3dot2-sensors-18-03081} ------------------------ We know that not all the sidelobes need to be suppressed in some applications if prior knowledge is acquired in advance. Thus, steerable sidelobes are needed. As mentioned in the Introduction, several methods for designing window functions with steerable sidelobes have been proposed. Steerable sidelobes were first proposed by Tseng et al. \[[@B22-sensors-18-03081],[@B23-sensors-18-03081]\], for which the IDFT is needed to evaluate the temporal weights. Zhong et al. \[[@B24-sensors-18-03081]\] evaluated explicit expressions of windows with steerable sidelobes using the sum of the cosine terms. However, the improvement in the SLFOR was not sufficiently investigated mathematically. In \[[@B25-sensors-18-03081]\], Magdy combined the objectives of a high SLFOR and many steerable sidelobe dips but seemingly did not consider the PSL and MW. In this paper, steerable sidelobes are easily obtained by making dips in the mask vector. Other spectral characteristics can also be effectively controlled by designing the stop criterion of the iteration. To demonstrate the performance of the proposed method, four typical CA windows, i.e., $CA_{1}$, $CA_{2}$, $CA_{5}$, and $CA_{6}$, are chosen as the initial windows of the algorithm. Windows with steerable sidelobes are presented in [Figure 3](#sensors-18-03081-f003){ref-type="fig"}. The intervals $q \in \left\lbrack {0.5,0.7} \right\rbrack \cup \left\lbrack {1.0,1.2} \right\rbrack\left( {{rad}/{sample}} \right)$ are set to be suppressed by setting $\Gamma\left( q \right) = 0$. [Figure 3](#sensors-18-03081-f003){ref-type="fig"} and [Table 2](#sensors-18-03081-t002){ref-type="table"} show that the PSL, MW and SLFOR of these windows are effectively maintained while having steerable sidelobes over designated intervals. Thus, due to the flexibility of the mask vector and the initial windows, we can design new window functions with steerable sidelobe properties in the frequency domain while maintaining the good properties of the initial windows as much as possible. 3.3. Non-Rectangular Applications {#sec3dot3-sensors-18-03081} --------------------------------- As stated in the Introduction, window functions should be specially designed according to specific applications. For instance, the well-known LFM waveform is widely used in radar and sonar systems. However, the rectangle-like spectrum of the LFM yields sinc-like range sidelobes after pulse compression \[[@B29-sensors-18-03081]\]. Thus, windows need to be applied to the kernel of the MF to suppress the range sidelobes. When the spectrum of the LFM is chosen as the initial window, the proposed method, in which the MW, PSL, SLFOR, and steerable sidelobes can be designed, provides more design freedom to control the range sidelobes. Assuming that the radar-transmitted waveform is an LFM signal, we obtain the echoes of three aerial targets with different radar cross sections (RCSs), the approximate locations of which are known, and additive Gaussian noise is added, the input SNR of which is 30 dB with respect to the strongest echo power. The relative distance between $Target_{0}$ and $Target_{1}$ is approximately 50 bins, and $Target_{2}$ is approximately 300 bins away from $Target_{0}$. However, two of the targets have relatively low RCSs and are covered by the sidelobes of the strong one when using the rectangular window. Thus, we can generate a window with a low sidelobe level over these regions by using the proposed method. The characteristics of the spectral windows in the time domain are shown in [Figure 4](#sensors-18-03081-f004){ref-type="fig"}. The original properties of the Hamming and Chebyshev windows are shown to have changed, while the sidelobes of the Chebyshev window function change such that they are no longer equal. In addition, we still cannot see $Target_{1}$ and $Target_{2}$ in [Figure 4](#sensors-18-03081-f004){ref-type="fig"}b using the two traditional windows. However, the window generated using the proposed methods considers the shape of the LFM spectrum. It has low ISRs over the specified regions with a narrow MW. Thus, the three targets can be detected clearly and accurately, as shown in [Figure 4](#sensors-18-03081-f004){ref-type="fig"}b. 3.4. ISAR Experiments {#sec3dot4-sensors-18-03081} --------------------- To test the practicability of the proposed method, experiments on some real measured data of a Yak-42 airplane, recorded using a C-band (5.52 GHz) inverse synthetic aperture radar (ISAR) experimental system, are conducted. This system transmits 400 MHz chirp signals with a 25.6 $\mathsf{\mu}$s pulse width, and the target's echoes are dechirped and I/Q-sampled at 10 MHz \[[@B30-sensors-18-03081],[@B31-sensors-18-03081]\]. A graph of Yak-42 is presented in [Figure 5](#sensors-18-03081-f005){ref-type="fig"}a. The FFT is needed for both the range and azimuth focusing to obtain the ISAR image. However, the strong scattering points have high sidelobe levels, and some weak scattering points near the strong ones are easily masked when using a rectangular window. This problem is usually solved by adding Hamming, Chebyshev or other traditional windows to suppress the sidelobes, as shown in [Figure 5](#sensors-18-03081-f005){ref-type="fig"}b, but the effects are not obvious. The traditional window functions have a wide MW and high PSL, as mentioned above, which reduce the ISAR image qualities, including the resolution and the contrast. Considering the proposed methods in this paper, the high PSL should first be suppressed, and the MW should be made as narrow as possible. Then, the wings, head and tail of the plane may be covered by the sidelobes of the strong scatters on the plane body. Thus, we can design the mask vector according to the size of the Yak-42 plane in [Figure 5](#sensors-18-03081-f005){ref-type="fig"}a to suppress the sidelobes over the peak sidelobe region and relative regions among the body, wings, head, and tail of the plane. More scattering points of the plane can be seen by adding the generated windows, as shown in [Figure 5](#sensors-18-03081-f005){ref-type="fig"}b. The image qualities of the ISAR images are also listed in [Table 3](#sensors-18-03081-t003){ref-type="table"}. Here, we use the MW to represent the resolutions. A lower value denotes a better resolution. [Table 3](#sensors-18-03081-t003){ref-type="table"} reveals that the window function generated using the proposed methods yields the best ISAR image result, with a good resolution and high image contrast. 4. Results and Discussion {#sec4-sensors-18-03081} ========================= [Section 3](#sec3-sensors-18-03081){ref-type="sec"} presents several simulations and experiments to show the advantages of the proposed ISO-CA including the effectiveness, flexibility, versatility and practicability. First, all the spectral characteristics of the generated window functions, i.e., PSL, MW, and SLFOR, can be controlled. $CA_{1}$ to $CA_{4}$ are the generated windows with different MWs and PSLs. [Table 1](#sensors-18-03081-t001){ref-type="table"} shows that the properties of these windows are superior to those of traditional windows. For instance, the MWs of Hanning, Bartlett-Hanning, and $CA_{2}$ are the same, but the PSL of $CA_{2}$ is the smallest. Many similar examples are found in [Table 1](#sensors-18-03081-t001){ref-type="table"}. Conversely, it can be reasonably inferred that the MW of the CA window is narrower than those of traditional windows when the PSLs are the same. $CA_{5}$ to $CA_{8}$ are the generated windows with a fixed PSL and different SLFORs. The controllable spectral characteristics of the generated windows present the effectiveness of the ISO-CA method. In the meantime, on that basis, steerable sidelobes can be easily added; [Figure 3](#sensors-18-03081-f003){ref-type="fig"} presents the results if specified regions of the sidelobes should also be suppressed, which shows the flexibility. The versatility is presented by an example in the radar-signal processing field. As stated above, traditional windows do not apply well when the input is non-rectangular like the spectra of the LFM signal. The original spectral characteristics are changed, as shown in [Figure 4](#sensors-18-03081-f004){ref-type="fig"}a. However, the input of the ISO-CA method is defined by designers. Hence, the performance is better than that of the commonly used windows. The shadowing effect between strong and weak targets, which is common in remote sensing sensors, can also be eliminated efficiently by adding steerable sidelobes. To evaluate the practicability, we used measured ISAR data to test the generated CA windows. The ISAR images in [Figure 5](#sensors-18-03081-f005){ref-type="fig"}b show that, compared with the rectangular window, more scattering points of the wings and head of the plane can be seen by adding the Hamming, Chebyshev, and ISO-CA windows. However, the scattering points of the first two windows melt together due to the large MW, which is difficult for further recognition. However, the MW of ISO-CA is close to that of the rectangular window, and steerable sidelobes decrease the shadowing effect of strong scattering points in the plane body. Thus, the ISAR image quality of the ISO-CA, including the resolution and the contrast, is the best. A window function is a basic signal processing tool that is needed in many sensor signal processing fields such as radar/sonar signal processing. Applications of window functions include spectral analysis/modification/re-synthesis, the design of finite impulse response filters, and beamforming and antenna design. Here, a simple demonstration for antenna design is given in [Figure 6](#sensors-18-03081-f006){ref-type="fig"}. We use the ISO-CA method to suppress the sidelobes within $30^{\circ}$ to $45^{\circ}$ and $- 45^{\circ}$ to $- 30^{\circ}$ for anti-interference or anti-clutter and at the same time maintain the mainlobe as narrow as possible. Designers can also use the ISO-CA method in other sensor signal processing fields, and we do not list them all. The proposed window functions are substantially generated by solving an optimization problem. The core signal processors of modern sensors are generally the digital signal processor (DSP) and field programmable gate array (FPGA), and the FFT and IFFT are the necessary modules. Since the proposed ISO-CA method for solving the optimization problem contains only the FFT, IFFT, and simple multiplication, it is easy to implement in the DSP or FPGA. In addition, the linear convergence rate guarantees the reliability. 5. Conclusions {#sec5-sensors-18-03081} ============== In this paper, a novel method, i.e., the ISO-CA, for designing general window functions with flexible spectral characteristics is proposed. The matrix multiplication and left inverse calculation in the ISO-CA can be replaced with the FFT and IFFT to improve the efficiency of the iteration. Simulations and experiments show that the generated CA window functions yield better results overall than those of traditional windows and that steerable sidelobes can be acquired both easily and flexibly by designing an appropriate mask vector and stopping criterion. Meanwhile, the favourable properties of the initial windows can be effectively maintained using the proposed methods. Thus, the variation in the initial windows provides more design freedom for specific applications. In addition to the flexibility of ISO-CA, simulations show that the window functions generated by the proposed methods yield better results overall, achieving a narrow MW, a small PSL, and a controllable SLFOR. Experiments based on Yak-42 ISAR real measured data show the practicability of the method, with the ISAR image qualities being greatly improved using the generated window functions. Although the FFT and IFFT can improve the efficiency of the iteration and the recursion formula is also given in the [Appendix](#app1-sensors-18-03081){ref-type="app"}, a simple closed-form solution has not been derived so far. The computational complexity will be slightly high if many windows with different lengths are needed at the same time because the length of the window is fixed during the iteration. Generalizing the iteration solution to different lengths is also an interesting topic, and this will be our future work. Q.L. and T.L. provided the conceptualization; Y.S. and Q.L. performed the model derivation and iteration analysis; Y.S. and J.C. proposed the design algorithm; Y.S. and J.C. performed the simulations and experiments; Q.L. and T.L. validated the algorithm, simulations and experiments; Y.S. wrote the original draft. This research was funded by the 111 Project of China (Grant No. B14010), the Chang Jiang Scholars Programme (Grant No. T2012122), and the National Natural Science Foundation of China (Grant No. 61771050). The authors declare no conflict of interest. The following abbreviations are used in this manuscript: MWMainlobe WidthPSLPeak Sidelobe LevelSLFORSidelobe Fall-Off RateISO-CAInverse of the Shaped Output using Cyclic AlgorithmCACyclic AlgorithmLFMLinear Frequency ModulationMFMatched FilterDFT/IDFTDiscrete Fourier Transform/Inverse DFTFFT/IFFTFast Fourier Transform/Inverse FFTRCSRadar Cross SectionISARInverse Synthetic Aperture RadarDSPDigital Signal ProcessorFPGAfield programmable gate arraySVDSingular Value Decomposition To analyze the convergence of the ISO-CA, the relationship between $\mathbf{w}_{i}$ and $\mathbf{w}_{i + 1}$ is derived. Here, we assume that $\Gamma$ is fixed during the iteration to simplify the derivation, with it being easy to generalize when $\Gamma$ changes. The M-point DFT of $\mathbf{w}_{i}$ is:$$\mathbf{W}_{i} = \mathbf{F}_{MN}\mathbf{w}_{i}.$$ Then, we have:$$\mathbf{W}_{\Gamma} = \Gamma \odot \mathbf{W}_{i}.$$ According to the properties of the DFT, the product of two vectors in the frequency domain is equivalent to their circular convolution in the time domain. Thus, we can rewrite ([A2](#FD17-sensors-18-03081){ref-type="disp-formula"}) as:$$\mathbf{W}_{\Gamma} = \mathbf{F}_{M}\mathbf{C}^{\gamma}\begin{bmatrix} \mathbf{w}_{i} \\ 0 \\ \end{bmatrix}_{M \times 1},$$ where:$$\mathbf{C}^{\gamma} = \begin{bmatrix} {\gamma\left( 0 \right)} & {\gamma\left( {M - 1} \right)} & \cdots & {\gamma\left( 1 \right)} \\ {\gamma\left( 1 \right)} & {\gamma\left( 0 \right)} & \cdots & {\gamma\left( 2 \right)} \\ \vdots & \vdots & \ddots & \vdots \\ {\gamma\left( {M - 1} \right)} & {\gamma\left( {M - 2} \right)} & \cdots & {\gamma\left( 0 \right)} \\ \end{bmatrix}$$ and $$\left\lbrack {\gamma\left( 0 \right),\gamma\left( 1 \right),\cdots,\gamma\left( {M - 1} \right)} \right\rbrack^{T} = \mathbf{F}_{M}^{- 1}\Gamma.$$ Then, we have:$$\begin{matrix} {\mathbf{w}_{i + 1} = \mathbf{F}_{MN}^{\dagger}\mathbf{F}_{M}\mathbf{C}^{\gamma}\begin{bmatrix} \mathbf{w}_{i} \\ 0 \\ \end{bmatrix} = \frac{1}{M}\mathbf{F}_{MN}^{H}\mathbf{F}_{M}\mathbf{C}^{\gamma}\begin{bmatrix} \mathbf{w}_{i} \\ 0 \\ \end{bmatrix}} \\ {= \frac{1}{M}\left\lbrack {M \cdot \mathbf{I}_{NN},0} \right\rbrack\begin{bmatrix} \mathbf{C}_{NN}^{\gamma} & \cdots \\ \vdots & \ddots \\ \end{bmatrix}\begin{bmatrix} \mathbf{w}_{i} \\ 0 \\ \end{bmatrix}} \\ {= \mathbf{C}_{NN}^{\gamma}\mathbf{w}_{i}} \\ \end{matrix}.$$ where $\mathbf{C}_{NN}^{\gamma}$ is the first N rows and N columns of $\mathbf{C}^{\gamma}$. Assuming that we can obtain the expected spectral characteristics after x iterations, we have:$$\mathbf{w}_{x} = \left( \mathbf{C}_{NN}^{\gamma} \right)^{x}\mathbf{w}_{0},$$ where $\mathbf{w}_{0}$ is the initial window function. We can rewrite ([A7](#FD22-sensors-18-03081){ref-type="disp-formula"}) as:$$\mathbf{w}_{x} = \left( {\mathbf{U}_{C}\Sigma_{C}\mathbf{V}_{C}^{H}} \right)^{x}\mathbf{w}_{0} = \mathbf{U}_{C}\Sigma_{C}^{x}\mathbf{U}_{C}^{H}\mathbf{w}_{0},$$ where $\mathbf{C}_{NN}^{\gamma} = \mathbf{U}_{C}\Sigma_{C}\mathbf{V}_{C}^{H}$ is the singular value decomposition (SVD) and $\mathbf{U}_{C} = \mathbf{V}_{C}$ because $\mathbf{C}_{NN}^{\gamma} = \left( \mathbf{C}_{NN}^{\gamma} \right)^{H}$ according to ([A4](#FD19-sensors-18-03081){ref-type="disp-formula"}) and ([A5](#FD20-sensors-18-03081){ref-type="disp-formula"}). Then, the rate of convergence is defined as:$$\begin{matrix} {\beta = \frac{{\parallel {\mathbf{w}_{i + 1}\mspace{600mu} - \mspace{600mu}\mathbf{w}_{x}} \parallel}_{2}}{{\parallel {\mathbf{w}_{i}\mspace{600mu} - \mspace{600mu}\mathbf{w}_{x}} \parallel}_{2}}} \\ {= \frac{{\parallel {\mathbf{U}_{C}\Sigma_{C}^{i}\left( {\Sigma_{C}\mspace{600mu} - \mspace{600mu}\Sigma_{C}^{x - i}} \right)\mathbf{U}_{C}^{H}\mathbf{w}_{0}} \parallel}_{2}}{{\parallel {\mathbf{U}_{C}\Sigma_{C}^{i}\left( {\mathbf{I}_{NN}\mspace{600mu} - \mspace{600mu}\Sigma_{C}^{x - i}} \right)\mathbf{U}_{C}^{H}\mathbf{w}_{0}} \parallel}_{2}}} \\ {= \frac{\mathbf{w}_{0}^{H}\mathbf{U}_{C}\left( {\Sigma_{C}\mspace{600mu} - \mspace{600mu}\Sigma_{C}^{x - i}} \right)^{2}\Sigma_{C}^{2i}\mathbf{U}_{C}^{H}\mathbf{w}_{0}}{\mathbf{w}_{0}^{H}\mathbf{U}_{C}\left( {\mathbf{I}_{NN}\mspace{600mu} - \mspace{600mu}\Sigma_{C}^{x - i}} \right)^{2}\Sigma_{C}^{2i}\mathbf{U}_{C}^{H}\mathbf{w}_{0}}} \\ \end{matrix}$$ According to the interlacing theorem for singular values, we obtain:$$\sigma_{C} \leq \sum\limits_{n = 0}^{M - 1}{\gamma\left( n \right)} = \Gamma\left( 0 \right) = 1,$$ where $\sigma_{C}$ denotes the eigenvalue of $\Sigma_{C}$. Considering the matrix from ([A9](#FD24-sensors-18-03081){ref-type="disp-formula"}):$$\begin{array}{l} {\left( {\mathbf{I}_{NN} - \Sigma_{C}^{x - i}} \right)^{- 2}\Sigma_{C}^{- 2i}\left( {\Sigma_{C} - \Sigma_{C}^{x - i}} \right)^{2}\Sigma_{C}^{2i}} \\ {= \left( {\mathbf{I}_{NN} - \Sigma_{C}^{x - i}} \right)^{- 2}\left( {\Sigma_{C} - \Sigma_{C}^{x - i}} \right)^{2}} \\ \end{array}$$ the eigenvalue of matrix ([A11](#FD26-sensors-18-03081){ref-type="disp-formula"}) can be expressed as $$\sigma_{\beta} = \frac{\left( {\sigma_{C} - \sigma_{C}^{x - i}} \right)^{2}}{\left( {1 - \sigma_{C}^{x - i}} \right)^{2}},\sigma_{C} \leq 1,$$ By inference, when the items of the mask vector $\Gamma$ are all ones, we obtain $\sigma_{C} = 1$ and thus $\sigma_{\beta} = 1$. In this case, $\mathbf{w}_{i + 1} \equiv \mathbf{w}_{i}$. Otherwise, $0 < \sigma_{\beta} < 1$. According to the generalized Rayleigh quotient, we can obtain from ([A9](#FD24-sensors-18-03081){ref-type="disp-formula"}) and ([A12](#FD27-sensors-18-03081){ref-type="disp-formula"}) that:$$0 < \beta < 1$$ Thus, the iteration of the ISO-CA has linear convergence rates \[[@B32-sensors-18-03081]\]. ![Windows with different MWs and PSLs (a) in the time domain and (b) in the frequency domain.](sensors-18-03081-g001){#sensors-18-03081-f001} ![Windows with different SLFORs (a) in the time domain and (b) in the frequency domain.](sensors-18-03081-g002){#sensors-18-03081-f002} ![Window functions with steerable sidelobes.](sensors-18-03081-g003){#sensors-18-03081-f003} ![(a) Properties of different windows based on a non-rectangular input; (b) output of the MF for different window functions.](sensors-18-03081-g004){#sensors-18-03081-f004} ![(a) Graph of Yak-42; (b) ISAR images obtained using different windows (contour plot).](sensors-18-03081-g005){#sensors-18-03081-f005} ![Antenna pattern with different weights.](sensors-18-03081-g006){#sensors-18-03081-f006} sensors-18-03081-t001_Table 1 ###### Property Comparison. Window Parameter MW ($\mathbf{\times \mathbf{\pi}}$ rad/Sample) PSL (dB) SLFOR (dB/oct) ------------------ ---------------- ------------------------------------------------ ---------- ---------------- Rectangular \- 0.0137 −13.3 6 Triangular \- 0.0195 −26.5 12 Hamming \- 0.0195 −42.6 6 Hanning \- 0.0215 −31.5 18 Bartlett-Hanning \- 0.0215 −35.9 12 Bohman \- 0.0254 −46.0 24 Blackman \- 0.0254 −58.1 18 Chebyshev \- 0.0254 −80.0 0 Kaiser $\beta = 2.0$ 0.0137 −18.5 6 $\beta = 3.0$ 0.0156 −24.0 $\beta = 4.0$ 0.0176 −30.3 Tukey $r = 0.25$ 0.0156 −13.6 18 $r = 0.50$ 0.0176 −15.1 $r = 0.75$ 0.0195 −19.4 Gaussian $\alpha = 2.5$ 0.0215 −43.7 6 $\alpha = 3.0$ 0.0234 −56.7 $\alpha = 3.5$ 0.0273 −71.8 $CA_{1}$ \- 0.0176 −34.0 6 $CA_{2}$ \- 0.0215 −49.2 $CA_{3}$ \- 0.0234 −62.6 $CA_{4}$ \- 0.0254 −74.2 $CA_{5}$ \- 0.0156 −19.9 6 $CA_{6}$ \- 0.0156 −19.9 12 $CA_{7}$ \- 0.0176 −19.9 18 $CA_{8}$ \- 0.0195 −19.9 24 sensors-18-03081-t002_Table 2 ###### Property Comparison (steerable sidelobes). Window (Steerable Sidelobes) $\mathbf{\mathbf{CA}_{1}}$ $\mathbf{\mathbf{CA}_{4}}$ $\mathbf{\mathbf{CA}_{5}}$ $\mathbf{\mathbf{CA}_{6}}$ ----------------------------------------------------- ---------------------------- ---------------------------- ---------------------------- ---------------------------- MW ($\times \mathsf{\pi}\mspace{720mu}$ rad/sample) 0.0176 0.0215 0.0156 0.0157 PSL (dB) −31.7 −46.8 $- 19.3$ $- 19.8$ ISL (interval 1, dB) −75.6 −86.8 −66.1 −76.9 ISL (interval 2, dB) −81.8 −92.1 −72.2 −82.3 SLFOR (dB/oct) 6 6 6 12 sensors-18-03081-t003_Table 3 ###### Image Quality Comparison. Image Quality Rectangular Hamming Chebyshev ISO-CA -------------------- ------------- --------- ----------- -------- Range Resolution 8.07 11.42 15.73 10.47 Azimuth Resolution 5.39 7.49 10.18 6.46 Contrast 4.01 3.79 4.08 5.26	{ "pile_set_name": "PubMed Central" }
Introduction ============ Rational design and controlled synthesis of perovskite materials have been extensively studied in recent years due to their tunable band gaps, superior charge-transfer properties, and potential applications in optoelectronic fields such as solar cells,[@cit1] photocatalysis,[@cit2] photodetectors,[@cit3] light-emitting diodes (LEDs),[@cit4] optically pumped lasers,[@cit5]etc.[@cit6],[@cit7] Additionally, the high optical gain properties of these perovskite materials can extend their use as optical and photonic devices.[@cit8] To date, several challenges remain in the development of organic--inorganic hybrid perovskite systems. For example, their water stability is still relatively low, which has largely restricted their practical applications. On the other hand, due to their quantum-size effects compared with bulk materials, great attention has been focused on the fabrication of micro-/nano-scale perovskite crystals.[@cit9] However, both facile and large-scale synthesis of high-quality low-dimensional perovskites with novel photo-electronic performance is still relatively limited. Generally, the optical properties of molecular materials are governed by pure chromophores (single molecule states) and their arrangement/packing fashion (aggregate states). In this context, the organic--inorganic hybrid perovskite structure is an idealized host--guest system for the formation of highly ordered chromophores and the uniform orientation of transition dipole moments for the construction of micro-/nano-scale optical devices. In such hybrid perovskite materials, organic cations play an important role in the light harvesting process. Herein, a rigid lead chloride inorganic matrix was selected as the typical 2D perovskite host layer; the acridine (AD) molecule was selected as the guest unit due to its excellent optical properties and wide applications for coloration and cell cycle determination.[@cit10] The AD molecule exhibits several polymorphic forms and has a tendency to be protonated in the presence of organic/inorganic acids.[@cit11] In addition, strong electrostatic and hydrogen bond interactions exist between the host--guest components. The larger steric hindrance of AD cations distributed between the rigid lead chloride inorganic matrixes can effectively prevent water erosion and maintain long-term stability under humid conditions. In this work, it is demonstrated that a one-dimensional (1D) single-crystalline perovskite micro-belt can be immediately isolated using a mixture of AD and lead chloride in an acid aqueous solution ([Scheme 1](#sch1){ref-type="fig"}). This is achieved without any organic solvent or expensive alkyl halide. The obtained organic--inorganic hybrid perovskite material \[(AD)Pb~2~Cl~5~\] (OIHP-AD) can retain its emission for months in an aqueous suspension, thus overcoming the typical instability of hybrid perovskite materials in water. Moreover, the OIHP-AD micro-belt, which possesses high optical gain, can serve as an efficient 1D waveguide material with a low waveguide loss coefficient. Therefore, this work demonstrates a facile, fast, and easily scalable synthesis for 1D organic--inorganic hybrid perovskite photonic materials with high water stability, which may have potential as micro-devices for optical and photonic applications. ![The facile synthesis process of OIHP-AD perovskite crystals in aqueous solution at room temperature.](c9sc00162j-s1){#sch1} Results and discussion ====================== Single crystals suitable for X-ray analysis (Fig. S1 in the ESI[†](#fn1){ref-type="fn"}) were obtained using a mixture of AD and PbCl~2~ (1 : 2 molar ratio) in water with addition of hydrochloric acid under hydrothermal conditions and heated at 120 °C for 12 h. Single-crystal X-ray diffraction analysis revealed that \[(AD)Pb~2~Cl~5~\] crystallizes in a triclinic system with the space group P1\[combining macron\] (Table S1[†](#fn1){ref-type="fn"}), which is slightly different from those of typically reported organic--inorganic hybrid perovskite structures.[@cit4] The asymmetric unit of OIHP-AD consists of one crystallographically independent Pb ion, two half chlorine ions and half a protonated AD cation. As shown in [Fig. 1](#fig1){ref-type="fig"}, the crystal structure of OIHP-AD exhibits an alternating arrangement of the organic AD cation layers and the inorganic anion layers constructed by corner-sharing (PbCl~6~) octahedra. Each organic AD layer derived from the (001) crystallographic plane is fixed between adjacent inorganic layers through electrostatic, C--H···Cl and N--H···Cl hydrogen bonds. In addition, π···π stacking between the parallel phenyl and pyridyl rings of AD in an offset fashion with centroid--centroid separations of 3.674 Å can be observed. In such an organic--inorganic hybrid system, the rigid lead chloride inorganic layer can afford a confined and ordered environment, minimizing the aggregation-caused quenching and increasing the optical/thermal stability of AD chromophores. Alternatively, the micro-belt structure of the OIHP-AD (Fig. S2[†](#fn1){ref-type="fn"}) can be obtained by a facile precipitation process in an aqueous solution: solutions of AD and PbCl~2~ were separately dissolved with hydrochloric acid and mixed together with stirring. As a result, it could be easily scaled up to the gram level for the 1D OIHP-AD nanocrystals (see the Experimental section, ESI[†](#fn1){ref-type="fn"}). Powder X-ray diffraction (PXRD) patterns for the samples isolated under hydrothermal conditions and room temperature correspond well with the simulated data, which illustrates the single phase and high purity of OIHP-AD (Fig. S3 and S4[†](#fn1){ref-type="fn"}). The difference in the intensity ratio between some PXRD peaks can be attributed to the preferred orientation of the samples during the measurement. Additionally, the PXRD peaks for the sample obtained under hydrothermal conditions present narrower and stronger peaks than those from room temperature, suggesting a relatively low crystalline degree for the latter. ![Ball-and-stick/polyhedron view of the OIHP-AD perovskite crystal structure with an alternate arrangement of the inorganic layers and the organic AD layers.](c9sc00162j-f1){#fig1} The SEM image ([Fig. 2a](#fig2){ref-type="fig"}) shows that the OIHP-AD sample at room temperature displays a stronger tendency to assemble into a 1D belt-like structure with the length and width in the ranges of 150--500 μm and 1--4 μm, respectively. Due to the high length/diameter aspect ratio of these 1D micro-belts (Fig. S5[†](#fn1){ref-type="fn"}), a bright yellow self-supporting OIHP-AD film of several centimeters can be easily achieved ([Fig. 2b](#fig2){ref-type="fig"}). Under UV irradiation (λ~ex~ = 365 nm), the OIHP-AD film exhibits a bright green emission arising from the highly orientated 1D micro-belt ensemble ([Fig. 2c](#fig2){ref-type="fig"} and S6[†](#fn1){ref-type="fn"}). To characterize the optical properties of the OIHP-AD films, their UV-vis-NIR absorption, photoluminescence (PL) spectra, PL quantum yield (PLQY) and lifetime were further measured. [Fig. 2d](#fig2){ref-type="fig"} presents the UV-vis-NIR absorption and emission spectra of the OIHP-AD film, which displays strong absorption ranging from 200 to 500 nm and a broad absorption peak at 857 nm in the NIR region. However, the pristine AD only exhibits sharp absorption ranging from 200 to 450 nm (Fig. S7a[†](#fn1){ref-type="fn"}). ![(a) SEM image of the 1D OIHP-AD micro-belts. Photographs of the self-supporting OIHP-AD film (about 0.8 × 3 cm^2^) under (b) daylight and (c) UV (365 nm). (d) UV-vis-NIR absorption (black) and fluorescence emission spectra (red, excitation at 400 nm) of OIHP-AD in the solid state at room temperature. (e) Time-resolved fluorescence decay of OIHP-AD in the solid state.](c9sc00162j-f2){#fig2} Under ambient conditions, the solid-state luminescence spectrum of AD presents weak blue fluorescence emission at 451 nm (λ~ex~ = 330 nm, Fig. S7a[†](#fn1){ref-type="fn"}) with a PLQY of 1.43%. This can be ascribed to a strong fluorescence quenching effect between the AD molecules. In contrast, the hybrid 1D OIHP-AD exhibits longer emission wavelength (maxima at 533 nm, λ~ex~ = 400 nm) and higher PLQY (7.45%) compared with the pristine AD. The distinct color change from blue to green (red-shift of 82 nm) can be assigned to π···π stacking interactions between adjacent AD moieties together with C--H···Cl and N--H···Cl hydrogen bonds between heterogeneous layers. Moreover, time-resolved fluorescence decay gives a fluorescence lifetime of 2.87 ns for AD and 3.81 ns for OIHP-AD ([Fig. 2e](#fig2){ref-type="fig"} and S7b[†](#fn1){ref-type="fn"}). The prolonged fluorescence of the OIHP-AD hybrid can be related to the fact that the AD dye is highly restricted and stabilized within the negatively charged inorganic matrix. Benefitting from the confinement effect of the inorganic layers, the molecular thermal vibration and nonradiative relaxation process can be effectively suppressed. The serious aggregation-caused quenching of the chromophores can also be minimized due to the isolation of the OIHP layers. To better understand the optical properties of OIHP-AD, the band structure, density of states (DOS) and electron-density distribution were determined using periodic density functional theory (DFT) calculations.[@cit12] The computation shows that OIHP-AD (Fig. S8[†](#fn1){ref-type="fn"}) exhibits a low band gap of 2.06 eV, which may correspond to the fluorescence emission around 533 nm (2.33 eV). Frontier orbital analysis further shows that the electron densities of the highest occupied molecular orbital (HOMO) are predominantly localized on the lead chloride inorganic layer, while the electron-density distributions of the lowest unoccupied molecular orbital (LUMO) are observed on the organic AD cation layers ([Fig. 3a and b](#fig3){ref-type="fig"}), suggesting potential energy/electron transfer during the photoexcitation process. ![Molecular orbitals of the HOMO (a) and LUMO (b) for OIHP-AD. (c) Total/partial electronic density of states (TDOS and PDOS) for the DFT optimized structure of OIHP-AD.](c9sc00162j-f3){#fig3} Further total electronic density of states (TDOS) and partial electronic density of states (PDOS) analyses ([Fig. 3c](#fig3){ref-type="fig"}) reveal that the valence bands (VB) are mainly contributed by the 3p orbitals of Cl and 2p orbitals of C atoms in AD, whereas the conduction bands (CB) are mainly derived from the 6p orbitals of Pb. Therefore, the photoexcitation and emission of OIHP-AD system can be obtained from the lead chloride inorganic layer to the AD organic layers during fluorescence process. Such an electronic structure and photo-physical properties of OIHP-AD may improve electron--hole separation during the charge transfer process, which could correspond to the enhanced PL lifetime in experiments. As is typical for hybrid perovskite materials, physicochemical stability is a critical issue that restricts their practical application. Among various factors, temperature, moisture and water are considered as major challenges. Thermal gravimetric analysis (TGA) reveals that the as-prepared OIHP-AD remains stable upon heating to higher temperatures (ca. 230 °C) compared to the pure AD (111 °C). The weight loss of 25.17% observed from 190 to 300 °C is very close to the calculated value of 23.34%, which corresponds to the AD cation component. With further heating, the remaining substance decomposes quickly until heating to 800 °C (Fig. S9[†](#fn1){ref-type="fn"}). [Fig. 4a](#fig4){ref-type="fig"} presents the temperature-dependent PL spectra of OIHP-AD. It was observed that the intensity at 533 nm exhibits a slightly decreasing trend from ambient conditions (293 K) to 373 K and decreases obviously upon further increase in temperature from 373 to 473 K. It is worth noting that the PL intensity of OIHP-AD can recover its original position ([Fig. 4b](#fig4){ref-type="fig"}) after returning to the initial temperature, confirming the stability of the temperature-dependent luminescence. Moreover, the reversible change in PL can be repeated at least 6 times during heating--cooling cycles between 293 and 373 K, indicating that the 1D OIHP-AD can potentially serve as a luminescent switch for sensing the temperature. ![(a) PL spectra of OIHP-AD measured at temperatures from 293 to 473 K. (b) Reversible variation of the PL intensity of OIHP-AD between 293 and 373 K.](c9sc00162j-f4){#fig4} The measurement of the moisture stability shows that the OIHP-AD self-supporting film can remain unchanged after exposure to a steam atmosphere (90% relative humidity) for 24 h (Fig. S10[†](#fn1){ref-type="fn"}). To check the water stability of OIHP-AD, 100 mg of powdered OIHP-AD was added to 20 mL deionized water at room temperature, giving rise to a yellowish suspension ([Fig. 5a](#fig5){ref-type="fig"}), which emits a strong cyan light under UV (365 nm) irradiation ([Fig. 5b](#fig5){ref-type="fig"}). As shown in [Fig. 5c](#fig5){ref-type="fig"}, the fluorescence emission of the OIHP-AD suspension is mainly located at 477 nm (λ~ex~ = 400 nm), which is distinctly blue-shifted compared with that of its solid state. It is worth noting that after a long aging time (60 days), there is no obvious decrease in the fluorescence intensity. The time-resolved PL measurement shows that the PL lifetime of the OIHP-AD suspension is 15.03 ns ([Fig. 5d](#fig5){ref-type="fig"}); further, the PL quantum yield (PLQY) is 58.79%. After being centrifuged from the suspension, 1D nano-scaled wires can be observed (SEM, [Fig. 5e](#fig5){ref-type="fig"}) with the length of 0.5--1.5 μm and width of 100 nm. The PXRD pattern of these nanowires matches well with the simulated pattern of OIHP-AD, indicating that the crystal structure of OIHP-AD nanowires remains unchanged even after soaking in water for more than 60 days ([Fig. 5f](#fig5){ref-type="fig"}). It is believed that the blue shift and enhanced PLQY of the 1D nano-scale wires relative to the OIHP-AD self-supporting film can be assigned to the deaggregation of the OIHP-AD along with the decrease in the size of the nanocrystals, as the luminescence of the hybrid nanoparticles is highly dependent on the size, shape, and morphology.[@cit7a] Additionally, when the 1D microbelts were soaked in other typical solvents (such as ethanol, acetone, acetonitrile and chloroform), there is nearly no decrease in the length and width of the 1D microbelts. ![Photographs of the as-obtained OIHP-AD nanocrystal aqueous suspension under (a) ambient conditions and (b) UV light (365 nm) irradiation. (c) Stability of OIHP-AD dispersed in water. The inset shows PL spectra measured at different times (excitation at 409 nm). (d) Time-resolved fluorescence decay of OIHP-AD dispersed in water measured at room temperature. SEM image (e) and PXRD patterns (f) of OIHP-AD after being soaked in water for 60 days.](c9sc00162j-f5){#fig5} To date, although numerous efforts have been made to achieve water-stable perovskite micro- and nano-particles with low dimension (0D, 1D and 2D),[@cit4a],[@cit7],[@cit9],[@cit13] most are focused on the external encapsulation of metal oxides,[@cit14],[@cit15] polymers,[@cit16],[@cit17] mesoporous inorganic oxides[@cit18],[@cit19] and metal--organic framework[@cit20] matrixes. Only recently, Han\'s and Yang\'s groups demonstrated that pyridine additives with various alkyl-chain substituents or alkyl ammonium cations on the perovskite surface can substantially improve the stability of perovskite solar cells.[@cit21],[@cit22] Different from these surface modifications of the perovskite mentioned above, in this work, the protonated AD cations can act as counter ions evenly distributed between inorganic laminates. Due to their steric hindrance, the coordination capability of AD decreases and thus restricts the reaction with lead ions. Meanwhile, the AD cations are joined together by strong electrostatic, hydrogen bond and π···π stacking interactions, affording denser crystal packing and effective organic water-resisting layers to prevent corrosion due to water molecules to maintain their long-term stability under humid conditions and in water. Therefore, this work may provide a new strategy for the facile synthesis of long-term stable organic--inorganic hybrid perovskite materials by introducing organic cations with larger steric hindrance. Due to the high physicochemical stability and excellent optical properties of the 1D OIHP-AD micro/nanocrystals, the upconversion, polarized photoemission and optical waveguide performances are highly desirable for new photonic applications. Encouraged by the broad absorption of OIHP-AD in the NIR region, long-wavelength-excited fluorescence emissions were thus further measured. Upon excitation with a femtosecond pulsed laser at 800 nm, the OIHP-AD exhibits short wavelength fluorescence with an emission peak close to the emission bands obtained by excitation with 400 nm light ([Fig. 6a](#fig6){ref-type="fig"}). It can be observed that OIHP-AD displays a nearly linear relationship between the PL intensity and the pump power ([Fig. 6b](#fig6){ref-type="fig"}), indicating that the optical gain can be effectively improved by the incorporation of dye cations into the organic--inorganic hybrid perovskite system. ![(a) Up-conversion fluorescence spectra of OIHP-AD excited using an 800 nm laser under different pump powers. (b) The changes in intensity with increasing pump powers. (c) Polarized fluorescence spectrum of OIHP-AD at various angles (0--360°). (d) Fluorescence intensity of the OIHP-AD as a function of the polarizer rotation angle.](c9sc00162j-f6){#fig6} The highly ordered arrangement of AD cations within the rigid lead chloride inorganic matrix further motivates us to study the polarization properties of the perovskite material. The polarized fluorescence spectrum was recorded at the tip of the single micro-belt by rotating the polarizer at different polarization angles. [Fig. 6c and d](#fig6){ref-type="fig"} present the relationship between the polarization photoemission intensity of a single micro-belt and the polarization angle. It can be found that the maximum photoemission intensity values for the 1D OIHP-AD micro-belt appear at the cross angles of about 0° and 180°, whereas the minimum ones occur at 90° and 270°. The emission dichroic ratio R~d~ and polarization anisotropy σ were measured to be about 2.75 and 0.93, respectively, where R~d~ = I~max~/I~min~ and σ = (R~d~ -- 1)/(R~d~ + 1).[@cit23] The polarization anisotropy value for the OIHP-AD 1D micro-belt is higher than that for most reported 1D micro-/nano-sized systems.[@cit24] These results confirm that the incorporation of dye cations into the perovskite rigid matrix forms a well-oriented and uniformly ordered structure that could potentially act as a polarized luminescent material. When irradiated under unfocused UV light (400 nm) on a PL microscope, the individual OIHP-AD micro-belts show bright green spots on the tips, but relatively weak emission from the belt's main bodies ([Fig. 7a](#fig7){ref-type="fig"}). This observation can be related to the strong optical waveguide effect for the 1D micro-belt, in which the photons transfer from the main body to the ends. This is the intrinsic performance for several 1D/2D luminescent micro/nanostructures: the fluorescence is highly confined inside the crystal and a low-loss optical signal can be obtained during light propagation. Based on such easily observed optical waveguide behavior, further spatially resolved PL imaging ([Fig. 7b](#fig7){ref-type="fig"}) and spectroscopy ([Fig. 7c](#fig7){ref-type="fig"}) from both the tips and excitation positions were performed by changing the excitation laser beam (400 nm) at different local positions of an individual OIHP-AD micro-belt. During light propagation, the generated photons are confined and propagate along the length of the 1D micro/nanostructure. The emission intensity at the tip decreases almost exponentially with the increase of the propagation distance, while it remains constant at the excited points. To evaluate the waveguide activity of this optical waveguide material, the optical-loss coefficient (R) was calculated. [Fig. 7d](#fig7){ref-type="fig"} shows the intensity ratio between the tips and bodies (I~tip~/I~body~) against the propagation distance, which reveals a nearly single-exponential decay. An optical loss coefficient R of 0.004 dB μm^--1^ can be obtained by using the function I~tip~/I~body~ = A exp(--RD), where A is a constant and D is the distance between the excited site and the emitting tip.[@cit25] The excellent optical waveguide properties of this perovskite micro-belt are comparable with those of reported inorganic nanowire and nanofiber waveguides.[@cit26] This reveals that the optical waveguide efficiency is related to the morphology, crystallinity and luminescent properties of the waveguide.[@cit24d],[@cit27] Here, the relatively low PLQY of the OIHP-AD micro-belt has extremely low optical loss. This phenomenon is similar to the case of the donor--acceptor molecular crystal optical waveguide material.[@cit28] Conversely, the high optical waveguide performance of OIHP-AD results from the spatial confinement effect of the layer-by-layer nanostructure, which may afford an opportunity to achieve high-quality resonant cavities for micro-lasers. ![(a) PL microscope images of the OIHP-AD micro-belt under UV (365 nm) light. (b) PL images obtained from a single OIHP-AD micro-belt by exciting at different positions (images 1--7). (c) Spatially resolved PL spectra from the tip of the OIHP-AD micro-belt for different distances between the excitation spot and the tip of the micro-belt shown in PL images 1--7. (d) Optical-loss coefficient (R) of the OIHP-AD micro-belt.](c9sc00162j-f7){#fig7} Conclusions =========== In summary, a new 1D single-crystalline micro-belt perovskite material can be fast fabricated by the supramolecular self-assembly of AD cations and lead chloride anionic layers in aqueous solution. The highly regular arrangement of protonated AD in the host--guest system confers enhanced fluorescence performance (such as higher PLQY and lifetime compared with the pristine AD) since the chromophore could be tightly fixed and locked in the rigid framework through non-covalent interactions, thus restricting the vibrations and nonradiative loss. The solid-state perovskite and self-supporting film present high water stability, which can be related to the electrostatic and hydrogen bonding interactions between the host--guest components and the larger steric hindrance of the AD cations distributed between the rigid lead chloride inorganic layers. This can vastly inhibit the ingression of water molecules into the perovskite. Spatially resolved photoluminescence spectra confirm that OIHP-AD micro-belts exhibit a low waveguide loss coefficient (0.004 dB μm^--1^). Therefore, this study not only provides a facile way to develop a new type of low-dimensional organic--inorganic hybrid perovskite material with high water stability but also enables their potential applications in upconversion, polarized photoemission and optical waveguides for the construction of future optical communication micro-devices. Conflicts of interest ===================== There are no conflicts to declare. Supplementary Material ====================== Supplementary information ###### Click here for additional data file. Crystal structure data ###### Click here for additional data file. This work was supported by the National Natural Science Foundation of China (Grant No. 21771021, 21822501, 21601080 and 21771097), the Beijing Nova Program (xx2018115), the Fundamental Research Funds for the Central Universities, and Analytical and Measurements Fund of Beijing Normal University. [^1]: †Electronic supplementary information (ESI) available. CCDC [1862717](1862717). For ESI and crystallographic data in CIF or other electronic format see DOI: [10.1039/c9sc00162j](10.1039/c9sc00162j)	{ "pile_set_name": "PubMed Central" }
Introduction ============ Ultrasonography (US) is a commonly performed examination for shoulder pain, recommended by experts as the first-choice technique to evaluate various rotator cuff diseases and nonrotator cuff diseases \[[@b1-usg-17055]-[@b4-usg-17055]\]. When US is performed by an experienced radiologist, its diagnostic sensitivity and specificity for detecting rotator cuff tears are comparable to those of magnetic resonance imaging (MRI) \[[@b5-usg-17055]\]. The advantages of US include not only excellent diagnostic accuracy but also high resolution and the capability of dynamic evaluation \[[@b6-usg-17055]\]. To assess all possible causes of shoulder pain, it is better to follow a standardized protocol and to perform a comprehensive evaluation of the shoulder than to conduct a focused examination. Moreover, a proper dynamic study can enhance the diagnostic quality of US, especially when the pathology is not revealed by a static evaluation. The purpose of this article is to review the common indications for dynamic US of the shoulder, and to present the basic techniques and characteristic US findings. Long Head of the Biceps Tendon Subluxation ========================================== The long head of the biceps tendon (LHBT) can be assessed with the patient in the neutral position, with his or her elbow flexed and the dorsum of the hand placed on the ipsilateral thigh. The LHBT is visible as a cord-like hyperechoic structure that lies within the bicipital groove between the greater and lesser tubercle of the proximal humerus and is covered by the transverse humeral ligament. In normal shoulders, the LHBT is secured by the tendon sheath and pulley, as well as the transverse humeral ligament, so that it is not subluxated or dislocated during internal or external rotation of the shoulder \[[@b2-usg-17055],[@b6-usg-17055]-[@b8-usg-17055]\]. Dynamic evaluation for subluxation or dislocation of the LHBT was first introduced in 1995, by Farin et al. \[[@b9-usg-17055]\]. They found that maximal external rotation of the shoulder was helpful for diagnosing transient subluxation of the LHBT, and the sensitivity of the dynamic assessment was 86%. The authors suggested that a dynamic study for LHBT subluxation should be included in routine shoulder US. LHBT subluxation and dislocation are frequently reported with a shallow bicipital groove and rotator cuff tears, especially subscapularis tendon tears \[[@b9-usg-17055],[@b10-usg-17055]\]. A bicipital groove less than 3 mm deep is regarded as shallow \[[@b9-usg-17055]\]. The diagnosis of subluxation or dislocation can be made when the LHBT overlies the wall of the bicipital groove or moves out of the groove in the short-axis view, either in the neutral or external rotation position \[[@b9-usg-17055]-[@b11-usg-17055]\] ([Fig. 1](#f1-usg-17055){ref-type="fig"}). Skendzel et al. \[[@b12-usg-17055]\] reported LHBT subluxations accompanied by partial tears of the LHBT, and postulated that LHBT subluxation associated with surface irregularities should raise suspicion of a partial-thickness tear of the LHBT. Intra-articular Entrapment of the LHBT ====================================== Intra-articular entrapment of the LHBT, known as "hourglass biceps," was first described by Boileau et al. \[[@b13-usg-17055]\]. This is a novel mechanical impairment of biceps tendon movement that presents as pain and locking of the shoulder. The sliding motion of the LHBT in the limited space of the bicipital groove is blocked by severe swelling of the intra-articular LHBT during shoulder abduction, subsequently causing buckling of the intra-articular LHBT and leading to pain in the anterior shoulder. US criteria for intra-articular entrapment of the LHBT were suggested by Pujol et al. \[[@b14-usg-17055]\], and include a 10% increase in the diameter of the intra-articular LHBT or tendon buckling that is visible during shoulder abduction ([Fig. 2](#f2-usg-17055){ref-type="fig"}). The sensitivity of the dynamic evaluation (50%) was not satisfactory, but the specificity was 100%. Subcoracoid Impingement ======================= Subcoracoid impingement or coracoid impingement is a rarely diagnosed, but well-known cause of anterior shoulder pain \[[@b15-usg-17055],[@b16-usg-17055]\]. The known etiologies of subcoracoid impingement are anatomic variations of the scapulae (coracoid process), ossifications of the subscapularis tendon ([Fig. 3](#f3-usg-17055){ref-type="fig"}, [Video clip 1](#SD1-usg-17055){ref-type="supplementary-material"}), ganglion cysts, and other osseous deformities caused by surgery or trauma \[[@b17-usg-17055]\] that produce excessive pressure on the LHBT or subscapularis tendon. The main diagnostic criterion for subcoracoid impingement in imaging studies is coracohumeral distance, although discordant opinions about its diagnostic value have been reported in the literature \[[@b18-usg-17055],[@b19-usg-17055]\]. The coracohumeral distance is known to be smaller in symptomatic patients than in asymptomatic volunteers \[[@b18-usg-17055]\]. Because the coracohumeral distance on MRI is known to be smaller in women than in men, a sex-adjusted criterion (11.5 mm in men) was proposed by Giaroli et al. \[[@b19-usg-17055]\], but it showed poor predictive value. Recently, subcoracoid bursitis and impingement were observed with dynamic US during internal/external rotation of the shoulder \[[@b20-usg-17055]\], which is one of the well-known appearances of subacromial impingement. Because anatomical or biomechanical properties are different in each patient, real-time observation of the mechanical blockage ([Fig. 4](#f4-usg-17055){ref-type="fig"}, [Video clip 2](#SD2-usg-17055){ref-type="supplementary-material"}) might have the potential to be able to diagnose this mechanical condition, although further validation is needed. Subacromial Impingement ======================= Subacromial impingement is the most common and well-recognized indication for dynamic US in the shoulder. Osteophytes, or an abnormal shape of the acromion, subacromial spurs, and the acromioclavicular joint, are common causes of compression at the rotator cuff and the overlying subacromial-subdeltoid bursa \[[@b21-usg-17055]\]. Dynamic evaluation can be done by shoulder abduction or flexion with the probe placed at the end of the acromion in the coronal plane or in the sagittal plane. The two important points to be checked for subacromial impingement are the humeral head depression and tendon/bursal impingement \[[@b22-usg-17055]\]. Because humeral head depression is essential to make enough space for the rotator cuff to slide beneath the acromion, the center of the humeral head normally moves inferiorly in the latter half of the cycle during shoulder abduction. When the humeral head does not move inferiorly or abnormally moves superiorly, the space for the rotator cuffs and the subacromial-subdeltoid bursa decreases and subacromial impingement can occur. Bureau et al. \[[@b23-usg-17055]\] observed proximal migration of the humeral head with dynamic US, and classified this osseous impingement as the most severe (grade 3) form of subacromial impingement ([Table 1](#t1-usg-17055){ref-type="table"}). They also described soft tissue (bursa or tendon) encroachment beneath the acromion, possibly combined with failure of humeral head depression, as grade 2 impingement ([Fig. 5](#f5-usg-17055){ref-type="fig"}, [Video clips 3](#SD3-usg-17055){ref-type="supplementary-material"}-[5](#SD5-usg-17055){ref-type="supplementary-material"}). The dynamic evaluation of subacromial impingement on US is commonly focused at the end of the acromion. However, subacromial impingement not only occurs beneath the acromion, but also at all possible locations below the coracoacromial arch, including anywhere between flexion and abduction of the shoulder ([Fig. 6](#f6-usg-17055){ref-type="fig"}). The coracoacromial ligament is the central part of the coracoacromial arch, which can be the main causative structure for subacromial impingement and has recently received attention as part of ultrasonographic evaluations \[[@b22-usg-17055],[@b24-usg-17055],[@b25-usg-17055]\]. The coracoacromial ligament can be visualized by placing one end of the probe at the acromion and the other end at the coracoid process ([Figs. 6](#f6-usg-17055){ref-type="fig"}, [7A](#f7-usg-17055){ref-type="fig"}). Subacromial impingement beneath the coracoacromial ligament can be also visualized by turning the probe 90° from the long-axis view of the coracoacromial ligament, which is slightly anterior from the acromion, and along the direction of the supraspinatus tendon ([Fig. 7B](#f7-usg-17055){ref-type="fig"}) \[[@b22-usg-17055]\]. The thickness and length of the coracoacromial ligament is not different between normal subjects and those who have subacromial impingement syndrome, but superior displacement of the coracoacromial ligament is significantly increased during shoulder abduction-internal rotation (throwing motion) \[[@b24-usg-17055]-[@b26-usg-17055]\]. Adhesive Capsulitis =================== Adhesive capsulitis or frozen shoulder is a frequently encountered disease in the shoulder, causing pain and a limited range of motion. It is more common in individuals with diabetes and perimenopausal women \[[@b22-usg-17055]\]. The ultrasonographic findings of adhesive capsulitis are hypoechoic changes and hypervascularity in the rotator interval \[[@b27-usg-17055]\], and thickening of the coracohumeral ligament (mean thickness of 3.0 mm, compared to 1.4 mm in asymptomatic patients) \[[@b28-usg-17055]\] and the inferior glenohumeral ligament (mean thickness of 4.0 mm compared to 1.3 mm in the asymptomatic contralateral shoulder) \[[@b29-usg-17055]\]. However, the diagnosis of adhesive capsulitis still substantially relies on the radiologist's observation of limited external rotation or abduction during routine shoulder US. The dynamic US findings of adhesive capsulitis were reported in 1993 by Ryu et al. \[[@b30-usg-17055]\]. They found that continuous limitation of supraspinatus movement beneath the acromion and continuous visualization of the supraspinatus tendon during shoulder abduction were useful criteria that could diagnose adhesive capsulitis with 92% accuracy. When the axillary pouch is stiff and cannot be stretched to let the shoulder abduct, the patient will try to compensate by scapulothoracic rotation to raise the arm. However, the glenohumeral joint relationship is fixed, and the supraspinatus tendon is persistently visible at the lateral aspect of the acromion ([Fig. 8](#f8-usg-17055){ref-type="fig"}, [Video clip 6](#SD6-usg-17055){ref-type="supplementary-material"}). If there is no visible tendon pathology or soft tissue impingement, we can more confidently make the diagnosis of adhesive capsulitis using this dynamic evaluation. Acromioclavicular Joint Instability =================================== Acromioclavicular (AC) joint instability is another indication for dynamic US. For traumatic injuries of the AC joint, the classification mainly relies on the AC joint space and the extent of the ligament injury \[[@b31-usg-17055]\]. Because a normal AC joint can show variable obliqueness and step-offs, AC joint space measurement on US can be problematic, and the criterion for widening (6 mm) established on radiographs \[[@b32-usg-17055]\] cannot be directly used on US \[[@b33-usg-17055]\]. Therefore, a comparison with the contralateral side is recommended to diagnose abnormal widening of the AC joint \[[@b34-usg-17055]\], and a relative measurement (AC index=AC joint space on the uninjured side/AC joint space on the injured side) is suggested \[[@b35-usg-17055]\]. The normal AC index is 1, and the AC index is lower in more widened and severely injured AC joints. A dynamic evaluation procedure for AC joint injuries was introduced by Peetrons and Bedard in 2007 \[[@b34-usg-17055]\]. By placing the palm at the contralateral shoulder, in the so-called cross-arm maneuver, the AC distance is decreased in the injured AC joint ([Fig. 9](#f9-usg-17055){ref-type="fig"}), and becomes widened again with the change of position to neutral (hands on the ipsilateral thigh). An uninjured AC joint shows minimal change (less than 1 mm) in the cross-arm maneuver \[[@b34-usg-17055]\]. Osteoarthritis is another cause of AC joint instability, and the joint space can be severely decreased, with the cross-arm maneuver showing a "kissing" appearance. However, we can easily differentiate osteoarthritis with chronic instability from acute AC joint injury by joint space narrowing, subchondral bone changes, and the presence of osteophytes in the neutral position, Joint Effusion and Synovial Hypertrophy ======================================= It is important to identify joint effusion for the diagnosis of septic arthritis or inflammatory arthritis in the glenohumeral joint. Joint effusion is commonly detected in the posterior recess of the glenohumeral joint and tendon sheath of LHBT by communication with the glenohumeral joint \[[@b36-usg-17055]\]. However, joint effusion can be invisible in the posterior recess in neutral position, even if the joint is distended with 8-12 mL of fluid \[[@b37-usg-17055]\]. Most likely, the fluid is pooled in the axillary pouch because of gravity when the patient is in sitting position. External rotation of the shoulder increased the sensitivity from 17% to 100% for detecting fluid in the glenohumeral joint in the posterior recess ([Fig. 10](#f10-usg-17055){ref-type="fig"}). Differential Diagnosis of Cystic Lesions at the Spinoglenoid Notch ================================================================== The spinoglenoid notch is the groove between the glenoid and the base of the scapular spine, where the suprascapular nerve and suprascapular vessels run. This region should be evaluated during routine shoulder US for a possible paralabral cyst \[[@b2-usg-17055],[@b6-usg-17055],[@b22-usg-17055],[@b36-usg-17055]\], because a paralabral cyst at this location can entrap the suprascapular nerve and cause pain or weakness of the shoulder. Occasionally, we can see the suprascapular vessels and nerves as tiny hypoechoic structures in the spinoglenoid notch \[[@b36-usg-17055]\]. Sometimes the vessels are engorged or distended with blood, especially in the external rotation position of the shoulder, and can be confused with a paralabral cyst or ganglion cyst \[[@b38-usg-17055]\]. Usually, the internal rotation position of the shoulder (cross-arm maneuver) decreases suprascapular varicosity ([Fig. 11](#f11-usg-17055){ref-type="fig"}), whereas a true paralabral cyst or ganglion cyst does not change with internal rotation of the shoulder \[[@b6-usg-17055],[@b22-usg-17055]\]. Conclusion ========== Dynamic real-time observation is a major strength of US, especially when the pathology is not revealed by a static evaluation. We do not perform all of these dynamic studies as part of routine shoulder US. However, familiarity with these dynamic maneuvers and indications, and their proper application, will significantly improve the diagnostic value of shoulder US. No potential conflict of interest relevant to this article was reported. Supplementary Material ====================== ###### Video clip 1. A 19-year-old man with subscapularis tendon ossification and subcoracoid impingement. Long axis view of the subscapularis tendon shows slightly thickened subcoracoid bursa, superficially located to the tendon. There are two ossifications in the subscapularis tendon. During internal rotation of the shoulder, the subscapularis tendon does not fully glide under the coracoid process due to the ossifications and subcoracoid impingement occurs. Note the hypoechoic soft tissue bulging contour by the subcoracoid bursa and the subscapularis tendon (<https://doi.org/10.14366/usg.17055.v001>). ###### Video clip 2. A 69-year-old woman with subcoracoid impingement with soft tissue involvement. On the long axis view of the subscapularis tendon, the thickened subcoracoid bursa is located superficial to the subscapularis tendon. During internal rotation of the shoulder, pooling of the fluid in the subcoracoid bursa is noted at the lateral aspect of the coracoid process. There is a smooth gliding of the subscapularis tendon underneath the coracoid process and the bursa, without significant impingement of the tendon (<https://doi.org/10.14366/usg.17055.v002>). ###### Video clip 3. A 31-year-old man without subacromial impingement. During the dynamic evaluation of subacromial impingement, there is no significant soft tissue or osseous impingement. The supraspinatus tendon and the greater tubercle show smooth passage underneath the acromion (<https://doi.org/10.14366/usg.17055.v003>). ###### Video clip 4. A 45-year-old woman with grade 2 subacromial impingement. Long axis view of the supraspinatus tendon between acromion and greater tubercle shows subacromial-subdeltoid bursal thickening above supraspinatus tendon surface. During shoulder abduction, subacromial-subdeltoid bursal fluid pooling at the lateral aspect of acromion is seen. The humeral head normally moves inferiorly (<https://doi.org/10.14366/usg.17055.v004>). ###### Video clip 5. A 50-year-old woman with grade 3 subacromial impingement. Long axis view of supraspinatus tendon between acromion and greater tubercle shows severe tendinosis of supraspinatus tendon and adjacent thickened subacromialsubdeltoid bursa. During the dynamic examination, the osseous impingement occurs. The humeral head does not sufficiently move inferiorly and the supraspinatus tendon and subacromial-subdeltoid bursa cannot be passed underneath the acromion (<https://doi.org/10.14366/usg.17055.v005>). ###### Video clip 6. A 54-year-old man with adhesive capsulitis. Long axis view of the supraspinatus tendon during shoulder abduction shows limitation of supraspinatus movement beneath the acromion and continuous visualization of the supraspinatus tendon. There is no significant subacromial soft tissue impingement (<https://doi.org/10.14366/usg.17055.v006>). ![An 84-year-old man with biceps tendon subluxation.\ A. In the neutral position, the biceps long head tendon (arrows) shows a split tear and the medial portion of the tendon partially overlies the medial wall of the bicipital groove formed by the lesser tubercle (L), although the biceps tendon is located in the groove. The depth of the bicipital groove is normal. B. In the external rotation position, the medial portion of the torn biceps long head tendon (arrows) is displaced more medially over the lesser tubercle (L). Note that the subscapularis tendon is completely torn at the footprint of the lesser tubercle (arrowheads).](usg-17055-f1){#f1-usg-17055} ![A 71-year-old woman with anterior shoulder pain.\ A, B. The long-axis view and the short-axis view of the long head of the biceps tendon (LHBT) in the neutral position of the shoulder are shown. The LHBT is thickened, with a hypoechoic appearance suggesting tendinopathy. A small amount of effusion (arrowheads) is noted in the biceps tendon sheath. The diameter of the intra-articular tendon was 2.8 mm. C, D. The long-axis view and the short-axis view of the LHBT in the abduction position of the shoulder are shown. The intra-articular LHBT has a more curved appearance (arrows) and is elevated from the humeral head (H). The diameter of the intra-articular tendon was 3.5 mm, with an increase of 25% compared to the neutral position. A, acromion.](usg-17055-f2){#f2-usg-17055} ![A 19-year-old man with subscapularis tendon ossification and subcoracoid impingement.\ A. The long-axis view of the subscapularis tendon shows slightly thickened subcoracoid bursa (arrows), superficially located above the subscapularis tendon (arrows). There are two ossifications in the subscapularis tendon (asterisks). B. During internal rotation of the shoulder, the subscapularis tendon does not fully glide under the coracoid process due to ossifications (asterisks), and subcoracoid impingement occurs. Note the bulging contour of the hypoechoic soft tissue by the subcoracoid bursa and the subscapularis tendon (arrowheads). C, coracoid process; LT, lesser tubercle of the humerus.](usg-17055-f3){#f3-usg-17055} ![A 69-year-old woman with subcoracoid impingement with soft tissue involvement.\ Long-axis view of the subscapularis tendon during internal rotation of the shoulder shows pooling of the fluid in the subcoracoid bursa (arrows) at the lateral aspect of the coracoid process. C, coracoid process; LT, lesser tubercle of the humerus.](usg-17055-f4){#f4-usg-17055} ![Dynamic evaluation of subacromial impingement.\ A. Normal dynamic ultrasonography of a 31-year-old man is shown. No visible soft tissue or osseous impingement was observed during shoulder abduction. B. A 45-year-old woman showed grade 2 subacromial impingement. Subacromial-subdeltoid bursal thickening (arrows) is visible above the supraspinatus tendon surface and gathers outside of the acromion during shoulder abduction. C. A 50-year-old woman showed grade 3 subacromial impingement. The supraspinatus tendon shows severe tendinosis and is impinged between the greater tubercle of the humerus and the acromion. The humeral head does not sufficiently move inferiorly, and the supraspinatus tendon and the subacromial-subdeltoid bursa cannot be passed underneath the acromion. The tendon (arrowheads) and bursal tissues (arrows) are impinged and protruded superficially between the acromion and the greater tubercle of the humerus. A, acromion; GT, greater tubercle of the humerus.](usg-17055-f5){#f5-usg-17055} ![Schematic drawing of the shoulder showing the coracoacromial ligament (CAL) seen from above.\ The CAL is located just anterior to the aspect of the acromion connecting the coracoid process and the acromion. Subacromial impingement can occur not only below the acromion, but also below the CAL, by elevation of the arm in any direction (arrows) between flexion and abduction of the shoulder.](usg-17055-f6){#f6-usg-17055} ![A 19-year-old man with subacromial impingement syndrome by the coracoacromial ligament.\ A. The coracoacromial ligament (arrows) is visible as a hyperechoic linear structure connecting the acromion (A) and the coracoid process (C) on the long-axis view, with the probe placed at the anterolateral shoulder in an oblique coronal plane. B. The short-axis view of the coracoacromial ligament (arrows) shows a plate-like structure covering the supraspinatus tendon and the subacromial bursa. Arm elevation midway between flexion and abduction of the shoulder provoked pain and bunching up of the underlying subacromial bursa (arrowheads) and the surface of the supraspinatus tendon.](usg-17055-f7){#f7-usg-17055} ![A 54-year-old man with adhesive capsulitis.\ The long-axis view of the supraspinatus tendon during full shoulder abduction shows limited supraspinatus movement beneath the acromion and continuous visualization of the supraspinatus tendon. There is no significant subacromial soft tissue impingement. A, acromion; GT, greater tubercle of the humerus.](usg-17055-f8){#f8-usg-17055} ![A 37-year-old woman with chronic acromioclavicular (AC) joint injury.\ A. A long-axis view of the AC joint in resting position shows joint space widening and step-off between the acromion (A) and the clavicle (C). B. The AC joint space is narrowed and step-off is decreased by the cross-arm maneuver (placing the palm on the contralateral shoulder).](usg-17055-f9){#f9-usg-17055} ![A 70-year-old man with glenohumeral joint synovitis.\ A. A long-axis view of the posterior glenohumeral joint shows a small amount of effusion (arrowheads) in neutral position, while the patient is being examined in the sitting position. B. External rotation of the shoulder joint more clearly revealed joint effusion and synovial hypertrophy (arrowheads) gathered at the posterior recess. H, head of humerus; G, glenoid.](usg-17055-f10){#f10-usg-17055} ![A 70-year-old man with a dilated vein in the spinoglenoid notch (SGN).\ A. A long-axis view of the posterior glenohumeral joint shows posterior glenohumeral joint recess and the spinoglenoid notch, which is located at just lateral of the glenoid. During external rotation of the shoulder, the distended suprascapular artery and veins (arrows) are seen as multiple anechoic structures in the SGN that can mimic a paralabral cyst. B. A color Doppler image shows vascularity within the dilated suprascapular vessels. C. Internal rotation of the shoulder collapses the suprascapular vein. We can distinguish the dilated vessels from a paralabral cyst, which does not collapse by internal rotation of the shoulder. H, head of humerus; G, glenoid.](usg-17055-f11){#f11-usg-17055} ###### Ultrasonographic classification of subacromial impingement Grade Pain provocation with shoulder motion Ultrasonographic finding ------- --------------------------------------- ---------------------------------------- 0 No No visible anatomic impingement 1 Yes No visible anatomic impingement 2 Yes Bursa or tendon impingement 3 Yes Superior migration of the humeral head Modified from Bureau et al. AJR Am J Roentgenol 2006;187:216-220, with permission of American Roentgen Ray Society \[[@b23-usg-17055]\].	{ "pile_set_name": "PubMed Central" }
Introduction {#S1} ============ Cutaneous T-cell lymphomas (CTCL) are a heterogeneous group of non-Hodgkin lymphomas; the majority of cases are mycosis fungoides and/or Sezary syndrome; the remainder include a variety of subtypes which vary greatly in clinical behavior. All forms of CTCL are generally considered to be incurable. While mycosis fungoides may have a relatively indolent course (ref. [@R1]), once the disease has progressed beyond early stages, it can behave in a highly aggressive manner. Sézary syndrome (circulating lymphoma cells greater than 1,000/mm^3^ and greater than 10% of peripheral blood leukocytes) has a median survival time of 31 months (ref. [@R2]). Although there are numerous therapies available to treat CTCL, including newer biological therapies, CTCL is highly responsive to immune manipulation (ref. [@R3])(ref. [@R4]--[@R7]). Responses to interferon α, interferon γ and interleukin-12 highlight that CTCL is influenced by host immune status (ref. [@R8])(ref. [@R9]). Conventional treatments are at best temporizing with aggressive disease (ref. [@R10]), and high dose chemotherapy with autologous HSCT for MF and SS has had disappointing results (ref. [@R11]). For many diseases, allogeneic HSCT is an ideal form of immunotherapy; however, the role for allogeneic HSCT in CTCL is not well defined. Allogeneic transplant is successful in part due to the GVL effect of the donor graft, independent of the conditioning regimen (ref. [@R12]). Several small series describe durable remissions after allogeneic HSCT after both RIC and myeloablative conditioning (ref. [@R13], [@R14])(ref. [@R15]). DLI has also been reported to be effective in producing a GVL effect but only when combined with several other modalities of therapy including chemotherapy (ref. [@R14]). Nevertheless, long-term remissions, particularly in the setting of RIC allogeneic transplant, and the observation of responses with withdrawal of immunosuppression, support the presence of a potent GVL effect (ref. [@R14])(ref. [@R15]). We present outcomes of 12 consecutive patients who underwent allogeneic HSCT using, most commonly, a reduced intensity conditioning regimen for this rare lymphoid malignancy at our institution. The high response rate, durable remissions, and direct GVL induction with DLI all support a potent GVL effect in patients with MF and SS from allogeneic HSCT. Methods {#S2} ======= A retrospective review was performed of the University of Pennsylvania bone marrow transplant database to identify all patients who underwent allogeneic HSCT for cutaneous T cell lymphoma. Twelve patients were identified who were transplanted between 2004 to 2010. A chart review was performed to obtain data about diagnosis, staging treatment, transplantation and outcomes. This study was reviewed and approved by the institutional review board at the University of Pennsylvania. Patient Selection {#S3} ----------------- Patients were referred for transplant because of poor prognosis or progression after conventional therapy. They were generally heavily pretreated, having received a median of 8 non-chemotherapy, and 2 chemotherapy-based treatment modalities before being transplanted. Treatments are listed in [Table 1](#T1){ref-type="table"}. Only 3 patients were in CR at the time of transplant, while 1 had MRD detectable by flowcytometry in the bone marrow, 4 had chemo-responsive active disease, and 4 had progressive active disease. Transplantation {#S4} --------------- Seven of the patients (Patients \#1 -- \#7) had HLA-identical sibling donors, while unrelated donors were identified for the rest. Of these, donors for 4 (Patients \#8, Patients \#10 -- \#12) were well matched by high resolution molecular typing at HLA A, B, C, DR and DQ, while Patient \#9 had a single antigen mismatch. RIC with fludarabine (120 mg/m2 over 4 days) and busulfan (6.4 mg/kg over 2 days) or fludarabine (same dose) and cyclophosphamide (3 gm/m2) was used in 10 cases. Conventional myeloablative conditioning with Cyclophosphamide (120 mg/kg over 2 days) and total body irradiation (TBI) 1200 cGy fractionated in 6 doses over 3 days was used in 2 young patients with excellent performance status and active disease prior to transplantation (see [Table 1](#T1){ref-type="table"}). GVHD prophylaxis consisted of a calcineurin inhibitor and methotrexate in all patients. Statistics {#S5} ---------- Overall survival was measured as time from HSCT to death from any cause. Event-free survival was measured as time from HSCT to relapse, progression or death from any cause, whichever occurred first. Probabilities of OS and EFS, and survival curves, were generated using Kaplan-Meier estimates (ref. [@R16]). Statistical analysis was performed using Statview. Results {#S6} ======= Patient characteristics {#S7} ----------------------- Patient characteristics, diagnosis and stage are shown in [Table 1](#T1){ref-type="table"}. Patients were staged using the ISCL/EORTC TNM staging system (ref. [@R17]). There were 5 male and 7 female patients with a median age at diagnosis of CTCL of 49 yrs (range 40--61 yrs). Of the 12 patients, 5 had MF (stages IIB, IIB, IIIB, IVA2, IVB; 2 with nodal transformation), 4 had SS (one stage IVA1, three IVA2; 1 with nodal transformation), and 3 had Gamma-delta T-cell lymphoma (all T3b). Median time to transplantation from diagnosis was 3.3 yrs (range 0.5--7 yrs), with a median age at transplantation of 53 yrs (range 41--61 yrs). The median follow up is 15 months (range 1 to 45 months) for all patients and 24 months (range 13 to 45 months) for patients currently alive. Engraftment {#S8} ----------- Neutrophil engraftment occurred with an ANC \>500 at a median of 13 days (range 10--34 days) after HSCT. At day 100 after HSCT in 10 evaluable pts, median whole blood chimerism was 97%. Patient \#1 was diagnosed with CNS PTLD within 100 days of HSCT and lost her graft after rapid tapering of immunosuppression. Patient \#8 had decreasing donor chimerism 15 months after HSCT, and received DLI with subsequent achievement of 100% donor chimerism. Individual chimerism trends are described in [Table 3](#T3){ref-type="table"}. Response and Survival Outcomes {#S9} ------------------------------ The outcomes of all patients are shown in [Figure 1](#F1){ref-type="fig"}. Two patients (one with RIC and the other with myeloablative conditioning) died within the first 100 days from sepsis with active disease. At day 100, 8 of 10 evaluable patients were in complete remission (CR); transplant had induced and maintained CR in 6 pts with active disease. Three patients remain alive in continuous CR 16, 22 and 45 months after transplant. Four patients relapsed after achieving CR between 4 and 13 months after HSCT, but two of these re-achieved and maintained CR, one after tapering immunosuppression and the other after receiving DLI (without additional chemotherapy). Thus, of the 10 patients who survived to 100 days, 5 are alive in CR at last contact and 1 is alive with active disease. The median EFS for all pts was 5.3 months ([Figure 2A](#F2){ref-type="fig"}). This underestimates achievement of CR, since two patients were able to re-achieve and maintain CR as mentioned above. Kaplan-Maier estimated 2-year OS was 58%, and estimated median overall survival was 37 months ([Figure 2B](#F2){ref-type="fig"}). GVHD {#S10} ---- 9 of 12 patients developed acute GVHD, with 4 developing Grade III--IV GVHD. GVHD was the cause of death in 1 patient. Acute and chronic GVHD are detailed in [Table 2](#T2){ref-type="table"}. GVL Activity {#S11} ------------ Patient \#4 received DLI (2.2 × 10\^8 nucleated cells/kg) at time of relapse (after tapering off immunosuppression) 13 months after HSCT, and achieved CR that is persistent 26 months since DLI. Patient \#12 was on Tacrolimus at the time of relapse, and achieved CR with discontinuation of immune suppression. She remains in CR 8 months later. This provides clear evidence for a GVL effect, as neither of these patients received any accompanying chemo/systemic therapy that might explain achievement and maintenance of CR. Of note, patient \#7 received DLI (1.5 × 10\^8 nucleated cells/kg) 6 months after HSCT, with minimal effect on disease progression. Patient \#8 received DLI (10\^7 nucleated cells/kg) 15 months after HSCT for decreasing donor chimerism (no evidence of relapse), with a subsequent achievement of 100% donor chimerism and continued CR. Discussion {#S12} ========== Although CTCL is often an indolent lymphoma, with numerous effective treatment options, it is generally incurable, and most patients eventually fail chemotherapy and biological therapies and will die of their disease (ref. [@R18]). The prognosis is especially poor for patients with transformed CTCL and for patients who fail conventional therapy (ref. [@R19]). Newer treatments like denileukin diftitox (ref. [@R20]), vorinostat (ref. [@R3]) and romidepsin (ref. [@R7]), have recently shown promising results. High dose therapy and autologous HSCT has been disappointing, and early relapses are common (ref. [@R11]). The role for allogeneic HSCT in CTCL is unclear. A recent retrospective study from the EBMT included a heterogeneous group of patients who received both myeloablative and RIC regimens with estimated 3-year survival of 54% (ref. [@R21]). Three other small reports further support the role of allogeneic HSCT in CTCL. Molina et al. described 8 patients with MF/SS who were treated with RIC:SCT and 6 patients remained alive without disease 33--108 months after transplant (ref. [@R13]). Duvic et al. described outcomes in 19 recipients of an intensive though reduced intensity conditioning regimen that included total skin electron beam irradiation followed by allogeneic transplant, with 11 patients achieving sustained remissions, and several achieving stable disease after a combination of DLI, immunosuppression withdrawal, and/or chemotherapy (ref. [@R14]). Delioukina et al. also showed similarly encouraging results for RIC:SCT in 11 patients with CTCL, and observed no further relapses/deaths after the first 18 months after HSCT (ref. [@R15]). We report our experience using, most commonly, a reduced intensity conditioning regimen followed by allogeneic stem cell transplant in 12 consecutive patients with extensively pretreated advanced CTCL. Response rates were high. Treatment related mortality was low. 2 patients who did not respond died from infectious complications within 100 days of transplant. GVHD accounted for death in one patient. The most common cause of death was disease progression in 5 patients. 42% of our patients were alive and disease-free a median duration of 22 months after transplant. While this percentage seems lower than other reports, it is difficult to compare patients across different retrospective studies. Furthermore, our outcomes provide strong and direct support the presence of a GVL effect in CTCL; 1 patient achieved CR after DLI alone for relapse and a second achieved sustained remission after withdrawal of immunosuppression. Despite the limitations associated with this retrospective review of small numbers of patients, our outcomes support the presence of a meaningful GVL effect associated with RIC allogeneic stem cell transplant. With small numbers of patients it was not possible to associate the occurrence of graft versus host disease with a GVL effect. Our experience shows that RIC HSCT can provide long-term disease control in patients with advanced CTCL otherwise refractory to immunotherapy and chemotherapy. Ideally, prospective studies will further define the role of allogeneic HSCT in this disease. However, given the limited transplant related mortality, consideration for earlier transplant should be given, with a focus on the use of newer agents for achieving minimal residual disease (MRD) prior to transplant. Interventions to decrease the risk of relapse should be considered, including the possibility of maintenance immunotherapy and/or planned DLI. This work was supported in part by grants from The Leukemia & Lymphoma Society (7000-02) and NIH (K24 CA11787901) (DLP). Conflicts of Interest Alain H. Rook - Therakos (Speaker's bureau), HY Biopharma (Consultancy) Ellen J Kim -- TenX, Biocryst, Genmab, Glouchester, Celgene (Research funding), Eisai (Consultancy) Other authors have no relevant conflicts of interest ![](nihms323066f1){#F1} ![](nihms323066f2){#F2} ![](nihms323066f3){#F3} ###### Patient characteristics (sorted chronologically according to date of transplant) Pt Sex/Age Time Dx to SCT (yrs) Diagnosis Prior treatment (non-chemotherapy Prior-treatment (chemotherapy) Highest stage prior to HSCT Disease status prior to HSCT Conditioning regimen ---- --------- ---------------------- ---------------------------------------- ------------------------------------------------------------------------ --------------------------------------------------------- ----------------------------- ---------------------------------- ---------------------- 1 F/57 2.8 SS w/o large cell transformation (LCT) B, C, EB, ECP, INF, Me, PUVA, S, Sar, Tacrolimus, TSEB Alemtuzumab IVA1 Active disease, progressive Flu/Cy 2 M/52 1.1 MF w/o LCT \- CHOP, HyperCVAD IIB CR Flu/Bu 3 M/41 1.6 Gamma-delta T-cell lymphoma (GDTL) B, Hydroxychloroquine, S, TSEB Denileukin Diftitox, HyperCVAD, Methotrexate T3b Active disease, chemo-responsive Flu/Mel 4 F/43 0.8 MF w/o LCT S, INF, TSEB \- IIB MRD (marrow) Cy/TBI 5 F/48 0.8 GDTL B, S Alemtuzumab, Cytarabine, Cyclophosphamide, Methotrexate T3b Active disease, chemo-responsive Cy/TBI 6 M/56 5.9 MF with LCT B, C, EB, Imiquimod, INF, M, S, TSEB, PUVA Alemtuzumab, Denileukin Diftitox, ICE, Zanolimumab IVA2 Active disease, progressive Flu/Bu 7 F/52 7.0 MF with LCT C, INF, Me, S, Tazarotene, TSEB, V CHOP IVB CR Flu/Bu 8 M/61 3.7 MF w/o LCT B, C, INF, PUVA, S, Sar, Tretinoin, TSEB, UVB Alemtuzumab IIIB CR Flu/Bu 9 F/44 3.9 GDTL B, EB, S, V Cyclophosphamide, Denileukin Diftitox T3b Active disease, progressive Flu/Bu 10 M/61 0.5 SS with LCT S, TSEB, UVB CHOP, HyperCVAD, Alemtuzumab IVA2 Active disease, progressive Flu/Bu 11 F/61 6.8 SS w/o LCT B, ECP, INF, M, PUVA, S, Sar, TSEB, V Alemtuzumab IVA2 Active disease, chemo-responsive Flu/Bu 12 F/54 6.8 SS w/o LCT B, ECP, Forodesine, INF, Pegfilgastrim, PUVA, Romidepsin, S, Tretinoin Alemtuzumab IVA2 Active disease, chemo-responsive Flu/Bu LCT = Large cell transformation MF = Mycosis Fungoides GDTL = Gamma-delta T-cell lymphoma SS = Sézary Syndrome TRM = Transplant-related mortality B = Bexarotene C = Carmustine (topical) EB = Localized electron beam radiotherapy ECP = Extracorporeal photopheresis INF = Interferon-alpha/gamma Me = Mechlorethamine (topical) PUVA = Psoralen + Ultraviolet light A photochemotherapy S = Steroids Sar = Sargramostim TSEB = Total skin electron beam radiation treatment UVB = Ultraviolet light B phototherapy V = Vorinostat ###### Post-transplant outcomes Pt Non-relapse/non- GVHD toxicity Acute GVHD (Grade, Organ, and Stage) Chronic GVHD Disease Status at 100 days Relapse (Y/N) Treatment(s) since relapse Current status Time SCT to last follow up (months) ---- --------------------------------------------------------------------------------- ----------------------------------------------------- ------------------------------------------------------------------------------------------- ---------------------------- --------------- ------------------------------------------------------------------------------------------------------------------ -------------------------------------------------------------------------------- ------------------------------------- 1 CNS PTLD (responded to Rituximab), CMV antigenemia, Pneumonitis \- \- CR Y Denileukin Diftitox, Alemtuzumab Expired (progression) 37 2 Widely metastatic melanoma[\](#TFN18){ref-type="table-fn"} Grade II (Gut Stage 1) \- CR Y \- Expired (progression and metastatic melanoma)[\](#TFN18){ref-type="table-fn"} 7 3 Sinusitis, Osteoporosis, Herniated disk, Hearing loss, Steroid-induced glaucoma Grade III (Gut Stage 3) Severe chronic GVHD of Mouth, Musculoskeletal system and Lungs (Bronchiolitis obliterans) CR N \- Alive, CR (continuous) 45 4 Steroid-induced DM Grade II (Skin Stage 3, Gut Stage 1, Liver Stage 0) \- CR Y DLI with re-achievement of CR. Alive, CR 39 5 Bacterial sepsis \- \- Expired Y \- Expired (Early TRM) 2 6 \- Grade III (Skin Stage 3, Liver Stage 2) \- Prog-ression Y Tapering of immunosuppression (not effective), nodal XRT, Cyclophosphamide Expired (progression) 5 7 \- Grade I (Skin Stage 1) \- Relapse Y Tapering of immunosuppression, DLI (ineffective), Vorinostat, Topical carmustine, TSEB, Interferon gamma-1b, ICE Alive, Active disease (chemo-responsive) 27 8 PICC-associated UE DVT, hypothyroidism, Herpes zoster Grade III (Skin Stage 3, Liver Stage 0) \- CR N \- Alive, CR (continuous) 22 9 Cholecystitis, bacterial sepsis Grade IV (Gut Stage 4, Liver Stage 4-VOD) \- CR N \- Expired in CR (GVHD) 6 10 Bacterial sepsis Grade I (Skin Stage 2) \- Expired Y \- Expired (Early TRM) 1 11 Hypothyroidism \- \- CR N \- Alive, CR (continuous) 16 12 \- Grade I (Skin Stage 2) \- CR Y Tapering of immunosuppression with re-achievement of CR Alive, CR 13 Patient \#2 had a remote history of localized, resected melanoma, which resurfaced as widely metastatic disease after HSCT. Death was due to rapid progression of both CTCL and melanoma. CMV: Cytomegalovirus DVT: Deep Venous Thrombosis ICE: Ifosfamide, Carboplatin, Etoposide PTLD: Post-Transplant Lymphoproliferative Disease TRM: Transplant Related Mortality VOD: Veno-Occlusive Disease ###### Post-transplant donor chimerism (for patients surviving to 100 days) \# Day 100 6 months 1 year 2 years ---- -------------- ------------------- ------------------- ---------------------- 1 15 NE (lost graft) NE (lost graft) NE (lost graft) 2 90 (BM) NE (Died month 7) NE (Died month 7) NE (Died month 7) 3 100 100 100 100 4 93 (BM) 85 88 (BM) 100 6 100 (Day 60) NE (Died month 5) NE (Died month 5) NE (Died month 5) 7 85 99 100 100 8 97 (BM) 98 94 NE (F/U not reached) 9 100 NE (Died month 6) NE (Died month 6) NE (Died month 6) 11 97 97 95 NE (F/U not reached) 12 97 (BM) 97 100 NE (F/U not reached) Listed chimerism is whole blood chimerism unless mentioned otherwise BM = Bone marrow NE = Not evaluable	{ "pile_set_name": "PubMed Central" }
Data from the PrEP in Breastfeeding Study are available by contacting the International Clinical Research Center at the University of Washington (<icrc@uw.edu>). Introduction {#sec008} ============ Women in Africa are disproportionately affected by HIV, with the greatest rates of new HIV infections among women of child-bearing age \[[@pmed.1002132.ref001]\]. Pregnancy and the early postpartum period are characterized by heightened HIV risk associated with up to 2-fold increased HIV acquisition risk \[[@pmed.1002132.ref002]--[@pmed.1002132.ref005]\]. Moreover, vertical HIV transmission to the breastfeeding infant is a potential serious consequence of maternal acute HIV seroconversion \[[@pmed.1002132.ref006]\]. Antiretroviral pre-exposure prophylaxis (PrEP) with emtricitabine (FTC)-tenofovir disoproxil fumarate (TDF) co-formulation or TDF alone is a highly effective strategy to reduce the risk of sexual acquisition of HIV \[[@pmed.1002132.ref007]--[@pmed.1002132.ref012]\]. The recent approval of FTC-TDF PrEP by some regulatory authorities in Africa will accelerate PrEP rollout in this region \[[@pmed.1002132.ref013],[@pmed.1002132.ref014]\]. With expanded access to PrEP, women who are breastfeeding may be prescribed PrEP. However, only limited data are available to assess the safety of PrEP use during breastfeeding. Currently, the United States Centers for Disease Control and Prevention guidelines for PrEP permit preconception use of PrEP after discussion of the risk-benefit balance involved, but have identified the need for additional data on infant drug exposure and safety during maternal FTC-TDF PrEP use during pregnancy and postpartum breastfeeding \[[@pmed.1002132.ref015],[@pmed.1002132.ref016]\]. We investigated whether tenofovir and emtricitabine are excreted into human milk and then absorbed by the breastfeeding infant in clinically significant concentrations in breast milk when the drugs are taken as PrEP by the lactating HIV-uninfected mother. Methods {#sec009} ======= Ethics Statement {#sec010} ---------------- The study protocol was approved by the University of Washington Human Subjects Review Committee, the Uganda National Council of Science and Technology, the Uganda National Drug Authority, the Kenya Medical Research Institute Scientific and Ethics Unit, and the Kenyan Pharmacy and Poisons Board. All women provided written informed consent and the fathers of infants provided written permission to enroll the infant. Population and Study Design {#sec011} --------------------------- This was a prospective, open-label, single-arm study of daily oral FTC-TDF PrEP among 50 HIV-uninfected lactating women and their breastfeeding infants, conducted between January and June 2015 at two clinical research sites in Thika, Kenya and Kampala, Uganda (ClinicalTrials.gov Identifier: NCT02776748). Eligible mothers were HIV seronegative and breastfeeding a singleton healthy infant, of legal age to provide written informed consent, had adequate renal function defined by normal creatinine levels and estimated creatinine clearance ≥60 mL/min, and were not infected with hepatitis B virus. Eligible infants were HIV-uninfected, aged 1--24 wk, born at term to eligible women, and had no serious infections or active clinically significant medical problems. Recruitment into the study was stratified by infant age, with half ≤12 and half 13 to 24 wk, to allow assessment of PrEP pharmacokinetics in breast milk among newborns and infants ages 3--6 mo. Study Procedures {#sec012} ---------------- Consenting HIV-uninfected breastfeeding women were followed with daily directly observed therapy (DOT) oral FTC-TDF PrEP administered at the study clinic for ten consecutive days. The ten-day schedule was chosen to attain drug concentration at steady-state levels (estimated to be achieved after five half-lives) sufficient to address the research question while minimizing potential undue infant drug exposure. PrEP was discontinued thereafter, and no medication was administered directly to infants. Co-formulated FTC and TDF were dosed at 200 mg daily and 300 mg daily, respectively; these doses are the standard doses for prevention and treatment of HIV infection. Maternal blood and breast milk samples were obtained concurrently (i.e., within 30 min of each other) regardless of the timing of food intake (i.e., non-fasting) on the seventh and tenth days. Peak samples were obtained 1--2 h after administration of maternal DOT PrEP, and trough samples were obtained at the end of the dosing interval (i.e., 23 to 24 h after DOT PrEP dose). A single infant blood sample was obtained after the seventh maternal DOT PrEP dose. All collected blood samples were centrifuged immediately at 2,000 relative centrifugal force for 15 min at room temperature, and blood plasma was aliquoted into 1 ml cryovials immediately. Breast milk was obtained by manual expression by the women and aliquoted into 1 ml cryovials immediately. All blood and breast milk specimens were stored below -80°C until testing. During daily follow-up, mothers completed a short quantitative interview about infant well-being, breastfeeding patterns, adverse events, and concomitant medication use. Both the mother and infant were monitored for adverse effects, and the severity of clinical symptoms was scored using the Division of AIDS Table for Grading the Severity of Adult and Pediatric Adverse Events \[[@pmed.1002132.ref017]\]. Laboratory Analytic Methods {#sec013} --------------------------- Tenofovir and emtricitabine concentrations in plasma were quantified via previously validated liquid chromatographic-tandem mass spectrometric (LC-MS/MS) methods at the Clinical Pharmacology Analytical Laboratory at the Johns Hopkins University School of Medicine. Furthermore, LC-MS/MS methods for tenofovir and emtricitabine quantification in whole breast milk were developed and validated in accordance with the recommendations included in the US Food and Drug Administration Guidance for Industry: Bioanalytical Method Validation guidelines \[[@pmed.1002132.ref018]\]. Briefly, tenofovir and emtricitabine were isolated from whole breast milk via protein precipitation and quantified from a blood plasma calibration curve. The lower limits of quantification for tenofovir in plasma and breast milk were 0.31 ng/mL and 1 ng/mL, respectively; and for emtricitabine in plasma and breast milk were 0.31 ng/mL and 5 ng/mL, respectively. Quantification of Infant Drug Exposure {#sec014} -------------------------------------- The primary measure of infant drug exposure through maternal breast milk was the concentrations of tenofovir and emtricitabine in infant plasma. Secondary measures were (1) maternal plasma and whole breast milk tenofovir and emtricitabine concentrations; (2) milk to maternal plasma concentration ratios (M/P); and (3) infant plasma drug-to-milk concentration ratio. To contextualize the clinical significance of the measured drug concentrations, we estimated two additional infant indices. First, we calculated the infant drug dose received from breast milk per day (infant dose), computed as the product of breast milk tenofovir and emtricitabine concentrations and the estimated volume of breast milk consumed by the infant daily. Daily amount of breast milk consumed by the infant was assumed to be 150 mL/kg/day, the standardized milk consumption of the mean milk intake of a fully breastfed infant \[[@pmed.1002132.ref019],[@pmed.1002132.ref020]\]. Second, we calculated infant dose fraction, which is the drug dose a fully breastfed infant would ingest from maternal milk as a fraction of the infant's weight-adjusted therapeutic dose. This was computed from infant daily dose from breast milk and the weight-adjusted recommended therapeutic pediatric doses as infant dose fraction (%) = infant dose from breast milk\100/infant therapeutic dose \[[@pmed.1002132.ref020],[@pmed.1002132.ref021]\]. The respective therapeutic doses for tenofovir and emtricitabine were considered to be 6 mg/kg; these doses were derived from published doses considered be effective for prevention of mother to child transmission of HIV for age equivalent populations and have not shown safety concerns \[[@pmed.1002132.ref022]--[@pmed.1002132.ref025]\]. All outcome measures were evaluated separately for tenofovir and emtricitabine concentrations and stratified by the timing of sample collection (i.e., trough or peak). Statistical Analysis {#sec015} -------------------- The primary outcome was the proportion of infants with detectable steady-state concentrations of tenofovir and emtricitabine in plasma, overall, and stratified by infant age (i.e., ≤12 wk or 13--24 wk). Data were summarized as medians and interquartile ranges (IQR) for continuous variables and proportions for categorical outcomes. For one peak maternal plasma record, the tenofovir concentration (1,040.0 ng/mL) was out of the assay analytic range (0.31--1,000.0 ng/mL). This record was imputed to the upper limit of the assay analytic range. For concentrations below the assay limit of detection, a value of one-half of the detection limit was used in summary calculations for continuous variables; where \>3 samples are below the lower of limit of quantification, the proportion of samples with undetectable levels are instead presented. Mann--Whitney U* test was used to compare the distribution of infant peak concentrations, daily dose from milk, and drug exposure index between the two infant age strata. All analyses were conducted in SAS version 9.4, SAS Institute Inc., Cary, North Carolina, US. Results {#sec016} ======= Population Characteristics and Follow-Up {#sec017} ---------------------------------------- Of the 50 mother--infant pairs enrolled, 24 (48%) infants were ≤12 weeks of age, and median (IQR) weight at study entry was 5 kg (4.3--6.0) for the ≤12 wk group and 6.6 kg (6.0--7.1) for the 13--24 wk group. Infants were breastfed for median of 15 (IQR 12--18) times daily during the week prior to study participation ([Table 1](#pmed.1002132.t001){ref-type="table"}); median daily proportion of infant food intake derived from breastfeeding in the week prior to study entry was 100%. During study follow-up, the median (IQR) daily frequency of infant breastfeeding was 15 times (12--18) overall, 16 (14--19) for the ≤12 wk and 14 (12--17) for the 13--24 wk infant age groups. 10.1371/journal.pmed.1002132.t001 ###### General characteristics. ![](pmed.1002132.t001){#pmed.1002132.t001g} Characteristic All infants (n = 50) Infant age ≤12 wk (n = 24) Infant age 13--24 weeks (n = 26) ------------------------------------------------------- ------------------------ ------------------------------ ------------------------------------ Infant age in weeks 13 (9--19) 9 (6--10) 19 (17--21) Birth weight in kg 3.4 (3.0--3.5) 3.3 (3.0--3.7) 3.4 (2.8--3.5) Infant weight at screening in kg 6.0 (5.0--6.7) 5 (4.3--6.0) 6.6 (6.0--7.1) Maternal age in years 25 (22--28) 24 (22--28) 26 (22--28) Infant length in cm 58 (55--61) 55 (52--58) 60 (58--62) Average daily frequency of breastfeeding, past week 15 (12--18) 16 (8--25) 15 (6--20) Median proportion of infant feed due to breastfeeding 100% (100--100) 100% (100--100) 100% (100--100) Maternal creatinine clearance in mL/min 107 (93--120) 109 (95--120) 105 (93--119) Maternal serum creatinine in mg/dL 0.64 (0.58--0.71) 0.60 (0.57--0.68) 0.66 (0.58--0.72) Maternal AST 21 (19--24) 22 (19--24) 20 (19--24) Maternal ALT 19 (14--23) 19 (14--23) 22 (15--27) Statistics are median (interquartile range) for continuous covariates and percent for binary variables. ALT, Alanine transaminase; AST, Aspartate aminotransferase. Overall, 499 of 500 (\>99%) daily DOT PrEP doses were taken by the mother, and 439 of 450 (98%) expected samples for pharmacokinetic analysis were collected: 195 maternal plasma (98 for peak and 97 for trough); 195 breast milk (98 for peak and 97 for trough); and 49 infant plasma samples. Peak maternal blood, breast milk, and infant blood samples were obtained after a median (IQR) of 63 (60 to 68), 70 (65 to 77), and 80 (45 to 90) min after maternal DOT PrEP dose, respectively, whereas maternal trough samples were obtained a median of 23 h (IQR 23 to 24) from the previous maternal DOT PrEP dose. Tenofovir and Emtricitabine Concentrations in Maternal Plasma and Breast Milk {#sec018} ----------------------------------------------------------------------------- In maternal plasma, tenofovir was detected at concentrations consistent with steady-state use, and breast milk tenofovir concentrations were considerably lower than those in maternal plasma ([Fig 1](#pmed.1002132.g001){ref-type="fig"}). The median (IQR) time-averaged peak steady-state concentrations of tenofovir in maternal plasma and breast milk were 152.0 ng/mL (IQR 56.9 to 321.0) and 3.2 ng/mL (2.3 to 4.7), respectively, resulting in a median peak milk/plasma (M/P) ratio of 0.03 (0.01 to 0.05) ([Table 2](#pmed.1002132.t002){ref-type="table"}). Similarly, median (IQR) time-averaged trough steady-state concentrations of tenofovir in maternal plasma and breast milk were 51.9 ng/mL (IQR 40.7 to 59.6) and 3.3 ng/mL (2.3 to 4.4), respectively, representing a trough median M/P ratio of 0.07 (IQR 0.05 to 0.08). ![Box plot of maternal and infant tenofovir concentrations.\ Non-fasting maternal blood and breast milk samples were obtained concurrently (i.e., within 30 min) at the seventh and tenth visits (corresponding to seventh and tenth maternal DOT PrEP doses). A single infant blood sample was obtained after the seventh maternal DOT PrEP dose. Peak maternal blood, breast milk, and infant blood samples were obtained a median (IQR) of 63 (60 to 68), 70 (65 to 77), and 80 (45 to 90) min after the maternal DOT PrEP dose, respectively. Trough samples were obtained at close of the dosing interval, a median of 23 h (IQR range 23 to 24) after the previous maternal DOT PrEP. One outlier peak maternal plasma tenofovir concentration (1,040 ng/ml) was out of the assay analytic range (0.31--1,000.0 ng/mL). This record was imputed to the upper limit of the assay analytic range and was included in the computation of the displayed summary estimate. Middle box line represents the median. Upper box line represents the 75th percentile and the lower box line represents the 25th percentile. The top whisker denotes the maximum data value or the third quartile plus 1.5 times the interquartile range, whichever is smaller. The lower whisker denotes the minimum data value or the third quartile plus 1.5 times the interquartile range, whichever is larger. The notches display the 95% confidence interval around the median. Small circles represent outlier data points (i.e., observations that are as extreme as ±1.5 of interquartile range). Only 3 of 49 infant plasma samples had quantifiable tenofovir concentration in plasma (infants aged 11 and 13 wk \[both 0.9 ng/mL\] and 17 wks \[17.4 ng/mL\]). NA, not applicable; BLQ, below assay limit of quantification for tenofovir: \<0.31 ng/mL in plasma and \<1 ng/mL in whole milk.](pmed.1002132.g001){#pmed.1002132.g001} 10.1371/journal.pmed.1002132.t002 ###### Tenofovir concentrations and infant exposure. ![](pmed.1002132.t002){#pmed.1002132.t002g} Variable All infants Infant age ≤12 wk Infant age 13--24 wk p-value ------------------------------------------------------------------------------------------------------------------------------------ ------------------------ ------------------------ ------------------------ ----------- *Peak* [\](#t002fn002){ref-type="table-fn"} *n = 98 n = 49 n = 49 Maternal plasma concentration in ng/mL 152.0 (56.9--321.0) 140.5 (53.3--327.5) 165.5 (58.4--309.0) Breast milk concentration in ng/mL 3.2 (2.3--4.7) 3.8 (2.7--6.9) 2.9 (2.1--3.8) M/P concentration ratio 0.03 (0.01--0.05) 0.03 (0.02--0.07) 0.02 (0.01--0.04) Proportion of infant plasma samples with concentration below the lower limit of quantification[†](#t002fn003){ref-type="table-fn"} 94% (46/49) 96% (23/24) 92% (24/25) Infant daily dose from breast milk in μg/kg 0.47 (0.35--0.71) 0.57 (0.41--1.04) 0.44 (0.32--0.56) 0.06 Infant dose fraction[ⱡ](#t002fn004){ref-type="table-fn"} \<0.01% (\<0.01--0.01) \<0.01% (0--0.02) \<0.01% (\<0.01--0.01) 0.06 Trough** [\](#t002fn002){ref-type="table-fn"} *n = 97 n = 48 n = 49** Maternal plasma concentration in ng/mL 51.9 (40.7--59.6) 54.1 (45.7--62.3) 46.0 (39.4--57.2) Breast milk concentration in ng/mL 3.3 (2.3--4.4) 3.5 (2.3--6.8) 3.2 (2.3--3.8) M/P concentration ratio 0.07 (0.05--0.08) 0.07 (\<0.01--0.31) 0.07 (\<0.01--0.11) Infant daily dose from breast milk in μg/kg 0.49 (0.34--0.66) 0.52 (0.05--0.08) 0.49 (0.05--0.08) 0.11 Infant dose fraction[ⱡ](#t002fn004){ref-type="table-fn"} \<0.01% (\<0.01--0.01) \<0.01% (\<0.01--0.01) \<0.01% (\<0.01--0.01) 0.11 Unless stated, statistics are median (interquartile range). n are for samples tested, with each woman providing a maximum of two of respective records (i.e, one on day 7 and another on day 10), while each infant provided one record. \Peak maternal blood, breast milk, and infant blood samples were obtained after a median (IQR) of 63 (60 to 68), 70 (65 to 77), and 80 (45 to 90) min after maternal DOT PrEP dose, respectively, while maternal trough samples were obtained a median of 23 h (IQR 23 to 24) from the previous maternal DOT PrEP dose. ^†^ n* = 49, a single infant plasma sample was obtained: 24 samples for ≤12 wk age group and 25 samples for 13--24 wk age group. Only 3 of 49 infant plasma samples had detectable tenofovir concentration in plasma (Infants aged 11 and 13 wk \[both had 0.9 ng/mL\] and 17 wk \[17.4 ng/mL\]). ^ⱡ^ Infant dose fraction, represents the daily amount of tenofovir dose an infant would be expected to ingest from breast milk as a percentage of the proposed therapeutic daily dose (6 mg/kg). p-Values are from Mann--Whitney U test testing the null hypothesis that the two infant age groups are drawn from the same distribution. M/P, milk to maternal plasma ratio; Lower limit of quantification was \<0.31 ng/mL in plasma and \<1 ng/mL in whole breast milk. Maternal plasma emtricitabine concentrations were also consistent with steady-state use, and emtricitabine concentrations in breast milk were more similar to plasma concentrations than had been seen for tenofovir ([Fig 2](#pmed.1002132.g002){ref-type="fig"}). The median (IQR) time-averaged peak steady-state concentrations of emtricitabine in maternal plasma and breast milk were 267.5 ng/mL (103.0 to 1370.0) and 212.5 ng/mL (140.0 to 405.0), respectively, representing a median peak M/P ratio of 0.63 (0.31 to 1.43) ([Table 3](#pmed.1002132.t003){ref-type="table"}). Similarly, the median (IQR) time-averaged trough steady-state concentrations of emtricitabine in maternal plasma and breast milk were 84.4 ng/mL (68.5 to 99.7) and 183.0 ng/mL (113.0 to 250.0), respectively, representing a median trough M/P ratio of 2.1 (IQR 1.67 to 2.81). Overall, in contrast to maternal plasma concentrations, there was less variability in concentration of both tenofovir and emtricitabine in breast milk (median \[range\] trough to peak breast milk concentration ratio; 1.0 \[0.7 to 1.3\] for tenofovir and 0.8 \[0.4 to 1.3\] for emtricitabine). ![Box plot of maternal and infant emtricitabine concentrations.\ Non-fasting maternal blood and breast milk samples were obtained concurrently (i.e., within 30 min) at the seventh and tenth visits (corresponding to seventh and tenth maternal DOT PrEP doses). A single infant blood sample was obtained after the seventh maternal DOT PrEP dose. Peak maternal blood, breast milk, and infant blood samples were obtained a median (IQR) of 63 (60 to 68), 70 (65 to 77), and 80 (45 to 90) min after the maternal DOT PrEP dose, respectively. Trough samples were obtained at close of the dosing interval, a median of 23 h (IQR range 23 to 24) after the previous maternal DOT PrEP. Middle box line represents the median. Upper box line represents the 75th percentile, and the lower box line represents the 25th percentile. The top whisker denotes the maximum data value or the third quartile plus 1.5 times the interquartile range, whichever is smaller. The lower whisker denotes the minimum data value or the third quartile plus 1.5 times the interquartile range, whichever is larger. The notches display the 95% confidence interval around the median. Small circles represent outlier data points (i.e., observations that are as extreme as ±1.5 of interquartile range). NA, not applicable; BLQ, below assay limit of quantification for emtricitabine: \<0.31 ng/mL in plasma and \<5 ng/mL in whole milk.](pmed.1002132.g002){#pmed.1002132.g002} 10.1371/journal.pmed.1002132.t003 ###### Emtricitabine concentrations and infant exposure. ![](pmed.1002132.t003){#pmed.1002132.t003g} Variable All infants Infant age ≤12 wk Infant age 13--24 wk p-value -------------------------------------------------------------------------- ----------------------- ---------------------- ----------------------- ----------- *Peak* [\](#t003fn002){ref-type="table-fn"} *n = 98 n = 49 n = 49 Maternal plasma concentration in ng/mL 267.5 (103.0--1370.0) 236.5 (93.6--1380.0) 533.0 (115.0--1370.0) Breast milk concentration in ng/mL 212.5 (140.0--405.0) 208.0 (139.5--377.5) 215.0 (149.0--431.0) M/P concentration ratio 0.63 (0.31--1.43) 0.70 (0.31--1.76) 0.59 (0.31--1.14) Infant plasma concentration[ⱡ](#t003fn003){ref-type="table-fn"} in ng/mL 13.2 (9.3--16.7) 16.6 (13.2--20.9) 10.5 (7.1--13.2) \<0.01 Infant plasma/milk concentration ratio 0.05 (0.03--0.08) 0.07 (0.04--0.10) 0.05 (0.02--0.06) 0.12 Infant daily dose from breast milk in μg/kg 31.9 (21.0--60.8) 31.2 (20.9--56.6) 32.3 (22.4--64.7) 0.94 Infant dose fraction[†](#t003fn004){ref-type="table-fn"} 0.5% (0.3--1.0) 0.5% (0.3--0.9) 0.5% (0.4--1.1) 0.94 Trough** [\](#t003fn002){ref-type="table-fn"} *n = 97 n = 48 n = 49** Maternal plasma concentration in ng/mL 84.4 (68.5--99.7) 82.8 (69.3--101.0) 84.8 (68.2--97.5) Breast milk concentration in ng/mL 183.0 (113.0--250.0) 187.5 (95.6--256.0) 183.0 (125.0--250.0) M/P concentration ratio 2.10 (1.67--2.81) 2.36 (1.48--2.83) 2.08 (1.69--2.81) Infant daily dose from breast milk in μg/kg 27.5 (17.0--37.5) 28.1 (14.3--38.4) 27.5 (18.9--37.5) 0.58 Infant dose fraction[†](#t003fn004){ref-type="table-fn"} 0.5% (0.3--0.6) 0.5% (0.2--0.6) 0.5% (0.3--0.6) 0.58 Unless stated, statistics are median (interquartile range); n are for samples tested, with each woman providing a maximum of two of respective records (i.e, one on day 7 and another on day 10), while each infant provided one record. \Peak maternal blood, breast milk, and infant blood samples were obtained after a median (IQR) of 63 (60 to 68), 70 (65 to 77), and 80 (45 to 90) min after maternal DOT PrEP dose, respectively, while maternal trough samples were obtained a median of 23 h (IQR 23 to 24) from the previous maternal DOT PrEP dose. ^ⱡ^ n* = 49, a single infant plasma sample was obtained: 24 samples for ≤12 wk age group and 25 for 13--24 wk age group. Emtricitabine was unquantifiable in 2 of 49 infant plasma samples. ^†^Infant dose fraction (also called exposure index) represents the daily amount of emtricitabine dose an infant would ingest from breast milk as a percentage of the proposed pediatric therapeutic daily dose (6 mg/kg). p-values are from Mann--Whitney U test testing the null hypothesis that the two infant age groups are drawn from the same distribution. M/P milk to maternal plasma ratio; Lower limit of quantification was \<0.31 ng/mL in plasma and \<5 ng/mL in whole breast milk. Infant Exposure to Tenofovir and Emtricitabine from Maternal Breast Milk {#sec019} ------------------------------------------------------------------------ Overall, after seven consecutive maternal daily DOT FTC-TDF PrEP, tenofovir was undetectable in 46 of 49 (94%) infant plasma samples; the three infants with detectable tenofovir also had detectable emtricitabine. These three infants were ages 11, 13, and 17 wk (plasma concentrations 0.9, 0.9, and 17.4 ng/mL, respectively, and body weight 6.4, 5.8, and 6.2 kg, respectively), and their maternal milk tenofovir concentrations were modestly greater than the overall median (6.57, 3.64, and 4.05 ng/mL, respectively, versus median 3.2 ng/mL). There were no other notable unique characteristics between these three mother--infant pairs and the others. The median amount of tenofovir dose estimated to be ingested by an infant from breast milk was 0.47 μg/kg (IQR 0.35 to 0.71), translating into \<0.01% (i.e., 12,500-fold lower) of the proposed pediatric tenofovir therapeutic daily dose (6 mg/kg) \[[@pmed.1002132.ref024],[@pmed.1002132.ref025]\]. Specifically, a 5-kg body weight infant would be expected to ingest a total tenofovir dose of 2.35 x 10^−3^ mg daily from breast milk compared to the proposed therapeutic daily dose of 30 mg. Emtricitabine was detectable in 47 of 49 (96%) infant plasma samples. The median (IQR) emtricitabine concentration in infant plasma was 13.2 ng/mL (9.3 to 16.7) overall, approximately 5% of breast milk concentrations: 16.6 ng/mL for infants aged ≤12 wk and 10.5 ng/mL in infants 13--24 wk. Based on the steady-state concentrations, the estimated median dose of emtricitabine expected to be ingested by the infant per day from breastfeeding was 31.9 μg/kg (IQR 21.0 to 60.8), translating into 0.5% (i.e., 200-fold lower) of the proposed pediatric emtricitabine therapeutic daily dose (6 mg/kg) \[[@pmed.1002132.ref022],[@pmed.1002132.ref024]\]; the estimated doses were similar in the two infant age groups ([Table 3](#pmed.1002132.t003){ref-type="table"}). Specifically, a 5-kg body weight infant would ingest from breastfeeding a total daily emtricitabine dose of 0.16 mg compared to the recommended therapeutic dose of 30 mg per day. Safety and Tolerance {#sec020} -------------------- FTC-TDF was well tolerated by study mothers and infants. Over the ten-day maternal FTC-TDF PrEP dosing period, clinical symptoms recorded on ≥2 occasions were abdominal pain, diarrhea, and nausea in three (6%), two (4%), and three (6%) women, respectively (abdominal pain and nausea were concurrent in two women). In two infants (4%), diarrhea was reported on two visits during the study duration. These symptoms in both mother and infant were mild and resolved in 2--3 d. Of 50 women, 48 completed a safety kidney function screen at exit. Calculated creatinine clearance was \>90 mL/min at baseline and exit for all women (median: serum creatinine \[0.64 versus 0.66 mg/dL\] and creatinine clearance \[107 versus 101 ml/min\]). Discussion {#sec021} ========== In this prospective study of daily, directly observed doses of daily oral FTC-TDF PrEP in HIV-uninfected breastfeeding women, the estimated infant doses from breastfeeding and the resultant infant plasma concentrations for both tenofovir and emtricitabine were far below what would result from the proposed pediatric doses. Based on breast milk concentration measurements, the estimated daily tenofovir and emtricitabine doses ingested by the infant through breastfeeding were 12,500-fold and 200-fold, respectively, lower than the proposed daily pediatric dose for prophylaxis against vertical HIV acquisition. Thus, infants had low exposures to tenofovir and emtricitabine, which would not be expected to pose substantial safety risk to infants of mothers who use PrEP during breastfeeding. To our knowledge, this is the first study to directly assess infant drug exposure via breast milk of mothers using FTC-TDF PrEP. We implemented an intensive daily maternal DOT PrEP dosing schedule to remove variability due to adherence. Daily oral PrEP offers an effective female-controlled option to reduce the risk of sexual HIV acquisition for women who are pregnant or breastfeeding, with the advantage relative to other prevention methods that it does not require cooperation of sexual partners. These data provide important empirical evidence to inform the discussion and assessment of risk-to-benefit balance of initiating or continuing PrEP during breastfeeding and are informative for evidence-based clinical guidelines. Although we were unable to implement a full concentration--time pharmacokinetic profile approach, our data collected at steady state demonstrate minimal variation in the concentrations of tenofovir and emtricitabine in breast milk, indicating that infants acquired consistent drug dosing throughout the day via breast milk. Thus, our findings suggest that PrEP can be safely used during breastfeeding with minimal infant exposure. Our study provides both novel and complementary findings to the Agence Nationale de Recherche sur le Sida (ANRS) 12109 study \[[@pmed.1002132.ref026]\], a pharmacokinetic study that assessed tenofovir and emtricitabine exposure in five HIV-infected Ivorian women dosed at 400 mg FTC--600 mg TDF at the start of labor and 200 mg FTC--300 mg TDF daily for 7 d postpartum. In the 16 breast milk samples obtained in that study, simulated peak median infant tenofovir and emtricitabine daily doses from breast milk were 4.2 μg/kg and 146 μg/kg, respectively, which represented 0.03% and 2% of the respective therapeutic oral infant doses. Notably, these estimated infant doses are larger than those we found in this study (estimated infant daily doses from breast milk: tenofovir = 0.47 μg/kg and emtricitabine = 31.9 μg/kg). One explanation is the difference in dosing and sampling schedules between the two studies. Alternatively, the difference could mean that infants are exposed to far smaller tenofovir and emtricitabine concentrations from breast milk based on direct plasma measure in our study than anticipated from the simulated doses in the ANRS study. Importantly, infant plasma drug concentrations were not directly measured in that study. For breastfeeding women taking oral TDF, breast milk will exclusively contain tenofovir in an unesterified anionic form, and, due to its poor oral bioavailability, negligible tenofovir concentrations would be expected to be absorbed by the infant from breastfeeding, consistent with our findings. In contrast, emtricitabine, which has excellent bioavailability, would be expected be absorbed to some degree from breast milk by the infant, as has been observed with the structurally similar lamivudine and which was seen in this study. Although emtricitabine concentrations were quantified in infant plasma, the concentration we observed in this study was a small fraction (\~0.5%) of the infant therapeutic daily doses used to treat HIV. For most drugs, including tenofovir and emtricitabine, the dose below which there is no clinical effect in infants is unknown. A dose fraction index (exposure of 10% weight-adjusted therapeutic pediatric dose) has been proposed as a safety threshold for infant exposure to maternal drugs from breast milk, below which the degree of exposure to the drug in breast milk is considered clinically unimportant \[[@pmed.1002132.ref027]\]. In this study, we found infant plasma tenofovir and emtricitabine concentrations to be only \<0.01% and 0.5% of the respective proposed therapeutic pediatric doses. Accordingly, for TDF-based PrEP use during lactation, the small concentrations of tenofovir and emtricitabine absorbed by infants from maternal breast milk observed in our study are likely to be of limited clinical consequence. Our results must be interpreted in light of the following limitations. First, we only tested for plasma tenofovir and emtricitabine concentrations, not their pharmacologically active intracellular derivatives, tenofovir-triphosphate and emtricitabine-diphosphate concentrations, respectively. Second, we collected only a single infant sample to minimize venipunctures for children. In sparse data pharmacokinetic situations like our study, in which the traditional full drug concentration--time profile approach is not applicable, daily DOT and a steady-state sampling provide an adequate approach to address our key research question. Importantly, there was minimal variation in the concentrations of tenofovir and emtricitabine in breast milk, demonstrating that infants were exposed to consistent tenofovir and emtricitabine dosing throughout the day via breast milk. Third, quantifying the volume of milk intake was not feasible, so we used the standard assumption of 150 mL/kg/day breast milk intake of a fully fed infant. Fourth, we only tested peak and trough maternal concentrations, which limits the precision of our M/P estimates throughout a dosing interval. Fifth, maternal blood and breast milk samples were obtained concurrently (i.e., within 30 min of each other). It is possible that a lag in blood-to-breast milk excretion could impact the observed breast milk concentrations. Other studies have used a similar approach when sampling breast milk, and this lag was not readily evident in a previous animal model of FTC-TDF breast milk excretion in two macaques \[[@pmed.1002132.ref026],[@pmed.1002132.ref028]\]. Similar to our findings, minimal variations in breast milk concentrations of nucleoside/nucleotide reverse transcriptase inhibitor concentrations have previously been reported in HIV-infected breastfeeding women, potentially owing to the slow elimination from breast milk \[[@pmed.1002132.ref029]\]. Thus, given the steady-state sampling approach in our study, any potential lag in blood-to-breast milk excretion would be expected to have minimal effects on our findings. In conclusion, in this prospective study among HIV-uninfected breastfeeding African women using DOT FTC-TDF PrEP, nursing infants were exposed to lower tenofovir and emtricitabine concentrations from breastfeeding than the proposed pediatric therapeutic doses. These data provide evidence suggesting that this PrEP regimen can be safely used during breastfeeding, which is informative for clinical guidelines for women who are at substantial risk of HIV during pregnancy and the postpartum period. Supporting Information {#sec022} ====================== ###### Data underlying Figs [1](#pmed.1002132.g001){ref-type="fig"} and [2](#pmed.1002132.g002){ref-type="fig"}. (XLSX) ###### Click here for additional data file. ###### STROBE checklist. (DOCX) ###### Click here for additional data file. ###### Study protocol. (PDF) ###### Click here for additional data file. The authors thank the study participants and the study teams at the University of Washington, the Partners in Health Research and Development, Thika, Kenya, and the Kasangati-Infectious Diseases Institute, Kampala, Uganda. BLQ : below limit of quantification DOT : directly observed therapy FTC : emtricitabine IQR : interquartile range LC-MS/MS : liquid chromatography-tandem mass spectrometry M/P : maternal plasma concentration ratio PrEP : pre-exposure prophylaxis TDF : tenofovir disoproxil fumarate [^1]: I have read the journal\'s policy, and the authors of this manuscript have the following competing interests: FTC-TDF was donated by Gilead Sciences. CWH reports a prior contract from Gilead Sciences outside the submitted work and a patent pending, both managed by Johns Hopkins. MM has grant funding via the NIH. GJS has research grants from NIH (unrelated), CDC (unrelated), Thrasher Foundation (unrelated), Bill and Melinda Gates Foundation (sponsor), royalties from UpToDate (unrelated), and salary support from the University of Washington. All authors declare no other conflicts of interest. [^2]: Conceived and designed the experiments: KKM CWH JMB.Performed the experiments: KKM CWH MM JMB.Analyzed the data: KKM CWH MM JMB.Contributed reagents/materials/analysis tools: KKM CWH MM CLC JMB.Wrote the first draft of the manuscript: KKM CWH JMB.Contributed to the writing of the manuscript: KKM CWH NRM MM ETK GJS KN NBS TRM GM AS CLC JMB.Enrolled patients: KKM NRM ETK KN NBS TRM GM.Agree with the manuscript's results and conclusions: KKM CWH NRM MM ETK GJS KN NBS TRM GM AS CLC JMB.All authors have read, and confirm that they meet, ICMJE criteria for authorship.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== Thorough knowledge and understanding of pulp chamber and root canal system anatomy are essential for successful root canal therapy. Familiarity with variations in tooth anatomy and characteristic features in various racial groups can aid location and negotiation of canals \[[@B1]\]. Missed extra root canals are major reason for endodontic failure \[[@B2]\]. Pulp canal system is complex with branching and divisions throughout the root length. Vertucci (1984) classified the root canal configurations of human permanent teeth into various types ranging from single to three separate distinct canals \[[@B3]\]. Permanent maxillary canines are more commonly single rooted, single canal teeth. Presence of two root canals is a rarity \[[@B4]\]. Majority of them join in apical third and exit as single apical foramen \[[@B5]\]. This case report presents a permanent maxillary canine having two root canals exiting as single foramen. 2. Case Report {#sec2} ============== A 35-year-old female patient reported to dental clinic with a chief complaint of spontaneous pain from four days in maxillary left anterior region. Subjective symptoms include sharp, severe, continuous, throbbing pain and are aggravated by taking hot foods and relieved by medication. Past medical history was noncontributory. Oral examination revealed a deep carious lesion involving maxillary left canine (23, FDI System). The teeth was asymptomatic to palpation and tested negative using electrical pulp tester and gave an exaggerated severe response to cold test. No mobility was seen and tested negative for percussion. Radiographic examination revealed an abnormal root canal morphology. A diagnosis of acute irreversible pulpitis of 23 was established and endodontic treatment was planned. Following local anesthesia with 2% lidocaine, the tooth was isolated with rubber dam \[Hygienic; Coltene Whaledent\], and an endodontic access was made on palatal side with \#1014 round bur and endo-Z carbide bur. The vital pulp tissue was extirpated and initially two canal orifices were located. Working length of 31 mm was measured by using K flex files \#15 of length 31 mm ([Figure 1](#fig1){ref-type="fig"}). The palatal canal joined the buccal canal in apical third of root canal (Vertucci type II canal configuration). Crown down root canal preparation was done; the coronal and middle thirds were prepared using Gates Glidden drills \#1--3 \[Tulsa Dental, Dentsply\] and apical preparation by hand K flex files \[Dentsply\] to size \#40. Copious irrigation with 5.25% sodium hypochlorite, 17% EDTA were performed after use of each file. 2% chlorhexidine was used as final irrigant. The canals were dried with paper points and obturated with gutta-percha and AH Plus Sealer ([Figure 2](#fig2){ref-type="fig"}) \[Dentsply De Trey GmbH, Konstanz, Germany\] using a lateral condensation technique ([Figure 3](#fig3){ref-type="fig"}). The patient was asymptomatic during the 3-month follow-up. 3. Discussion {#sec3} ============= Debridement of root canal to remove pulpal remnants, bacteria, and their byproducts before obturation is primary requisite for successful endodontic treatment. Being unable to locate and fill a canal results in failure of root canal therapy. Therefore, it is imperative to have knowledge of anatomic variations as endodontic success is related to canal debridement \[[@B6]\]. The diagnostic difficulty and possible canal superimposition on radiographic examination should be kept in mind when examining such cases. When locating extra canals, identification of periodontal ligament space that often projects onto root surface resembling a canal should be differentiated. Vertucci (1984) classified root canals according to number of canals present and their configuration into eight types \[[@B7]\]. Anatomic anomaly observed at first appointment should be checked for similar anomaly of tooth on the other side \[[@B8]\]. In the present case, radiographic image showed bilateral Vertucci type II \[2-1\] configuration. Çalişkan et al. \[[@B9]\] studied root canal number, configuration, and ramifications of permanent teeth in Turkish population. They reported percentage of Vertucci type III \[1-2-1\] and type V \[1-2\] as 4.35 and 2.17, respectively. Alapati et al. \[[@B10]\] and Onay and Ungor \[[@B11]\] reported a maxillary right canine with type II canal configuration and Weisman \[[@B12]\] also reported a birooted maxillary left canine. In the present case, the maxillary canines had an unusual root length of 31 mm which necessitated use of lengthier k flex files of 31 mm removing silicone directional stopper before use. They showed canal configuration of type II similar to that reported by Alapati et al. and Onay and Ungor. Two distinct canal orifices were located in labial/palatal direction which joined in apical third, forming a type II configuration. Teeth with type II configuration during treatment may pose problems. The canal that is in line with the main passage is usually amenable to adequate enlarging and obturation procedures; the preparation and filling of the other canal are often extremely difficult \[[@B13]\]. A thorough knowledge of root canal anatomy and operator skill are essential for endodontic success. Careful clinical examination with radiographs from several different angles may lead to suspicion or identification of additional canals and leads to higher possible success \[[@B3]\]. 4. Conclusion {#sec4} ============= Several variations exist in the root canal system and clinicians should be aware of the variations for complete infection removal and prevention of reinfection. Special care with careful endodontic exploration, different angle radiographs, and magnification with surgical microscope aids in detection and treatment of extra canals. Conflict of Interests ===================== The authors declare that there is no conflict of interests regarding the publication of this paper. ![Radiograph showing working length of maxillary canine with two separate canals.](CRID2015-780908.001){#fig1} ![Radiograph showing master cone gutta-percha.](CRID2015-780908.002){#fig2} ![Postoperative radiograph obturated with gutta-percha and AH Plus Sealer.](CRID2015-780908.003){#fig3} [^1]: Academic Editor: Malka Ashkenazi	{ "pile_set_name": "PubMed Central" }
Introduction ============ Inflammation is controlled by a balance of proinflammatory and antiinflammatory signals, resulting in the development of an immune response followed by temporally released proresolution factors that lead to inflammation switching off and injured tissues returning to normal physiology.^[@B1]^ In this setting, attention is traditionally placed on phagocytes such as polymorphonuclear leukocytes (PMNs) and monocyte-derived macrophages, with comparatively less weight place on the importance of lymphocytes in acute inflammation and its resolution. Given their roles in early defense to bacteria and viruses, innate-type lymphocytes including B1 cells merit further exploration for their potential roles in host defense and restorative physiology, as it is becoming clear that diminished innate lymphocyte function or enhanced lymphocyte death by apoptosis, for instance, has being postulated to play a central role in the pathogenesis of burn injury^[@B2],[@B3]^ and sepsis,^[@B4],[@B5]^ respectively One of the major impetuses for this current investigation stemmed from our previous observations^[@B6]^ and those made by others^[@B7][@B8][@B9]--[@B10]^ showing lymphocytes repopulating sites of tissue injury as inflammation abates, suggesting that lymphocytes might help to switch off acute inflammation. Investigating this possibility in experimental peritonitis, T and B lymphocytes, normal residents of the naive peritoneum, were found to regulate the severity of the early onset phase of acute inflammation by elaborating anti-inflammatory cytokines and dampening PMN influx. However, once PMNs begin to accumulate, resident lymphocytes disappear in response to PGD~2~ working through its DP1 receptor. As inflammation resolves, a unique profile of lymphocytes begin to repopulate the cavity. Repopulating lymphocytes, however, do not bring about resolution but replenish resolving tissues with the necessary cellular players (CD3, B1 cells, natural killer \[NK\] and γ/δ T cells as well as CD4^+^/CD25^+^ T cells) to control future innate immune-mediated responses. We provide relevance of these findings to human health by showing the absence of repopulating lymphocytes in nonresolving inflammation, which predisposes to secondary infection, resulting in severe inflammatory responses. Although lymphocytes in adaptive immunity are well understood, their role in innate immunity and resolution is highlighted here, as is their functional control by lipid mediators, of which there is a growing body of evidence. For instance, in addition to PGD~2~,^[@B11]^ lipoxins and aspirin-triggered epi-lipoxins inhibit human T-cell TNFα secretion,^[@B12]^ while docosahexanoic acid--derived protectin D1 blocks T-cell migration and TNFα and IFNγ secretion and promotes apoptosis in human T cells.^[@B13]^ Thus, from these current studies and those published by others,^[@B12],[@B13]^ we highlight the regulatory role played by n-3 and n-6 polyunsaturated fatty acid metabolites on lymphocyte function during acute inflammation and its resolution. Methods ======= Animal maintenance, induction of inflammation human peritonitis sampling ------------------------------------------------------------------------ Hematopoietic prostaglandin D~2~ (PGD~2~) synthase (hPGD~2~S) knockout mice were generated as previously described.^[@B14]^ All other animals were bred under standard conditions and maintained in a 12-hour/12-hour light/dark cycle at 22 (±1)°C and given food and tap water ad libitum in accordance with United Kingdom Home Office regulations. Peritonitis was induced by the intraperitoneal injection of either type A zymosan (1 mg for all experiments unless otherwise stated), group B streptococcus (GBS), or LPS (1 mg/kg; Sigma-Aldrich, St Louis, MO), and cells were enumerated by haemocytometer at time points stated in "Results" by sterile phosphate-buffered saline (PBS) washout. Ethical approval (P/03/136A) for collection of human peritonitis samples was obtained from St Bartholomew\'s and the Royal London Hospitals from patients with end-stage renal failure undergoing peritoneal dialysis. For pharmacologic rescue experiments, BW245C (DP1 agonist^[@B15],[@B16]^) or 15(R)-15-methyl PGD~2~ (DP2 agonist^[@B17]^) in 100 μL of PBS (pH 7.2)/BSA 0.1% was injected at equal doses at 30 minutes prior to and after the zymosan. Trypsinization of peritoneal cavity ----------------------------------- To determining the fate of peritoneal T and B cells, cavities of mice bearing a 4-hour zymosan-induced peritonitis was lavaged with sterile PBS to remove accumulated inflammatory cells and edema. A total of 5 mL prewarmed 5% trypsin was then added to the peritoneal cavity for 10 minutes, followed by an equal volume complete medium to acquire cells adhered to the peritoneal lining/greater omental lymphoid organ. Cells were then analyzed for composition by fluorescence-activated cell sorter (FACS). FACS analysis and cytokine/chemokine analysis --------------------------------------------- Cytokines were measured by enzyme-linked immunosorbent assay (ELISA) according to the manufacturer\'s instructions (eBioscience, San Diego, CA). FACS was carried out on a Becton Dickinson Facscalibur with data analyzed by CellQuest (BD, Franklin Lakes, NJ). Leukocytes were incubated with antibodies for 30 minutes to either CD3/CD19 (Serotec, Oxford, United Kingdom), B cells (Ly220; Serotec), CD5 (BD PharMingen, San Diego, CA), MAC-1/CD11b (BD PharMingen), NK and γ/δ cells (gift from Dr T. Hussell, Kennedy Institute, London, United Kingdom), GR1 (BD PharMingen), or F4/80 (Caltag Laboratories, Burlingame, CA) using respective isotype antibodies as controls (Serotec) and compensated as appropriate for dual labeling. For apoptosis, cells were incubated with annexin V/propidium iodide (BD PharMingen) and analyzed on Becton Dickinson Facscalibur with data analyzed by CellQuest. Leukocyte separation and analysis --------------------------------- Contents of resolving-phase peritoneal cavities of wild-type animals were isolated, and macrophages were separated from remaining lymphocytes by adherence to the bottom of 6-well tissue-culture plates. Nonadherent cells were removed and used to isolate T and B cells as well as NK and γ/δ T cells for transfer back in to gp91^phox^ knockout mice using FACS and relevant antibodies to confirm that their composition and ratios reflects that present in situ at resolution. Resolving-phase lymphocytes were enriched at a concentration of 10^6^/mL and 0.5 mL injected into gp91^phox^ knockout mice. In addition, one of the problems with identifying discrete populations of cells such as those found at sites of inflammation by FACS from a larger mixed cell population is that the fluorescence of one cell type after labeling with a fluorescent antibody may be masked by the natural fluorescence of others. In order to confirm the cell types identified using cell-surface antigen markers, peritoneal lymphocytes and macrophages were also isolated and put back into the FACS with their forward- and side-scatter signatures compared against specifically labeled cells. Thus, after lymphocytes were removed from 6-well plates, adherent cells, mainly macrophages, were eluted with Versene, washed with 2% fetal calf serum (FCS) in Hanks balanced salt solution (HBSS) and resuspended in Dulbecco modified Eagle medium (DMEM). These cells were further depleted of contaminant T and B cells using magnetic beads coated with rat monoclonal antibodies to mouse CD3 or B220 (Dynal Biotech, Paisley, United Kingdom). T and B lymphocytes were isolated using the Dynal mouse B-cell (or T-cell) negative isolation kit according to the manufacturer\'s instructions (Dynal Biotech). In brief, a mixture of rat monoclonal antibodies with specificity to all mouse non-B cells (or non-T cells when isolating T cells) cells were added to cell suspensions and incubated for 20 minutes at 4°C. Cells coated with the added monoclonal antibodies were then removed with magnetic beads coated with sheep polyclonal antibody to rat Ig. Purity of the cells were regularly greater than 95%. Bacterial culturing ------------------- The clinical GBS isolate, NCTC10/84 (serotype V) was grown in Todd Hewitt Broth (THB) without agitation at 37°C to an OD~600~ of 0.4, equivalent to 10^8^ cfu/mL. Bacteria collected by centrifugation were washed with sterile PBS. Mice were inoculated intraperitoneally with 3 × 10^7^ cfu NCTC in 30 μL PBS. For survival experiments, mice were inoculated by intraperitoneal injection with 5 × 10^7^ cfu NCTC in 0.3 mL PBS. Results ======= Biphasic trafficking of lymphocytes during acute inflammation ------------------------------------------------------------- We^[@B6]^ and others^[@B7][@B8][@B9]--[@B10]^ have previously observed lymphocytes repopulating sites of injury as inflammation resolves, but without attributing a functional relevance to their reappearance. Trafficking of lymphocytes during resolution is apparent in inflammation associated with human chronic ambulatory peritoneal dialysis where, as inflammation decreases ([Figure 1](#F1){ref-type="fig"}A,B), lymphocyte numbers increase ([Figure 1](#F1){ref-type="fig"}C). Clinical assessment of these patients was based on patients presenting with abdominal pain, cloudy dialysate, and leukocyte count of more than 100/mm^3^. In all cases, peritonitis resolved by day 5 as determined by the appearance of a clear dialysate and abatement of abdominal symptoms. Similar results were evident in resolving experimental murine peritonitis ([Figure 1](#F1){ref-type="fig"}D). Experiments were therefore carried out to establish the role these cells play in innate immune-mediated inflammation by characterizing, in the first instance, the profile of lymphocyte populations throughout the time course of an acute zymosan-induced peritonitis. In the naive murine peritoneum (0 hours), lymphocytes constitute about 50% of the total cell population, with the remaining being resident macrophages. Of the CD3 cells, CD4^+^ and CD8^+^ cells were found as well as lower numbers of CD4^+^/CD25^+^, γ/δ T cells, and NK cells ([Figure 1](#F1){ref-type="fig"}E-J). However, the majority of lymphocytes in the naive cavity are B cells constituting about 70% to 80% of the total lymphocyte population labeling positively for CD19 as well as B220 ([Figure 1](#F1){ref-type="fig"}K). Of these B cells, about 80% are B220^low^/CD5^+^/MAC-1^low^, indicative of a B1 phenotype, with the remainder being B220^high^/MAC-1^−^ B2 cells. As inflammation initiates (1-4 hours), T and B cells disappear but repopulate the peritoneum again between 12 and 24 hours ([Figure 1](#F1){ref-type="fig"}D). Notably, there were more CD4^+^/CD25^+^, γ/δ T cells, and NK cells found during resolution ([Figure 1](#F1){ref-type="fig"}H-J) as well as B1 cells expressing higher levels of MAC-1 than in the naive state ([Figure 1](#F1){ref-type="fig"}K), collectively referred to hereafter as resolution-phase lymphocytes. Figure S6 (available on the Blood website; see the Supplemental Materials link at the top of the online article) shows FACS analysis of lymphocyte cell-surface labeling. As inflammation peaks between 6 and 12 hours in this model and subsequently resolves ([Figure 1](#F1){ref-type="fig"}D), it is argued that lymphocytes repopulate the peritoneum during or just after resolution occurs. Indeed, experimentally enhancing the severity of the inflammatory response within the peritoneum by injecting 3 separate doses of zymosan (0.1, 1.0, and 10 mg), therefore prolonging resolution, is associated with reduced lymphocyte repopulation ([Figure 1](#F1){ref-type="fig"}L). Whether this is a delay or suppression is unclear at this stage. However, from the data presented here, we suggest that lymphocytes repopulate sites of inflammation once resolution occurs. Therefore, this questions what controls lymphocyte influx, which we suggest may be factors released by stromal and/or hematopoietic cells once restitution processes are under way. If so, lymphocyte repopulation is an active process and, until resolution occurs, lymphocyte repopulation is suppressed. Thus, from these experiments we show a shift in lymphocyte populations from the naive to a resolving state constituting more innate-type lymphocytes as well as a different phenotype of B1 cells. ![Biphasic trafficking of lymphocytes during acute inflammation. (A-C) Patients undergoing chronic ambulatory peritoneal dialysis and who developed peritonitis that resolved were found to have lymphocytes at the time of recovery of clinical symptoms. Pursuing this observation in murine zymosan-induced peritonitis (D), T and B cells native to the naive peritoneal cavity were found to disappear within hours of stimulus injection. Once inflammation resolves, lymphocytes repopulate the peritoneum but comprise more (E-J) CD4^+^/CD25 and γ/δ T cells as well as NK cells, than is present in the naive cavity (0 hours) as well as (K) MAC-1^+^ B cells. (L) Experimentally enhancing the severity of the inflammatory response within the peritoneum by injecting 3 separate doses of zymosan (0.1, 1.0, and 10 mg) prolonged resolution and delayed lymphocyte repopulation (72 hours), suggesting that lymphocytes repopulate only after resolution occurs. n = 6 to 8 animals per group; \*P ≤ .05; \\P ≤ .01, as determined by analysis of variance (ANOVA), followed by Bonferroni t test, with data expressed as means plus or minus SEM.](zh80070817490001){#F1} Lymphocytes are dispensable for resolution but mediate responses to superinfection ---------------------------------------------------------------------------------- To examine the role of lymphocytes in acute inflammation, zymosan was injected into the peritoneal cavity of lymphocyte-deficient RAG1^−/−^ mice. Inflammation at onset was greatly exaggerated in these animals, twice that in wild-type ([Figure 2](#F2){ref-type="fig"}A), with the principal cell type being PMNs ([Figure 2](#F2){ref-type="fig"}B). This exaggerated response in RAG1 knockout mice was associated with decreased exudate IL-10 and elevated TNFα levels ([Figure 2](#F2){ref-type="fig"}C). Surprisingly, inflammation in wild-type and RAG1^−/−^ mice resolved uniformly from 24 hours onward ([Figure 2](#F2){ref-type="fig"}A,B), suggesting that lymphocytes have no role in switching off acute inflammation and that regardless of how many PMNs traffic to sites of tissue injury, resolution pathways are sufficiently adept at dealing with their disposal. Pursuing the idea that lymphocytes may protect against secondary infection, in a second experiment, RAG1^−/−^ and wild-type mice were injected with a sublethal dose of GBS 48 hours after zymosan injection. Thus, live bacteria were introduced into the inflamed cavity as inflammation resolved, and its effects determined 24 hours later ([Figure 2](#F2){ref-type="fig"}D). Wild-type mice that got zymosan followed by bacteria displayed fewer signs of illness compared with those that received GBS alone, which exhibited 50% mortality by 24 hours, with the remaining animals dying by 48 hours. However, injecting GBS into RAG1^−/−^ mice 48 hours after receiving zymosan showed an approximate doubling of inflammatory cell accumulation compared with wild-type mice treated in the same way ([Figure 2](#F2){ref-type="fig"}D). This resulted in a lower bacterial load in the plasma of GBS-treated RAG1^−/−^ ([Figure 2](#F2){ref-type="fig"}E) but substantially accelerated mortality due to the concomitant hyperinflammatory response ([Figure 2](#F2){ref-type="fig"}F). Results from these studies suggest that inflammation in the resolving phase exerts greater defense against bacterial infection and lethality than do naive tissues. In addition, in terms of controlling initial leukocyte trafficking in response to nonspecific stimuli, protection is conferred by resident lymphocytes, with resistance to secondary infection exerted by repopulating, resolution-phase lymphocytes. Therefore, lymphocytes modulate host responses to injury/infection but are not required to bring about resolution (ie, clear PMNs and macrophages from inflamed sites). The different profile and proportion of lymphocyte repopulating during resolution is necessary to combat secondary infection. ![Lymphocytes control early onset of innate inflammation but are dispensable for its resolution. (A,B) Although inflammation doubled in RAG1^−/−^ during the onset phase in response to zymosan, coincident with (C) an imbalance of IL-10 versus TNFα, inflammation normalized with that of wild-type mice from 24 hours onward, suggesting no role for lymphocytes in actively bringing about resolution (ie, clearing PMNs or macrophages). (D) However, RAG1^−/−^ and wild-type mice were injected with GBS during resolution (48 hours after zymosan injection), resulting in enhanced leukocyte accumulation in RAG1^−/−^ but not wild-type mice 24 hours later. (E) This was associated with reduced bacterial colonization in plasma but (F) increased mortality in RAG1^−/−^ mice as a result of the concomitant hyperinflammatory response. We therefore argue that lymphocytes are not required for bringing about resolution and propose that their reappearance hails the end of the inflammatory event and an attempt at restorative physiology. Their role in this setting is in protecting against secondary infection or injury with B cells, CD4^+^/CD25^+^ cells, and γ/δ T cells as well as NK cells playing a likely role in this setting. n = 6 to 8 animals per group, with experiments repeated on 2 separate occasions to confirm original findings. \*P ≤ .05; \\P ≤ .01, as determined by ANOVA, followed by Bonferroni t test, with data expressed as means plus or minus SEM.](zh80070817490002){#F2} Peritoneal lymphocytes disappear in response to PGD~2~ STOPPED -------------------------------------------------------------- Investigating how lymphocytes disappear, we recorded equivalent numbers of T and B lymphocytes in the peritoneal cavity of hPGD~2~S knockout mice at 6 hours as in the naive cavity (0 hours) of wild-type ([Figure 3](#F3){ref-type="fig"}A-E). hPGD~2~S metabolizes cyclooxygenase-derived PGH~2~ to PGD~2~,^[@B18]^ which activates 2 G-protein--coupled receptors, DP1 and DP2 (CRTH2).^[@B19],[@B20]^ PGD~2~ is further dehydrated to the cyclopentenone PG, 15-deoxy-Δ^12,14^-PGJ~2~ (15d-PGJ~2~),^[@B21],[@B22]^ with levels of PGD~2~ and 15d-PGJ~2~ peaking within the first few hours of acute peritonitis coincident with lymphocyte disappearance.^[@B11]^ Incidentally, the existence of 15d-PGJ~2~ in mammalian systems has been hotly debated over the years, but in these experiments, we used liquid chromatography--mass spectrometry--mass spectrometry (LC-MS-MS) to definitively confirm its presence in resolving inflammation.^[@B11]^ Adding BW245C (DP1 agonist^[@B15],[@B16]^) to hPGD~2~S knockout mice reduced lymphocytes at 6 hours, with the clearance effects of DP1 on B cells ([Figure 3](#F3){ref-type="fig"}F). 15(R)-15-methyl PGD~2~ (DP2 agonist^[@B17]^) had no effect on lymphocyte numbers in knockout mice, indicating that DP1 receptor activation is responsible for B-cell disappearance early in acute inflammation. The fate of CD3^+^ cells remains less clear. For instance, CD3 cells were examined for their adherence to the greater omental lymphoid organ, the so-called leukocyte disappearance reaction typical of peritoneal macrophages during acute peritonitis.^[@B23]^ But, addition of 5% trypsin to a 4-hour inflamed cavity for 10 minutes recovered displaced macrophages but not lymphocytes (Figure S7). Another possibility was that CD3 cells underwent programmed cell death in response to 15d-PGJ~2~, a highly reactive electrophile and, in our experience, a potent inducer of lymphocyte apoptosis. Although CD3 cells from hPGD~2~S^−/−^ mice remained low for annexin V/propidium iodide labeling up to 6 hours, on a percentage basis there was a trend toward an increase in annexin V labeling within the equivalent population in wild-type mice. However, this did not reach significance due to the diminished numbers of peritoneal CD3 cells in wild-type mice available for analysis (data not included). Therefore, we can suggest that within a few hours of inducing an inflammatory response, peritoneal resident B cells disappear in a PGD~2~/DP1-dependent manner, but that the fate of CD3^+^ cells remains unclear. ![PGD~2~ controls the clearance of peritoneal resident lymphocytes. In response to inflammatory stimuli, lymphocytes in the peritoneum disappear between 6 and 24 hours. (A-E) However, lymphocyte numbers in hPGD~2~S knockout mice at 6 hours (■) were found to be equivalent to that present in the naive cavity of wild-type mice (□), suggesting a role for either PGD~2~ and/or 15d-PGJ~2~ in the initial clearance of lymphocytes. (F) Adding back BW245C (DP1 receptor agonist) to hPGD~2~S knockout mice caused a reduction in B cells. Attempts made to identify the fate of CD3 cells generated inconclusive results, with data suggesting that they may die locally by programmed cell death (data not included). n = 8 animals per group. \*P ≤ .05 as determined by Bonferroni t test, with data expressed as means plus or minus SEM.](zh80070817490003){#F3} Deficiency of repopulating lymphocytes in nonresolving chronic granulomatous disease ------------------------------------------------------------------------------------ Finally, the relevance of these findings to human inflammatory diseases was determined by examining the role of lymphocytes in gp91^phox^ knockout mice, an experimental model of human chronic granulomatous disease caused by defects in the phagocyte respiratory burst oxidase, which generates microbicidal superoxide.^[@B24],[@B25]^ Hence, patients with chronic granulomatous disease lack antimicrobial capacity and the ability to combat bacterial and fungal infections. Moreover, there is the associated occurrence of inflammatory granulomas in lung, liver, and skin, which in some instances may arise from sterile stimuli, suggesting that their formation may be due to incomplete degradation of inflammatory debris and/or impaired resolution of inflammation.^[@B26],[@B27]^ gp91^phox^ knockout mice were injected intraperitoneally with sterile zymosan and found to have elevated leukocyte numbers compared with controls, with inflammation failing to resolve ([Figure 4](#F4){ref-type="fig"}A). See also Figure S8 for comparison of cell types in both animals at resolution. Fewer lymphocytes were found at 48 to 96 hours in gp91^phox^ knockout mice, the time frame of resolution and lymphocyte repopulation in wild-type mice ([Figure 4](#F4){ref-type="fig"}B). Lymphocytes obtained from the resolution phase (72 hours) of normal strain-matched wild-type controls which were therefore composed of B1, NK, and γ/δ T cells as well as CD4^+^/CD25^+^ cells, were transferred back to gp91^phox^ knockout mice (72 hours) and challenged with LPS. Inflammation was reduced in knockout mice that received resolution-phase lymphocytes compared with gp91^phox^ mice alone ([Figure 4](#F4){ref-type="fig"}C). These results suggest that during ongoing, nonresolving inflammation, the absence of lymphocytes may account for susceptibility to superinfection and the associated hyperinflammatory response. ![Absence of repopulating lymphocytes during nonresolving inflammation. (A) Zymosan was injected into the peritoneal cavity of pg91^phox^ knockout mice, which, when compared with controls, showed a more aggressive inflammatory response that failed to resolve. (B) FACS analysis of cell types present during resolution revealed a progressive repopulation of lymphocytes during resolution that was lower in pg91^phox^ knockout mice. (C) Lymphocytes obtained from the resolution phase (72 hours) of normal strain-matched wild-type controls and comprising B1 cells, NK cells, and γ/δ T cells, as well as CD4^+^/CD25^+^ cells, were transferred back into the peritoneal cavity of gp91^phox^ knockout mice (72 hours) and subsequently challenged, intraperitoneally, with LPS. Inflammation was reduced in gp91^phox^ knockout mice that received resolution-phase lymphocytes compared with gp91^phox^ mice alone. \*P ≤ .05; \\P ≤ .01 as determined by ANOVA, followed by Bonferroni t test, with data expressed as means plus or minus SEM.](zh80070817490004){#F4} Discussion ========== Here, evidence is presented that lymphocytes play a pivotal role in controlling the onset of innate immune-mediated inflammation by regulating cytokine synthesis as well as host susceptibility to secondary infection. Analysis of lymphocyte subsets in the naive peritoneal cavity of mice revealed that B cells constitute about 70% to 80% of the total lymphocyte population, of which the majority have a B1 phenotype, with the remainder being B2 cells. CD3^+^ cells as well as NK and γ/δ cells make up the remaining 20% of lymphocytes. This profile differs to that found at resolution, which comprises more innate-type lymphocytes and B1 cells expressing MAC-1. Although we did not discern which lymphocyte or combination of lymphocytes bestowed protection at onset or at resolution, in a separate study, peritoneal CD3 T cells and in particular B220^+^ B cells were found to elaborate high levels of IL-10 in a DP1-dependent manner,^[@B11]^ most likely explaining the reduction in IL-10 in lymphocyte-deficient RAG1^−/−^ mice and subsequent increase in PMN influx in these animals. Thus, given the relative proportion of T cells versus B1 cells in the inflamed cavity and the ability of B cells to synthesis comparatively high levels of IL-10, we suspect that B lymphocytes may one of the predominant cell types modulating acute inflammatory responses to nonspecific stimuli. As PMNs begin to accumulate in response to zymosan, B cells disappeared in a PGD~2~-dependent manner, as there was equivalent numbers of B cells in the inflamed cavity of hPGD~2~S^−/−^ at 6 hours as there was in the naive cavity of wild-type mice. This accumulation of B cells in hPGD~2~S knockout mice was reversed by BW245C, a DP1 receptor agonist, with DP2(CRTH2) playing no role in this setting. While the mechanism of PGD~2~-dependent B-cell clearance is unknown, a Toll-like receptor (TLR)--mediated transient down-regulation of integrins and CD9 on B1 cells was shown to be required for detachment of these cells from the local peritoneal matrix and their subsequent efflux from the inflamed cavity.^[@B28]^ Whether BW245C alters CD9 expression needs further investigation, but given that B cells disappear concomitantly with PGD~2~ synthesis, it is possible that DP1 activation may play a role in regulating this pathway of B1 cell detachment and efflux. The fate of CD3 cells, on the other hand, is less clear. While their attachment to the peritoneal lining can certainly be excluded, there is the possibility that these cells may die locally by programmed cell death in response to 15d-PGJ~2~, which is synthesized concomitant with their disappearance,^[@B11]^ and in our experience, a potent inducer of lymphocyte apoptosis. However, future detailed work is required to definitively identify whether they die locally or clear via draining lymphatics. Thus, PGD~2~ exerts a dual role on resident B cells at least---regulating their inflammatory cytokine release as well as their efflux from the inflamed peritoneal cavity. The mechanism by which lymphocytes exert their protective effects in these experiments is unclear at this stage. Certainly, there is a cytokine imbalance favoring proinflammatory TNFα but reduced IL-10 in the inflamed peritoneum of RAG mice, thereby potentially explaining the enhanced influx of PMNs in these mice compared with wild-type. Indeed, we have shown that both T and B lymphocytes are capable of elaborating inflammatory cytokines, which, on balance, in the inflamed peritoneum at least, serve to limit PMN influx.^[@B11]^ However, in addition to cytokines/chemokines, cell adhesion molecules are also central to facilitating PMN adhesion and accumulation at sites of inflammation and, contrary to data presented here, there is evidence showing that lymphocytes trigger cell adhesion molecule expression. For instance, intracellular adhesion molecule 1 expression in Plasmodium-infected mice is reduced in the brain but not the lung of RAG1^−/−^ mice, while P-selectin expression is attenuated in both organs in these animals.^[@B29]^ Equally, T cells were shown to enhance the expression of TNFα-triggered endothelial cell adhesion molecule expression, with these effects varying between vascular beds.^[@B30]^ On this basis, it is difficult to reconcile the role circulating lymphocytes play in up-regulating cell adhesion molecule expression on the leukocyte/endothelial cell interface to that of the resident peritoneal T/B cells other than to highlight the different profile of lymphocytes present in the peritoneum (B1, B2, and small numbers of CD4/CD25 cells) that exert a predominantly protective effect in both the naive and postresolution state. The trigger for lymphocyte repopulation is unclear, but critical determinants of resolution such as PMN apoptosis or signals released by macrophages during phagocytosis of apoptotic leukocytes may play a central role. However, we have shown previously that inducible cyclo-oxygenase is expressed during and is essential for the resolution of acute inflammation,^[@B6],[@B31],[@B32]^ while others have reported that lipoxygenase-lipoxygenase interaction products of arachidonic as well as eicosapentaenoic and docosahexanoic acids dampen the severity of inflammatory onset and trigger resolution.^[@B33][@B34]--[@B35]^ Taking a closer look at whether cyclo-oxygenase or lipoxygenase play a role in lymphocyte repopulation, we found that not only is COX-2 expressed during the resolution phase of zymosan-induced peritonitis, but that its inhibition with either NS-398 or the nonselective COX inhibitor indomethacin impairs lymphocyte repopulation, in particular CD3^+^ cells. However, inhibition of lipoxygenase isoforms with baicalein was without effect. These data implicate a COX-2 metabolite in the recruitment of postresolution lymphocytes and therefore protection against host susceptibility to superinfection (J. Newsom, M. Stables, P.C.-N., G.B., and D.W.G., manuscript in preparation). Comparing repopulating lymphocytes with those in the naive cavity revealed more NK cells, γ/δ T cells, and CD4^+^/CD25^+^ cells, in addition to B1 cells with higher Mac-1 labeling than that found at onset. The functional relevance of increased MAC-1 expression on resolving B1 cells is not apparent at this stage, but may reflect a state of differentiation/activation or a specific requirement for migration back to the resolved peritoneum. However, repopulating lymphocytes have no role in actively bringing about resolution, but protect against superinfection. In these experiments, GBS was given into the resolving peritoneal cavities of wild-type as well as RAG1^−/−^ mice. Interestingly, the degree of inflammation in wild-type mice that received zymosan followed by GBS was not significantly greater than the level of inflammation in resolving wild-type mice not given GBS. In contrast, inflammation in RAG1^−/−^ mice that received zymosan followed by GBS was almost twice that of inoculated wild-type mice. This suggests that resolution-phase lymphocytes confer protection against secondary infection. This was certainly confirmed by injecting live bacteria to RAG^−/−^ mice undergoing resolution, which subsequently died substantially faster than wild-type mice treated in the same manner ([Figure 2](#F2){ref-type="fig"}F). It is unknown why the proportion and profile of repopulating lymphocytes is different to that at onset. Perhaps as resolving tissues are physiologically altered as a consequence of the inflammatory event they underwent, host defense mechanisms need to be fundamentally different to guard against superinfection by recruiting more protective lymphocytes. On this theme, it is not really understood when acutely inflamed tissues revert back to their original state. Certainly, a population of macrophages (about 1 × 10^6^) were found to linger for at least 3 weeks after a zymosan peritonitis apparently resolved, supporting the idea that although the original response was acute and transitory in terms of PMN influx and efflux, its effects may be longer-lasting than originally believed, thereby requiring a different profile and greater proportion of lymphocytes to modulate future inflammatory events. This may explain why inflammation in the resolving phase conferred greater protection against GBS lethality than uninflamed or naive mice ([Figure 2](#F2){ref-type="fig"}D,F). Taking these findings to a more clinically relevant setting, it became clear that unlike wild-type mice, there was a deficit of repopulating lymphocytes in nonresolving gp91^phox^ knockout mice bearing zymosan-induced peritonitis. Replenishing gp91^phox^ knockout mice with resolution-phase lymphocytes taken from strain-matched controls and then challenging animals with LPS conferred protection compared with sham-operated gp91^phox^ knockout mice. Data from this study confirm that not only does lymphocyte repopulation fail to occur in nonresolving inflammation, but that resolving-phase lymphocytes protect against exaggerated inflammatory responses to superinfection ([Figures 2](#F2){ref-type="fig"}D,[4](#F4){ref-type="fig"}C). On this note, defects in innate lymphocyte functioning have been suggested to lead to secondary infections associated with burn injury,^[@B36]^ while increased lymphocyte apoptosis contributes to the pathogenesis of sepsis,^[@B4],[@B5]^ underlining the crucial role lymphocytes play in host defense against nonspecific injury. Moreover, as NK cells were originally described for their ability to lyse tumor cells,^[@B37]^ and γ/δ T cells have well-known tumor surveillance properties,^[@B38],[@B39]^ their absence from sites of nonresolving inflammation may be one of the predisposing factors to the development of inflammation-related cancer.^[@B40]^ Thus, one of the hazards of ongoing acute inflammation and consequently failed lymphocyte repopulation may be increased susceptibility to superinfection and even cancer as a result of lymphocyte apoptosis, lymphocyte immunoparalysis, or, as presented here, a failure of protective lymphocytes to repopulate after resolution. In summary, we report a biphasic role for lymphocytes during innate immune-mediated inflammation, summarized in [Figure 5](#F5){ref-type="fig"}. The first phase controls PMN trafficking with lymphocytes then vacating the peritoneal cavity in response to PGD~2~ activating its DP1 receptor. The second phase is characterized by lymphocyte repopulation occurring after inflammation begins to resolve in a different proportion and profile to that of the naive state. Importantly, repopulating lymphocytes have no role in bringing about resolution, but protect against secondary infection. ![A summary of the scheme of events that occurs in acute inflammation with reference to lymphocyte trafficking. As inflammation ensues resident lymphocytes begin to disappear, with B1 cells clearing via draining lymphatics and the fate of CD3 cells remaining unclear at this stage. Once inflammation begins to resolve, lymphocytes repopulate the site of injury in a profile different to that in the naive state.](zh80070817490005){#F5} The online version of this article contains a data supplement. The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked "advertisement" in accordance with 18 USC section 1734. The authors would like to acknowledge Dr Tracy Hussell (Kennedy Institute for Rheumatology, London, United Kingdom) for supplying antibodies to NK and γ/δ T cells and Dr Y. Urade of the Osaka Bioscience Institute (Japan) for supplying hPGD~2~S knockout mice. D.W.G. is a Wellcome Trust funded Career Development Fellow and R.R. is a Kidney Research UK Funded Clinical Research Fellow. Authorship ========== Contribution: D.W.G. designed research and wrote the paper along with R.R., who carried out the research. T.L., G.B., J.B., and P.C.N. carried out experiments and provided essential experimental material. M.H., D.F., and M.M.Y. supplied essential experimental tools. D.W.G. and P.C.-N. analyzed data. Conflict-of-interest disclosure: The authors declare no competing financial interests. Correspondence: Derek W. Gilroy, Centre for Clinical Pharmacology and Therapeutics, Division of Medicine, 5 University Street, University College London, London WC1E 6JJ, United Kingdom; e-mail: <d.gilroy@ucl.ac.uk>.	{ "pile_set_name": "PubMed Central" }
###### Key questions What is already known? ====================== - Poor facility birth experience is prevalent especially in countries where childbirth-related mortality is high, and facility-based childbirth is low. - Little is known on how facility birth experience of care attributes influences women's choice to deliver in a health facility or not. What are the new findings? ========================== - Women associated the most utility to good health system conditions, including having a qualified birth attendant, drugs and supplies and a clean and conducive birth environment. - Poor facility culture, including an unclean birth environment with no privacy and unclear user fee, was the experience of care attribute associated with the most disutility. What do the new findings imply? =============================== - At this point in time, enhancing the delivery environment in Gombe would have the most positive impact on women preferences for facility-based childbirth. - Interventions designed to improve respectful maternity care require greater understanding of women's preferences. INTRODUCTION {#s1} ============ High quality, facility-based care at the time of childbirth remains the main strategy to address the continuing high burden of maternal mortality and morbidity, especially in low-income and middle-income countries (LMICs).[@R1] The provision of respectful and dignified care, including during childbirth, is an important component of quality of care and is the right of every woman.[@R3] Any mistreatment during facility-based childbirth is a violation of women's fundamental human right. It has also been highlighted as contributing to women not seek care at all or for subsequent deliveries, delaying seeking care, or discouraging others from delivering in health facilities.[@R3] The recognition that experience of care during childbirth has an impact on the use of maternal and newborn health (MNH) services has made the elimination of mistreatment and the provision of respectful maternity care a public health issue of global importance.[@R3] Advances in the methods for measuring experience of care have allowed the assessment of mistreatment during facility-based childbirth; with widespread mistreatment reported where it has been assessed.[@R9] For example, in Nigeria, where a high burden of maternal mortality persists, and where the coverage of facility delivery remains relatively low, women have reported mistreatment due to improper examination and lack of supportive care,[@R3] neglect or abandonment, lack of confidentiality, physical abuse, verbal abuse and mistreatment related to health system constraints.[@R11] In a recent study in Gombe State, where we conducted this study, about two-thirds of women interviewed reported experiencing at least one form of mistreatment during their facility-based childbirth.[@R15] Half of all women (50%) reported mistreatment related to health system conditions and constraints; for example, staff shortages, supply constraints, unconducive birth environment and perceptions of unreasonable requests by health workers. Similarly, almost half (46%) of all women reported mistreatment related to poor rapport between themselves and the healthcare provider; for example, denial of birth companion and not being allowed to eat, drink or move around during labour and childbirth. Reports of mistreatment related to sexual abuse, stigma and discrimination were uncommon.[@R15] Being able to define and assess mistreatment provides the opportunity to design effective interventions across a range of attributes including professional standards of care, rapport between women and providers, health system conditions and constraints, physical abuse, verbal abuse, sexual abuse and discrimination, to improve facility birth experience.[@R3] However, to date there is limited evidence on how these different attributes of experience of care during childbirth interact to influence women's choice of whether or not to deliver in a health facility. This study investigates how these attributes influence a woman's choice for a place of delivery in Gombe State, northeast Nigeria. Through a discrete choice experiment (DCE), we aim to determine the relative value (utility) rural women in northeast Nigeria place on a range of attributes of experience of care during facility-based birth. Methods {#s2} ======= Study context {#s2-1} ------------- This study was conducted in Gombe state in northeast Nigeria. MNH indices are suboptimal for Nigeria as a whole but there is also considerable regional variation, with the northwest and northeast regions having lower utilisation of healthcare compared with southern regions of the country.[@R16] This is true of Gombe State which, relative to national estimates, has lower coverage of at least four antenatal care visits (44% vs 57%), lower coverage of facility delivery (28% vs 39%) and higher infant mortality rate (90/1000 live births vs 70/1000 live births).[@R17] Almost 98% of formal health services in Gombe State are provided through government at three levels---primary, secondary and tertiary levels.[@R18] Primary health services are delivered through dispensaries, health posts, health centres and primary health centres, and secondary services are provided through the state specialist hospital and general hospitals, which also serve as referral centres. Tertiary services are provided through the Federal medical centre.[@R18] The majority of healthcare workers in Gombe State are lower cadre for example, Community Health Extension Workers (CHEWS), Junior CHEWS and Health officers.[@R20] Skilled healthcare providers including medical doctors and nurses/midwives constitute only 4% and 27% of the health workforce, respectively.[@R20] The majority of women deliver in primary healthcare facilities, attended to by the lower cadre healthcare workers.[@R20] Study design {#s2-2} ------------ ### Identification of attributes and levels {#s2-2-1} In healthcare, patients or clients often have strong preferences for treatment and health service options, and these can affect service utilisation.[@R22] In this study, conducted in March 2018, we used DCE to elicit women's preferences for a facility-based birth based on different attribute levels presented in [table 1](#T1){ref-type="table"}. DCEs are health economic tools, widely used to understand user preferences in healthcare.[@R24] DCEs ask respondents to choose their preferred service from a set of hypothetical alternatives over several choice tasks. Studying how respondents choose across repeated scenarios allows researchers to quantitatively elicit the key drivers of decision making.[@R25] ###### Attributes used in discrete choice experiment on respectful maternity care attributes influencing women's stated preferences for facility-based childbirth Attributes Attribute levels --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Failure to meet standards of care Lack of informed consent and confidentiality: birth attendant may not ask for your permission before performing any medical examination or procedure and may discuss your personal information openly with others. Physical examinations and procedures: birth attendant may not give you pain relief as necessary (eg, during examination) or for the stitching of episiotomy. Neglect and abandonment: birth attendant may leave you unattended, you may deliver without assistance. Meet professional standards of care: Birth attendant will take care of you throughout labour and delivery, will ask for your consent, provide pain relief when needed. Poor rapport with providers Poor communication: birth attendant may not explain what will happen to you or your baby and may not encourage you to ask questions or answer your questions. Lack of supportive care: birth attendant will not empathise or show genuine interest in your well-being. Loss of autonomy: birth attendant will not allow a companion to stay with you and will not allow you to eat, drink, move about or choose your preferred birth position. Good rapport with providers: Birth attendant will receive you with open arms, comfort and encourage you, respond to you in friendly way, allow you to have a companion and to choose your preferred birth position. Health system constraints Staffing constraints: qualified birth attendant not available to assist with your delivery. Drugs and supply constraints: drugs and supplies needed for delivery not available. Poor facility culture: birth environment unclean, smelly, with mosquitoes, with no privacy, user fee not clear. Good health system conditions: qualified birth attendant present to assist you and drugs and supplies needed for delivery available, clean and conducive birth environment. Physical and verbal abuse Physical and verbal abuse: birth attendant may hit, slap or put restraints on you, shout or insult you. No physical and verbal abuse: birth attendant will not hit, slap or put restraints on you and will not shout or insult you. Sexual abuse Sexual abuse: birth attendant may touch your body parts or private parts inappropriately. No sexual abuse: birth attendant will not touch your body parts or private inappropriately. Stigma and discrimination Discrimination: birth attendant may discriminate against you because you are poor, from a village, not educated or because of your religion, tribe or disease condition. No discrimination: birth attendant will not discriminate against you because you are poor, from a village, not educated or because of your religion, tribe or disease condition. Attributes and attribute levels derived from literature review,[@R3] and revised based on qualitative findings. The first step in designing a DCE is to select the key service attributes---or characteristics---which may be important to users. To do this, we conducted a comprehensive review of the literature to identify attributes of respectful maternity care. Data bases searched included PubMed, Google Scholar, EMBASE, CINAHL and EBSCO. We further searched the reference list of the identified articles and reached out to experts in MNH to identify additional literature. From the results of the review, we selected the revised typology of mistreatment by Bohren et al,[@R3] based on a systematic review of 65 studies from 34 countries. The typology builds on earlier work of Bowser and Hill[@R29] and revised the dimensions of mistreatment to include seven domains: physical abuse, sexual abuse, verbal abuse, stigma and discrimination, failure to meet professional standards of care, poor rapport between women and health providers and health system conditions and constraints. We used the typology to design a qualitative study to investigate the relevance of the dimensions in the study setting, and to derive attribute levels by investigating how the different dimensions manifest for each attribute. The qualitative study included in-depth interviews with 31 women and four focus groups with 32 women (eight women per focus group). The qualitative study participants were women who had recently delivered in a health facility, purposively sampled from the communities. The qualitative study was conducted in December 2017. The qualitative data were analysed using thematic content analysis, with a manifest approach,[@R30] in which the data analysis focused on what women said about their experience during labour and delivery. In-depth interviews and focus groups, alongside the literature review findings, informed our selection of six attributes and a total of 18 attribute levels relevant to the context (three attributes of four levels each and three attributes of two levels), these are highlighted in [table 1](#T1){ref-type="table"}. Further, the qualitative study provided us with locally appropriate expressions and language translation, enhancing respondents ease of comprehension.[@R31] Experimental design and construction of choice sets {#s2-3} --------------------------------------------------- From the number of attributes and attribute levels decided, a full factorial design would have consisted of 729 (3^4^×3^2^) possible alternatives---too many for a survey, and tedious for the respondents to handle.[@R27] Therefore, we developed 16 choice sets based on a fractional factorial orthogonal main effects design from a design catalogue, that ensured the inclusion of levels proportionally (level balance) with no correlation between levels of different attributes (orthogonal).[@R32] We constructed an unlabelled choice experiment of three choice alternatives, with each set consisting of two unlabelled facility alternatives and home delivery, an example is highlighted in [table 2](#T2){ref-type="table"}. We decided to include a home delivery option to avoid bias in estimating parameters in a forced choice design.[@R33] ###### Example of discrete choice experiment choice task as shown to women who had a facility-based childbirth We would like you to imagine you are deciding where you are going to deliver your next baby, tell us which of the options in each scenario you will prefer to go to (or not) for your delivery. There are no right or wrong answers, we are only interested in knowing what is important to you regarding facility delivery. Please tick your preferred choice. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------- Birth attendant may not ask for your permission before performing any medical examination or procedure and may discuss your personal information openly with others. Birth attendant may not give you pain relief as necessary (eg, during examination) or for the stitching of episiotomy. Birth attendant may not explain what will happen to you or your baby and may not encourage you to ask questions or answer your questions. Birth attendant will not empathise or show genuine interest in your well-being. Qualified birth attendant not available to assist with your delivery. Drugs and supplies needed for delivery not available. Birth attendant may hit, slap or put restraints on you, shout or insult you. Birth attendant will not hit, slap or put restraints on you and will not shout or insult you. Birth attendant may touch your body parts or private parts inappropriately. Birth attendant will not touch your body parts or private inappropriately. Birth attendant may discriminate against you because you are poor, from a village, not educated or because of your religion, tribe or disease condition. Birth attendant will not discriminate against you because you are poor, from a village, not educated or because of your religion, tribe or disease condition. ![](bmjgh-2019-002135ilf01.jpg){#img1} ![](bmjgh-2019-002135ilf02.jpg){#img2} ![](bmjgh-2019-002135ilf03.jpg){#img3} The choice sets were reviewed and validated for content in collaboration with a group of health workers (doctors, nurses and midwives) working in Gombe, followed by a pilot test with 40 women with similar characteristics as the target sample (women with recent facility-based childbirth). None of the women that took part in the pilot participated in the main study. The pilot exercise involved completing the DCE and answering questions afterwards regarding the exercise, including clarity of instructions, understanding and relevance of the choice sets and ease in answering. Following the pilot, minor modifications were made which included small changes to the wording used to describe the attribute levels, and to introduce the choice sets. For example, participants from the pilot suggested we use 'hospital' rather than 'health facility' when describing the choice sets. The final DCE tool was incorporated into a larger study instrument consisting of questions on sociodemographic information, and experience of care during institutional delivery. The final questionnaire was programmed in CSPro.[@R34] DCE study sample {#s2-4} ---------------- The final DCE was nested within an ongoing measurement, learning and evaluation project in Gombe State.[@R35] As part of that project, a total of 1889 birth observations were carried out in ten primary health facilities across three time points: June 2016, March 2017 and August 2017.[@R36] The health facilities were spread across six of the 11 local government areas in Gombe State. Six of the health facilities were located in urban settings and four were located in rural settings. Subsequently, in August 2018, a simple random sample of 450 observed women was taken and these women were followed up at home to ask what they recalled about their facility birth experience. Of the 450 eligible and selected women, 426 (95%) women were successfully interviewed in March 2018 at their homes, while 24 (5%) of the eligible women selected could not be reached or were unable to participate. According to commonly used rules of thumb for DCE sample size calculation, the sample size of 426 women was enough to guarantee precision in the estimation of all model parameters.[@R37] Orme posit that the sample size N required for main effects depends on the number of choice task (t), the number of alternatives (a) and the number of cells (c), according to the following equation N\>500 c/(t×a).[@R37] Based on this equation, an approximate sample of 56 participants would have been sufficient to model our preference data. Lancsar and Louviere suggested that a sample of 20 respondents per questionnaire version as adequate to estimate reliable DCE models.[@R39] However, we recruited much larger sample to allow for more variability between respondents and to allow for other post hoc analysis. Data collection {#s2-5} --------------- Data were collected using personal digital assistants, and interviews took about an hour to complete. Before data collection, data collectors and supervisors received 5 days of training on data collection and the study tools. Model specification {#s2-6} ------------------- The discrete choice data were analysed based on the random utility model.[@R25] We specified our analytical model around a utility maximising framework. This assumes, given alternatives to choose from, a respondent i (i=1, ..., N) will choose the one alternative that yields the maximum utility among the choice bundle (j=1, 2, 3, ...) at the moment of choice. The utility of the respondent is defined by a deterministic or observable component and a random error component: $$U_{ij} = v_{ij} + \varepsilon_{ij}$$ Where U~ij~ represents the utility of respondent,v~ij~ the observable component and ε~ij~ the random error term with standard statistical properties. Following from [equation (1)](#E1){ref-type="disp-formula"} the probability of a respondent selecting a specified place of delivery is modelled. The probability of a choosing a place of delivery is determined by the indirect utility function for the respondent i from choice j in choice set s, assuming this is linear and additive and of the form: $$V_{ijs} = X_{ijs}\beta + \varepsilon_{ijs}$$ Where V~ijs~ represent the utility derived from a choice, and X~ijs~β the utility component and ε~ijs~ as the random component. The vector X~ijs~ is specified below, where β~1-6~ represent the design attributes of the choice experiment and β~0~ the constant. $$\begin{matrix} {X_{ijs}\beta_{j} =} & {\beta_{0} + \beta_{1}{failure\_ to\_ meet\_ standard\_ of\_ care}_{\mathit{j}} +} \\ & {\beta_{2}{poor\_ rapport\_ with\_ provider}_{\mathit{j}} + \beta_{3}{health}} \\ & {{\_ system\_ constraints}_{\mathit{j}} + \beta_{4}physical\_\&} \\ & {{\_ verbal\_ abuse}_{\mathit{j}} + \beta_{5}{sexual\_ abuse}_{\mathit{j}}} \\ & {+ \beta_{6}stigma\_{\_ discrimination}_{\mathit{j}}} \\ \end{matrix}$$ Data analysis and model estimation {#s2-7} ---------------------------------- We analysed the DCE data using STATA V.15. We conducted a dominance test of internal validity, presenting an additional choice set to all the respondents where one of the hypothetical health facilities is more favourable, but do not exclude participants if they fail this test.[@R40] We first estimated standard multinomial logit model (MNL) ([online supplementary table S1](#SP1){ref-type="supplementary-material"}) to provide a benchmark for more detailed analysis, and which were used as starting values for estimating a mixed multinomial logit model (MMNL) ([online supplementary table S2](#SP1){ref-type="supplementary-material"}). Further, we fitted MMNLs to the DCE data from rural women investigated, using 500 Halton draws. We used the MMNL to avoid restrictions due to the independence of irrelevant alternatives assumption required to interpret findings from the MNL.[@R41] 10.1136/bmjgh-2019-002135.supp1 When estimating MMNLs all parameter estimates may be treated as random, and in a model where more than one parameter or all parameters are estimated as random, there is no requirement that the distribution be the same.[@R43] McFadden and Train have shown that mixed multinomial logit does not embody any theoretical restrictions on the distribution of preferences or the choice model.[@R44] Also, with mixed multinomial logit, we can accurately approximate any choice model, with any distribution of preferences.[@R44] Therefore, all attributes levels were effects coded, specified as random components and a multivariate normal distribution---a generalisation of the one-dimensional normal distribution to more than one variable was assumed. Assuming a multivariate normal distribution to represent the distribution of a multivariate random variable that is made up of multiple random variables that can be correlated with each other was a rational choice because this allows for correlation which introduces error dependence across the alternatives in each choice situation.[@R45] Furthermore, assuming a random effect to be respondent specific induces correlation across choice situations, thus accounting for the dependence structure in unobserved utility among the repeated choices of a respondent due to the panel structure of the data.[@R45] Findings are reported below in line with Strengthening the Reporting of Observational Studies in Epidemiology statement.[@R47] Preference heterogeneity {#s2-8} ------------------------ An important output from the main effects MMNL I estimation is the SD associated with each parameter estimate, which indicate the distribution about the corresponding mean preference weight, also the preference variability among women around the mean. Some of the differences between the parameter estimates in the multinomial logit and MMNLs further indicate that preferences vary among women in Gombe. Respondents characteristics are constant across alternatives for example, a woman's ethnicity does not change because she is considering delivering in a health facility as opposed to home delivery.[@R43] Respondents characteristics are likely to influence their choice decisions, but they are not part of the attribute description of alternatives and not a direct source of utility.[@R43] One way to predict how respondents' characteristics influenced their choices is to extend [equation (2)](#E2){ref-type="disp-formula"} to allow attribute weights to vary with respondent characteristics, through the inclusion of interaction terms between attribute and individual characteristics.[@R48] Therefore, to understand the preference (taste) variability among women the MMNL (model I) was extended with interaction terms between attribute levels with significant SD and women sociodemographic characteristics likely to influence women's behaviours or predisposition to mistreatment, as highlighted in [equation](#E4){ref-type="disp-formula"} (4) below (and MMNL model II). This approach often leads to models fit improvement. It is also easy to interpret the parameters related to covariates in a relative sense both within and between alternatives---holding all else equal.[@R43] Revelt and Train have shown that entering demographics into the MMNL itself is a more direct and accessible way to hypothesis testing, as such, this approach has been widely used in DCE studies.[@R49] $$\begin{matrix} {X_{ijs}\beta_{j}} & {= \beta_{0} + \beta_{1}failure\_ to\_ meet\_ standard\_ of\_ care_{\mathit{j}}} \\ & {+ \beta_{2}poor\_ rapport\_ with\_ provider_{\mathit{j}} + \beta_{3}} \\ & {health\_ system\_ constraints_{\mathit{j}} + \beta_{4}} \\ & {physial\_\&\_ verbal\_ abuse_{\mathit{j}} + \beta_{5}sexual\_ abuse_{\mathit{j}} +} \\ & {\beta_{6}stigma\_\&\_ discrimination_{\mathit{j}} + \beta_{7}female \ast} \\ & {neglect\_ abandonment_{\mathit{j}}\;...... + \beta_{26}upperSES \ast} \\ & {no\_ sexual\_ abuse_{\mathit{j}}} \\ \end{matrix}$$ Where β~1-6~ represent the design attributes of the choice experiment and β~7--26~ the parameters for the interaction terms that were introduced for the women sociodemographic characteristics variables. The sociodemographic characteristics used included ethnicity coded (0=others, 1=Fulani), age coded (0= ≤29 years, 1= ≥30 years), education coded (0=low-level education, 1=high-level education) and socioeconomic status (SES) coded (0=low SES (lower 60%), 1=upper SES (upper 40%). Results {#s3} ======= Demographic characteristics of women {#s3-1} ------------------------------------ The final sample used in the analysis comprised 425 respondents; one case was dropped due to incomplete data. Descriptive statistics for the study sample are displayed in [table 3](#T3){ref-type="table"}. The sample was young, with 60% of the respondents being between the ages of 20 and 29 years, and just 3% were between the ages of 40 and 49 years. The majority belonged to the Fulani ethnic group 60%, followed by Hausa. Respondents were mostly married (95%) and Muslims (98%). Almost half (45%) had no formal education. Only about 7% failed the dominance test by choosing to deliver in a less favourable hypothetical health facility, suggesting acceptable internal validity.[@R51] ###### Demographic characteristics of women in discrete choice experiment on facility birth experience of care attributes influencing women's stated preferences for facility-based childbirth Characteristics and delivery context n=425 (%) -------------------------------------- ----------- Age \<20 years 11 20--29 years 60 30--39 years 26 40--49 years 3 Ethnicity Fulani 60 Hausa 17 Kanuri 7 Others 16 Religion Christian 2 Muslim 98 Marital status Married 95 Single/widowed 5 Education level None 45 Primary 23 Secondary and postsecondary 32 Parity Primigravida 0 Multigravida 100 Period of birth Day time (07:00 to 19:59) 55 Night time (20:00 to 06:59) 45 Day of delivery Weekdays 78 Weekend 22 Model estimates of preferences for attribute-levels of respectful maternity care {#s3-2} -------------------------------------------------------------------------------- [Figure 1](#F1){ref-type="fig"} and [online supplementary table S2](#SP1){ref-type="supplementary-material"} highlight the utility parameter estimates from main effects MMNL for preferences for facility birth experience of care attributes levels. The positive sign of the estimated coefficients indicates that the attribute level has a positive effect on preferences for facility-based childbirth. Conversely, a negative coefficient indicates a negative effect on preference. The estimated coefficients all followed anticipated direction of effect; implying women derived higher levels of utility from attribute levels which were considered better ex ante. ![Positive values indicate positive effect on preference, negative values indicate negative effect on preferences. \\Significance level at 1% and \* significance level 5%.](bmjgh-2019-002135f01){#F1} In the main effects model (model I), most estimated coefficients were statistically significant. At least one attribute level was significant in each of the attributes included in the study, indicating all attributes contributed to the decisions made by women when stating their preferences for a place of delivery. The most important attribute level that influenced the likelihood of women choosing a facility-based childbirth was good health system condition, including having qualified birth attendant present during childbirth, drugs and supplies needed for delivery available, clean and conducive birth environment (coefficient: 0.59), followed by absence of sexual abuse (coefficient: 0.26). Absence of physical and verbal abuse during labour and delivery was the third most preferred attribute level among women (coefficient: 0.16) ([figure 1](#F1){ref-type="fig"} and [online supplementary table S2](#SP1){ref-type="supplementary-material"}). For rural women in Gombe, poor facility culture including unclean birth environment, with no privacy and unclear user fee was associated with the most disutility, therefore, the most negative impact on preferences for facility-based childbirth (coefficient: −0.31), closely followed by sexual abuse (coefficient: −0.26) and staffing constraints (coefficient: −0.23). The likelihood of the respondents being discriminated against because of a disease condition or because they were poor, not educated or from a village had the least influence on their choice of place of delivery. Home delivery was also associated negative utility ([figure 1](#F1){ref-type="fig"} and [online supplementary table S2](#SP1){ref-type="supplementary-material"}). Preference heterogeneity {#s3-3} ------------------------ The influence of sociodemographic characteristics on preferences highlighted in [online supplementary table S2](#SP1){ref-type="supplementary-material"} (interaction model II) suggested variation in preferences for some facility birth experience of care attributes among rural women in Gombe. Women of lower SES were more likely to choose to deliver in a hypothetical health facility where they are less likely to be neglected or abandoned. Absence of mistreatment related to sexual abuse had less effect on preference for place of delivery among women with low education (ie, no formal education or only primary school education) and those that identified as non-Fulani by ethnicity. However, the preference variability observed could not be explained by observed characteristics of women. Considerable variation (indicated by SD in MMNL II) still remained even after including ethnicity, age, education and SES variables. Discussion {#s4} ========== In this study, we explored preferences for facility birth experience of care attributes among rural women with recent facility-based childbirth in northeast Nigeria. An in-depth understanding of how women value different facility birth experience of care attributes is of global interest since it could provide the foundation for developing and designing appropriate interventions that could improve utilisation. Our findings suggest that hypothetical facility birth experiences are influential in women's decision making about place of delivery. Based on the magnitude and direction of the estimated attribute-level coefficients, women associated varying degrees of utility and disutility to hypothetical facility birth experience of care attributes levels. We found that women associated the most utility to good health system conditions, including having a qualified birth attendant, drugs and supplies, and a clean and conducive birth environment. Women expressed strong preference to deliver in a health facility where they were less likely to be sexually, physically or verbally abused, as absence of sexual abuse and absence of physical and verbal abuse during labour and delivery were the second and third most important attribute levels for rural women in this study. But the experience of care attribute they associated the most disutility to was poor facility culture, including an unclean birth environment with no privacy and unclear user fee. Our findings are consistent with studies from Ethiopia, Ghana, Kenya, Nigeria and the Philippines which have reported a strong association between improvements in structural quality and the availability of commodities with a substantial increase in health facility utilisation.[@R53] Interestingly, the utility associated with drug and supply constraints was negative but did not reach statistical significance, perhaps a reflection of women in Gombe having coping strategies for persistent shortages of drugs and supply by sourcing these outside of the health system, rather than an indication of not caring about drugs and supply constraints.[@R15] Studies in many LMICs have also reported a widespread prevalence of physical, verbal and sexual abuse during facility-based birth.[@R3] Women have reported incidences of being treated differently due to their SES or disease condition,[@R3] and incidences of being denied their preferred birth position, or not allowed to eat, drink or move.[@R3] These negative experiences during facility-based birth have been found to undermine women's trust in the health system, and negatively impacted their subsequent decision to deliver in the health facility.[@R56] In an earlier study in Gombe State, only 3% of women reported mistreatment related to physical abuse, and less than 1% reported mistreatment related to sexual abuse during labour and delivery.[@R15] Nonetheless, we found that absence of physical abuse and absence of sexual abuse influenced women's hypothetical choice for place of delivery. Our findings suggest that the low prevalence of a particular negative facility birth experience, for example, sexual abuse or physical abuse does not negate its impact. We found few of the women's observable characteristics could explain the variability in preferences for the attribute levels. These included ethnicity, SES and level of education. Women with low education or women of low SES were more concerned with the likelihood of experiencing certain forms of mistreatment when making a choice for a place of delivery, than women with more years of education or high SES. It is probable that differential experiences of women from different communities with their respective health facilities influenced the value they associated with attributes where heterogeneity was observed. Kruk et al argued that respondents draw on their experiences or that of their friends or families when conducting DCE exercises.[@R54] Variability in preference for facility birth experience of care attributes has been observed elsewhere. A study in urban Beirut found that women preferred their husbands to accompany them during labour and delivery, while women in the rural areas preferred to be cared for by healthcare providers alone.[@R59] Understanding how preferences vary between women could improve interventions by making them more responsive and oriented towards patient-centred care. This study had strengths and limitations. The results represent the predominantly Fulani, Muslim, married study participants who accessed the study health centres. These women had their last birth in a health facility and had therefore experienced facility-based respectful maternity care, or the lack of it, so our findings may not apply to women with no facility delivery experience. Further, in this context and other similar settings, the decision to deliver in a health facility may not rest exclusively with women, as household heads, and in-laws may also have a say. In this study we used qualitative methods to develop and refine our hypothetical scenarios, but still they may not have reflected real-world options for place of delivery. Nonetheless, in previous studies, DCEs has been shown to predict real-world choices with a reasonable degree of sensitivity and external validity.[@R51] Implications for policy and research {#s4-1} ------------------------------------ Access, cost, sociocultural barriers and other contextual issues continue to be important determinants of utilisation of MNH services in LMICs.[@R3] Here we underline the additional importance of respectful maternity care as a determinant of women's decision to deliver in a health facility. The renewed global drive to eliminate mistreatment and improve respectful maternity care as part of the quality of care in health facilities has given the evaluation of the experience of care a new impetus.[@R3] Strategies toward improving respectful maternity care could be made more efficient with additional understanding of the utility and disutility women associate with attributes of care. For example, evidence on the extent to which absence of physical abuse during labour or having a birth companion mattered to women could help decision-makers to develop strategies that are more context-appropriate and relevant to women, rather than transplanting interventions because they worked elsewhere. This study reinforces the importance that women attribute to both structural and process quality of care. In a recent study in Gombe State, most women in the qualitative study suggested that the possibility of receiving drugs, injections and skilled assistance during childbirth was a sufficient motivation to deliver in a health facility again.[@R15] However, women also described the various coping mechanisms they applied in anticipation of mistreatment, for example, bypassing the facility closest to their home[@R61] or delaying going to the health facility to limit how much time they spent in an unconducive birth environment. Such coping mechanisms could have several negative implications, for example, on birth outcomes and on increased out of pocket expenditure for health due to increased transport costs. Catastrophic health expenditure has been increasing worldwide, and more people around the world are pushed into extreme poverty yearly due to out-of-pocket expenses.[@R62] Nigeria's drive towards achieving the sustainable development goals must include addressing causes that increase out of pocket spending for health and perpetuates poverty. To achieve Universal Health Coverage, the Nigerian government is planning to build, upgrade or refurbish about 10 000 health facilities across the country.[@R63] Our study has shown that women prioritise good health system conditions in their decision making and these new health facilities need to be adequately staffed as well as equipped and stocked with essential supplies. Otherwise, women are very likely to continue bypassing facilities. As part of these strategies, improving the enabling environment for health staff, including building their capacity to deliver respectful maternity care, could improve interactions between providers and women.[@R63] Inevitably, addressing these problems also means more resources, difficult to achieve in a low resource setting. Hence, there is a need to rationally consider effective trade-offs when designing interventions to improve respectful maternity care, trade-offs that are likely to be context specific and may change overtime.[@R65] A strategy of focusing first on addressing mistreatment in areas where most benefits can be obtained quickly while keeping an eye on the whole picture is worth considering. Conclusions {#s5} =========== In this study we used a DCE to explore how hypothetical facility birth experience influenced rural woman's choice for place of delivery in northeast Nigeria. Women associated the most utility to good health system conditions, including having a qualified birth attendant, drugs and supplies and a clean and conducive birth environment. Conversely, respondents were less likely to choose to deliver in a hypothetical health facility where the health facility culture was poor, including unclean birth environment with no privacy and unclear user fee. The results further underline the important relationship between facility birth experience and utilisation. The study findings suggest achieving universal health coverage will require not only that health services are available, but are available to a sufficient quality that women would want to access them. The authors would like to thank Data Research and Mapping Consult for Coordinating the data collection, the Gombe State Primary Health Care Development Agency, Gombe State Ministry of Health, our partners Society for Family Health and Pact Nigeria and study participants for their cooperation and participation in the study. We would like to thank Dr Gilbert Abotisem Abiiro, Department of Planning, Faculty of Planning and Land Management, University for Development Studies, Wa, Ghana and Dr Umar Adamu Usman DRMC Nigeria, Eng Nura Umar Sambo DRMC, Nigeria, for their contribution to this study. Handling editor: Seye Abimbola Contributors: NU, TM, conceptualised and designed the study; NU, conducted the analysis and drafted the manuscript; MMQ, JE, ZH and AS revised the manuscript and provided critically important intellectual content; all authors read and approved the final manuscript. Funding: This work was supported by IDEAS---Informed Decisions for Actions to improve maternal and newborn health (<http://ideas.lshtm.ac.uk>), which is funded through a grant from the Bill & Melinda Gates Foundation to the London School of Hygiene & Tropical Medicine. (Gates Global Health Grant Number: OPP1149259). Competing interests: None declared. Patient and public involvement: Patients and/or the public were involved in the design, or conduct, or reporting, or dissemination plans of this research. Refer to the Methods section for further details. Patient consent for publication: Not required. Ethics approval: This research was conducted with approval from the Federal Ministry of Health Abuja, Nigeria, the State Ministry of Health Gombe State, Nigeria and the London School of Hygiene & Tropical Medicine (reference 12181). The free and informed consent of all interviewees was obtained. Provenance and peer review: Not commissioned; externally peer reviewed. Data availability statement: Data are available upon request. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.	{ "pile_set_name": "PubMed Central" }
Vol. 200, No. 8, October 18, 2004. Pages 1063--1074. Dr. Eun-Yi Moon was inadvertently excluded as an author on the article. Her affiliation appears below. Department of Pathology, Harvard University, Cambridge, MA 02115	{ "pile_set_name": "PubMed Central" }
Introduction {#Sec1} ============ Neuromyelitis optica spectrum disorder (NMOSD), formerly referred to as NMO or Devic's disease, is an astrocyte-specific central nervous system (CNS) disorder characterized by relapsing optic neuritis and longitudinally extensive transverse myelitis which is a secondary phenomenon \[[@CR1]--[@CR5]\]. Astrocyte damage is much more striking than myelin and neuron damage in NMOSD compared with multiple sclerosis, and autoimmune astrocytopathy is the primary pathology of NMOSD \[[@CR5]\]. Approximately 70--90% of NMOSD patients are seropositive for disease-specific autoantibodies targeting the water channel protein aquaporin-4 (AQP4) \[[@CR6]--[@CR8]\], and AQP4 is mainly localized to astrocytic end-feet that are adjacent to microvessels at the blood-brain barrier (BBB) throughout the CNS \[[@CR9], [@CR10]\]. The presence of AQP4-immunoglobulin G (AQP4-IgG) is a serological or laboratory hallmark of NMOSD that is used to distinguish NMOSD from relapsing-remitting multiple sclerosis (RRMS) \[[@CR11]\]. There is growing evidence that NMO-IgG/AQP4-IgG is pathogenic in NMOSD \[[@CR3], [@CR12]--[@CR15]\] by a mechanism involving complement-dependent cytotoxicity (CDC) \[[@CR16], [@CR17]\] and antibody-dependent cell-mediated cytotoxicity (ADCC) \[[@CR18], [@CR19]\], which lead to astrocyte damage and an inflammatory response, causing oligodendrocyte injury, demyelination, and neurological deficits. Neutrophils play an important role in the formation of NMOSD lesions \[[@CR20], [@CR21]\]. The severity of NMOSD lesions is increased in mice with neutrophilia and reduced in mice with neutropenia \[[@CR21]\]. The pathogenetic implication of neutrophils in NMOSD was demonstrated in a previous report of a patient whose first NMOSD episode was worsened by inadvertent administration of granulocyte-colony stimulating factor (G-CSF) \[[@CR22]\]. Experimental NMOSD models are needed to elucidate the underlying pathophysiological mechanisms of this disease and to test candidate therapeutic drugs. Other groups and our group have fully confirmed that intracerebral injection of NMO-IgG and human complement (hC) in mice are sufficient to induce NMOSD-like lesions \[[@CR23]--[@CR26]\]. Tanshinone IIA (TSA, Fig. [1](#Fig1){ref-type="fig"}a), an active natural compound extracted from Salvia miltiorrhiza Bunge (Fig. [1](#Fig1){ref-type="fig"}a) \[[@CR27]\], has been clinically used to treat cardiovascular \[[@CR28], [@CR29]\] and cerebrovascular \[[@CR30]--[@CR32]\] diseases. These protective effects were attributed at least in part to its anti-inflammatory properties \[[@CR28]--[@CR31]\]. TSA can accelerate the resolution of inflammation by promoting neutrophil transmigration and apoptosis in zebrafish \[[@CR33]\]. TSA also efficiently ameliorates rheumatoid arthritis in mice by inhibiting neutrophil infiltration and activation and by promoting neutrophil apoptosis in the ankle joints \[[@CR34]\]. Fig. 1TSA administration notably suppresses astrocyte damage in NMOSD lesions at 24 h after brain injection. aSalvia miltiorrhiza Bunge and the chemical structures of TSA. b The protocol of the animal experiment. The purple arrow indicates the time point of NMO-IgG and hC injection. The black arrows represent the time points of TSA or vehicle treatment. The red arrow indicates the end point of the animal model experiment. c Representative immunofluorescence staining of AQP4 (top-left, green) and GFAP (bottom-left, red) in TSA- or vehicle-treated mice. d Reduction in the area of astrocyte marker immunoreactivity. Vehicle, n = 15; pre-TSA, n = 15; post-TSA, n = 5. \*p \< 0.05, \\\*p \< 0.001. The data are representative of three independent experiments. One-way ANOVA followed by Tukey's multiple comparison tests was performed Here, we utilized an NMOSD mouse model to evaluate the effect of TSA on NMOSD and the underlying active mechanism. Our results indicated that prophylactic administration of TSA significantly suppressed astrocyte damage and demyelination in NMOSD mice, and its protective effect was correlated with the degree of neutrophil apoptosis within lesions. We further confirmed that TSA indeed accelerates neutrophil apoptosis under inflammatory stimuli in vitro, and this proapoptotic effect was also observed in human neutrophils. These results suggest that TSA might serve as a prophylactic treatment for human neutrophil-associated inflammatory disorders. Methods {#Sec2} ======= Reagents {#Sec3} -------- TSA (purity: ≥ 97%, catalog T4952), lipopolysaccharide (LPS, purity ≥ 97%, catalog L6143), and DMSO were purchased from Sigma-Aldrich. DyLight 488-conjugate donkey anti-rat IgG (catalog 712-545-153), DyLight 594-conjugate donkey anti-rat IgG (catalog 712-585-153), DyLight 488-conjugate donkey anti-rabbit IgG (catalog 711-545-152), and DyLight 594-conjugate donkey anti-goat IgG (catalog 705-585-147) were obtained from Jackson ImmunoResearch. Animals {#Sec4} ------- Adult female C57BL/6 mice, 8--10 weeks of age, were purchased from Beijing HFK Bioscience Co., Ltd. The mice were maintained in standard housing cages under specific pathogen-free conditions. All procedures were approved by the Committee for Research and Animal Ethics of Shaanxi Normal University and were conducted in accordance with the US Public Health Service's Policy on the Humane Care and Use of Laboratory Animals. Investigators involved in pathological staining and analysis were blinded to the experimental groups during the experiments. Isolation of NMO-IgG {#Sec5} -------------------- The total IgG from NMOSD patient sera was purified as previously described \[[@CR24], [@CR25]\]. Serum was obtained from five patients (P1--P5) with an established diagnosis of NMOSD and strong AQP4 autoantibody titers (AQP4-IgG titers ≤ 1:100). The clinical details of the patients with NMOSD were previously described \[[@CR24]\]. Patient serum IgG was purified with protein-A resin (GeneScript, catalog: L00210) and eluted with 0.1 M glycine buffer (pH 2.8) and then neutralized in Tris buffer (1.0 M, pH 9.0). Finally, the samples were concentrated using Amicon ultra centrifugal filter units (100 kDa, Merck Millipore, catalog: UFC910008) to obtain NMO-IgG (15 mg/mL). Informed consent was obtained from all participants, and the study was approved by the Shaanxi Normal University Institutional Review Boards and Ethics Committee and was performed in accordance with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Induction of mouse NMOSD models and administration of TSA {#Sec6} --------------------------------------------------------- First, mice in the prophylactic group (pre-TSA, n = 20) were administered TSA (736 μg/kg/day, dissolved in PBS with 2% DMSO), and mice in the control group mice (n = 20) were administered vehicle (PBS with 2% DMSO) by intraperitoneal injection (i.p.) every 8 h for 3 days (Fig. [1](#Fig1){ref-type="fig"}b). Then, the NMOSD model was induced based on previous studies \[[@CR24], [@CR25]\], and TSA or vehicle was administered by i.p. until 16 h after NMOSD model induction. In addition, therapeutic group (post-TSA) mice were administered an equal dose of TSA at 8 h and 16 h after injection of NMO-IgG plus hC (n = 10). Briefly, all animals were anesthetized with 10% chloral hydrate (3 mL/kg, i.p.) and mounted on a stereotactic frame (RWD Life Science). Following a midline scalp incision, a burr hole with a diameter of 1 mm was made in the skull 2 mm to the right of the bregma. A 33-gauge needle attached to a 25-μL gas-tight glass syringe (Hamilton) was inserted 3 mm deep to infuse 6 μL of NMO-IgG and 4 μL of hC (Innovative Research, catalog: IPLA-CSER) in a total volume of 10 μL (at 1 μL/min). After 24 h, the brains were processed into frozen sections. Immunofluorescence staining {#Sec7} --------------------------- Brain sections (8 μM) were permeabilized with ice-cold [acetone](http://cn.bing.com/dict/search?q=acetone&FORM=BDVSP6&mkt=zh-cn) for 7 min. After being blocked with 5% BSA (Sigma-Aldrich) in PBS, the sections were incubated with primary antibodies against AQP4 (1:100, rabbit polyclonal antibody, Santa Cruz Biotechnology, catalog: sc-20812), glial fibrillary acidic protein (GFAP, 1:1000, goat polyclonal antibody, Abcam, catalog: ab53554), myelin basic protein (MBP, 1:200, rat monoclonal antibody, clone: 12, Merck Millipore, catalog: MAB1965), CD45 (1:100, rat monoclonal antibody, clone 30-F11, BD Pharmingen, catalog 550539), Ly6G (1:100, rat monoclonal antibody, clone 1A8, BD Pharmingen, catalog 551459), and ionized calcium binding adapter molecule 1 (Iba-1, 1:500, rabbit polyclonal antibody, Wako, catalog 019-19741) at 4 °C overnight. The sections were washed three times with ice-cold PBS and incubated with the corresponding secondary antibody at room temperature for 60 min. Finally, the slides were covered with Fluoroshield mounting medium with DAPI (Abcam, catalog: ab104139). The results were visualized using a fluorescence microscope (Leica DM6000B). Coronal brain sections through the needle tract were selected, and areas were defined by hand and quantified using the ImageJ software (NIH). Analysis of neutrophil apoptosis in vivo {#Sec8} ---------------------------------------- To analyze the extent of neutrophil apoptosis in the lesions, brain tissues were assessed by terminal deoxynucleotidyl transferase (TdT)-mediated biotin-dUTP nick end labeling (TUNEL) staining using an in situ cell death detection kit (Roche Diagnostics, catalog 11-684-795-910) combined with Ly6G immunofluorescence staining according to the manufacturer's instructions. The results were visualized by a Leica DM6000B fluorescence microscope. Murine BMNE isolation and culture {#Sec9} --------------------------------- Murine bone marrow-derived neutrophil (BMNE) isolation was performed as previously described \[[@CR35]\]. Briefly, femurs and tibias were removed and flushed with Hank's balanced salt solution (HBSS) without Ca^2+^/Mg^2+^. The BM cells were suspended in HBSS and overlaid on top of a 2-layer Percoll (GE Healthcare, catalog 17089101) gradient (72% and 65% in HBSS) and then centrifuged at 1200×g for 30 min at 25 °C. BMNEs were recovered at the interface of the 65--72% fractions, stained using antibody against APC-CD45 (clone 30-F11, eBioscience, catalog 17-0451-82), PE-Ly6G (clone 1A8-Ly6G, eBioscience, catalog 12-9668-80), Alexa Flour 488-CD11b (clone: M1/70, BioLegend, catalog 101219), and confirmed by flow cytometry (purity = 90.93%, Fig. S[1](#MOESM1){ref-type="media"}a). The neutrophils were washed twice, and then cultured in RPMI 1640 complete medium at 37 °C with 5% CO~2~. Human primary neutrophil and PBMC isolation {#Sec10} ------------------------------------------- Primary human neutrophils were obtained from anticoagulated venous blood (treated with EDTA) from healthy volunteers (male, n = 2; female, n = 3; 24--29 years old) who had not taken any drugs for at least 2 weeks before the experiments. Informed consent was obtained from all participants, and the study was approved by the Shaanxi Normal University Institutional Review Boards and Ethics Committee and was performed in accordance with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The blood was used within 2 h of drawing from the donor. Human neutrophils were isolated by density gradient centrifugation (500×g for 30 min) using Polymorphprep (Axis-Shield) \[[@CR36]\]. After centrifugation, the upper band of peripheral blood mononuclear cells (PBMCs) and the lower band of polymorphonuclear cells (PMNs) were harvested. Residual of erythrocytes were eliminated using red blood cell lysing buffer (BD Bioscience, catalog 555899) at room temperature for 5 min. Then, the process was ceased by adding RPMI 1640 basic medium, and the mixture was centrifuged at 250×g for 10 min at 4 °C. The purity of the isolated neutrophils was \> 99%, as determined by flow cytometry using FITC-conjugated anti-human CD15 antibody (clone: HI98, BioLegend, catalog 301903) staining (Fig. S[1](#MOESM1){ref-type="media"}b). Freshly isolated cells were then cultured in RPMI 1640 complete medium at 37 °C with 5% CO~2~. Neutrophil apoptosis analysis by flow cytometry {#Sec11} ----------------------------------------------- Human or murine neutrophils (1 × 10^6^ cells/mL/well) were plated onto 24-well plates and cultured with 100 ng/mL LPS in the absence or presence of TSA (1 μM, 5 μM, and 10 μM) at 37 °C with 5% CO~2~. The cells were harvested at the indicated time points and washed twice with cell staining buffer. Then, the cells were suspended in 500 μL of binding buffer and incubated with 5 μL of APC-Annexin V and 10 μL of propidium iodide (PI) (BioLegend, catalog 640932) at room temperature for 15 min in the dark. Flow cytometry analysis was performed using a NovoCyte flow cytometer (ACEA). At least 3 × 10^4^ events were recorded and analyzed using the FlowJo software (Tree Star). In all analyses, cell debris was eliminated by appropriate forward and side scatter gating. Western blotting {#Sec12} ---------------- The mice were anesthetized and perfused with 40 mL of PBS. Then, the hemispheres containing the lesions were homogenized in NP40 cell lysis buffer (Thermo Fisher, catalog: FNN0021) containing EDTA-free complete protease inhibitor cocktail tablets (Roche, catalog 11873580001) and a phosphatase inhibitor cocktail (04906845001) at 4 °C. For in vitro culture cells, neutrophils were harvested at the indicated time points. The cells were washed twice with ice-cold PBS and lysed with NP40 cell lysis buffer containing EDTA-free complete protease inhibitor cocktail tablets and a phosphatase inhibitor cocktail at 4 °C. Cell lysates were centrifuged at 16,000g for 20 min at 4 °C. After centrifugation, the cell lysis supernatant was collected, mixed with SDS loading buffer, and boiled for 10 min. Total protein (30 μg) was resolved on 12% sodium dodecyl sulfate-polyacrylamide gels (SDS-PAGEs) and transferred to 0.45 μM polyvinylidene fluoride (PVDF) membranes (Merck Millipore) using a wet transfer method. Then, the membranes were washed twice with Tris-buffered saline with 0.05% Tween-20 (TBST) and blocked with 5% skim milk (Difco) for 1 h at room temperature followed by incubation with rabbit anti-phosphorylated p65 (p-p65) polyclonal antibody (1:1000, Cell Signaling Technology, catalog 3033 T), rabbit anti-total p65 (t-p65) monoclonal antibody (1:1000, Cell Signaling Technology, catalog 8242 T), rabbit anti-phosphorylated p38 mitogen-activated protein kinase (p-p38 MAPK) monoclonal antibody (1:1000, Cell Signaling Technology, catalog 9211S), rabbit anti-total p38 MAPK (t-p38 MAPK) monoclonal antibody (1:1000, Cell Signaling Technology, catalog 9212S), rabbit anti-phosphorylated extracellular-regulated protein kinases 1/2 (p-Erk1/2) monoclonal antibody (1:1000, Cell Signaling Technology, catalog 4377), rabbit anti-total Erk1/2 (t-Erk1/2) polyclonal antibody (1:1000, Cell Signaling Technology, catalog 9102), rabbit anti-caspase-3 polyclonal antibody (1:1000, Proteintech, catalog 19677-1-AP), and rabbit anti-β-actin polyclonal antibody (1:1000, Bioss, catalog: bs-0061R) overnight at 4 °C. The membranes were washed three times with TBST and incubated with horseradish peroxidase-conjugated goat anti-rabbit (1:3000, Bioss, catalog: bs-0295G-HRP) secondary antibody. Then, the target proteins were visualized with ECL blotting reagents (GE Healthcare, catalog: RPN2109) by a digital gel image analysis system (Tanon 4600). Statistical analysis {#Sec13} -------------------- Statistical analysis was performed using the GraphPad Prism 5.0 software. The values are presented as the mean ± standard error of the mean (SEM). When two groups were compared, the nonparametric Mann-Whitney test was used. One-way analysis of variance (ANOVA) followed by Tukey's posttest was performed to analyze results from more than two groups. A p value \< 0.05 was deemed statistically significant. Results {#Sec14} ======= TSA administration dramatically decreases astrocyte damage in NMOSD mice {#Sec15} ------------------------------------------------------------------------ To assess astrocyte damage 24 h after brain injection, AQP4 and GFAP immunofluorescence staining was performed. We found extensive loss of AQP4 and GFAP immunoreactivity in the basal ganglia and corpus callosum around the injection site 24 h after intracerebral injection of NMO-IgG and hC (Fig. [1c](#Fig1){ref-type="fig"}). The reduction in the area of AQP4 immunoreactivity was comparable to the reduction in the area of GFAP immunoreactivity in each group (4.47 ± 0.21 mm^2^ versus 4.34 ± 0.23 mm^2^ in vehicle-treated mice, 2.96 ± 0.26 mm^2^ versus 2.67 ± 0.24 mm^2^ in pre-TSA-treated mice, 3.45 ± 0.32 versus 3.35 ± 0.30 in post-TSA-treated mice) (Fig. [1](#Fig1){ref-type="fig"}d). Reactive gliosis was generally observed around the lesion (Fig. [1](#Fig1){ref-type="fig"}c bottom). Compared with the vehicle group, the two groups of TSA-treated NMOSD mice had dramatically reduced loss of AQP4 and GFAP, and the pre-TSA group displayed more obvious reduction in AQP4 and GFAP loss than the post-TSA group (Fig. [1](#Fig1){ref-type="fig"}d right and Fig. [1](#Fig1){ref-type="fig"}e right). These results suggest that TSA has a potent protective effect on astrocytes after NMO-IgG-induced hC activation in the brain, and prophylactic administration seems to be a better strategy than posttreatment. TSA suppresses demyelination in NMOSD mice {#Sec16} ------------------------------------------ To analyze the alterations in myelin integrity in NMOSD mice upon TSA treatment, we performed MBP immunofluorescence staining of brain tissue. The injection site in all mice displayed disorganized myelinating fibers (Fig. [2](#Fig2){ref-type="fig"}a). The area of demyelination in TSA-treated NMOSD mice was notably reduced compared with that in mice treated with vehicle (from 1.16 ± 0.07 to 0.72 ± 0.10 mm^2^ for pre-TSA or 0.88 ± 0.03 mm^2^ for post-TSA Fig. [2](#Fig2){ref-type="fig"}b). Interestingly, the suppressive effect of TSA on demyelination appeared to be more effective in the prevention group than in the posttreatment group, but there was no significant difference between the pre-TSA and post-TSA groups (Fig. [2](#Fig2){ref-type="fig"}). This finding indicates that TSA may play an important role in maintaining myelin integrity after intracerebral injection of NMO-IgG and hC. Fig. 2TSA administration inhibits the loss of myelination in NMOSD lesions at 24 h after brain injection. a Representative immunofluorescence staining of MBP. Lower magnification: top. Higher magnification: bottom. Yellow line, needle tract; white lines delimit the lesion. b Quantification of the extent of demyelination is expressed as the absolute reduction in area (mm^2^) of MBP in NMOSD lesions. Vehicle, n = 15; pre-TSA, n = 15; post-TSA, n = 5. \*p \< 0.05, \\\*p \< 0.001. The data are representative of three independent experiments. One-way ANOVA followed by Tukey's multiple comparison tests was performed TSA restricts neuroinflammation in the brains of NMOSD mice {#Sec17} ----------------------------------------------------------- We showed that intracerebral injection of NMO-IgG and hC caused extensive neuroinflammation. High numbers of CD45^+^ cells were observed around the injection site at 24 h after intracerebral injection of NMO-IgG and hC (Fig. [3](#Fig3){ref-type="fig"}a top). In addition, we found that neutrophils were the dominant cell type among these inflammatory cells (Ly6G^+^, red; Fig. [3](#Fig3){ref-type="fig"}a middle). Compared with mice treated with vehicle, TSA-treated mice exhibited a profound reduction in the number of CD45^+^ cells (Fig. [3](#Fig3){ref-type="fig"}b top). Pre-TSA-treated mice also exhibited reduced infiltration of Ly6G^+^ neutrophils (Fig. [3](#Fig3){ref-type="fig"}b middle). Furthermore, mild microglia/macrophages (Iba^+^, green; Fig. [3](#Fig3){ref-type="fig"}a bottom) activation was observed around the NMOSD lesions, but there were no significant differences in the numbers of microglia/macrophages in the three groups (Fig. [3](#Fig3){ref-type="fig"}b bottom). These results suggest that TSA is capable of inhibiting inflammation in the acute stage of NMOSD mice for at least 24 h, and prophylactic TSA administration seems suppress neutrophils inflammation. Fig. 3TSA administration markedly ameliorates neuroinflammation in NMOSD lesions at 24 h after brain injection. a Representative immunofluorescence staining of CD45^+^ leukocytes (top, red), Ly6G^+^ neutrophils (middle, red), and Iba^+^-activated microglia and macrophages (bottom, green) in the NMOSD lesion-side hemisphere at 24 h after brain injection. Yellow lines indicate needle tract; white lines delimit the lesion. b Inflammatory cells' quantification is expressed as the absolute area (mm^2^) \[CD45^+^ leukocytes (top) and Ly6G^+^ neutrophils (middle)\] or the number of inflammatory cell subsets \[Iba-1^+^-activated microglia and macrophages (bottom)\] in the NMOSD lesion-side hemisphere. Vehicle, n = 15; pre-TSA, n = 15; post-TSA, n = 5. \*p \< 0.05, \\\*p \< 0.001. The data are representative of three independent experiments. One-way ANOVA followed by Tukey's multiple comparison tests was performed TSA accelerates recruited neutrophils apoptosis in the brains of NMOSD mice {#Sec18} --------------------------------------------------------------------------- To clarify whether the protective effect of TSA after injection of NMO-IgG and hC is mediated by promoting the resolution of inflammation, tissues were immunofluorescently stained for TUNEL and the dominant inflammatory cell (neutrophil)-specific marker (Ly6G). We observed that abundant levels of neutrophils were colocalized with TUNEL in NMOSD lesions at 24 h after brain infusion (Fig. [4](#Fig4){ref-type="fig"}a). Interestingly, the frequency of apoptotic neutrophils (Ly6G^+^ TUNEL^+^) was substantially higher in mice pretreated with TSA than in mice treated with vehicle (44.33% ± 2.50% versus 32.29% ± 1.75%; Mann-Whitney test, p = 0.0079, n = 15; Fig. [4](#Fig4){ref-type="fig"}b). Thus, these results indicate that TSA-induced neutrophil apoptosis is partially responsible for the beneficial effects of TSA on the mouse NMOSD model. Fig. 4TSA promotes apoptosis of recruited neutrophils in NMOSD lesions at 24 h after brain injection. Ly6G and TUNEL immunofluorescence double staining was performed to determine the extent of spontaneous neutrophil apoptosis in NMOSD lesions at 24 h after brain injection. a Representative immunofluorescence staining of Ly6G (top, red) and TUNEL (middle, green). Yellow lines indicate the needle tract; white lines delimit the lesion; purple rectangles indicate magnified field. White arrows indicate apoptotic neutrophils (Ly6G^+^ TUNEL^+^, yellow); purple arrows indicate nonapoptotic neutrophils (Ly6G^+^ TUNEL^−^, red). b Quantification of neutrophil apoptosis was defined as the ratio of Ly6G^+^ TUNEL^+^ cells in Ly6G^+^ cells in the NMOSD lesion-side hemisphere (n = 15, \\p \< 0.01). The data are representative of three independent experiments. Mann-Whitney tests was performed TSA induces mouse neutrophil apoptosis via caspase-dependent pathways in vitro {#Sec19} ------------------------------------------------------------------------------ To investigate whether TSA can directly induce neutrophil apoptosis, freshly isolated mouse BMNEs were incubated with increasing concentrations of TSA in the presence of 100 ng/mL LPS for 48 h, and then viability and apoptosis were analyzed by flow cytometry. Annexin V/PI staining showed that the frequency of apoptotic neutrophils was strikingly suppressed in response to inflammatory stimuli (from 34.2 ± 1.0 to 14.4 ± 2.0%), and the life span of neutrophils was prolonged (Fig. S[2](#MOESM1){ref-type="media"}a, b). The addition of TSA to neutrophil cultures potently inhibited LPS-induced neutrophil survival and induced neutrophil apoptosis in a concentration-dependent manner. After incubation with TSA (5 μM and 10 μM) in the presence of LPS for 21 h, the frequency of apoptotic neutrophils was significantly elevated (Fig. [5](#Fig5){ref-type="fig"}a, b). In addition, western blotting further demonstrated that TSA promoted the cleavage of procaspase-3 in a dose-dependent manner (Fig. [5](#Fig5){ref-type="fig"}c, d). Collectively, these data indicate that TSA can promote mouse neutrophil apoptosis under inflammatory conditions via caspase-dependent pathways in vitro. Fig. 5TSA promotes mouse neutrophil apoptosis via caspase-dependent pathways in vitro. Fresh mouse BMNEs (1 × 10^6^ cells/mL/well) were cultured in RPMI 1640 complete medium at 37 °C with 5% CO~2~ and stimulated with 100 ng/mL LPS for 21 h in the absence or presence of TSA (1 μM, 5 μM, and 10 μM). a, b Neutrophil apoptosis was assessed by flow cytometry after APC-Annexin V/PI labeling. For each condition, 3 × 10^4^ events were recorded within the neutrophil gate. a Representative flow cytometry plots showing Annexin-V/PI binding. b The apoptosis rate of mouse BMNEs (n = 5, \\\*p \< 0.001). c Western blotting was performed to analyze procaspase 3 (33 kDa) and cleaved caspase 3 (17 kDa and 19 kDa) expression. d Quantification of the relative expression of procaspase 3 and cleaved caspase 3 (n = 5, \*p \< 0.05, \\p \< 0.01). β-actin was used as an internal control. The data are representative of three independent experiments. One-way ANOVA followed by Tukey's multiple comparison tests was performed TSA also induces human neutrophil apoptosis via caspase-dependent pathways in vitro {#Sec20} ----------------------------------------------------------------------------------- In human neutrophil cultures, we also observed that inflammatory stimuli (100 ng/mL LPS) could dramatically promote neutrophil survival and suppress neutrophil apoptosis (from 74.3 ± 3.1 to 33.3 ± 0.8%) (Fig. S[2](#MOESM1){ref-type="media"}c, d). After culture with TSA (5 μM and 10 μM) in the presence of LPS for 21 h, the frequency of apoptotic neutrophils was strongly elevated (Fig. [6](#Fig6){ref-type="fig"}a, b top). Interestingly, TSA had no significant effect on PBMC viability and apoptosis (Fig. [6](#Fig6){ref-type="fig"}a, b bottom). TSA selectively induced apoptosis in neutrophils but not PBMCs. Similarly, western blotting also showed that TSA significantly enhanced the cleavage of procaspase-3 in a dose-dependent manner (Fig. [6](#Fig6){ref-type="fig"}c, d). Taken together, these results suggest that TSA also promotes human neutrophil apoptosis via caspase-dependent pathways in vitro. Fig. 6TSA accelerates human neutrophil apoptosis via caspase-dependent pathways in vitro. Human primary neutrophils and PBMCs (1 × 10^6^ cells/mL/well) were plated in 24-well plates at 37 °C with 5% CO~2~ and stimulated with 100 ng/mL LPS for 21 h in the absence or presence of TSA (1 μM, 5 μM, and 10 μM). a, b Flow cytometry analysis of cultured neutrophils and PBMCs was performed after APC-Annexin V/PI staining. a Representative flow cytometry plots showing Annexin-V/PI binding (human neutrophils, top; human PBMCs, bottom). b Quantification of the rate of apoptotic cells (human neutrophils, top; human PBMCs, bottom; n = 5, \\\*p \< 0.001). c The cleavage of procaspase-3 was analyzed by western blotting. d Quantification of the relative expression of procaspase 3 and cleaved caspase 3 (n = 5, \*p \< 0.05, \\p \< 0.01, \\\*p \< 0.001). The expression of target proteins was normalized to β-actin. The data are representative of three independent experiments. One-way ANOVA followed by Tukey's multiple comparison tests was performed TSA suppresses activation of the NF-κB signaling pathway in vivo {#Sec21} ---------------------------------------------------------------- To clarify the mechanism of TSA and promote its clinical transformation, it is necessary to evaluate other reported actions of this drug. Previous studies have demonstrated that TSA can inhibit activation of the NF-κB \[[@CR37]--[@CR39]\] and p38 MAPK \[[@CR37], [@CR38], [@CR40]\] signaling pathways. Thus, the expression of key molecules of the NF-κB (p-p65 and t-p65) and p38 MAPK (p-p38, t-p38, p-Erk1/2, and t-Erk1/2) signaling pathways were examined in the lesion-side hemispheres from the three groups by western blotting. We found that the relative level of p-p65 to t-p65 was dramatically reduced in the TSA-treated group compared to the control group, and the pre-TSA group displayed a more obvious effect than the post-TSA group (Fig. [7](#Fig7){ref-type="fig"}a, b). However, the expression of p-p38, t-p38, p-Erk1/2, and t-Erk1/2 was not changed compared to that of the control group (Fig. [7](#Fig7){ref-type="fig"}a, c, d). These results indicate that TSA can inhibit activation of the NF-κB signaling pathway in vivo, which may be related to the therapeutic effect of TSA on NMOSD mice. Fig. 7TSA suppressed activation of the NF-κB signaling pathway. Western blotting was performed to analyze the levels of p-p65, t-p65, p-p38, t-p38, p-Erk1/2, and t-Erk1/2 in the hemispheres containing the lesions from TSA- or vehicle-treated NMOSD mice at the end of the experiment. a Representative western blots showing p-p65, t-p65, p-p38, t-p38, p-Erk1/2, t-Erk1/2, and β-actin. b Quantification of the expression of p-p65 relative to t-p65. c Quantification of the expression of p-p38 relative to t-p38. d Quantification of the expression of p-Erk relative to t-Erk (n = 5). β-actin was used as an internal control. n = 5, \*p \< 0.05, \\\*p \< 0.001. One-way ANOVA followed by Tukey's multiple comparison tests was performed Discussion {#Sec22} ========== Here, we confirmed that neutrophils are the dominant inflammatory cells during the acute phase in relatively early NMOSD model lesions (24 h). TSA can efficiently reduce NMO-IgG-induced damage in the mouse brain, accompanied by an increased frequency of apoptotic neutrophils. These results suggest that the possible mechanisms by which TSA exerts protective effects in this NMOSD model may be by promoting the resolution of inflammation by accelerating neutrophil apoptosis. To validate the effect of TSA on neutrophils in the NMOSD model, we performed in vitro experiments. We observed that the inflammatory milieu dramatically suppressed neutrophil apoptosis and prolonged the neutrophil lifespan, but TSA partially alleviated this suppression and significantly accelerated apoptosis in primary neutrophils isolated from mouse BM or human peripheral blood. This finding is nearly consistent with a previous finding that TSA can promote the elimination of inflammation by accelerating the process of neutrophil reverse migration and spontaneous apoptosis in a zebrafish sterile injury model \[[@CR33]\]. However, the experimental design and methods used in our study make it difficult to discern whether the reduction in neutrophils is partially caused by reverse migration. This is an important question, and future studies are warranted. The inflammatory cells recruited into NMOSD lesions mainly include neutrophils, macrophages, and eosinophils, with relatively few T lymphocytes \[[@CR21], [@CR41]\]. During the pathogenesis of NMOSD, these inflammatory cells represent temporal dynamic changes. Neutrophils are the first cell type to be recruited into NMOSD lesions, which occurs only a few hours after intracerebral injection of NMO-IgG and hC \[[@CR21]\]. Within 12 h, there is perivascular complement activation, with the loss of AQP4 and early myelin loss \[[@CR21], [@CR42]\]. By 24 h, neutrophils enter the lesion and accumulate around small vessels \[[@CR21], [@CR42]\], where AQP4 is highly expressed. At the chronic phase (7 days), few neutrophils remain in the lesions, and macrophages dominate and infiltrate extensively into the white matter \[[@CR21], [@CR42]\]. Perivascular eosinophils are also observed at 7 days. However, the role of macrophages and eosinophils in NMOSD lesions is unclear \[[@CR21]\]. For these reasons, 24 h postinjection was selected as the detection time point in the present study. Neutrophils are the most abundant cell type in the innate immune system and have been implicated in the pathogenesis of NMOSD \[[@CR19], [@CR21], [@CR23]\]. Under physiological conditions, circulating neutrophils have a very short half-life because they constitutively undergo apoptosis and are functionally quiescent \[[@CR43], [@CR44]\]. However, the neutrophil lifespan is dynamically influenced during the course of the inflammatory response and depends on the balance between signals that promote survival and those that accelerate apoptosis \[[@CR45]--[@CR47]\]. After recruitment to injured sites, the lifespan of neutrophils is markedly prolonged by the inflammatory microenvironment, which allows effective phagocytosis and probably causes excessive tissue damage. The ideal result is that neutrophils will undergo spontaneous apoptosis after their physiological function (e.g., elimination of damaged tissue components) has been fulfilled in the injured sites. In contrast to necrosis, apoptosis renders neutrophils unresponsive to extracellular stimuli, prevents the release of toxic constituents, and facilitates their recognition and clearance by macrophages \[[@CR48], [@CR49]\]. Thus, neutrophil apoptosis has emerged as a crucial control point in determining the outcome of the inflammatory response. Blocking neutrophil apoptosis produces persistent inflammation, whereas accelerating neutrophil apoptosis enhances the resolution of inflammation to restrain excessive tissue injury and avoid persistent chronic inflammation \[[@CR44]\]. As a lipid-soluble compound, TSA can penetrate the BBB and enter the brain parenchyma, partially avoiding the difficulty faced by water-soluble drugs. Sucher et al. reported that TSA was detectable in the brain within 5 min of injection, which is almost the same time it was detected in the circulation. In addition, the concentration of TSA in the brain reached a peak at 60 min, decreased slowly over several hours, and was undetectable after 8 h \[[@CR50]\]. Currently, therapeutic strategies for NMOSD mainly focus on limiting the deleterious effects of AQP4-IgG binding \[[@CR51]\], complement activation \[[@CR17], [@CR24], [@CR52], [@CR53]\], IgG production \[[@CR54]\], neutrophil activation \[[@CR21]\], and eosinophil activation \[[@CR55]\]. Thus, screening small molecule compounds to target these deleterious effects will be a potential approach for NMOSD therapy \[[@CR19]\]. Conclusion {#Sec23} ========== In summary, TSA can act as a prophylactic treatment that reduces NMO-IgG-induced damage in the mouse brain by promoting the resolution of inflammation by inducing neutrophil apoptosis. TSA can also restrict activation of the NF-κB signaling pathway in vivo, which may be involved in the effect of TSA on NMOSD mice. Thus, I believe that further studies are needed to elucidate these mechanisms in the future. Collectively, these findings strongly support the notion that TSA has potent anti-inflammatory functions and may serve as a promising therapeutic agent for human neutrophil-associated inflammatory disorders, such as NMOSD. Supplementary information ========================= {#Sec24} Additional file 1: Fig. S1. Purity test for neutrophils isolated from mouse BM or human peripheral blood. Fig. S2. Inflammatory stimuli dramatically suppresses neutrophil spontaneous apoptosis. NMOSD : Neuromyelitis optica spectrum disorder AQP4 : Aquaporin-4 TSA : Tanshinone IIA GFAP : Glial fibrillary acidic protein MBP : Myelin basic protein CNS : Central nervous system BBB : Blood-brain barrier RRMS : Relapsing-remitting multiple sclerosis IgG : Immunoglobulin G CDC : Complement-dependent cytotoxicity ADCC : Antibody-dependent cell-mediated cytotoxicity hC : Human complement G-CSF : Granulocyte-colony stimulating factor Iba-1 : Ionized calcium binding adapter molecule 1 BMNE : Bone marrow-derived neutrophil PBS : Phosphate buffer saline HBSS : Hank's balanced salt solution PBMCs : Peripheral blood mononuclear cells PMNs : Polymorphonuclear cells SDS-PAGE : Sodium dodecyl sulfate-polyacrylamide gel PVDF : Polyvinylidene fluoride TUNEL : Terminal deoxynucleotidyl transferase (TdT)-mediated biotin-dUTP nick end labeling PI : Propidium iodide MAPK : Mitogen-activated protein kinase Erk : Extracellular-regulated protein kinase Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Ye Gong and Ya-ling Zhang contributed equally to this work. Supplementary information ========================= Supplementary information accompanies this paper at 10.1186/s12974-020-01874-6. None. YPY and LJC conceived and designed the experiments. YG, YLZ, ZW, HHS, YCL, AWL, LLT, WLZ, and XLD performed the experimental work. YPY, LJC, YG, YLZ, and ZW analyzed the data. YG, YLZ, ZW, and YF wrote the manuscript. All authors discussed the results. YG and YLZ contributed equally to this work. The author(s) read and approved the final manuscript. This study was supported by the Natural Science Foundation of China (81571596, 81601044, and 81771279) and the Fundamental Research Funds for the Central Universities (2018TS088, GK201701009, and GK201603110). Please contact the author for data requests. All procedures were approved by the Committee for Research and Animal Ethics of Shaanxi Normal University and conducted in accordance with the US Public Health Service's Policy on Human Care and Use of Laboratory Animals. Not applicable. The authors declare that they have no competing interests.	{ "pile_set_name": "PubMed Central" }
![](hosplond73603-0007){#sp1 .631} ![](hosplond73603-0008){#sp2 .632}	{ "pile_set_name": "PubMed Central" }
Specifications TableTableSubject areaBiologyMore specific subject areaMesenchymal stromal cells and tissue regenerationType of dataTable, graphs and microphotographsHow data was acquiredIllumina Human v2 microRNA panel was used to obtain a list of evidently expressed microRNAs. Capillary-like structures were assayed in 24 h under the light microscope (Leica). Concentrations of HGF and angiopoetin-1 in lysates of MSCs transfected with pre-miR-92a (Ambion), anti-miR-92a or pre-miR negative control \#1 were measured by ELISA kits (R&D Systems).Data formatAnalyzedExperimental factorsMSC nucleofected with pre-miR-92a, anti-miR-92a or scramble oligos; MSC conditioned medium applied to HUVEC, recombinant human HGF addedExperimental featuresmicroRNA content in adipose-derived mesenchymal stromal cells was analysed. Adipose-derived MSCs were nucleofected with pre-miR-92a, anti-miR-92a or scramble oligos. HGF and angiopoietin-1 were measured in cell lysates by ELISA. Conditioned medium of transfected cells was applied to HUVEC and tube formation was assessed. Extracellular vesicles were depleted by ultracentrifugation.Data source locationLomonosov Moscow State University, Moscow, RussiaData accessibilityData with this article Value of the data {#s0005} ===================== •List of microRNAs expressed in adipose-derived MSCs will help to develop new experiments to study functional activities of these cells.•Overexpression or down-regulation of miR-92a by nucleofection does not affect MSC viability.•Our data point to HGF as a new target of miR-92a. 1 Data {#s0010} ====== Using human microRNA v2 panel (Illumina) we have identified 586 miR species, which were evidently expressed in MSCs (see [Table 1](#t0005){ref-type="table"}). We selected miR-92a as a one of the most abundant angio-miRs expressed in MSCs and confirmed its expression by real-time PCR [@bib1]. Then, we overexpressed or down-regulated its content using nucleofection. We examined viability of transfected cells, which was about 90% and did not differ between cells transfected with pre-miR-92a, anti-miR-92a or scramble oilgos ([Fig. 1](#f0005){ref-type="fig"}). We also analyzed the content of HGF and angiopoietin-1 in these cells and found that intracellular content of HGF was 2.6 times lower in MSCs transfected with pre-miR-92a comparing to scramble transfected cells; however, angiopoietin content within MSCs did not change significantly (see [Fig. 2](#f0010){ref-type="fig"}). We collected conditioned medium of MSCs transfected with pre-miR-92a, anti-miR-92a or scramble oilgos, applied it to HUVEC and analyzed their viability. Conditioned medium of transfected MSCs did not affect the viability of HUVEC (see [Fig. 3](#f0015){ref-type="fig"}), which was about 90%. Addition of recombinant HGF but not angiopoietin-1 to the conditioned medium of MSCs transfected with pre-miR-92a restored its ability to stimulate the tube formation by HUVEC (see [Fig. 4](#f0020){ref-type="fig"}). We also examined if the suppressive effect of conditioned medium of MSCs, which overexpress miR-92a, could be mediated by a direct transfer of this microRNA to endothelial cells by extracellular vesicles. We removed these vesicles from conditioned medium by ultracentrifugation and analyzed the effect of cleared medium on tube formation. Removal of extracellular vesicles completely abrogated the ability of conditioned medium to induce tube formation (see [Fig. 5](#f0025){ref-type="fig"}). Data interpretation and discussion can be found in [@bib1]. 2 Experimental design, materials and methods {#s0015} ============================================ 2.1 Cell culture {#s0020} ---------------- MSCs were isolated from subcutaneous fat tissue of healthy young donors using enzymatic digestion as previously described [@bib2]. All donors gave their informed consent and the local ethics committee approved the study protocol. Cells were cultured in AdvanceSTEM Mesenchymal Stem Cell Media containing 10% AdvanceSTEM Supplement (HyClone), 1% antibiotic--antimycotic solution (HyClone) at 37 °C in 5% CO~2~ incubator. Cells were passaged at 70% confluency using HyQTase solution (HyClone). For the experiments, MSCs cultured up to 3rd--4th passages were used. Human umbilical vein endothelial cells (HUVEC) were isolated from human umbilical cord vein as previously described [@bib3]. Cells were cultured on gelatin-coated plastic in endothelial growth medium (EGM-2, Lonza) and used for experiments at 3--4 passages. 2.2 miRNA isolation, hybridization and real-time PCR {#s0025} ---------------------------------------------------- Total RNA was extracted from MSCs with Ambion® mirVana™ miRNA Isolation Kit according to manufacture instruction. 200 ng of total RNA was processed according to MicroRNA Assay Guide (Illumina) using Human v2 microRNA panel (1146 probes). Data acquisition and analysis of evidently expressed microRNA were performed by GenomeStudio software (Illumina) using gene expression module. MiRs with detection p value\<0.05 were considered as evidently expressed. MiR-92a in MSCs was detected using miRVana qRT-PCR miRNA detection kit (Ambion), according to manufacturer׳s protocol. Reverse transcription was performed during 30 min at 37 °C using 25 ng of RNA and hsa-miR-92a RT-Primer (Ambion). Real-time PCR was performed using hsa-miR92a -- qRT-PCR assay primer (Ambion) and ready-to-use reaction mix, containing DNA polymerase, SYBR Green and ROX (Evrogen) in 7500 Fast Real-time PCR system (Applied Biosystems). 2.3 Cell viability {#s0030} ------------------ Viability of MSCs transfected with pre-miR-92a, anti-mi-92a or scramble oligos was assessed 48 h post-transfection. HUVEC viability was analyzed after 24 h incubation in the conditioned medium of MSCs. To assess the viability cells were trypsinized and counted using Trypan blue staining on Countess Automated Cell Counter. 2.4 in vitro tube formation assay {#s0035} --------------------------------- HUVEC were seeded in 48-well plates coated with growth factor reduced Matrigel (BD Bioscience, 150 µl per well) in concentration 2×10^4^ cells per well and MSC conditioned media (300 µl per well) were added [@bib4]. Three wells were used for each sample of conditioned medium. Supplement-free serum-free endothelial basal medium (EBM-2, Lonza) was utilized as a negative control; endothelial growth medium (EGM-2, Lonza) with 10% of FBS served as positive control. Plates were placed into CO~2~-incubator at 37 °C and capillary-like structures were assayed in 24 h under the light microscope (Leica). Total length of tubular structures was counted in 5 random fields of view per well (objective 10×) using MetaMorph 5.0 software (Universal Imaging). To evaluate a significance of HGF and angiopoietin-1 for tube formation we also supplemented the conditioned medium of MSCs overexpressing miR-92a with recombinant growth factors (R&D). To examine an impact of extracellular vesicles on angiogenic action of MSC conditioned medium, we removed them by ultracentrifugation as described in [@bib3]. 2.5 ELISA of HGF and angiopoietin-1 {#s0040} ----------------------------------- Concentrations of HGF and angiopoetin-1 in lysates of MSCs transfected with pre-miR-92a (Ambion), anti-miR-92a or pre-miR negative control \#1 were measured using ELISA kits (R&D Systems) according to manufacturer instructions. Appendix A. Supplementary material {#s0055} ================================== Supplementary material This study was supported by subsidy \#14.607.21.0045 from the Russian Ministry of Education and Science (GrantID RFMEFI60714×0004). Supplementary data associated with this article can be found in the online version at <http://dx.doi.org/10.1016/j.dib.2015.12.021>. ![MSC viability. MSC transfected with pre-miR-92a, anti-miR-92a or scramble oilgos were grown for 48 h post-transfection. The portion of viable cells were calculated as a number of Trypan-blue excluding cells per 100 MSCs using Countess Cell Counter, Invitrogen.](gr1){#f0005} ![Angiogenic factors content in MSCs 48 h after transfection with either scramble oligos (scramble), pre-miR-92a (pre-miR92a) or anti-miR-92a (anti-miR92a). The content of HGF and angiopoietin-1 in cell lysates was measured by ELISA and normalized to cell counts. Presented are data of 3 experiments. \*p\<0.05 vs. scramble transfected cells.](gr2){#f0010} ![HUVEC viability in the conditioned medium of MSC transfected with pre-miR-92a, anti-miR-92a or scramble oilgos. The portion of viable cells was calculated as a number of Trypan-blue excluding cells per 100 cells.](gr3){#f0015} ![Capillary-like tubules formation in the presence of MSC conditioned medium. Representative microphotographs of capillary-like tubules formed in growth medium without serum (negative control), in the presence of 10% FBS (positive control), in the presence of conditioned medium of MSCs transfected either with scramble oligos (scramble), anti-miR-92a (anti-miR92a), pre-miR-92a (pre-miR92a). Conditioned medium of MSCs overexpressing pre-miR-92a with addition of recombinant human HGF and HGF alone was also used. Graph represents morphometric analysis of capillary-like structures total length per view field. Presented are data of 3 experiments. \*p\<0.05 vs. conditioned medium of MSCs transfected with pre-miR-92a.](gr4){#f0020} ![Capillary-like tubules formation in the presence of MSC conditioned medium. Representative microphotographs of capillary-like tubules formed in growth medium without serum (negative control), in the presence of 10% FBS (positive control), in the presence of conditioned medium of MSCs transfected either with scramble oligos (scramble), pre-miR-92a (pre-miR92a) or anti-miR-92a (anti-miR92a) or in the presence of conditioned medium from transfected cells after removing extracellular vesicles by ultracentrifugation.](gr5){#f0025} ###### List of microRNAs evidently expressed in adipose-derived MSC. Table 1 TargetID Mean SD ---------------------------------------- ---------------- ----------- hsa-miR-1280 23567.3 480.0 hsa-miR-21 23136.3 781.4 hsa-let-7a 22779.0 692.8 hsa-miR-214 22123.8 804.3 solexa-3927-221 21213.3 1085.0 hsa-miR-221 20160.3 954.3 hsa-miR-23a 19960.5 719.2 hsa-miR-199a:9.1 19815.5 1912.1 hsa-miR-720 19392.3 513.3 hsa-let-7g 19323.3 974.2 hsa-miR-125b 18808.8 643.9 hsa-miR-24 18341.5 1204.0 HS_22.1 18187.5 961.4 hsa-let-7f 18114.5 609.4 hsa-miR-320d,hsa-miR-320b,hsa-miR-320a 18100.8 428.0 hsa-miR-1274b 17902.8 2536.3 hsa-miR-26a 17701.5 1134.3 hsa-let-7e 17673.3 2166.1 hsa-miR-768-3p:11.0 17642.0 2731.8 hsa-miR-100 17579.3 1745.4 hsa-let-7i 17241.8 804.4 hsa-miR-923 16997.3 3498.3 hsa-let-7b 16807.8 585.4 hsa-miR-222 16276.8 1295.1 hsa-miR-125a-5p 16050.5 1068.0 hsa-miR-199a-3p,hsa-miR-199b-3p 15980.3 1402.7 hsa-miR-92a 15764.3 650.1 hsa-let-7c 15695.5 2136.7 hsa-miR-151-5p 15560.0 1490.1 hsa-miR-23b 15545.5 2236.8 hsa-miR-1308 15318.5 2330.0 hsa-miR-196a 15272.8 645.1 hsa-miR-16 15181.3 1449.2 hsa-miR-25 15137.8 1294.5 hsa-miR-193a-5p 15061.8 978.5 hsa-miR-27a 15018.3 2589.6 solexa-555-1991 14762.5 897.6 hsa-miR-143 14714.8 1900.5 hsa-miR-15b 14409.3 884.1 HS_100 14373.3 1144.9 hsa-let-7d 14286.3 1618.1 hsa-miR-1274a 13864.0 1486.9 hsa-miR-145 13787.3 1564.7 hsa-miR-31 13704.5 1530.4 hsa-miR-574-3p 13679.8 1696.2 hsa-miR-152 13586.8 2963.6 hsa-miR-26b 13480.8 1235.7 hsa-miR-1260 13472.3 1928.7 hsa-miR-1246 13409.8 2074.7 hsa-miR-1826 13393.8 3269.9 HS_192.1 13391.0 2645.2 solexa-2952-306 13256.3 953.8 hsa-miR-424 13215.3 4095.2 hsa-miR-191 13163.8 1851.0 hsa-miR-197 12943.0 1094.8 hsa-miR-98 12607.5 1564.1 hsa-miR-199a-5p 12574.5 3478.1 HS_204.1 12545.0 1184.5 hsa-miR-29a 12474.3 2501.4 HS_263.1 12208.3 2000.0 hsa-miR-224 11944.3 1349.4 hsa-miR-10b 11895.3 1703.7 hsa-miR-27b 11848.5 3310.9 hsa-miR-1201 11518.3 2715.9 hsa-miR-20a 11454.3 2086.0 hsa-miR-30d 11199.5 2248.6 hsa-miR-503 11092.3 1588.7 solexa-3464-254 11059.3 391.3 hsa-miR-132 10687.5 1492.4 hsa-miR-768-5p:11.0 10639.3 1341.5 hsa-miR-10a 10583.8 1824.0 hsa-miR-155 10402.0 2429.4 HS_243.1 10376.3 2551.2 hsa-miR-193b 10359.5 2138.7 hsa-miR-664 10341.0 1762.0 hsa-miR-432 10037.0 2265.1 hsa-miR-195 9968.0 2337.5 hsa-let-7d 9928.8 783.0 hsa-miR-29b-1 9900.3 1879.9 hsa-miR-34a 9880.8 2109.6 hsa-miR-99b 9711.8 2675.7 hsa-miR-30c 9600.3 2441.4 hsa-miR-134 9303.3 934.1 hsa-miR-93 9285.0 592.1 hsa-miR-382 9096.3 1240.3 hsa-miR-92b 8970.5 814.7 hsa-miR-151-3p 8963.0 1436.3 hsa-miR-484 8832.0 1093.8 hsa-miR-28-5p 8679.8 2117.9 hsa-miR-455-3p 8657.8 1939.9 hsa-miR-423-3p 8573.5 1260.9 hsa-miR-424 8397.5 1168.2 solexa-51-13984 8340.0 850.3 hsa-miR-409-3p 8295.8 866.6 HS_284.1 8291.0 1037.4 hsa-miR-374a 8283.5 2769.2 hsa-miR-30e 8194.8 2534.1 hsa-miR-1228 7994.8 492.6 hsa-miR-29b 7992.3 774.7 hsa-miR-148a 7992.0 3801.0 hsa-miR-185 7845.3 1517.6 hsa-miR-15a 7627.3 3472.2 hsa-let-7b 7564,8 881.2 hsa-miR-365 7544.3 1540.2 hsa-miR-493 7387.3 1028.6 hsa-miR-324-5p 7379.0 1880.9 hsa-miR-361-5p 7259.5 2303.2 hsa-miR-544 7162.3 3061.8 hsa-miR-324-3p 7112.3 1953.1 hsa-miR-30b 7072.0 1144.3 hsa-miR-379 6963.3 771.4 hsa-miR-342-3p 6835.3 1553.5 hsa-miR-99a 6792.0 3415.3 hsa-miR-22 6725.0 3846.5 hsa-miR-328 6488.0 1575.3 hsa-miR-151:9.1 6460.3 2527.1 hsa-miR-146a 6456.3 2689.7 solexa-539-2056 6418.3 3114.5 HS_96 6396.5 551.2 hsa-miR-196b 6232.5 647.2 hsa-miR-181b 6187.5 2869.8 solexa-9029-92 6172.0 299.1 hsa-miR-423-5p 6119.0 1363.2 hsa-miR-615-3p 6081.8 332.8 hsa-miR-708 6033.8 2431.3 hsa-miR-411 5970.5 1412.8 hsa-miR-130b 5941.5 1460.8 hsa-miR-146b-5p 5919.0 3049.7 hsa-miR-329 5866.5 1477.3 solexa-8048-104 5741.5 2872.2 hsa-miR-22 5670.8 2076.8 hsa-miR-130a 5667.3 853.1 hsa-miR-181a 5661.5 4037.6 hsa-miR-199b-5p 5543.8 1838.6 hsa-miR-337-3p 5513.0 1579.9 hsa-miR-28-3p 5509.5 1213.1 hsa-miR-140-3p 5425.3 814.1 hsa-miR-1275 5317.8 1528.2 hsa-miR-218 5297.0 488.0 hsa-miR-30a 5277.8 2734.4 hsa-miR-106b 5075.3 1656.1 hsa-miR-425 4907.8 3170.9 hsa-miR-886-5p 4860.3 632.8 hsa-miR-17 4835.8 558.0 solexa-8211-102 4810.8 3885.4 solexa-499-2217 4718.8 514.6 hsa-miR-103 4697.0 1083.3 hsa-miR-574-5p 4654.8 1382.9 hsa-miR-594:9.1 4525.3 1320.1 hsa-miR-939 4440.3 524.5 hsa-miR-889 4371.3 835.2 hsa-miR-877 4359.5 2022.5 hsa-miR-145 4353.5 2399.0 hsa-miR-576-5p 4334.8 688.7 hsa-miR-744 4311.3 520.4 solexa-578-1915 4307.3 785.1 HS_188 4275.3 399.0 hsa-miR-331-3p 4249.8 507.8 hsa-miR-370 4245.3 941.4 hsa-miR-7 4233.0 887.0 hsa-miR-129-5p 4125.3 289.4 hsa-miR-127-3p 4077.5 1046.9 hsa-miR-625 4070.8 855.9 HS_108.1 4007.3 1646.7 HS_29 3980.5 1256.6 hsa-miR-518a-3p 3974.5 439.7 hsa-miR-542-3p 3970.3 2890.8 hsa-miR-1249 3959.0 1498.5 hsa-miR-106a 3835.0 825.1 hsa-miR-193b 3768.3 1340.9 hsa-miR-24-2 3718.0 1956.8 hsa-miR-532-3p 3716.8 1867.2 hsa-miR-128 3681.0 888.9 hsa-miR-628-5p 3673.0 1407.6 hsa-miR-29c 3658.8 2366.8 hsa-miR-663 3592.3 588.6 hsa-miR-625 3586.5 745.3 hsa-miR-204 3579.8 443.0 hsa-miR-30c-1 3551.8 1099.1 hsa-miR-376c 3492.5 589.6 hsa-miR-221 3461.0 2028.3 hsa-miR-539 3454.3 466.6 hsa-miR-30a 3437.3 1699.3 hsa-miR-548b-5p 3402.5 343.6 hsa-miR-886-3p 3352.5 649.1 hsa-miR-7-1 3338.0 3726.7 hsa-miR-1300 3277.5 2082.2 hsa-miR-450a 3205.5 2247.8 hsa-miR-628-3p 3185.8 1065.8 hsa-miR-15b 3160.3 764.0 hsa-miR-130b 3152.0 1059.8 hsa-miR-372 3144.3 302.8 hsa-miR-433 3129.8 378.9 hsa-miR-335 3084.5 1863.3 hsa-miR-126 3052.3 703.7 hsa-miR-675 3024.3 591.5 hsa-miR-487b 3016.0 534.6 hsa-miR-16-2 3011.0 1345.6 hsa-miR-126 2964.3 2488.8 hsa-miR-299-5p 2726.3 738.4 hsa-miR-30e 2707.3 1939.7 hsa-miR-10a 2704.0 604.5 hsa-miR-31 2686.3 2011.9 hsa-miR-517a 2681.8 439.7 solexa-4793-177 2568.5 298.2 hsa-miR-421 2521.0 728.3 hsa-miR-342-5p 2514.5 439.7 hsa-miR-654-5p 2511.8 472.9 hsa-miR-940 2479.5 212.6 HS_239 2468.8 749.4 hsa-miR-143 2447.8 213.5 hsa-miR-138 2432.8 1123.8 hsa-miR-485-3p 2414.8 1197.9 hsa-miR-450b-5p 2403.8 1373.2 hsa-miR-137 2398.5 2711.8 HS_244 2385.8 156.3 hsa-miR-378 2350.5 611.8 HS_275 2310.3 174.3 hsa-miR-500 2307.8 1487.3 hsa-miR-499-5p 2305.3 367.1 hsa-miR-148b 2279.5 2879.8 hsa-miR-664 2117.5 660.3 hsa-miR-301a 2088.8 190.6 hsa-miR-214 2069.5 1221.0 hsa-miR-181a-2 2052.5 1471.7 hsa-miR-335 1995.0 1939.3 HS_38.1 1993.3 175.6 hsa-miR-29a 1978.5 1742.2 hsa-miR-362-3p 1967.5 421.5 hsa-miR-23a 1926.3 661.7 HS_152 1902.3 74.4 HS_303_a 1883.8 254.7 hsa-miR-548d-5p 1872.8 480.8 hsa-miR-339-5p 1872.3 419.9 hsa-miR-502-3p,hsa-miR-500 1848.5 321.7 hsa-miR-560:9.1 1840.8 249.8 hsa-miR-189:9.1 1812.8 638.2 hsa-miR-125b-1 1798.3 557.2 hsa-miR-585 1794.8 360.7 hsa-miR-210 1787.5 616.7 hsa-miR-494 1746.5 766.3 hsa-miR-769-5p 1733.3 264.7 hsa-miR-532-5p 1717.5 465.2 HS_94 1704.8 219.0 hsa-miR-493 1696.3 786.1 hsa-miR-517c 1695.5 633.4 hsa-miR-565:9.1 1675.3 493.9 hsa-miR-454 1663.3 527.2 hsa-miR-323-3p 1660.8 616.3 solexa-9655-85 1649.5 120.6 hsa-miR-92b 1644.3 727.9 hsa-miR-1287 1643.0 272.5 hsa-miR-212 1632.3 810.2 HS_199 1621.5 211.5 hsa-miR-1307 1608.3 381.4 hsa-miR-154 1606.0 957.6 hsa-miR-1301 1584.5 285.1 hsa-miR-1284 1539.0 114.2 hsa-miR-485-5p 1535.8 381.7 hsa-miR-194 1534.8 88.7 hsa-miR-624 1526.8 622.9 hsa-miR-519b-3p 1509.8 95.1 HS_113 1492.0 41.2 HS_71.1 1466.5 89.0 hsa-miR-216b 1444.8 389.1 hsa-miR-1285 1442.0 443.3 HS_150 1428.5 143.1 hsa-miR-27b 1396.0 892.5 HS_268 1390.5 100.5 hsa-miR-133a 1384.5 203.6 HS_156 1361.5 984.0 HS_139 1354.8 105.8 HS_19 1354.0 97.7 hsa-miR-551a 1344.8 195.1 hsa-miR-505 1342.5 835.9 HS_184 1299.3 63.8 hsa-miR-1271 1277.0 806.7 hsa-miR-491-5p 1243.0 206.9 HS_114 1242.8 196.3 hsa-miR-381 1231.0 278.1 hsa-miR-627 1210.3 356.3 hsa-miR-181a 1200.3 46.4 hsa-miR-660 1192.0 1533.2 hsa-miR-1224-3p 1182.8 115.5 hsa-miR-449b 1178.3 39.1 hsa-miR-140-5p 1161.0 840.4 hsa-miR-1228 1158.3 214.7 hsa-miR-501-3p 1156.0 167.3 hsa-miR-548a-3p 1153.3 66.9 hsa-miR-1299 1153.0 185.4 hsa-miR-296-3p 1151.8 410.7 HS_32 1139.5 123.6 hsa-miR-1254 1130.5 125.3 hsa-miR-200c 1126.8 208.4 HS_81 1122.8 90.8 hsa-miR-671:9.1 1109.5 455.8 hsa-miR-34c-3p 1103.8 866.9 hsa-miR-1234 1097.5 302.5 hsa-miR-34b 1097.3 369.2 hsa-miR-125a-3p 1073.3 88.8 hsa-miR-1296 1071.8 354.0 hsa-let-7e 1068.8 429.2 hsa-miR-571 1065.8 110.0 hsa-miR-149 1061.5 533.1 HS_166.1 1056.8 305.9 hsa-miR-19b 1041.0 80.3 hsa-miR-298 1038.0 120.1 hsa-miR-1233 1037.5 183.0 hsa-let-7f-1 1036.5 333.0 hsa-miR-362-5p 1034.8 270.1 hsa-miR-99b 1034.5 449.7 hsa-miR-105 1021.0 259.3 hsa-miR-23b 999.8 338.4 hsa-miR-409-5p 997.5 728.5 hsa-miR-629 988.5 336.3 hsa-miR-369-5p 986.8 304.0 hsa-miR-602 983.8 75.8 hsa-miR-875-5p 982.3 130.0 hsa-miR-495 974.8 217.1 hsa-miR-584 969.3 310.9 hsa-miR-18a 957.5 786.4 hsa-miR-501-5p 957.3 246.6 hsa-miR-187 955.5 126.8 hsa-miR-1257 948.5 41.4 hsa-miR-25 947.5 141.3 hsa-miR-452:9.1 946.0 343.0 hsa-miR-1267 940.0 146.5 hsa-miR-505 926.8 575.6 hsa-miR-642 919.0 295.5 HS_10 918.5 780.9 hsa-miR-18a 916.8 353.2 hsa-miR-615-5p 912.5 284.8 hsa-miR-517 908.5 116.3 hsa-miR-26b 904.0 564.8 hsa-miR-890 896.5 71.4 hsa-miR-452 892.3 397.7 hsa-miR-431 889.5 179.5 hsa-miR-186 874.8 415.7 HS_303_b 863.8 241.4 hsa-miR-671-3p 862.3 430.5 hsa-miR-192 861.8 991.3 hsa-miR-609 859.8 29.5 hsa-miR-194 859.3 1047.4 HS_97 846.8 133.1 hsa-miR-369-3p 845.0 191.3 hsa-miR-942 844.5 187.6 HS_116 842.3 321.3 hsa-miR-34c-5p 831.8 186.6 hsa-miR-603 824.5 574.6 hsa-miR-27a 816.8 557.3 hsa-miR-766 809.5 249.5 HS_85.1 806.3 108.5 HS_260 797.8 173.7 hsa-miR-29c 797.8 204.0 hsa-miR-223 796.0 418.9 hsa-miR-106b 790.0 334.5 hsa-miR-550 783.3 208.7 hsa-miR-30d 781.8 93.6 hsa-miR-195 781.3 98.8 hsa-miR-654-3p 771.8 204.8 hsa-miR-665 762.8 156.5 HS_208 757.5 72.5 hsa-miR-138-1 757.3 398.7 hsa-miR-302d 750.3 319.6 solexa-15-44487 749.5 58.3 hsa-miR-554 738.8 82.8 solexa-9081-91 734.5 42.2 hsa-miR-1204 720.5 64.3 hsa-miR-193a-3p 718.0 357.6 hsa-miR-549 713.8 393.5 hsa-miR-507 711.5 58.1 hsa-miR-497 710.3 339.4 HS_110 706.0 133.1 hsa-miR-361-3p 703.5 321.3 hsa-miR-663b 700.3 68.3 HS_250 697.8 120.2 hsa-miR-483-3p 694.8 312.6 hsa-miR-486-5p 685.0 276.9 hsa-miR-1305 684.8 58.5 hsa-miR-425 682.5 305.4 hsa-miR-545 677.8 116.9 hsa-miR-491-3p 675.8 62.8 HS_52 670.8 54.0 hsa-miR-573 658.5 20.5 hsa-miR-34a 654.8 305.1 hsa-miR-1286 651.3 43.5 HS_209.1 644.3 291.1 hsa-miR-598 643.8 240.9 hsa-miR-371-5p 643.3 68.0 hsa-miR-641 635.0 14.9 hsa-miR-525-3p 634.5 123.4 hsa-miR-655 633.5 296.8 hsa-miR-801:9.1 631.8 58.4 HS_78 626.8 69.8 hsa-miR-33a 625.3 43.6 hsa-miR-16-1 620.8 296.5 hsa-miR-652 619.0 213.5 hsa-miR-411 616.8 79.5 hsa-miR-1245 614.3 280.4 hsa-miR-302b 612.5 278.1 hsa-miR-518c 611.3 38.9 hsa-miR-1263 611.0 39.9 HS_262.1 606.0 33.8 hsa-miR-516a-3p,hsa-miR-516b 600.5 85.7 hsa-miR-513a-5p 598.5 29.9 hsa-miR-502-5p 594.5 192.1 hsa-miR-302b 590.5 107.0 hsa-miR-518d-3p 578.3 59.0 solexa-7534-111 577.5 222.6 hsa-miR-583 576.8 100.9 HS_24 573.5 315.9 hsa-miR-376a 573,5 183.2 hsa-miR-136 572.5 530.1 HS_20 571.3 46.5 hsa-miR-346 570.5 183.7 hsa-miR-576-3p 568.8 221.2 solexa-7509-112 564.0 40.9 HS_279_a 555.8 50.5 hsa-miR-659 548.5 83.4 hsa-miR-188-5p 546.5 120.0 solexa-9578-86 545.5 223.2 hsa-miR-376a:9.1 540.3 236.5 HS_186 536.8 60.1 hsa-miR-379 535.0 218.6 hsa-miR-1225-3p 534.8 118.3 HS_305_b 526.5 115.1 hsa-miR-26a-2 525.3 337.8 solexa-1460-671 519.8 182.3 hsa-miR-525-5p 516.3 24.6 solexa-9124-90 509.0 124.0 hsa-miR-376a 504.8 315.1 hsa-miR-1180 501.8 55.2 hsa-miR-668 500.5 89.6 hsa-miR-101 500.3 103.7 hsa-miR-330-3p 500.0 130.0 HS_160 495.0 74.7 hsa-miR-542-5p 490.5 43.0 hsa-miR-518a-5p,hsa-miR-527 487.5 7.3 solexa-826-1288 487.3 120.0 hsa-miR-99a 487.0 38.7 HS_17 486.3 259.7 hsa-miR-1184 485.5 153.5 hsa-miR-520d-5p 479.0 34.3 hsa-miR-550 478.5 101.5 HS_287 473.5 119.3 hsa-let-7c 468.5 189.5 hsa-miR-374a 465.5 45.5 hsa-miR-651 462.0 49.0 solexa-8926-93 461.3 293.0 hsa-miR-200b 456.8 100.3 hsa-miR-17 449.5 78.6 hsa-miR-19b-2 448.0 17.3 HS_252.1 444.8 12.3 hsa-miR-218-1 444.8 38.8 hsa-miR-616 437.5 247.5 hsa-miR-1226 434.5 133.6 hsa-miR-1225-5p 429.8 321.2 hsa-miR-21 428.3 357.3 hsa-miR-376b 423.0 55.8 hsa-miR-122 421.5 33.6 hsa-miR-30c-2 421.0 134.6 hsa-miR-647 418.8 69.6 hsa-miR-132 418.5 288.2 hsa-miR-367 416.3 17.6 HS_276.1 412.8 50.3 solexa-3022-299 409.0 26.0 hsa-miR-941 408.5 154.1 hsa-miR-299-3p 407.8 54.0 HS_60 406.5 89.3 hsa-miR-591 406.5 28.0 hsa-miR-1197 404.5 91.6 hsa-miR-340 403.8 242.4 HS_142.1 402.0 82.7 hsa-miR-139-5p 399.3 257.8 solexa-3126-285 398.8 184.2 HS_145.1 395.5 30.6 hsa-miR-1294 395.0 95.9 hsa-miR-30b 395.0 129.0 HS_105 394.3 44.1 hsa-miR-1268 393.0 68.3 hsa-miR-302c 392.8 15.3 HS_203 384.8 25.9 hsa-miR-629 384.5 213.8 hsa-miR-518e:9.1 384.3 29.2 HS_126 383.5 34.6 hsa-miR-410 382.5 217.5 HS_149 380.8 167.1 hsa-miR-545:9.1 378.8 140.1 HS_48.1 377.8 65.3 hsa-miR-1207-5p 376.3 32.7 hsa-miR-516a-5p 372.8 60.1 hsa-miR-377 369.0 77.0 HS_196.1 368.5 117.1 hsa-miR-182 368.0 92.5 hsa-miR-452 367.8 29.0 HS_27 363.5 15.6 hsa-miR-564 360.8 60.5 HS_219 351.8 63.6 HS_285 350.3 87.8 hsa-miR-1322 348.8 81.7 hsa-miR-1206 348.0 25.7 hsa-miR-1243 347.0 72.4 hsa-miR-614 345.3 39.1 HS_25 34.8 70.5 hsa-miR-92a-1 340.8 119.4 hsa-miR-296-5p 337.8 54.2 hsa-miR-1248 332.5 86.2 hsa-miR-144 331.5 23.9 HS_6 328.3 74.8 hsa-miR-128a:9.1 326.3 122.1 solexa-5874-144 326.3 35.0 hsa-miR-567 324.0 19.4 hsa-miR-202:9.1 323.3 131.5 hsa-miR-220b 323.0 101.0 HS_90 319.3 26.8 hsa-miR-653 316.3 14.3 hsa-let-7a 313.0 162.2 hsa-miR-107 310.8 109.9 HS_99.1 309.3 51.0 HS_201 307.5 105.4 hsa-miR-767-5p 306.5 38.6 hsa-miR-34b 305.0 43.5 hsa-miR-222 304.8 34.2 hsa-miR-935 303.8 45.1 HS_2 303.5 65.8 HS_122.1 299.5 16.6 hsa-miR-454 297.3 86.9 hsa-miR-181c 296.3 53.6 hsa-miR-380 294.0 38.9 hsa-miR-326 293.5 84.4 hsa-miR-339-3p 293.0 72.7 hsa-miR-338-5p 291.0 28.2 hsa-miR-100 285.8 63.8 hsa-miR-1231 283.5 33.4 HS_138 279.5 21.4 hsa-miR-887 278.3 36.9 hsa-miR-1273 276.0 127.5 hsa-miR-129-3p 275.5 107.2 hsa-miR-1237 274.0 116.0 HS_119 271.0 25.6 hsa-miR-646 268.5 109.8 HS_40 261.8 33.6 hsa-miR-1304 256.0 79.5 hsa-miR-384 255.5 17.2 hsa-miR-371-3p 251.8 143.6 hsa-miR-377 250.5 44.9 hsa-miR-760 250.5 47.3 hsa-miR-455-5p 249.0 129.0 hsa-miR-147b 248.3 21.7 hsa-miR-1323 248.0 22.9 hsa-miR-219-2-3p 246.8 24.5 hsa-miR-596 246.3 42.0 hsa-miR-581 242.0 28.9 hsa-miR-519c-3p 237.3 30.7 hsa-miR-9 234.3 46.0 HS_153 232.8 11.0 hsa-miR-1272 232.5 24.2 hsa-miR-769-3p 230.3 140.8 hsa-miR-643 227.3 25.3 hsa-miR-520a-3p 222.3 30.6 HS_23 221.8 25.9 hsa-miR-323-5p 221.8 55.4 hsa-miR-559 221.0 13.5 hsa-miR-345:9.1 220.8 110.5 hsa-miR-1297 220.5 15.5 hsa-miR-412 218.3 13.7 hsa-miR-624 218.3 18.0 hsa-miR-548c-5p 210.3 36.5 hsa-miR-563 210.3 34.9 hsa-miR-876-3p 208.0 29.9 hsa-miR-135b 207.8 42.0 HS_200 201.3 28.0 hsa-miR-1208 198.0 33.8 hsa-miR-383 196.8 80,2 hsa-miR-453 196.0 23.8 HS_141 195.0 25.1 hsa-miR-331-5p 193.8 11.0 HS_33 193.3 18.0 hsa-miR-548f 189.3 25.2 hsa-miR-206 188.8 17.0 hsa-miR-190 180.8 5.6 hsa-miR-548m 177.5 21.3 hsa-miR-592 171.8 14.8 hsa-miR-1262 159.0 13.3	{ "pile_set_name": "PubMed Central" }
Responsible Editor Jennifer Stewart Williams, Umeå University, Sweden Background {#S0001} ========== Maternal mortality has been reduced but remains high in sub-Saharan Africa (SSA) \[[1](#CIT0001)\]. Even though sub-Saharan African countries aim for zero avoidable maternal deaths, many struggle to achieve substantial progress \[[1](#CIT0001)--[5](#CIT0005)\]. To further improve quality of maternity care, the World Health Organization (WHO) has promoted the Maternal Near-Miss (MNM) approach since 2004 \[[6](#CIT0006)\]. An MNM is 'a woman who nearly died but survived a pregnancy-related complication during pregnancy, birth or within 42 days after termination or delivery' \[[6](#CIT0006)\]. Women with an MNM event often share the same characteristics and risk factors as women who died, such as underlying medical or pregnancy-related conditions and delays in reaching and obtaining adequate health care. Much can be learned with regard to the functioning of the health system, and failing of the system, by analysing MNM and maternal deaths. However, MNM occurs in larger numbers and may be less threatening for health-care workers to discuss since these can be regarded as great saves \[[7](#CIT0007),[8](#CIT0008)\]. Namibia has a high maternal mortality ratio of an estimated 385/100,000 livebirths in 2013 \[[9](#CIT0009)\]. Even though it has a very large surface area, it is one of the least densely populated countries in the world with 2.8 people per square kilometer \[[10](#CIT0010)\]. Annually, there are around 75,000 births \[[10](#CIT0010)\]. To analyse and improve maternity care, Namibia is planning to apply the MNM approach. Adoption of a recognized set of MNM criteria will allow comparison across settings and is therefore preferred. Reports from other SSA countries indicated that the WHO MNM criteria (Appendix 1) may not be suitable for use in district hospitals in low-income settings, due to limited availability of laboratory tests, blood products and management options leading to under-registration of MNM \[[11](#CIT0011),[12](#CIT0012)\]. Recently, a Delphi study was published proposing MNM criteria deemed suitable for hospitals in low-income settings in SSA \[[13](#CIT0013)\]. The SSA criteria added several clinical criteria such as eclampsia, sepsis and uterine rupture and lowered the threshold for blood transfusion from five to two units of red cells (Appendix 1). For Namibia, classified by the World Bank as a sub-Saharan upper middle-income country, it is unclear which set of MNM criteria would be appropriate to use \[[14](#CIT0014)\]. The national referral hospital in Windhoek, the capital of Namibia, has all the facilities to apply the WHO MNM criteria, but in district hospitals, several laboratory tests to identify organ failure or management options are not available. Nevertheless, Namibian district hospitals are generally better supplied compared to most district hospitals in low-income settings in SSA. The aim of this study was to establish MNM criteria suitable for use in all Namibian hospitals by applying both the WHO and the adapted SSA MNM criteria to a cohort of women in four Namibian hospitals. Methods {#S0002} ======= Study setting and design {#S0002-S2001} ------------------------ Participating facilities were the national referral hospital, Windhoek Hospital Complex (Hospital A), a large regional hospital, Rundu Intermediate Hospital (Hospital B), and two smaller district hospitals, Gobabis and Okahandja State Hospital (Hospital C and D). The hospitals were selected based on their characteristics, to obtain a representative sample of Namibian hospitals. Namibia has only three regional hospitals of which one is part of the hospital complex in the capital (Hospital A). Hospital B and the third regional hospital (not included) are similar in terms of number of births, available resources and catchment area. Of the 31 district hospitals, approximately half have more than 2000 births annually. Staff and resources are allocated accordingly. We selected two district hospitals representing a higher and lower-burden facility. Since Namibia is a large but sparsely populated country with long distances between health-care facilities, Hospital C was chosen since this facility is located in one of the least densely populated districts. Hospital A is located in the capital, has almost 14,000 births annually, and is the only tertiary facility in Namibia for the entire population of 2.3 million \[[10](#CIT0010)\]. There are three consultant obstetrician and gynaecologists. The Intensive Care Unit (ICU) has advanced equipment including ventilators and a dialysis machine. Sophisticated laboratory tests including pH and lactate measurements are available to identify and manage organ failure. Hospital B is located near the Angolan border, has just over 6,000 births annually and serves a population of approximately 350,000 \[[10](#CIT0010)\]. There is one consultant gynaecologist, an ICU with mechanical ventilation and a well-supplied blood bank. The district hospitals have around 2500 and 1200 births, respectively, and there are no ICU facilities or consultants. Hospital C has basic haematology and chemistry laboratory tests available on site. Hospital D has no laboratory tests on site. Both district hospitals have a reliable supply of red cells for blood transfusion but no other blood products. For this prospective cross-sectional study we included all pregnant women or within 42 days after termination of pregnancy or birth, regardless of gestational age, fulfilling at least one WHO or SSA MNM criterion. Furthermore, they had to present between 1 March 2018 and 31st of May 2018 to the participating facilities \[[6](#CIT0006),[13](#CIT0013)\]. Maternal deaths, defined as death of a woman while pregnant or within 42 days of termination of pregnancy or birth, from any cause related to or aggravated by the pregnancy or its management, but not from accidental or incidental causes, were excluded \[[6](#CIT0006)\]. MNM identification criteria and sample size {#S0002-S2002} ------------------------------------------- The WHO and SSA criteria are presented in Appendix 1. The SSA criterion 'Laparotomy other than caesarean section' was modified to 'Laparotomy other than caesarean section (CS) or ectopic pregnancy', since we anticipated an over-reporting of ectopic pregnancies that were not necessarily an MNM. The threshold of transfusion of two units of blood was adopted from the SSA MNM criteria. The WHO MNM criterion defining MNM from haemorrhage has a threshold of five units of blood, so inclusion from two units of blood and above would allow further analysis of the women who received two, three or four units of blood to identify which threshold would most accurately define severe haemorrhage without over-reporting mild or moderate haemorrhage. In the proposed SSA criterion 'severe pre-eclampsia with ICU admission', ICU was not clearly defined. For our study, we decided to regard the ICU as a ward where mechanical ventilation and administration of continuous vasoactive drugs were possible, together with the continuous presence of a medical doctor. At the time of design of the study, the performance of the SSA criteria was not yet assessed in clinical setting, rendering sample size calculations difficult. A large global study performed by the WHO found a prevalence of approximately 15 MNM per 1000 live births in countries with a maternal mortality ratio similar to that of Namibia \[[6](#CIT0006),[15](#CIT0015)\]. Considering the lower threshold for transfusion and the addition of clinical criteria, we expected the SSA criteria to identify approximately twice as many MNM. To identify at least five MNM in the district hospitals, we set our study period at 3 months, with the possibility to extend it with another 3 months in case of insufficient inclusions in hospital C and D. Data collection and analysis {#S0002-S2003} ---------------------------- Doctors and nurses working in participating facilities, involved in the care of pregnant and/or postpartum women were trained in MNM identification and data collection prior to the study by members of the research team. A manual with information on data collection was present in each participating facility. MNM identification and data collection were supported in hospital A through daily visits to the maternity wards, gynaecology wards, acute care and ICU by AB, AD, AD, SH. On the respective wards, the admission books were screened for possible missed MNM. The register of the blood bank was screened for missed MNM by screening blood units delivered to the above-mentioned wards. Daily rounds and screening of blood bank registers were not feasible in the other hospitals due to budget and logistical restrictions. In hospitals B, C and D a doctor and two nurses were in charge of the project. These staff members were supervising MNM identification and data collection and were in close contact with the research team for support. For the identified MNM, data were collected anonymously from medical records using a structured data collection tool, including socio-demographic characteristics, contributing factors, maternal and neonatal outcome, the main cause of MNM and long-term complications. All collected data were verified with medical records by either SH, LK, CL, FM, SM, who are medical doctors with at least 3 years of experience providing obstetric care in Namibia. We collected the total number of live births and maternal deaths that occurred within the facility during the study period. Data were double entered and crosschecked in Epidata version 4.2. Data analysis consisted of frequencies of demographic and clinical variables. Data analysis was performed with SPSS version 22. All results are reported as numbers (n) and frequencies. Confidence intervals (CI) were calculated using the Clopper-Pearson exact test. Severe maternal outcome (SMO) was defined as the number of MNM and maternal deaths combined. The MNM ratio was defined as the number of maternal near-miss per 1000 live births and the SMO ratio by the number of women with SMO per 1000 live births \[[6](#CIT0006)\]. The confidence interval was calculated using Poisson distribution. In order to establish MNM criteria which will identify all MNM in Namibian hospitals without over-reporting minor morbidities we did the following: for the women identified as an MNM by one of the unique criteria, medical records were reviewed by SH, LK, CL, FM, SM. Based on their clinical judgement it was decided for each unique criterion if it should be used for MNM identification in Namibia. This study was reviewed and approved by the research unit of the Namibian Ministry of Health and Social Services. We followed the STROBE reporting guidelines. Results {#S0003} ======= In the three-month study period, 194 women were identified to be MNM according to the WHO and/or SSA criteria. Of these 194 MNM, 31 were identified through the blood bank register and ward registers of hospital A. Nine women were included by staff, but upon verification with medical records, it appeared these did not fulfil any MNM criterion and were excluded. Four of these nine women were included for 'severe pre-eclampsia with ICU admission', however these women were not admitted to an ICU department. Two women received one unit of blood for transfusion, whereas the threshold of the MNM criterion was at least two units of blood. Two women were included for sepsis but did not reach any of the clinical thresholds in the definition of the SSA criterion, Appendix 1. One woman was included for uterine rupture. However, only a pending uterine rupture was found with the lower uterine segment thin but intact. Characteristics of the identified MNM are presented in [Table 1](#T0001). There were 32 (16.5%) teenage pregnancies and 60 (30.9%) women were primiparous. Chronic anaemia was present in 32 (16.5%) women. Obstructed labour was present in 18 (9.3%) women and 71 (36.6%) women gave birth by caesarean section. Miscarriage was present in 31 women, of whom 28 needed surgical evacuation, one women received medical treatment and two women had a spontaneous complete abortion.10.1080/16549716.2019.1646036-T0001Table 1.Characteristics of all maternal near-miss women.Characteristics(n = 194)%Age \<203216.5%21--3413067.0%≥353216.5%Parity Para 06030.9%Para 1--311458.8%≥4189.3%Unknown21.0%Gestational age^a^ ≤12 weeks3216.5%13--26 weeks2512.9%27--36 weeks4623.7%≥37 weeks8041.2%Unknown115.7%Previous CS No15579.9%Yes3216.5%Unknown73.6%HIV status Positive189.3%Negative14474.2%Unknown3216.5%Chronic anemia No11860.8%Yes3216.5%Unknown4422.7%Other comorbidities Astma10.5%GDM21.0%Chronic hypertension31.5%TB10.5%Nephrotic syndrome10.5%Unknown63.1%Pregnancy outcome NVD5729.4%Vacuum31.5%CS7136.6%Laparotomy uterine rupture21.0%Miscarriage3116.0%Ectopic157.7%Undelivered147.2%TOP10.5%[^1] [Table 2](#T0002) shows the number of live births, MNM, MD, MNM ratio and SMO ratio for each hospital. Hospital A had the most live births (3515) in the three-month study period, most MNM (150) and the highest MNM ratio (43/1000 live births). The six maternal deaths in hospital A were due to tuberculosis (2), hepatitis E (2), ruptured uterus (1) and septic miscarriage (1). The three maternal deaths in hospital B were due to complicated malaria (2) and a woman who died upon arrival due to an unclear cause. The SMO ratio for all hospitals was 35/1000 live births, 95% CI 30--40.10.1080/16549716.2019.1646036-T0002Table 2.For each hospital number of live births, maternal near-miss, maternal deaths, MNM ratio and the SMO ratio.FacilitiesLivebirthsMNMMDMNM/\ 1000 LB95% CISMO/\ 1000 LB95% CIHospital A351515064336--504437--51Hospital B13263032315--312516--33Hospital C66412^a^0188--28188--28Hospital D2676^a^0224--40224--40Total577219493429--383530--40[^2] Underlying causes of MNM and MNM criteria {#S0003-S2001} ----------------------------------------- The underlying cause of the MNM is presented in [Figure 1](#F0001). The commonest causes were obstetric haemorrhage (75, 38.7%, 95% CI 31.8--45.9), abortion-related complications (44, 22.7%, 95% CI 17.0--29.2) and hypertensive disorders (36, 18.6%, 95% CI 13.3--24.8).10.1080/16549716.2019.1646036-F0001Figure 1.Primary cause of maternal near miss. N = 194. ^a^Two women with eclampsia also had HELLP syndrome and are counted twice. [Table 3](#T0003) compares the two sets of MNM criteria and the frequency of each criterion present among the 194 MNM. The frequencies of each criterion separated for the four hospitals are presented in Appendix 2. Of the unique SSA criteria, eclampsia and sepsis were most frequently present with 26 women fulfilling each of these criteria. Eighteen women with eclampsia and 16 women with sepsis did not fulfil any WHO criterion and would have been missed by this set of criteria. Sepsis was caused by endometritis (6), septic miscarriage (3), pyelonephritis (2), wound sepsis (2), abdominal sepsis (2) and pneumonia (1). For these women, sepsis resolved within 24 h after starting antibiotics, wound treatment or evacuation of retained products of conception and therefore were assessed as not severe morbidity. There were two women with uterine rupture of whom one did not fulfil any other WHO criterion. All women with eclampsia and uterine rupture were assessed as severe morbidity. Four women with severe malaria had also either renal, hepatic or neurological dysfunction and this criterion did not contribute to the identification of MNM. The five women who fulfilled the criterion of 'laparotomy other than caesarean section or ectopic pregnancy' but no other (WHO) criteria were due to septic cervical ectopic pregnancy not suitable for vaginal approach ([Text box 1](#UT0001) case 3), laparotomy after caesarean section to perform B-lynch suture, perforated appendix, septic abdominal pregnancy and transabdominal approach to remove missed miscarriage in second trimester after failed medical induction. All five were assessed as having sustained severe morbidity. Of the unique WHO criteria, 'bilirubin \> 100 mmol/L' was most frequently present, in 13 women. At the time of study, a hepatitis E outbreak was ongoing and became the underlying disease in ten MNM and two maternal deaths.10.1080/16549716.2019.1646036-T0003Table 3.Frequencies of maternal near-miss by type of organ system dysfunction.MNM identification criteriaTotal per MNM criterion\ N, % (95% CI)MNM by SSA onlyMNM by WHO onlySevere maternal complications72, 37.3% (30.3--44.3) Eclampsia2618 Sepsis or severe systemic infection2616 Uterine rupture21 Pulmonary oedema72 Severe abortion complications83 Severe malaria40 Severe pre-eclampsia with ICU admission41 Laparotomy other than caesarean section or ectopic pregnancy125 Cardiovascular dysfunction21, 10.8% (6.8--16.1) Shock20 Cardiac arrest0 [Use of continuous vasoactive drugs]{.ul}3 0CPR0 [pH \< 7.1]{.ul}1 1[Lactate \>5 mEq/mL]{.ul}0 0Respiratory function16, 8.2% (4.8--13.0) Acute cyanosis0 Gasping0 Respiratory rate \>40 or \<6/min5 Intubation and ventilation for ≥60 minutes not related to anaesthesia11 Oxygen saturation \<90% for ≥60 minutes4 Renal dysfunction4, 2.1% (0.6--5.2) Oliguria non-responsive to fluids or diuretics4 Creatinine ≥300 µmol/l or ≥3.5 dL/dL1 [Dialysis for acute renal failure]{.ul}1 0Coagulation/hematologic dysfunction147, 75.8% (69.1--81.6) Failure to form clots6 Transfusion of ≥ 2 units of red blood cells140103 Acute thrombocytopenia (\<50,000 platelets/ml)12 Hepatic dysfunction13, 6.7% (3.6--11.2) Jaundice in the presence of pre-eclampsia0 [Bilirubin \>100 mmol/l or \>6.0dL/dL]{.ul}13 9Neurological dysfunction8, 4.1% (1.8--8.0) Loss of consciousness lasting 12 hours (GCS \< 10)6 Stroke0 Loss of consciousness and ketoacids in urine0 Uncontrollable fit1 New onset of paralysis1 Uterine dysfunction6, 3.1% (1.1--6.6) Hysterectomy following infection or haemorrhage6 Total19413310[^3] [Figure 2](#F0002) summarizes the number of units of blood given to MNM. SSA MNM criterion of at least two units of blood was fulfilled by 140 women. An estimation of the amount of blood loss was not available in medical records. Assessment of severity of haemorrhage was based on the vitals of the patients and interventions performed to stop the bleeding. Two or three units of red cells were given to 95 women for mild/moderate haemorrhage or chronic anaemia who did not fulfil any WHO criterion and were not considered to be MNM. Fourteen women received four units and were assessed to be MNM from severe haemorrhage. Even though these women did not reach the clinical thresholds of shock, they often had tachycardia and needed several interventions to stop the bleeding and to restore these clinical parameters ([Text box 1](#UT0001) case 3 and 4). Of the women identified through the register of blood bank, 20 women received two units of blood and eight women three.10.1080/16549716.2019.1646036-F0002Figure 2.Summary of number of women who received two or more units of blood transfusion. In white the women that did not fulfil another criterion of the WHO set and would have been missed by the WHO MNM criteria. In grey the number of women that did fulfil another criteria of the WHO MNM criteria. RCC, red cell concentrate. If only the WHO criteria would have been applied to our cohort 61 MNM would have been identified and if the SSA criteria were applied 184 MNM. In Textbox 1, several summaries are presented of MNM that would have been missed if the WHO MNM criteria were applied. Long-term outcome {#S0003-S2002} ----------------- Although 194 women survived the pregnancy-related complication, 15 (7.7%) women sustained a long-term complication. Six (3.1%) women underwent hysterectomy of whom five were haemorrhage and one infection related. Three (1.5%) women developed cardiomyopathy and three needed a colostomy: one woman had iatrogenic bowel perforation during caesarean section, one had an abdominal pregnancy with placental invasion into the sigmoid and one woman with a previous caesarean section had an obstructed bowel due to adhesions during her current pregnancy. Two (1.0%) women had a ruptured uterus due to obstructed labour of whom one developed a vesicovaginal fistula. One women with known chronic renal failure developed acute on chronic renal failure secondary to pre-eclampsia and was still in need of dialysis after discharge. The two women with a uterine rupture, two women who needed a colostomy and one woman with a haemorrhage-related hysterectomy had had one or more previous caesarean sections. Discussion {#S0004} ========== In the middle-income setting of Namibia, both the WHO and SSA MNM criteria cannot be applied without amendments. Even though the WHO developed the criteria to be applied across settings, our findings suggest it would result in an underestimation of MNM in Namibia, which corresponds with experiences from low-income settings \[[11](#CIT0011),[12](#CIT0012),[16](#CIT0016)\]. The SSA criteria were proposed for use in low-income settings. Application of the SSA criteria to our cohort resulted in an overestimation of MNM which suggests that both sets are less feasible to be applied to a middle-income setting such as Namibia. There are several other middle-income countries which used the MNM approach. However, they applied the WHO criteria or locally established criteria without validating them, and most of these studies only included referral or tertiary hospitals \[[16](#CIT0016)--[22](#CIT0022)\]. For example, a large national study from Brazil included 27 referral hospitals and used the WHO criteria \[[18](#CIT0018)\]. India developed its own MNM criteria and performed a pilot in six tertiary facilities \[[22](#CIT0022)\]. A recently published systematic review assessed the applicability of WHO criteria in sub-Saharan Africa and included 15 reports of which four reported findings from middle-income countries \[[16](#CIT0016),[17](#CIT0017),[23](#CIT0023)--[25](#CIT0025)\]. These four reports were all from tertiary facilities with availability of resources to apply the WHO criteria \[[17](#CIT0017),[23](#CIT0023)--[25](#CIT0025)\]. For MNM registration in all Namibian hospitals of all levels, we decided to adopt the WHO criteria with the addition of the criteria eclampsia, uterine rupture, laparotomy other than caesarean section or ectopic pregnancy' and a lower threshold of four units of blood. Even though eclampsia and uterine rupture are not part of the WHO criteria, high case fatality rates are reported for both conditions across settings \[[26](#CIT0026)--[28](#CIT0028)\]. Furthermore, these are commonly used in both high- and low-income countries to define MNM for local and national registrations and recently suggested as a standard outcome to define severe morbidity in maternal health-related research \[[11](#CIT0011),[29](#CIT0029)--[32](#CIT0032)\]. To our knowledge, the criterion 'Laparotomy other than caesarean section or ectopic pregnancy' is not commonly used to define MNM. However, in our cohort, several women would have been missed and this criterion, therefore, turned out to be a useful addition. We opted to change the threshold for blood transfusion to four units of red cells. The threshold of two or three units resulted in inclusion of many women with minor to moderate haemorrhage-related complications, as well as inclusion of women with chronic anaemia. At a threshold of five units of blood, women with severe haemorrhage would have been missed. The criterion sepsis led to overinclusion in our cohort of women who were not critically ill. The SSA set used a definition of sepsis developed for the non-obstetric population \[[33](#CIT0033)\]. Amendment of the definition of sepsis, with higher thresholds of clinical signs, might lead to better MNM identification. The remaining SSA criteria (pulmonary oedema, severe abortion complications, severe malaria and severe pre-eclampsia with ICU admission), appeared not to be essential to identify MNM, resulted in inclusion of women with minor morbidity or were unclear leading to incorrect inclusion. The prevalence of SMO is expected to correlate with the MMR of a country \[[6](#CIT0006)\]. The WHO multicountry survey showed an SMO ratio of 15.9/1000 livebirths in countries with a comparable MMR \[[15](#CIT0015)\]. By applying the SSA criteria we found a more than double SMO ratio, whereas by applying the above-suggested amendments we would have identified 93 MNM, which leads to an SMO ratio of 17.5/1000 live births. We recommend other middle-income countries to validate our criteria in their setting. A country's income level does not necessarily correlate with maternal outcome \[[34](#CIT0034)\]. For example, Namibia has a relatively high prevalence of severe maternal outcome. The income level is based on national income per person and does not take into account income inequities \[[14](#CIT0014)\]. However, a country's income level will, in most cases, affect the availability of resources. This poses the most important restriction in use of the WHO MNM criteria. Access to care does not seem to be a major issue contributing to poor maternal outcome in Namibia. The country has one of the best road networks in Africa and an effective referral system appears to be in place. This is reflected by the high antenatal care attendance rate of 96% and the fact that 87% of women give birth in health facilities, attended by a skilled health-care worker \[[9](#CIT0009)\]. A recent national maternal death review, of which the report is expected to be published next year, indicates that most deaths occur whilst the woman is in the health facility. Namibian health facilities, like most state-funded facilities in sub-Saharan Africa, have high turnover of staff, and are staffed mainly by junior staff with limited experience and supervision. Within the five-step 'obstetric transition' as described by the World Health Organization, Namibia is now in stage III, the typical state of a middle-income country that has largely overcome barriers for women to access care, where women are now indeed reaching health facilities, and hence where improved quality of care becomes the critical step to achieve a reduction of maternal mortality \[[35](#CIT0035)\]. Over 80% of the underlying causes of the MNM were direct pregnancy-related causes. We feel that this might be an overrepresentation, caused by the over-reporting of haemorrhage. A better representation of the underlying causes is expected to be found in the planned national MNM observational study with the amended criteria. Abortion-related complications were the second most common underlying cause of the MNM. In Namibia, women only have limited access to abortion and only for significant maternal or fetal indications. A high incidence of abortion-related MNM events is seen in other countries where women have limited access to safe abortion as well \[[36](#CIT0036)\]. A considerable proportion of our women had a previous caesarean section and over half of our women gave birth by caesarean section which correlates with other reports \[[15](#CIT0015)\]. A possible explanation for this high proportion could be our sick population. However, even in an MNM population, there are reports available indicating a high number of caesarean sections without medical indication and for some women, the MNM event was the direct result rather than the indication for caesarean section \[[37](#CIT0037)\]. In our cohort, a woman sustained iatrogenic bowel injury during birth by caesarean section and needed several laparotomies and went home with a colostomy. Our study has several limitations. We used a relatively small cohort and were unable to perform a sample size calculation. Possibly larger studies, such as the planned national MNM observational study, may permit additional conclusions with statistical significance. Availability of background data was very limited regarding maternal morbidity and mortality in Namibia. Objective data such as case fatality rate would have been useful to further support our decisions to amend the WHO criteria. Data were collected by local staff who already had a high workload and MNM might have been missed. Even though data collection done by full-time research staff might have been more accurate, we aimed for data collection mainly done by local staff as this would enable us to identify barriers and challenges for a planned routine national MNM data collection in all Namibian hospitals. The project was overall well received by staff members as they felt they could finally report 'the other side of the story, rather than only reporting maternal deaths'. The research team identified an additional 31 women, of whom the majority appeared to not be critically ill. This further supports the experience that our staff were motivated to report MNM. We found a higher prevalence of MNM in the capital. As the national referral centre, this was expected. However, this is most likely also caused by reporting bias since this was the only facility where data collection was supported by the research team. We suggest the assessment of long-term complications should become part of all MNM projects. Even though these women nearly died but survived a pregnancy-related complication, a considerable proportion left the hospital with significant disability. Conclusions {#S0005} =========== This report validated MNM criteria to be applied nationwide in hospitals of all levels in a sub-Saharan African country. Our findings suggest that a middle-income country such as Namibia needs MNM criteria 'in between' the WHO and SSA criteria. Context-specific MNM criteria are vital to help Namibia and other middle-income countries to identify MNM. Only with accurate MNM criteria can this tool be used for what it is designed for; to improve maternity care and to stop women dying of avoidable causes. We recommend other middle-income countries to validate our criteria in their setting when the WHO criteria are not feasible, rather than developing new criteria, as this will allow comparison of findings. 10.1080/16549716.2019.1646036-UT0001Text box 1.Summary of cases missed by WHO criteria sorted by assessment of research team as severe or not severe morbidity case. CS, caesarean section; G, gravida; P, parity; Hb, haemoglobin; WBC, white blood cells x 10^9^/L; RCC, red cell concentrate.Severe morbidityEclampsia(1) A 17-year old primigravida fitted eight times due to eclampsia at home and in a small district hospital 119 km away from hospital B. She was referred to hospital B were a CS was done. She recovered well.Uterine rupture(2) A 28-year old G3P2 with one previous CS was admitted in active labour in hospital A. An emergency CS was done and a uterine rupture found. A fresh stillborn baby was delivered. The woman developed a vesicovaginal fistula and received 1 unit of blood during admission.Laparotomy other than CS or ectopic(3) A 33-year old G3P2 presented to Hospital A with a septic cervical pregnancy. A hysterotomy, abdominal approach, was done to remove the pregnancy since a vaginal approach was deemed not to be feasible. To control the bleeding a bilateral ligation of the uterine arteries was done. The woman did not fulfil the criteria of shock. She received four units of blood and two units fresh frozen plasma.Transfusion of 2 or more units of blood(4) A 21-year old primigravida presented to hospital D with severe bleeding due to an incomplete miscarriage. Her Hb was 5.2 g/dL, she received four units of blood and the retained products of conception were removed.No severe morbiditySepsis(1) A 27-year old woman developed fever 1 day after her normal delivery in hospital A. She had a heart rate of 100/min, WBC of 9. She was not acutely ill, endometritis was suspected and she recovered well on intravenous antibiotics.Transfusion of 2 or more units of blood(2) A 28-year old G5P4 presented to hospital A with symptomatic anaemia at 32 weeks of gestation. Her Hb was 6.8 g/dL and she was transfused 2 RCCs.(3 A 17-year old primigravida gave birth in Hospital B with minimal blood loss. During antenatal care an Hb of 5.9 g/dl was found. Three units of blood were transfused postpartum. All authors contributed to the review and drafting of the study protocol. Data collection was done by local staff and supported by AB, AD, AD. Collected data were verified with medical records by SH, LK, CL, FM, SM. Data were summarized by SH and analysed by all authors. All authors contributed to the review and drafting of the paper and approved the final version. Acknowledgments =============== The authors would like to thank the Ministry of Health and Social Services of Namibia and all their staff who contributed to this study. We also would like to thank C.E. Gyllenbern from the University Medical Center Groningen for her help with implementing our research in the four hospitals. Disclosure statement {#S0006} ==================== No potential conflict of interest was reported by the authors. Ethics and consent {#S0007} ================== This study was reviewed and approved by the research unit of the Namibian Ministry of Health and Social Services. After stabilizing the woman, all data were collected from medical records without identification of the woman and study inclusion did not have any effect on clinical management. Therefore, individual informed consent from the woman was not needed. Paper context {#S0008} ============= The Maternal Near Miss (MNM) approach was introduced to improve maternal health. A recognized set of MNM identification criteria will facilitate comparisons across settings. 'WHO MNM criteria', designed to be applied across all settings, underestimate and 'sub-Saharan Africa MNM criteria', an adaptation for use in low-income settings, overestimate the prevalence of MNM in a middle-income country such as Namibia, where MNM criteria 'in between' may be more appropriate. Adaptations for middle-income countries were proposed. Data availability statement {#S0009} =========================== Data supporting the findings are available upon reasonable request. 10.1080/16549716.2019.1646036-UT0002Appendix 1.Comparison of WHO MNM criteria and SSA MNM criteria.WHO near-miss criteriaSub-Saharan Africa near-miss criteriaClinical criteriaClinical criteriaAcute cyanosisAcute cyanosisGaspingGaspingRespiratory rate \>40 or \<6/minRespiratory rate \>40 or \<6/minShock^a^ShockOliguria non-responsive to fluids or diuretics^b^Oliguria non-responsive to fluids or diureticsFailure to form clotsFailure to form clotsLoss of consciousness lasting \>12 hLoss of consciousness lasting 12 hoursCardiac arrestCardiac arrestStrokeStrokeUncontrollable fit / total paralysisUncontrollable fit / total paralysisJaundice in the presence of pre-eclampsiaJaundice in the presence of pre-eclampsiaLaboratory-based criteriaLaboratory-based criteriaOxygen saturation \<90% for ≥60 minutesOxygen saturation \<90% for ≥60 minutesPaO2/FiO2 \<200 mmHg Creatinine ≥300µmol/l or ≥3.5mg/dLCreatinine ≥300µmol/l or ≥3.5mg/dLBilirubin \>100mmol/l or \>6.0 mg/dL pH \< 7.1 Lactate \>5 mEq/mL Acute thrombocytopenia (\<50,000 platelets/ml)Acute thrombocytopenia (\<50,000 platelets/ml)Loss of consciousness and ketoacids in urineLoss of consciousness and ketoacids in urineManagement-based criteriaManagement-based criteriaUse of continuous vasoactive drugs Hysterectomy following infection or haemorrhageHysterectomy following infection or haemorrhageTransfusion of ≥5 units of red cellsTransfusion of ≥ 2 units of red cellsIntubation and ventilation for ≥60 minutes not related to anaesthesiaIntubation and ventilation for ≥60 minutes not related to anaesthesiaDialysis for acute renal failure Cardio-pulmonary resuscitationCardio-pulmonary resuscitation Laparotomy other than caesarean section or ectopic pregnancySevere maternal complicationsSevere maternal complications Eclampsia Sepsis or severe systemic infection^c^ Uterine rupture^d^ Pulmonary oedema Severe abortion complications^e^ Severe malaria^f^ Severe pre-eclampsia with ICU admission[^4] 10.1080/16549716.2019.1646036-UT0003Appendix 2.Frequencies of maternal near-miss by type of organ system dysfunction in the four hospitals Facilities ABCDTotalSevere maternal complications52164072, 37.3%Eclampsia1772026Sepsis or severe systemic infection^b^2222026Uterine rupture^c^11002Pulmonary oedema52007Severe abortion complications^d^71008Severe malaria^e^04004Severe pre-eclampsia with ICU admission13004Laparotomy other than caesarean section or ectopic pregnancy1200012Cardiovascular dysfunction1722021, 10.8%Shock^f^1622020Cardiac arrest00000[Use of continuous vasoactive drugs]{.ul}30003CPR00000[pH \< 7.1]{.ul}10001[Lactate \>5 mEq/mL]{.ul}00000Respiratory function1420016, 8.3%Acute cyanosis00000Gasping00000Respiratory rate \>40 or \<6/min41005Intubation and ventilation for ≥60 minutes not related to anaesthesia1010011Oxygen saturation \<90% for ≥60 minutes40004Renal dysfunction31004, 2.0%Oliguria non-responsive to fluids or diuretics^g^31004Creatinine ≥300µmol/l or ≥3.5mg/dL10001[Dialysis for acute renal failure]{.ul}10001Coagulation/hematologic dysfunction1122465147, 75.8%Failure to form clots42006Transfusion of ≥ 2 units of red blood cells1082165140Acute thrombocytopenia (\<50,000 platelets/ml)840012Hepatic dysfunction1210013, 6.7%Jaundice in the presence of pre-eclampsia00000[Bilirubin \>100mmol/l or \>6.0 mg/dL]{.ul}1210013Neurological dysfunction62008, 4.2%Loss of consciousness lasting 12 hours (GCS \< 10)42006Stroke00000Loss of consciousness and ketoacids in urine00000Uncontrollable fit10001New onset of paralysis10001Uterine dysfunction51006, 3.0%Hysterectomy following infection or haemorrhage51006Total1503095194[^5] [^1]: NVD, normal vaginal delivery; CS, Caesarean section; TOP, termination of pregnancy; GDM, gestational diabetes mellitus. ^a^Number of completed weeks at the end of pregnancy or on admission if undelivered. [^2]: ^a^Three MNM women from hospital C and one MNM woman from Hospital D were referred to hospital A and part of the total number of these hospitals. MNM, Maternal Near Miss; MD, Maternal Death; LB, livebirths; CI, Confidence Interval; SMO, severe maternal outcome. [^3]: In italics unique SSA criteria and [underlined]{.ul} unique WHO criteria. A woman could fulfil more than one MNM criterion. At the top of each section the total number of women with each organ dysfunction, followed by the proportion, calculated by the total divided by 194 MNM women and presented as percentage. The second and third column presents the number of women that only fulfilled criteria of either the WHO or SSA set and would have been missed if the other set was applied. MNM, Maternal near miss; CI, Confidence interval; SSA, sub-Saharan Africa; WHO, World Health organization, ICU, intensive care unit; CPR, cardiopulmonary resuscitation; GCS, Glasgow coma scale. [^4]: MNM, Maternal near-miss; SSA, sub-Saharan Africa; WHO, World Health organisation, ICU, intensive care unit; CPR, cardio pulmonary resuscitation; GCS, Glasgow coma scale. ^a^persistent systolic blood pressure \<80 mmHg or a persistent systolic blood pressure \<90 mmHg with a pulse rate at least 120 bpm, ^b^urinary output \<30 ml/hour over 4 hours or \<400 ml/24 hr, ^c^presence of fever (Temp \>38°C) AND confirmed or suspected infection AND at least one of the following: heart rate \>90, respiratory rate \>20, WBC \> 12 or \< 4 x 10^9^/L, ^d^complete rupture of uterus during labour confirmed by laparotomy, ^e^septic incomplete abortion, complicated gestational trophoblastic disease with anaemia, ^f^major signs of organ dysfunction and/or high level parasitaemia or cerebral malaria. [^5]: Italics unique SSA criteria and [underlined]{.ul} unique WHO criteria. A woman could fulfil more than one MNM criterion. At the top of each section the total number of women with each organ dysfunction, followed by the proportion, calculated by this total divided by 194 MNM women and presented as percentage. MNM, Maternal near miss; SSA, sub-Saharan Africa; WHO, World Health organisation, ICU, intensive care unit; CPR, cardio pulmonary resuscitation; GCS, Glasgow coma scale. ^b^presence of fever (Temp \>38°C) AND confirmed or suspected infection AND at least one of the following: heart rate \>90, respiratory rate \>20, WBC \> 12 or \< 4 x 10^9^/L; ^c^complete rupture of uterus during labour confirmed by laparotomy; ^d^septic incomplete abortion, complicated gestational trophoblastic disease with anaemia; ^e^major signs of organ dysfunction and/or high level parasitaemia or cerebral malaria; ^f^persistent systolic blood pressure \<80 mmHg or a persistent systolic blood pressure \<90 mmHg with a pulse rate at least 120 bpm; ^g^urinary output \<30 ml/hour over 4 hours or \<400 ml/24 hr.	{ "pile_set_name": "PubMed Central" }
All relevant data are within the manuscript. Introduction {#sec001} ============ The formation of memory has been linked to long-term potentiation (LTP) and long-term depression (LTD), or the lasting increase or decrease of the strength of synaptic connections \[[@pone.0225169.ref001]\], \[[@pone.0225169.ref002]\]. Through LTP and LTD, information can be stored and behavioral patterns can become fixed in the brain. LTP and LTD occur after a period of learning, which is dominated by sensitization and habituation \[[@pone.0225169.ref002]\]. Sensitization and habituation are two basic processes in memory that are controlled by the strength of synaptic connections. Sensitization increases the probability that a given stimulus will produce a downstream transmembrane potential by increasing the connectivity of excitatory synapses while habituation decreases the probability of an transmembrane potential by increasing the connectivity of inhibitory synapses \[[@pone.0225169.ref003]\], \[[@pone.0225169.ref004]\]. This has been demonstrated in experiments on the gill withdrawal reflex in Aplysia californica, where repeated stimulation caused a prolonged withdrawal of the gill \[[@pone.0225169.ref005]\], \[[@pone.0225169.ref006]\]. The process of altering behavior based on changes in synaptic connection strengths is known as synaptic plasticity and is considered to be the underlying mechanism of the formation of memory \[[@pone.0225169.ref005]\], \[[@pone.0225169.ref007]\]. So far, the main stream approach to model the formation of memory was based on the threshold models of individual neurons \[[@pone.0225169.ref008]\], \[[@pone.0225169.ref009]\], \[[@pone.0225169.ref010]\]. Nevertheless, these models did not reflect physiological reaction-diffusion mechanisms which are responsible for the conduction of excitation in the neuronal environment. A recent effort to incorporate reaction-diffusion effects to quantify changes in the synaptic strength of isolated biological synapses \[[@pone.0225169.ref011]\] and synaptic-like memristive elements \[[@pone.0225169.ref012]\] was a step in the right direction. However, it still did not help to elucidate the reaction-diffusion origin of sensitization and habituation. Another way to account for the spatial distribution of neuronal structures is the introduction of the concept of a meta-neuron. A meta-neuron consists of a relatively small group of several tens of neurons, which may be collectively involved in a particular macroscopic function. Such a structure includes Hodgkin-Huxley axons with added synaptic connections described by a set of gating equations \[[@pone.0225169.ref013]\]. In general, this approach may be considered as an adequate tool to describe large neuronal clusters, yet it entirely ignores the essential structural details which govern the balance between sensitization and habituation needed to process the information by a particular memory unit. A typical example of sensitization and habituation can be seen in the startle response in zebrafish. The startle circuit is comprised of an auditory sensory neuron, a Mauthner cell (motor neuron) that triggers startle movement, and an inhibitory neuron that prevents familiar stimuli from triggering a startle response \[[@pone.0225169.ref014]\], \[[@pone.0225169.ref015]\]. The auditory neuron connects to both the inhibitory neuron and Mauthner cell with two excitatory synapse regions. The inhibitory neuron also connects to the Mauthner cell with an inhibitory synapse region. As a new incoming stimulus is repeated, the excitatory synapses connecting to the inhibitory neuron strengthen, and inhibitory synapses connecting to the Mauthner cell strengthen as well, resulting in less frequent triggering of the startle response. Although the described above circuit may be useful for explanation of a simple startle response in zebrafish \[[@pone.0225169.ref014]\], it lacks sufficient complexity to relate to the abrupt disappearance of habituation (dishabituation) which, according to Groves and Thompson dual-process theory, occurs even if the strength of excitatory synapses in the sensory neuron remains constant \[[@pone.0225169.ref004]\]. More complex examples of habituation are described in Sokolov's comparator and Ramaswami's negative-image models \[[@pone.0225169.ref016]\], \[[@pone.0225169.ref017]\]. The system for the formation of the model in Sokolov's approach inhibits the excitatory process for recognized stimuli. When presented with unfamiliar stimuli, this inhibition ceases, resulting in dishabituation. Similarly, dishabituation may also occur in the negative-image model by inhibiting the negative image that negates incoming stimuli. Based on these examples we propose a novel method based on the reaction-diffusion approach that is capable of quantifying combined effects of sensitization, habituation and dishabituation by connecting just a few axons with several synapses of adjustable strength. This model incorporates a circuit of three Morris-Lecar-type neurons \[[@pone.0225169.ref018]\] linked by four synapses defined by Hebbian synaptic strength rules \[[@pone.0225169.ref019]\]. The circuit is connected to two distinct Morris-Lecar-type input cables to allow a separate stimulation of sensory and inhibitory neurons. We will further refer to this circuit as the reaction-diffusion memory unit (RDMU, [Fig 1](#pone.0225169.g001){ref-type="fig"}). ![Schematic diagram which illustrates the composition of the RDMU with two stimulating inputs A and B.\ Excitatory synapses C~1~ and C~2~ are shown as empty triangles. Inhibitory synapses C~3~ and C~4~ are represented by filled-in triangles. Synaptic junctions are denoted by empty squares and the neuronal branching point is marked with an empty circle. Arrows represent the direction of propagation of transmembrane potentials. The portion of the RDMU between C~1~ and C~3~ is the interneuron.](pone.0225169.g001){#pone.0225169.g001} Unlike conventional spiking threshold models, our reaction-diffusion approach eliminates the need for the use of purely phenomenological temporal delays associated with propagation of excitation from one neuron to another \[[@pone.0225169.ref020]\], \[[@pone.0225169.ref021]\]. Instead, these delays form naturally as a result of spatio-temporal evolution of excitation waves under the influence of different rates of cellular membrane polarization and re-polarization processes, various neuronal lengths, and altered strengths of synaptic connections between different neuronal fibers. Model {#sec002} ===== It is possible to illustrate each individual synaptic connection as shown in [Fig 2](#pone.0225169.g002){ref-type="fig"}. To quantify cases with an arbitrary number of such connections, we define a variable C as a conglomerate synapse given by the sum in [Eq (1)](#pone.0225169.e001){ref-type="disp-formula"} $$C = {\sum\limits_{i = 1}^{n}s_{i}}$$ where excitatory and inhibitory synaptic wieghts s~i~ between two specific neurons have positive and negative signs, respectively. The sign of the conglomerate synapse C tells whether the net effect of all connections in the bundle is excitatory (+) or inhibitory (-) ([Fig 2](#pone.0225169.g002){ref-type="fig"}). ![The summation of weights of individual synaptic connections to a conglomerate synapse C.](pone.0225169.g002){#pone.0225169.g002} Using synaptic connections determined by [Eq (1)](#pone.0225169.e001){ref-type="disp-formula"}, we can link neurons to elucidate the processes of sensitization, habituation, and dishabituation ([Fig 1](#pone.0225169.g001){ref-type="fig"}). In this figure, the initial propagation of the transmembrane potential is provided by stimulating inputs A and B. The further evolution of the wave from input A causes sensitization through the passage of excitatory synapse C~2~. The wave also propagates to the interneuron through excitatory synapse C~1~, and produces habituation via inhibitory synapse C~3~. The evolution of the wave from input B causes dishabituation by passing through inhibitory synapse C~4~. We study the RDMU mathematically by using a Morris-Lecar model with incorporated Hebbian conditions at the synaptic junctions, no flux boundary conditions at the ends of neurons, and additional diffusion terms at the sensory neuron's branching point \[[@pone.0225169.ref018]\] ([Fig 1](#pone.0225169.g001){ref-type="fig"}). The equations for the Morris-Lecar model are as follows: $$C_{0}\frac{\partial v}{\partial t} = - g_{L}\left( {v - v_{L}} \right) - M_{\infty}g_{Ca}\left( {v - v_{Ca}} \right) - g_{K}w\left( {v - v_{K}} \right) + D\frac{\partial^{2}v}{\partial x^{2}} + F\left( t \right)$$ $$\frac{\partial w}{\partial t} = \frac{\left. \left( W \right._{\infty} - w \right)}{\tau}$$ $$M_{\infty} = \frac{1}{2}\left( {1 + {\text{tanh}\left( \frac{v - v_{1}}{v_{2}} \right)}} \right)$$ $$W_{\infty} = \frac{1}{2}\left( {1 + {\text{tanh}\left( \frac{v - v_{3}}{v_{4}} \right)}} \right)$$ $$\tau = \frac{1}{\phi}{\text{sech}\left( \frac{v - v_{3}}{{2v}_{4}} \right)}$$ $$F\left( t \right) = \left\{ \begin{matrix} {I,\quad t \leq t_{dur}} \\ {0,\quad t > t_{dur}} \\ \end{matrix} \right.$$ where I and t~dur~ are the amplitude of the external current (stimulus) and its duration, respectively. Variables v and w represent the transmembrane voltage and dimensionless gating variable corresponding to the inhibitory response of the potassium channels. Parameters v~L~, v~Ca~, and v~K~ are equilibrium potentials for leakage, calcium, and potassium currents, respectively. Factors M~∞~ and W~∞~ are dimensionless constants which are determined by regulating voltages v~1~, v~2~, v~3~, and v~4~ \[[@pone.0225169.ref022]\]. By introducing specific time and spatial scales, one can define a set of dimensionless variables as follows: $$v^{} = \frac{v}{v_{Ca}},\mspace{360mu} v_{i}^{} = \frac{v_{i}}{v_{Ca}},\mspace{360mu} i = L,K,Ca,1,2,3,4$$ $$t^{} = t\frac{g_{Ca}}{C_{0}}$$ $$x^{} = \frac{x}{L_{0}}$$ $$L_{D} = \sqrt{\frac{D}{g_{Ca}}}$$ $$g_{i}^{} = \frac{g_{i}}{g_{Ca}},\mspace{360mu} i = L,K,Ca$$ $$\phi^{} = \phi\frac{C_{0}}{g_{Ca}}$$ Here v\, t\*, and x\** are dimensionless variables for transmembrane potential, time and spatial variables. Parameters g~L~\* and g~K~\* refer to dimensionless leakage and potassium conductance, respectively. Parameter g~Ca~\, which determines dimensionless calcium conductance, is equal to one. The value of L~D~ corresponds to the diffusion length. The value of L~0~ is the length of the main section of the RDMU, which is equal to the sum of the lengths of the sensory and motor neurons. The scales are given as follows: C~0~ = 10μF, v~Ca~ = 100mV, D* = 1μS•cm^2^, g~Ca~ = 10mS, and L~0~ = 1mm. For simplicity, we will further refer to dimensionless variables as v, w, t, x, g~K~, v~K~, etc., continuing with dimensionless Morris-Lecar equations in the following way: $$\frac{\partial v}{\partial t} = F\left( t \right) - g_{L}\left( {v - v_{L}} \right) - M_{\infty}\left( {v - v_{Ca}} \right) - g_{K}w\left( {v - v_{K}} \right) + \left( \frac{L_{D}}{L_{0}} \right)^{2}\frac{\partial^{2}v}{\partial x^{2}}$$ $$\frac{\partial w}{\partial t} = \frac{\left( {W_{\infty} - w} \right)}{\tau}$$ $$M_{\infty} = \frac{1}{2}\left( {1 + {\text{tanh}\left( \frac{v - v_{1}}{v_{2}} \right)}} \right)$$ $$W_{\infty} = \frac{1}{2}\left( {1 + {\text{tanh}\left( \frac{v - v_{3}}{v_{4}} \right)}} \right)$$ $$\tau = \frac{1}{\phi}{\text{sech}\left( \frac{v - v_{3}}{{2v}_{4}} \right)}$$ [Table 1](#pone.0225169.t001){ref-type="table"} summarizes the model dimensionless parameters which, unless stated otherwise, were used in all numerical experiments. These values are based on the dimensional values from \[[@pone.0225169.ref022]\]. 10.1371/journal.pone.0225169.t001 ###### The dimensionless parameters used to solve model Eqs ([14](#pone.0225169.e014){ref-type="disp-formula"})--([22](#pone.0225169.e023){ref-type="disp-formula"}). ![](pone.0225169.t001){#pone.0225169.t001g} ϕ g~Ca~ g~K~ g~L~ v~Ca~ v~K~ v~L~ v~1~ v~2~ v~3~ v~4~ I v~o~ w~o~ v^Thr^ ------- ----------- ---------- ---------- ----------- ---------- ---------- -------- -------- -------- -------- ----- ---------- ---------- ------------ 0.017 1 1.8 0.45 1 -0.84 -0.6 -0.012 0.18 0.02 0.30 1 -0.58 0.0177 -0.225 A set of boundary conditions for Eqs ([14](#pone.0225169.e014){ref-type="disp-formula"})--([18](#pone.0225169.e018){ref-type="disp-formula"}) includes no-flux conditions at the ends of the neurons ([Eq (19)](#pone.0225169.e019){ref-type="disp-formula"}), and Hebbian links between the pre- and postsynaptic values of transmembrane potential at each of the RDMU synapses ([Eq (20)](#pone.0225169.e020){ref-type="disp-formula"}) \[[@pone.0225169.ref019]\]. $$\frac{\partial v}{\partial x} = 0$$ $$v_{A,post} - v_{o} = C_{2}\left( {v_{A,pre} - v_{o}} \right) + C_{3}\left( {v_{B,pre} - v_{o}} \right)$$ $$v_{B,post} - v_{o} = C_{1}\left( {v_{A,pre} - v_{o}} \right) + C_{4}\left( {v_{B,pre} - v_{o}} \right)$$ Here C~i~, v~i,pre~, v~i,post~, and v~o~ are synaptic strengths, pre- and post-synaptic potentials of the i^th^ synapse and resting value of transmembrane potentials, respectively. The first of [Eq (20)](#pone.0225169.e020){ref-type="disp-formula"} describes the cumulative post-synaptic action of adjacent excitatory and inhibitory synapses C~2~ and C~3~, located at the beginning of the motor neuron. The second part describes the cumulative post-synaptic action of adjacent excitatory and inhibitory synapses C~1~ and C~4~, located at the beginning of the interneuron. The resting transmembrane potential, v~o~, as well as the resting value of the recovery variable, w~o~, are determined by the intersection of null-clines of the system of Eqs ([14](#pone.0225169.e014){ref-type="disp-formula"}) and ([15](#pone.0225169.e015){ref-type="disp-formula"}). The null-clines also define the excitation threshold, v^thr^, as shown in [Fig 3](#pone.0225169.g003){ref-type="fig"}. ![Phase portrait for the Morris-Lecar model.\ The solid line is the null-cline for v and the dashed line is the null-cline for w. The equilibrium values for v and w occur at the intersection of null-clines. The horizontal dot-dashed line has a value of w~o~ for all v, and is used to find v^thr^, which is the second intersection of v and w~o~.](pone.0225169.g003){#pone.0225169.g003} To complete the formulation of synaptic conditions one needs an additional boundary condition to warrant that each synapse acts as a unidirectional gate which prevents the backward flow of transmembrane potentials. This condition is defined by [Eq (22)](#pone.0225169.e022){ref-type="disp-formula"}: $$\left. \frac{\partial v}{\partial x} \right\|_{x = x^{-}} = 0$$ Here x^−^ is upstream with respect to the direction of the synaptic current. Finally, we analyze the branching point located at the end of input A, where the sensory axon diverges ([Fig 1](#pone.0225169.g001){ref-type="fig"}). At this point we need to modify [Eq (14)](#pone.0225169.e014){ref-type="disp-formula"} and consider two diffusion terms to account for cumulative two-dimensional effects comprised of two one-dimensional diffusion processes in the first (x) and the second (y) neuron branches, respectively: $$\frac{\partial v}{\partial t} = F\left( t \right) - g_{L}\left( {v - v_{L}} \right) - M_{\infty}\left( {v - v_{Ca}} \right) - g_{K}w\left( {v - v_{K}} \right) + \left( \frac{L_{D}}{L_{0}} \right)^{2}\left( {\frac{\partial^{2}v}{\partial x^{2}} + \frac{\partial^{2}v}{\partial y^{2}}} \right)$$ The rationale for considering a linear steady-state Hebbian rule ([Eq (20)](#pone.0225169.e020){ref-type="disp-formula"}) is based upon the observation that behavioral and, to some extent, cognitive memories are associated with neural oscillations within theta and partial gamma ranges below 20 Hz \[[@pone.0225169.ref023]\], \[[@pone.0225169.ref024]\]. Under these conditions one can consider only isolated stimulating currents ([Eq (7)](#pone.0225169.e007){ref-type="disp-formula"}) applied to the sensory neuron shown in [Fig 1](#pone.0225169.g001){ref-type="fig"}. Indeed, the transmembrane potentials induced by neuronal spikes in the hippocampus are on average 1-3ms in duration \[[@pone.0225169.ref025]\] and the intervals between successive spikes at frequencies below 20Hz are greater than 50ms. Therefore, the temporal evolution of the transmembrane potential resulted from a previous neuronal spike becomes completed well before the initiation of the next spiking activity. Accordingly, propagation of the transmembrane potentials in the RDMU branches evolves into transmission of the steady-state solitary pulses. Finally, since the propagation of transmembrane potentials is steady-state, a temporal derivative term in the Hebbian rule \[[@pone.0225169.ref019]\] can be omitted and resulting steady-state Hebbian links can be expressed as linear algebraic relations described by [Eq (20)](#pone.0225169.e020){ref-type="disp-formula"}. Numerical method and system parameters {#sec003} -------------------------------------- The system of Eqs ([14](#pone.0225169.e014){ref-type="disp-formula"})--([22](#pone.0225169.e023){ref-type="disp-formula"}) was solved numerically using an explicit finite difference method (See [S1 Appendix](#pone.0225169.s001){ref-type="supplementary-material"}). The dimensionless time and spatial steps were Δt = 2.5×10^−5^ and Δx = .01 for all experiments, respectively. Parameter $\left( \frac{L_{D}}{L_{0}} \right)^{2}$ in Eqs ([14](#pone.0225169.e014){ref-type="disp-formula"}) and ([22](#pone.0225169.e023){ref-type="disp-formula"}) was set to 0.01. Unless stated otherwise, the sensory neuron and both inputs spanned 25 spatial intervals each, while the motor neuron and interneuron individually consisted of 50 spatial intervals ([Fig 4](#pone.0225169.g004){ref-type="fig"}). At the initial time t = 0 an external stimulus I of amplitude one was applied for a duration of 5×10^4^Δt to nodes one through fifteen located at the beginning of inputs A and B. ![The diagram which depicts the spatial scales of the RDMU.](pone.0225169.g004){#pone.0225169.g004} Taking into consideration that the speed of transmembrane potentials in the brain is on average greater than 10m/s, parameters in [Table 1](#pone.0225169.t001){ref-type="table"} were set to reflect that the width of the excitation wave is much longer than a one millimeter total length of the sensory and motor neurons \[[@pone.0225169.ref026]\], \[[@pone.0225169.ref027]\]. Typical spatial and temporal evolutions of such waves are depicted in [Fig 5](#pone.0225169.g005){ref-type="fig"}. ![Transmembrane potential v and gating variable w as functions of spatial variable x.\ Upper panels show a spatial evolution of the excitation pulse in the sensory and motor neurons in the interval of time between 2.5 and 25. Lower left panel shows progression of the excitation pulse in the interneuron at time 25. Lower right panel illustrates temporal evolution of excitation at x = 0.75. Parameters C~1~, C~2~, and C~3~ are equal to 1, 1.8, and -0.2, respectively, for ideal propagation through all regions.](pone.0225169.g005){#pone.0225169.g005} Experimental protocol for numerical simulations {#sec004} ----------------------------------------------- We studied the propagation of solitary pulses originated by identical input stimuli, I, applied to both inputs of the RDMU. A series of numerical simulations has been performed in order to evaluate the RDMU's ability to reproduce the processes of sensitization, habituation and dishabituation. Depending on the values of synaptic strengths C~1~, C~2~, C~3~, and C~4~, the input stimuli propagated to the motor neuron and originated either sub-threshold or over-threshold responses, thus signifying the initiation of processes of sensitization, habituation and dishabituation. The synaptic strength boundary between sensitization and habituation (BSH) was computed iteratively with the value of C~4~ fixed at zero. At fixed values of C~2~ we incrementally adjusted the value of C~3~ until regimes changed from sensitization to habituation, preventing the propagation of the over-threshold stimulus in the motor neuron. After that, values of C~2~ were increased by a set of sufficiently small increments and the process was repeated until values of C~2~ were equal to 1.5 or values of \\|C~3~\\| exceeded 5, beyond which the BSH and BHDH curves become linearly proportional. The boundary between habituation and dishabituation (BHDH) was calculated in the same manner at different values of C~4~. Results {#sec005} ======= We determined whether the system was in sensitization, habituation, or dishabituation by comparing the maximum transmembrane potential to a threshold potential shown in [Fig 3](#pone.0225169.g003){ref-type="fig"}. The threshold potential was increased by 20% to account for wave propagation decay due to diffusion. BSH and BHDH curves were determined depending on whether the maximum transmembrane potential exceeded the modified threshold or remained below it. It was found that the differences between BSH and BHDH curves measured at 10 Δx from the end of the motor neuron and further away (30 Δx) did not exceed 5% and 16% at low and high values of C~3~, respectively. We chose to measure the magnitude of the transmembrane potentials closer to the end of the motor neuron at 10 Δx. The influence of relaxation parameter, potassium and leakage conductance on shape of the excitation pulse {#sec006} --------------------------------------------------------------------------------------------------------- One of the main parts of our numerical simulations was focused on investigating the influence of the relaxation parameter ϕ and the potassium and leakage conductances on the dynamics of excitation pulses in the RDMU. As expected, we found that the magnitude of ϕ significantly affected the rate of relaxation of recovery variable w, and therefore invoked considerable changes in the width and speed of the excitation pulse. Potassium and leakage conductances also contributed to changes of the width of the pulse in a noticeable way. [Fig 6](#pone.0225169.g006){ref-type="fig"} demonstrates various shapes of excitation pulses for different parameters ϕ, g~K~ and g~L~. One can observe that smaller values of ϕ cause prolongation of pulses. Similar changes occur due to the decrease of either g~K~ or g~L~. ![Various transmembrane potentials for different sets of parameters.\ The top left panel depicts the transmembrane potential for the parameters in [Table 1](#pone.0225169.t001){ref-type="table"}, while the top right, bottom left, and bottom right show transmembrane potentials for decreased ϕ, increased g~K~, and decreased g~L~, respectively.](pone.0225169.g006){#pone.0225169.g006} Boundary between sensitization and habituation. The influence of ϕ, g~K~, g~L~ and C~1~ {#sec007} --------------------------------------------------------------------------------------- We performed a series of numerical simulations to study the edge between sensitization and habituation processes in the RDMU, where the propagation of excitation waves is depicted in [Fig 6](#pone.0225169.g006){ref-type="fig"}. As shown in [Fig 7](#pone.0225169.g007){ref-type="fig"} the BSH can be adequately described by [Eq (23)](#pone.0225169.e025){ref-type="disp-formula"} (see also [Table 2](#pone.0225169.t002){ref-type="table"}) $$C_{3} = aC_{2}^{b} + c,\mspace{360mu} b > 1$$ 10.1371/journal.pone.0225169.t002 ###### Constants a, b, and c for curves depicted in Figs [7](#pone.0225169.g007){ref-type="fig"} and [8](#pone.0225169.g008){ref-type="fig"}. The values of these constants are determined from [Eq (23)](#pone.0225169.e025){ref-type="disp-formula"} using linear regression. ![](pone.0225169.t002){#pone.0225169.t002g} ϕ, g~K~, g~L~, C~1~ a b c Figure/curve shape --------------------------------- -------- ------ ------ -------------------- 0.0017, 2.5, 0.45, 0.8 -3.81 2.76 1.59 8, filled circle 0.0017, 1.8, 0.45, 0.8 -3.37 3.14 1.06 8, filled square 0.0017, 1.8, 0.3, 0.8 -3.39 3.60 0.33 7, filled square 0.017, 1.8, 0.3, 0.8 -3.36 3.47 0.59 7, open square 0.0017, 1.8, 0.3, 0.6 -21.11 5.50 0.57 7, filled diamond 0.017, 1.8, 0.3, 0.6 -18.08 4.08 2.10 7, open diamond ![Sensitization-habituation boundaries depicted as dependences of C~3~ on C~2~ for different values of ϕ and g~L~ (g~K~ = 1.8).\ Open and filled shapes correspond to ϕ = 0.017 and ϕ = 0.0017, respectively. Parameter C~1~ is equal to 0.8 for all curves except open and filled diamonds, where C~1~ = 0.6. Circles correspond to g~L~ = 0.45 while diamonds and squares relate to g~L~ = 0.3. Other parameters are fixed at values shown in [Table 1](#pone.0225169.t001){ref-type="table"}.](pone.0225169.g007){#pone.0225169.g007} It should be noted that shorter (slower) pulses with higher values of ϕ correspond to lower absolute values of the inhibitory synaptic strength \\|C~3~\\|, thus indicating that it is easier to counter play an excitatory action of the synapse C~2~ for higher magnitudes of relaxation parameter ϕ. Alternatively, it was found that a decrease of leakage conductance g~L~ resulted in an opposite shift of BSH towards higher values of \\|C~3~\\| associated with greater thresholds required to inhibit the RDMU at any given strength of C~2~ ([Fig 7](#pone.0225169.g007){ref-type="fig"}). As shown in [Fig 1](#pone.0225169.g001){ref-type="fig"}, the excitatory synapse C~1~ plays a role as some type of a gate which regulates the flow of transmembrane potentials between the sensory and interneuron branches of the RDMU. Specifically, it varies the transmembrane potential's diffusion flux, and therefore controls the amplitude of the excitation pulse which propagates through the inter-neuronal branch of the RDMU. At lower values of C~1~, as well as in case of lesser g~L~, we again observed a significant shift of the BSH towards higher values of \\|C~3~\\| ([Fig 7](#pone.0225169.g007){ref-type="fig"}). Specifically, at C~1~ = 0.6 and C~2~ = 0.65 the value of \\|C~3~\\| required for the suppression of a pulse in the motor neuron was more than three times greater than a corresponding value of \\|C~3~\\| necessary for the suppression of a similar pulse at C~1~ = 0.8. It should be noted that all BSH curves depicted in [Fig 7](#pone.0225169.g007){ref-type="fig"} are in agreement with approximation ([23](#pone.0225169.e025){ref-type="disp-formula"}), since values of b are greater than one ([Table 2](#pone.0225169.t002){ref-type="table"}). However, when the relaxation parameter ϕ and potassium conductance g~K~ increase simultaneously the BSH curves turn to nearly directly proportional changes between inhibitory and excitatory synaptic strengths C~2~ and C~3~ ([Fig 8](#pone.0225169.g008){ref-type="fig"}). ![Sensitization-habituation boundaries depicted as dependences of C~3~ on C~2~ for different values of ϕ and g~K~ (g~L~ = 0.45).\ Open and filled shapes correspond to ϕ = 0.017 and ϕ = 0.0017, respectively. Circles and squares correspond to g~K~ = 2.5 and g~k~ = 1.8, respectively. Other parameters are fixed at values shown in [Table 1](#pone.0225169.t001){ref-type="table"}.](pone.0225169.g008){#pone.0225169.g008} Boundary between sensitization and habituation. The influence of the length of RDMU branches {#sec008} -------------------------------------------------------------------------------------------- It has been demonstrated above that the propagation of excitation waves from the sensory to motor neuron may significantly depend on both the strengths of the excitatory synapses C~1~ and C~2~, as well as on the influence of the inhibitory interneuron synaptic connection C~3~. We also found that the lengths of the RDMU neurons can be additional important contributors into the balance between habituation and sensitization. Accordingly, the larger ratio of the interneuron's length to the total length of the sensory and motor neurons results in more significant shift of the BSH curve to the left, making it more difficult to inhibit the RDMU even at smaller values of C~2~ ([Fig 9](#pone.0225169.g009){ref-type="fig"}). ![Sensitization-habituation boundaries depicted as dependences of C~3~ on C~2~ for different interneuron lengths.\ Black diamonds denote interneurons with length 0.625 while squares relate to interneurons with length 0.5. Parameters ϕ and g~L~ are equal to 0.0017 and 0.3, respectively. Other parameters are fixed at values shown in [Table 1](#pone.0225169.t001){ref-type="table"}.](pone.0225169.g009){#pone.0225169.g009} Boundary between habituation and dishabituation. The influence of C~4~ {#sec009} ---------------------------------------------------------------------- To calculate the BHDH curves, we applied two stimuli through inputs A and B. As shown in [Fig 1](#pone.0225169.g001){ref-type="fig"}, input A connects directly to the sensory neuron while input B connects to the interneuron through inhibitory synapse C~4~. In this manner, C~4~ affects the BHDH curves by decreasing the responsiveness of the interneuron. We found that increasing the strength of C~4~ resulted in a shift of BHDH curves towards smaller values of C~2~, thus reflecting dishabituation of the motor neuron. This effect is more pronounced for lower values of g~L~, where the shift in C~2~ is greater, and the slopes of the BHDH curves are consistently shallower ([Fig 10](#pone.0225169.g010){ref-type="fig"}). ![Habituation-dishabituation boundaries depicted as dependences of C~3~ on C~2~ for different values of C~4~.\ Filled shapes represent g~L~ equal to 0.45 while empty shapes represent g~L~ equal to 0.3. Squares, circles, and diamonds stand for C~4~ equal to 0, 0.05, and 0.15, respectively. Dashed lines correspond to C~4~ equal to 0.225. Other parameters are fixed at values shown in [Table 1](#pone.0225169.t001){ref-type="table"}.](pone.0225169.g010){#pone.0225169.g010} As the strength of C~4~ further increases, the BHDH curves continue to shift to the left with steeper slopes, until the value of C~4~ is approximately 0.225, where the BHDH curves become vertical, as shown by dashed lines ([Fig 10](#pone.0225169.g010){ref-type="fig"}). Beyond this value, waves in the interneuron are unable to propagate to inhibitory synapse C~3~, resulting in complete dishabituation. Discussion {#sec010} ========== Utilizing a novel approach, we studied variations in the BSH dependences in response to changes of parameters of the reaction-diffusion model with Hebbian type synaptic junctions between neurons. It was found that longer transmembrane potential waves (lower g~L~), which propagate in the motor neuron, caused the BSH curves to shift towards sensitization. On the contrary, shorter waves (greater g~K~) triggered the opposite shift of the BSH curves towards habituation. Also, we observed that synaptic strength C~1~ is another important parameter which has a significant effect on the positioning of BSH. The value of C~1~ directly affects the transmembrane potential flux into the interneuron, thus changing its inhibitory influence on the RDMU. Specifically, it was found that different values of C~1~ either substantially reduced or increased the effectiveness of the inhibitory synapse C~3~, resulting in a state of the RDMU that is either significantly harder or easier to habituate. In addition to excitatory synapse C~1~, inhibitory synapse C~4~ also influences the RDMU through changing conditions for dishabituation. Indeed, just a small increase in C~4~ produces a notable shift in the BHDH curves towards lower values of excitatory strength in the synapse C~2~. There are two possible approaches for incorporating dishabituation in the RDMU. These two approaches can be derived from the existing concepts of superimposition of sensitization and reversal of habituation described in \[[@pone.0225169.ref028]\]. The first approach is to increase the responsiveness of the motor neuron through an additional strong stimulus, which can be accomplished by adding an additional sensory neuron. In contrast, the second approach is to inhibit the interneuron that causes habituation. While the first approach results in the intertwining of sensitization and dishabituation, since the two processes share the same mechanism, the second one allows sensitization and dishabituation to be further distinguished. There has been much debate about whether sensitization and dishabituation can be dissociated \[[@pone.0225169.ref028]\]-\[[@pone.0225169.ref032]\]. Once again, based on specific experimental procedures, these two processes could either occur through the same mechanisms \[[@pone.0225169.ref029]\], \[[@pone.0225169.ref030]\] or could have differing ones \[[@pone.0225169.ref028]\], \[[@pone.0225169.ref031]\], \[[@pone.0225169.ref032]\]. The classic dishabituation described in \[[@pone.0225169.ref004]\] could be an example of superimposed sensitization, where dishabituation and sensitization both result from direct stimulation to the habituated neuron. However, a more recent revision of this work emphasizes the reversal of habituation as another form of dishabituation \[[@pone.0225169.ref033]\], which we chose to model using additional input B ([Fig 1](#pone.0225169.g001){ref-type="fig"}). A possible future enhancement to the RDMU would be the ability to model more complex behaviors than described above. Often described alongside sensitization and habituation is training, the process by which a weak input (trained input) becomes able to excite a target neuron by being repeatedly paired with a strong stimulus (training stimulus) \[[@pone.0225169.ref005]\]. After sufficient stimulations, the strength of the synapses in the trained input increases enough to excite the target neuron independently of the training stimulus. At present, our RDMU is unable to model training because the strengths of synaptic connections are fixed. Additionally, the current RDMU may need to be modified to accommodate series of periodic stimulations, as excitations due to isolated stimuli may not be capable of producing training. Supporting information {#sec011} ====================== ###### Fig A1. Numerical mesh used to approximate the Morris-Lecar equations. Each node is one spatial interval Δx apart from adjacent nodes. Fig A2. Block diagram for solving the explicit grid Eqs. S1 (A1)--S1 (A24). At each time step these equations are solved in the following order: no-flux boundary conditions at the edges, Morris-Lecar equations at the inner grid points, modified-diffusion-term Morris-Lecar equations at the branching node and Hebbian and uni-directional no-flux boundary conditions at the synapses. (DOCX) ###### Click here for additional data file. We thank Andrei Starobin for the introduction of Eric Kandel\'s book \"In search of memory\", which played a significant role in shaping our approach to the problem. [^1]: Competing Interests:The authors have declared that no competing interests exist.	{ "pile_set_name": "PubMed Central" }
Introduction {#s1} ============ Although the etiology of primary biliary cirrhosis (PBC), primary sclerosing cholangitis (PSC) and autoimmune hepatitis (AIH) remains unknown, there are several case reports of association of these hepatic autoimmune conditions in the same patient \[[@R01]\]. The expression "overlapping syndrome" has been used to describe forms of auto-immune disease, generally AIH/PBC or AIH/PSC, that present typical characteristics of more than one condition in the same patient, occurring simultaneously or sequentially, and sometimes migrating from one to another clinical presentation \[[@R03]\]. However, overlapping between PBC/PSC is much less described. PBC is mostly prevalent among women, causing destruction of biliary ducts, resulting in progressive ductopenia and cirrhosis. AMA is considered a specific biomarker of PBC and some authors describe it as the serologic signature of the disease \[[@R07], [@R08]\]. PSC, in its turn, is also a chronic cholestatic liver disease of unknown etiology, typically marked by progressive inflammation and concentric fibrosis of intra- or extra-hepatic biliary ducts, causing cirrhosis, liver failure and high incidence of cholangiocarcinoma \[[@R09]\]. So far, there are no specific serological markers for PSC and AMA is virtually absent in PSC patients \[[@R12]\]. To our knowledge, there are only five PBC/PSC overlapping cases reported in the literature, none of them corresponding to small biliary duct PSC \[[@R15]\]. Here we describe a patient with clinical, biochemical and serological markers of PBC, whose biopsy was compatible with small duct PSC. Case Report {#s2} =========== A 48-year-old woman was referred to the hepatologist in order to investigate elevated levels of liver enzymes. Except for a treated systemic hypertension, mild obesity and a ten-year irregular use of amfepramone, she had no remarkable medical history. She also had no familiar history of liver disease or alcohol consumption. Physical examination revealed only mild hepatomegaly. Laboratory tests are summarized in [Table 1](#T1){ref-type="table"}. She presented a positive antinuclear antibody (ANA) test (titer \> 1/640) with a rim-like membranous and cytoplasmic speckled pattern, suggestive of anti-mitochondrial positivity ([Fig. 1](#F1){ref-type="fig"}). These ANA patterns are known to be associated to antibodies to gp210 and mitocondrial antigens, respectively. Indirect immunofluorescence (IIF) tests for anti-smooth muscle antibodies (SMA), anti-liver/kidney microsome 1 (LKM-1), and neutrophil cytoplasm antigens (ANCA) were negative. IIF-AMA on in-house rodent tissue preparations was performed as previously described \[[@R19]\] and was positive at 1/160. Anti-pyruvate dehydrogenase antibodies (anti-M2 fraction) were positive and detected by enzyme immunosorbent assay (ELISA-Orgentec, Mainz, Germany). There was also reactivity for anti-M2-E3 BPO, anti-gp210 and anti-Sp-100 (Euroline profile, Euroimmun, Lubeck, Germany). ###### Case Report: Laboratory Findings 29/04/04 28/05/04 05/06/06 31/08/06 31/08/06 --------------------------- ---------- ----------- ----------- ----------- ---------- AST (UI/mL) 98 (45) 79( 45) 42 (45) 116 (45) 26 (45) ALT (UI/mL) 125 (35) 68 (35) 38 (35) 167 (35) 14 (35) GGT (UI/mL) 227 (48) 249 (48) 408 (38) 383 (38) 12 (38) AP (UI/mL) 472 (147) 211 (147) 468 (147) 83 (147) Total Bilirub (mg/dL) 0.41 0.59 Dir. Bilirub (mg/dL) 0.06 0.17 INR 1.0 Albumin (g/L) 4.5 4.3 Gamma-glob (g/L) 1.6 Total cholesterol (mg/dL) 177 HDL -C (mg/dL) 52 Triglycerides (mg/dL) 116 IgM (mg/dL) 183 IgG (mg/dL) 1460 HBsAg Neg Anti-HBc Neg anti-HCV Neg ANA Pos anti-SMA Neg anti-LKM-1 Neg anti-SLA Neg p-ANCA Neg AMA M2 Pos Anti-Sp100 Pos Anti-gp210 Pos HBsAg, hepatitis B surface antigen; Anti-HBc, antibody against core antigen of hepatitis B virus; anti-HCV, antibody against hepatitis C virus; ANA: anti-nuclear antibody; SMA: Smooth muscle antibodies; Anti-LKM: Liver-kidney microsomal antibodies; Anti-SLA/LP: Antibodies against soluble liver antigen/liver pancreas; pANCA: Perinuclear antineutrophil cytoplasmic antibodies. AMA M2: Antibodies to mitochondrial E2-pyruvate dehydrogenase; Anti-Sp100: Antibodies to Sp100; Anti-gp210: Antibodies to gp210. ![Indirect immunofluorescence on HEp-2 cells (Bion Interprise Ltd) with human serum diluted 1/80. (A) Pattern nuclear envelope; (B) and cytoplasmic discrete speckled pattern, suggestive of antimitocondrial positivity; (C) Chromosome metaphase plate negative. Magnification x 400.](jocmr-04-429-g001){#F1} Liver biopsy revealed bridging portal fibrosis, lymphomononuclear infiltrate with lymphocytic interface hepatitis and marginal ductular reaction. Surprisingly, some of the portal tracts revealed small biliary ducts with concentric fibrosis ("onion skin" type) with duct obliteration ([Fig. 2](#F2){ref-type="fig"}, [3](#F3){ref-type="fig"}). Magnetic resonance cholangiography was then performed, with normal findings. Diagnosis of small duct PSC was done, and she was treated with ursodeoxycholic acid (UDCA) with progressive normalization of liver enzymes within 4 months. ![Portal fibrosis with portal-portal linking septa, and complete obliteration of bile duct which is replaced by dense fibrous whorls (Arrow). H and E, 40 x.](jocmr-04-429-g002){#F2} ![Small bile duct showing an "onion-skin" type of periductal fibrosis. H and E, 40 x.](jocmr-04-429-g003){#F3} Discussion {#s3} ========== The classically described overlapping syndrome is characterized by diagnosis of more than one form of autoimmune liver disease in the same patient \[[@R01]\]. If not correctly diagnosed and appropriately treated, its course can be more aggressive as compared to isolated forms \[[@R20], [@R21]\]. It is a matter of discussion if overlapping syndromes are isolated entities or different manifestations of a unique autoimmune liver disease \[[@R01], [@R02], [@R06]\]. Overlapping between PBC/AIH, AMA-negative PBC/AIH, and PSC/AIH, have been related in different series \[[@R03], [@R05], [@R22]\]. However, few cases of overlapping between PBC/PSC have been reported \[[@R15]\]. As both PBC and PSC are cholestatic diseases presenting with increased AP and GGT, overlapping of these two conditions could occasionally not be perceived. The patient herein reported presented an unremarkable clinical picture represented by the incidental finding of biochemical evidence of cholestatic disease, which might be suitable to both PBC and PSC. The serologic investigation initially performed was vigorously suggestive of PBC, with a positive a panel of autoantibodies specific for this disease, including antibodies to mitochondrial E2-pyruvate dehydrogenase (AMA M2), gp-210 and Sp-100. AMA may be considered one of the most useful parameters for the diagnosis of cholestatic liver diseases, since AMA is virtually absent in PSC patients compared to over 95% prevalence in PBC \[[@R07], [@R12], [@R25]\]. Other autoantibodies, such as anti-gp-210 and anti-Sp-100, are also specifically associated with PBC \[[@R26]\]. Anti-gp-210 has been associated to a more aggressive course of the disease \[[@R27], [@R28]\] whereas anti-Sp-100 is considered highly specific (specificity of 95%), despite its low sensitivity (around 30%) \[[@R27]\]. These PBC-associated autoantibodies elicit peculiar ANA patterns, such as the cytoplasmic reticule-like speckled pattern (associated with AMA) \[[@R30]\], the multiple nuclear dots pattern (associated to anti-Sp100 antibodies), the rim-like membranous pattern (linked to positivity to anti-gp210 antibodies), and the centromeric pattern \[[@R26]\]. Despite all these evidences for the diagnosis of PBC in the present case, the histological examination was not characteristic of the disease; instead, it showed typical onion-skin lesions in biliary ducts, with concentric fibrosis, compatible with PSC, a non-expected finding in face of the serological findings. The diagnosis of PSC is confirmed by the demonstration of typical lesions of the biliary tree, either by histological analysis or by imaging techniques (endoscopic cholangiography or MRI) \[[@R31], [@R32]\]. Patients who present with clinical, biochemical and histological features compatible with PSC, but have a normal cholangiogram, are classified as small duct PSC \[[@R31], [@R33]\]. The histological hallmark of PSC is represented by concentric fibrosis of bile ducts; however, this finding is detected in only 30 to 40% of patients with PSC \[[@R31], [@R34]\]. Onion-skin fibrosis can also be found in secondary sclerosing cholangitis, but the patient reported here had no identified cause to this condition \[[@R31]\]. Autoantibodies are often found in PSC patients, but none of them have enough sensitivity and/or specificity to be considered a diagnostic maker. Autoantibodies regularly described in PSC include ANA (8-77%), SMA (0-83%), and pANCA (26-94%), usually in low titers \[[@R12], [@R37]\]. Description of overlapping between PBC and PSC is very uncommon. After revision of literature, only five reports of PBC/PSC overlapping syndromes were identified \[[@R15]\]. MRI was normal in the present patient, leading to the diagnosis of small duct PSC/PBC overlapping. To the best of our knowledge, this is the first report of small duct PSC overlapping with PBC. As she presented positivity to gp210, prognosis is a matter of awareness, even in the presence of small duct PSC, a condition with better prognosis than large duct PSC \[[@R38], [@R39]\]. We have not enough follow up for evaluating prognosis in this case. After five years, response to treatment has been satisfactory. In conclusion, we describe a female patient with features of PBC and small duct PSC, a diagnosis that would remain undiscovered if histology was not performed. This finding has important and controversial clinical implications, such as the management of treatment and the need of screening for cholangiocarcinoma \[[@R32]\]. This case is emblematic of how complex diagnosis of liver autoimmune diseases can be, and reinforces the impression that these are polymorphic entities that can express different patterns of autoantibodies and histological findings, making the diagnosis of these conditions particularly challenging. Conflicts of Interest {#s4} ===================== There are no conflicts of interest.	{ "pile_set_name": "PubMed Central" }
The Accreditation Council for Graduate Medical Education (ACGME) requires that internal medicine residents participate in scholarly activities, including conducting original research, or preparing case reports or literature reviews.[@CR1] Promoting residents' scholarly activity is also a laudable goal of medical educators. However, achieving this goal can be challenging. Barriers include a lack of interest by residents and limited research infrastructure, as reported in a recent national survey of internal medicine program directors.[@CR2] Issues relating to faculty availability and expertise may also serve as impediments to successful resident research endeavors. In large programs, finding a sufficient number of research mentors may be difficult. In training programs with few research-intensive faculty members, opportunities for residents to participate in original research may be limited or absent. Due to these limitations, preparing case reports for poster presentation or publication remain an important way for residents to fulfill research requirements. Two articles in this issue of JGIM demonstrate practical and effective methods of promoting and assessing the scholarly efforts of residents. Importantly, these activities were studied in trainees from multiple programs, did not require substantial faculty expertise or funding, and could easily be implemented in internal medicine residency training programs in a variety of settings. In the first paper, Sridhar et al. studied the results of a case report writing workshop.[@CR3] The authors evaluated outcomes in participants who attended a 60--90-min workshop at various regional and national general internal medicine meetings from 2005--2008. The workshop was designed to help attendees prepare a case report for presentation or publication and included content in formatting, outlining, and reviewing specific journal requirements. Approximately two-thirds of the 214 workshop participants were students, residents, or fellows. After the workshop, participants reported a significant improvement in perceived competence to prepare a case report, and an increased likelihood to submit a case report for presentation or publication within the next year. Two years after the program began, a PubMed search of participants found that nine (seven residents, two faculty members) had successfully published a case report. The work of Sridhar and colleagues advances what is known about research in internal medicine residency programs. Given the simplicity of the methods and short-term success of program participants, residency programs may wish to recreate the workshop, using the content, format, and handouts described in the paper. Subsequently, residents could be encouraged to prepare case reports for presentation or publication. This approach not only fulfills ACGME requirements, but also has been shown by the authors to successfully promote the academic development and productivity of trainees. In the second paper, Willett and colleagues present the results of an observational study performed to identify key components of case report posters and test an evaluation tool.[@CR4] During their study, 45 faculty evaluators from 20 institutions reviewed 247 case report posters at four Society of General Internal Medicine (SGIM) meetings (three regional, one national) from 2006--2008. Using an evaluation tool incorporating SGIM and American College of Physicians poster judging criteria, a mean score was calculated for each case report poster. Presenters received immediate written and verbal feedback from the faculty evaluators, including a copy of the completed evaluation tool. The authors found three areas often requiring improvement in case report posters. These are: (1) clearly stating learning objectives, (2) linking conclusions to learning objectives, and (3) use of appropriate number of words. In regards to the evaluation tool, internal consistency was high, and scores were similar across the broad range of faculty evaluators. This project illustrates a reasonable approach for instructing trainees in case report poster preparation. The criteria for successful posters and how submissions will be judged are clearly demonstrated. Many internal medicine residency programs have "resident research day" or other local or regional venues for housestaff to present their research. The evaluation tool developed by Willett and colleagues could certainly be used in individual residency programs to judge case report poster submissions and provide valuable feedback to trainees. Such critical evaluation may also increase the likelihood of eventual acceptance of abstracts and manuscripts for presentation or publication. The authors of both papers should be commended for providing residency programs the means to meet ACGME requirements regarding the scholarly activity of residents. Used together, these projects can form an integrated approach to fostering high quality case report preparation by residents. The first step is an educational activity, such as the case report workshop developed by Sridhar et al. Next, case report poster submissions are solicited from residents. Last, submissions are evaluated using the tool described by Willet et al., and immediate feedback provided to trainees. These manuscripts should encourage clinician-educators about possibilities to enhance the amount and quality of scholarly activities by residents. The work of these investigators represents practical and effective strategies to increase residents' academic productivity, enhance opportunities for mentoring, and promote career development. Whether promotion committees will adequately recognize the important work of clinician-educators in fostering the academic endeavors of residents is unknown. However, these papers show us that case report posters and publications represent a feasible, inexpensive, and important educational opportunity for internal medicine residents.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#sec1-1} ============ Urinary retention and voiding dysfunction (VD) symptoms (hesitancy, straining to void, difficulty in starting micturition, diminished stream, and sensations of incomplete emptying of the bladder) are common and problematic features occurring after surgery for stress urinary incontinence (SUI), especially following a mid-urethral sling placement (retropubic or trans-obturator procedure). These obstructive complications often induce quality of life impairment after surgery. They can lead to urinary infections, and can also require self-catheterization or sling section. The prevalence of urinary retention varies from 2.2--16% after surgery for SUI.\[[@ref1]--[@ref3]\] Preexisting voiding dysfunction is known to be a predictive factor for post-surgery urinary retention. In 2003 Hong et al.,\[[@ref4]\] showed that the best predictive factor for obstructive complications was the maximum flow rate, with which it was shown to be directly correlated. It is thus important to screen women suffering from SUI for emptying symptoms before such surgery. The purpose of this study was to determine whether the completion of a VD questionnaire could have a good predictive value for uroflowmetry findings, in a population of SUI women showing no obvious etiologies of obstruction. MATERIALS AND METHODS {#sec1-2} ===================== Population {#sec2-1} ---------- We analyzed 415 SUI women from the urodynamic database, who had undergone a filling cystometry in the department, and whose uroflowmetry indicated a total urine volume greater than 200 ml. SUI was defined by a positive bladder stress test. Women with neurological disorders, pelvic organ prolapse, urge or mixed urinary incontinence, detrusor overactivity or previous SUI surgery were excluded. With these exclusion criteria, only 93 of the 415 women were eligible for this study. These 93 women were divided into two groups: Those who complained of VD symptoms (VDS group) according to the Bristol Female Lower Urinary Tract Symptoms (BFLUTS) questionnaire, and those without VD symptoms (non-VDS group). We investigated the parity, age, menopausal status, and concomitant anorectal disorders of each woman. Questionnaire {#sec2-2} ------------- The women were interviewed on the presence of emptying symptoms, and completed the BFLUTS questionnaire \[[Table 1](#T1){ref-type="table"}\], which is systematic in our department. We screened for five items related to voiding difficulties: hesitancy, slow or intermittent stream, straining, or sensations of incomplete emptying. For the purposes of data analysis, those women who answered "never" to all five items were classed in the non-VDS group. Those women for whom one or more of the answers was positive (i.e.: "occasionally, sometimes, most of the time or all the time"), were classified in the VDS group. ###### Bristol Questionnarire BFLUTS (urinary symptoms questionnaire): The five items concerning voiding dysfunction ![](IJU-28-313-g001) Urodynamics {#sec2-3} ----------- The clinical examination included a physical examination, urodynamic testing and a "free flow" uroflowmetry and filling cystometry at a rate of 50 ml per minute (Duet, Medtronic). The urodynamic evaluation methods complied with the standards recommended by the International Continence Society.\[[@ref5]\] The filling cystometry parameters studied were: compliance, detrusor overactivity (DO), detrusor sensation. None of the patients presented with abnormal compliance, DO, or abnormal detrusor sensation. Both a quantitative and a qualitative analysis of bladder outlet obstruction (BOO) was made by uroflowmetry. Quantitatively, a patient was considered to have BOO when a maximum flow rate of less than 15 ml/sec was observed for a urine volume greater than 200 ml,\[[@ref6]\] and/or the post-void residual was greater than 50 ml,\[[@ref5][@ref7]\] as measured with a bladder scan or by catheterization. The urine flow was also analyzed in terms of average flow rate, urine volume, voiding time and acceleration (defined as the maximum flow rate divided by the time taken to reach maximum flow rate). The qualitative analysis dealt with the flow patterns: A standard "bell-shaped" curve was a normal pattern. Three pathologic patterns were considered: Intermittent or continuous polyphasic curves, and flat curves \[Figures [1a](#F1){ref-type="fig"}--[c](#F1){ref-type="fig"}\]. We determined the sensitivity and specificity as well as the positive and negative predictive value of the VDS questionnaire using the uroflowmetry data. ![(a) Intermittent polyphasic curve (b) Continuous polyphasic curve (c) Flat curve](IJU-28-313-g002){#F1} Statistical analysis {#sec2-4} -------------------- A comparison of the clinical characteristics and uroflowmetry of the two groups was made using a Wilcoxon test for continuous data and a Fischer test for categorical data. Logistic regression models were used for multivariate analysis (Statview software). The sample size was not calculated before initialization of the study. Fisher\'s test was used because the sample size was small. All of the factors were considered simultaneously using logistic regression. Ethics {#sec2-5} ------ As the BFLUTS questionnaire and urodynamics were identical to those systematically used for the routine evaluation of SUI women in our department, the study did not require specific ethics committee approbation. RESULTS {#sec1-3} ======= Among the 93 women eligible for the study, 61% (57/93) complained of VD symptoms on the BFLUTS questionnaire (VDS group) and 39% (36/93) had no VD symptoms (non-VDS group). BOO on uroflowmetry was founded in 27 women (29%). Patients' characteristics are detailed in [Table 2](#T2){ref-type="table"}. No difference was observed between the two groups concerning age, parity, menopausal status, concomitant anorectal disorders and uroflowmetry parameters. In the VDS group, only 18 women (18/57, 32%) had an abnormal uroflowmetry. 75% (27/36) of the women in the non-VDS group had normal uroflowmetry, and 25% (9/36) an abnormal one without symptoms. Consequently, the sensitivity of emptying symptoms was 66.7% and specificity 40.9%. The positive predictive value of the VDS questionnaire was 32% and the negative predictive value was 75% \[[Table 3](#T3){ref-type="table"}\]. There was no statistical correlation between symptoms and the defined criteria of BOO on uroflowmetry (Fischer test: P=0.64). Among the 42 women who complained of a sensation of incomplete emptying, only one had a real post-void residual greater than 50 ml. In this SUI population without severe pelvic organ prolapse (Baden and Walker staging system ≤1), previous SUI surgery, or neurological disorders (the most frequent etiologies of voiding dysfunction in women) we found a 29% (29/93) rate of abnormal uroflowmetry. ###### Characteristics of women with and without voiding difficulties' symptoms in terms of age, menopausal status, parity, anorectal disorders and uroflowmetry parameters ![](IJU-28-313-g003) ###### Sensitivity, specificity, positive and negative predictive value of voiding dysfunction symptoms on BFLUTS ![](IJU-28-313-g004) Multivariate analysis (using multiple logistic regression method) demonstrated no correlation between symptoms and quantitative uroflowmetry parameters (maximum and average flow rate, acceleration, urine volume, time to maximum flow and voiding time), and no correlation was found between VDS and age, menopausal status, parity or concomitant anorectal disorders \[[Table 4](#T4){ref-type="table"}\]. ###### Logistic regression method for multivariate analysis of urodynamic parameters (maximal and average flow rate, time to void and acceleration), age, menopausal status, parity, concomitant anorectal disorders with respect to voiding difficulties' symptoms Univariate analysis of multiple factors with respect to voiding difficulties' symptoms ![](IJU-28-313-g005) DISCUSSION {#sec1-4} ========== Our results show that 25% of SUI women without obstructive symptoms, i.e. without an obvious cause for bladder outlet obstruction (no severe genital prolapse (\> Grade 1)), having had no previous surgery for SUI, and with no neurological disorders, have an abnormal uroflowmetry, and thus a higher risk of post-surgery obstructive disorders. Our study clearly demonstrates the poor predictive value of the VDS questionnaire, when compared with urodynamic data. VDS are not correlated with abnormal uroflowmetry parameters as defined on a free uroflowmetry. The prevalence of voiding dysfunction symptoms among women is not easy to determine, in contrast to storage symptoms, which are more often related and analyzed. The prevalence of VDS among elderly women was about 10.9%.\[[@ref8]\] Stress urinary incontinence and bladder outlet obstruction (BOO) can coexist in women; Bradley found 19 cases of BOO out of 104 (18%), defined on the basis of pressure flow studies among women presenting with SUI.\[[@ref9]\] In our study, we found a high prevalence of VDS (61%), probably because in the VDS group we included women with any level of positive response, even those who gave one 'occasionally' response to the VD items (BFLUTS questionnaire). Urodynamic evaluation of the voiding phase in SUI women before surgery is useful in order to predict the risk of VD and urinary retention, and to inform women about these potential complications. Surgical treatments of SUI using mid-urethral sling procedures are known to affect the voiding phase: The spontaneous flow curve changes to a more obstructive pattern in 40% of cases, and the maximum and mean flow rates decrease significantly after one year.\[[@ref10]\] The incidence of urinary retention after such procedures ranges from 2.2--16%.\[[@ref1]--[@ref3]\] Postoperative VD induces a decrease in global satisfaction after surgery, and leads to an impairment of the quality of life. Factors which correlate highly with a postoperative voiding dysfunction include a preoperative peak flow rate \< 15 ml/s (several studies), and an abnormal uroflow pattern or detrusor pressure \< 12 cm H~2~O.\[[@ref3][@ref11]--[@ref14]\] Therefore, screening for such factors among SUI women is critical. One of the main limitations of the current study is the lack of a standard definition for BOO in urodynamics for women. Since voiding cystometry had not been performed, BOO cannot be defined urodynamically in the current study. Farrar et al.,\[[@ref6]\] defined obstruction as Qmax \< 15 ml/s with a voided volume of 200 ml. Further study should assess this point. We chose this definition because it is based on "free flow" uroflowmetry, which is close to physiological micturition. We added the flow curve pattern criteria, because a pathological pattern with a normal maximal flow rate can reflect voiding difficulties. Moreover, in pressure flow studies in women, the transurethral catheter has a significant impact on the urine flow.\[[@ref14]--[@ref18]\] However, "free flow" uroflowmetry is a composite interaction between the detrusor pressure and the resistance produced by the urethra. Bladder outlet obstruction as well as impaired detrusor contractility can perturb uroflowmetry. In our population, we needed to detect voiding difficulties, whatever their origin, because among SUI women both these causes of voiding disorders can promote obstructive complications after surgery. Chassagne et al.,\[[@ref13]\] prospectively studied 35 clinically obstructed women and 124 control patients. They determined cutoff values for the pressure flow studies as: Qmax\<15 ml/s and Pdet. Qmax \> 20 cm H~2~O, with a sensitivity of 74.3% and a specificity of 91.1%. No information was provided regarding the clinical criteria used to select the "clinically obstructed" patients. Groutz et al.,\[[@ref14]\] combined Qmax \< 12 ml/s in "free flow" uroflowmetry with a high detrusor pressure at maximal flow (Pdet. Qmax \> 20 cm H~2~O) during detrusor pressure uroflow studies. Among the 38 women considered as obstructed within this definition, 29% had no obstructive symptoms. They concluded that micturition symptoms relevant to bladder outlet obstruction are non-specific, and that a complete urodynamic evaluation is essential. The correlation between obstructive symptoms and objective urodynamic findings is known to be poor, but has never been studied in this specific population of SUI women, for which all cases with a POP-Q \> Stage 1, previous surgery for SUI, or neurological disorders, were excluded. Indeed, we excluded the most common etiologies of urodynamic assessment of bladder outlet obstruction. Groutz et al.,\[[@ref15]\] reported 26% of prior anti-incontinence surgery and 24% of severe genital prolapse in a population of 38 women presenting with urodynamically assessed obstructive criteria. Wyndaele et al., recently confirmed the poor correlation between symptoms and urodynamic data, and a high prevalence of VDS in women, whether it be considered to have a pathological origin or not.\[[@ref19]\] The absence of voiding symptoms is not a guarantee of normal micturition, since we found that almost one-third of the women not complaining of any voiding difficulty were actually shown to have an abnormal uroflowmetry. Previous studies\[[@ref14]--[@ref21]\] have clearly demonstrated the lack of correlation between symptoms and urodynamic diagnoses, especially for obstructive symptoms. We found the same results, but in a particular population of SUI women, with no obvious reasons for obstruction. Even in this selected population, in which voiding dysfunctions should be less frequent than in the general population, we found that almost 30% of the patients had an abnormal uroflowmetry. CONCLUSION {#sec1-5} ========== No correlation was found between obstructive symptoms and BOO as defined on uroflowmetry, in a specific population of SUI women. Our results suggest that uroflowmetry may be necessary rather than multichannel urodynamics. Source of Support: Nil Conflict of Interest: None declared.	{ "pile_set_name": "PubMed Central" }
All relevant data are within the manuscript and its Supporting Information files. Introduction {#sec005} ============ The frequency of cardiovascular disease before the age of 55 is lower in women. At the age of 55, the lifetime risk of cardiovascular disease for males and females is similar. Females are, however, more likely to suffer from cerebrovascular disease or develop heart failure, whereas males are more likely to suffer from coronary artery disease as an initial event \[[@pone.0215302.ref001]\]. These observed differences in the manifestation of atherosclerotic cardiovascular disease led to research focusing on the role of sex hormones in the pathophysiology of atherosclerosis. It has been postulated that the lower frequency of premature coronary artery disease (P-CAD) in premenopausal women compared to men is primarily due to the protective effect of female sex hormones \[[@pone.0215302.ref002],[@pone.0215302.ref003]\]. Lower estrogen levels in females are associated with a higher risk of cardiovascular disease \[[@pone.0215302.ref002],[@pone.0215302.ref003]\]. In contrast, other studies have shown that progestogens added to standard hormonal replacement therapy consisting of estrogens increased the risk of cardiovascular events in postmenopausal women \[[@pone.0215302.ref004],[@pone.0215302.ref005]\]. It has also been shown that in men, lower levels of testosterone and estradiol were associated with a higher coronary artery calcium score in the Offspring and Third Generation Cohorts of the Framingham Heart Study \[[@pone.0215302.ref006]\]. P-CAD in first-degree relatives is a commonly known risk factor for coronary artery disease (CAD) \[[@pone.0215302.ref007]\]. This effect is due to both genetic and environmental factors \[[@pone.0215302.ref007]\]. As shown in previous studies, traditional risk factors for CAD, i.e., smoking, arterial hypertension, hypercholesterolemia, and abdominal obesity are also more prevalent in subjects with family history (FH) of P-CAD \[[@pone.0215302.ref008]\]. To the best of our knowledge, the relationship between gonadal and pituitary hormones and the FH of P-CAD has not been evaluated so far. The aim of our study was to compare wide range of biochemical marker levels including i.a. pituitary hormones (follicle-stimulating hormone, FSH, and luteinizing hormone, LH) and sex steroids (testosterone, estradiol, progesterone) between young healthy adult men with and without FH of P-CAD. Methods {#sec006} ======= The following analysis is a part of the MAGNETIC (Metabolic and Genetic Profiling of Young Adults with and without a Family History of Premature Coronary Heart Disease) project. The MAGNETIC project is a case-control study that aims to analyze the classical and genetic risk factors of CAD in healthy young adults with and without FH of P-CAD. The study design of the MAGNETIC project has been described previously \[[@pone.0215302.ref009]\]. In brief, the study sample was recruited between July 2015 and October 2017. The inclusion criteria were as follows: male sex, age ≥18 and ≤35 years old, angiographically documented P-CAD (myocardial infarction, coronary artery bypass grafting, or percutaneous coronary intervention before the age of 55 in men and 65 in women) in first-degree relatives (test group) or no P-CAD in first-degree relatives (control group). The exclusion criteria for both groups were as follows: age \<18 or \>35 years, failure to provide informed consent, and acute or chronic diseases requiring pharmacotherapy. Subjects with a positive FH of P-CAD were recruited in the following two ways: 1) from healthy subjects aged 18 to 35 years who were asked to participate in a screening appointment at the Silesian Centre for Heart Disease, who provided documented proof of P-CAD history in their first-degree relatives, and 2) from the healthy offspring of patients who were hospitalized at our center between 2010 and 2017 due to P-CAD, who gave permission to be contacted once the treatment was completed. The control group was recruited from healthy subjects aged 18 to 35 years who were asked to participate in a screening appointment at our center, who were confirmed not to have FH of P-CAD. In total, 411 male Caucasian subjects met the inclusion criteria and were included in the study. Of these participants, 223 (54.3%) had first-degree relatives with documented P-CAD (test group), and 188 (43.7) subjects reported no FH of P-CAD (controls). During the screening appointment, the recruited subjects, along with a qualified interviewer, filled out a detailed, standardized questionnaire encompassing sociodemographic information and questions regarding FH of cardiovascular diseases. Arterial blood pressure, weight and height were measured. Moreover, waist and hip circumference were measured using a measuring tape according to the recommendations of the World Health Organization \[[@pone.0215302.ref010]\]. Further details about the data collection procedures were described previously \[[@pone.0215302.ref009]\]. Peripheral blood was collected from each patient during the first recruitment visit between 7 am and 9 am., approx. 8 to 10 hours after the last meal. Fibrinogen content was determined in the blood collected in S-Monovette tubes with 3.1% sodium citrate (SarstedtAG&Co. KG. Germany) using a Siemens BCS XP analyzer (Siemens Healthcare, Germany). Biochemical and immunochemical measurements were determined in the patients' serum using a Cobas 6000 analyzer (Roche Diagnostics. USA). Serum was obtained from the peripheral blood collected in S-Monovette tubes with clotting activator (SarstedtAG&Co. KG. Germany). The collected blood was centrifuged at 1500 rpm for 10 minutes at 4°C to obtain serum. The following concentration of the following components were measured using methods specified by the manufacturer: glucose, total cholesterol, high-density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), triglycerides, creatinine, uric acid, lipoprotein(a) (Lp(a)), apolipoprotein A1, apolipoprotein B, thyroid-stimulating hormone, FSH, LH, estradiol, progesterone, total testosterone, sex hormone binding globulin (SHBG), high-sensitivity C-reactive protein (hsCRP), total protein, and albumin. The HbA1C concentration in whole blood collected in S-Monovette tubes with EDTA was also measured using the Cobas 6000 analyzer. The test kits and catalog numbers are shown in [S1 Table](#pone.0215302.s001){ref-type="supplementary-material"} (Supporting Information). The percentage of high density lipoprotein (HDL%) was calculated according to the following formula: HDL-C level (mmol/L) divided by total cholesterol level (mmol/L) x 100. Statistical analysis was performed using the computing environment R (R Development Core Team, 2005) \[[@pone.0215302.ref011]\]. The deviation from the normal distribution was investigated by analysis of the normal probability plots (Quantile-Quantile plots). The comparison of continuous variables was performed using the Student's t-test or the U-Mann--Whitney test where appropriate. The categorical variables were assessed using Fisher's exact test. The correlations between FSH, LH, testosterone, progesterone, SHBG and other laboratory parameters were assessed using the Spearman rank correlation test. We analyzed 27 variables for 411 patients. The 11097 element matrix contained 42 missing values (0.4%). The data on waist-to-hip ratio (WHR) were lacking for 10 patients, data on SBP and DBP measurements were lacking for 14 patients, information on smoking habits were missing for 2 patients, FSH was lacking for 1 patient and there was missing information on estradiol for 1 patient. We performed a univariable logistic regression analysis using complete observations to assess the association between analyzed variables and FH of P-CAD. Variables with p-values \<0.35 in the univariable model were included in the multivariable analysis \[[@pone.0215302.ref012]\], except for total cholesterol, apolipoprotein B and HDL-C as they were significantly corelated with LDL and HDL%, and glucose because it was correlated with HbA1C. Next logistic regression with backward selection procedure was performed. Decision to retain variable in the model was based on Akaike information criterion. To minimize the impact of the missing data on the multivariable logistic regression analysis, we used missForest (R package missForest), a state of the art statistical method for mixed-type data imputation that utilizes the random forest model based on the non-missing variables in the dataset to predict the missing values. This process is repeated for all variables, and the process is iteratively repeated until a terminating criterion is attained \[[@pone.0215302.ref013]\]. Sample size of 188 patients in the control group and 223 patients in the test group (at α = 0.05 and β = 0.8), enabled us to detect difference of 27.8% of particular standard deviation value between analyzed groups. For example, we were able to detect difference of 0.139 units or less between means for variables with standard deviation of 0.5 or lower that is for variables such as: apolipoprotein B, apolipoprotein A1 \[g/L\], HDL-C \[mmol/L\], glucose \[mmol/l\], HbA1C \[%\], TSH \[μlU/mL\] and progesterone \[ng/mL\]. Power analysis was performed using PASS software (NCSS Statistical Software, Kaysville, Utah). This study was conducted following the Declaration of Helsinki and good clinical practice guidelines. The study protocol has been approved by the Ethics Committee at Institute of Occupational Medicine and Environmental Health, Sosnowiec (Resolution no. 03/2013). Informed, written consent was obtained from all subjects enrolled in the study. Results {#sec007} ======= Study group characteristics {#sec008} --------------------------- Subjects with positive FH of P-CAD were slightly older (28.0 vs. 28.8 years, p = 0.03) and more frequently current smokers and obese, and they had a higher WHR compared with subjects without FH of P-CAD. Subjects with FH of P-CAD had higher levels of total cholesterol, LDL-C, apolipoprotein B, and lower levels of HDL-C ([Table 1](#pone.0215302.t001){ref-type="table"}, [Fig 1](#pone.0215302.g001){ref-type="fig"}). In the test group, there were also significantly higher fibrinogen levels, a trend toward higher hsCRP, and lower albumin levels, compared with the control group ([Table 1](#pone.0215302.t001){ref-type="table"}, [Fig 1](#pone.0215302.g001){ref-type="fig"}). Despite the fact that fasting glucose and HbA1c levels were within the normal range for all participants, the mean levels of glucose and HbA1c were significantly higher in subjects with positive FH ([Table 1](#pone.0215302.t001){ref-type="table"}, [Fig 1](#pone.0215302.g001){ref-type="fig"}). ![The association between analyzed variables and family history of premature coronary artery disease---univariable analysis.](pone.0215302.g001){#pone.0215302.g001} 10.1371/journal.pone.0215302.t001 ###### The comparison of demographics, anthropometric factors and laboratory findings between young healthy male adults with and without family history of premature coronary artery disease. ![](pone.0215302.t001){#pone.0215302.t001g} ----------------------------------------------------------------------------------------------------------------------------------------- Variable Total sample Family history of P-CAD p-value ------------------------------------------ --------------------------- ------------------------- -------------- -------------- ---------- N (%) 411 (100) 223 (54.3) 188 (45.7) Age \[years\] 28.5 ± 4.3 28.8 ± 4.5 28.0 ± 4.0 0.03 Current vs. never and former smoking (%) 103 (25.1) 71 (31.8) 32 (17.0) 0.0009 Physical examination BMI \[kg/m^2^\] 25.9 ± 4.0 26.6 ± 4.0 24.9 ± 3.8 \<0.0001 WHR 0.89 ± 0.09 0.91 ± 0.08 0.87 ± 0.08 \<0.0001 SBP \[mmHg\] 131.2 ± 14.1 131.0 ± 13.8 131.4 ± 14.5 0.72 DBP \[mmHg\] 80.0 ± 11.0 79.5 ± 11.5 80.6 ± 10.4 0.47 Laboratory test results Cholesterol \[mmol/L\] 5.1 ± 1.2 5.2 ± 1.2 4.9 ± 1.1 0.005 LDL-C \[mmol/L\] 3.2 ± 1.0 3.4 ± 1.0 3.0 ± 0.9 \<0.0001 Apolipoprotein B \[g/L\] 1.0 ± 0.3 1.01± 0.29 0.96 ± 0.29 0.08 HDL-C \[mmol/L\] 1.4 ± 0.4 1.4 ± 0.4 1.5 ± 0.4 0.13 Apolipoprotein A1 \[g/L\] 1.5 ± 0.3 1.5 ± 0.2 1.5 ± 0.3 0.27 HDL \[%\]\ 29.2 ± 9.5\ 27.9 ± 9.4\ 30.7 ± 9.5\ 0.002\ Triglycerides \[mmol/L\] 1.4 ± 1.4 1.5 ± 1.6 1.2 ± 1.1 0.03 Lp(a) \[nmol/L\] 42.8 ± 66.2 49.9 ± 72.7 34.5 ± 56.6 0.03 Uric acid \[μmol/L\] 349.8 ± 66.5 352.3 ± 65.3 346.8 ± 67.8 0.41 Albumin \[g/L\] 48.6 ± 2.5 48.3 ± 2.5 48.9 ± 2.5 0.03 hsCRP \[mg/dl\] 1.3 ± 1.3 1.4 ± 1.4 1.2 ±1.2 0.08 Glucose \[mmol/l\] 5.1 ± 0.5 5.2 ± 0.4 5.0 ± 0.5 \<0.0001 HbA1c \[%\] 5.05 ± 0.26 5.11 ± 0.25 4.97 ± 0.25 \<0.0001 Fibrinogen \[mg/dL\] 263.9 ± 58.0 271.7 ± 52.8 254.6 ± 62.5 \<0.0001 ----------------------------------------------------------------------------------------------------------------------------------------- Abbreviations: BMI--body mass index; DBP--diastolic blood pressure; HDL-C--high-density lipoprotein cholesterol; hsCRP--high-sensitivity C-reactive protein; LDL-C--low-density lipoprotein cholesterol; Lp(a)--lipoprotein(a); P-CAD--premature coronary artery disease; WHR--waist-to-hip ratio. Sex hormones {#sec009} ------------ There were no significant differences in terms of testosterone, FSH and estradiol levels between the young healthy male subjects with and without FH of P-CAD. On the other hand, subjects with positive FH had lower SHBG levels and higher LH and progesterone levels, compared with the controls ([Table 2](#pone.0215302.t002){ref-type="table"}). 10.1371/journal.pone.0215302.t002 ###### The comparison of hormone levels between young healthy male adults with and without family history of premature coronary artery disease. ![](pone.0215302.t002){#pone.0215302.t002g} Variable Total sample Family history of P-CAD p-value -------------------- ------------------------- ------------------------- ------------- ------------- ---------- Pituitary hormones TSH \[μlU/mL\] 1.9 ± 0.8 1.9 ± 0.8 2.0 ± 0.8 0.32 FSH \[mlU/mL\] 4.3 ± 2.7 4.3 ± 2.7 4.3 ± 2.8 0.92 LH \[mlU/mL\] 4.7 ± 2.0 4.9 ± 2.1 4.5 ± 1.8 0.02 Gonadal hormones Testosterone \[nmol/L\] 15.7 ± 5.8 16.0 ± 6.0 15.4 ± 5.6 0.52 Estradiol \[pg/mL\] 21.4 ± 10.9 21.6 ± 9.8 21.1 ± 12.1 0.23 Progesterone \[ng/mL\] 0.37 ± 0.31 0.5 ± 0.31 0.22 ± 0.23 \<0.0001 Other SHBG \[nmol/L\] 32.9 ± 15.7 30.9 ± 12.6 35.2 ± 18.5 0.007 Abbreviations: FSH--follicle-stimulating hormone; LH---luteinizing hormone; P-CAD--premature coronary artery disease; SHBG--sex hormone binding globulin; TSH---thyroid-stimulating hormone. The correlation of hormone levels with biomarkers associated with the risk of CAD {#sec010} --------------------------------------------------------------------------------- Both total testosterone and SHBG were inversely correlated with the following risk factors for CAD: BMI, WHR, total and LDL-C, apolipoprotein B, uric acid and fibrinogen levels. Additionally, a significant, positive correlation was found between testosterone levels and HDL-C, HDL%, and apolipoprotein A1 ([Table 3](#pone.0215302.t003){ref-type="table"}). We did not detect any significant correlation between risk factors for CAD and FSH or LH, except for a weak correlation between age and LH ([Table 3](#pone.0215302.t003){ref-type="table"}). Progesterone level was correlated with factors associated with insulin resistance, including WHR, fasting glucose and HbA1c levels. These correlations, although weak (r-- 0.17, r-- 0.14 and r-- 0.24, respectively), were highly significant (p = 0.0005, p = 0.005 and p\<0.0001, respectively), ([Table 3](#pone.0215302.t003){ref-type="table"}). 10.1371/journal.pone.0215302.t003 ###### The correlations between hormone levels and age, anthropometric measurements and other laboratory findings. ![](pone.0215302.t003){#pone.0215302.t003g} ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Variable Correlations with FSH Correlations with LH Correlations with estradiol Correlations with\ Correlations with testosterone Correlation with SHBG progesterone ----------------------------- ----------------------------------------------- ---------------------- ---------------------------------------------- ------------------------------------------------ -------------------------------------------------- -------------------------------------------------- -------------------------------------------------- Age 0.19[\\\](#t003fn004){ref-type="table-fn"} 0.03 -0.14[\\](#t003fn003){ref-type="table-fn"} -0.004 -0.22[\\\\](#t003fn005){ref-type="table-fn"} -0.10[\](#t003fn002){ref-type="table-fn"} Anthropometric measurements BMI -0.06 -0.09 -0.01 0.01 -0.41[\\\\](#t003fn005){ref-type="table-fn"} -0.46[\\\\](#t003fn005){ref-type="table-fn"} WHR 0.08 -0.01 -0.04 0.18[\\\](#t003fn004){ref-type="table-fn"} -0.26[\\\\](#t003fn005){ref-type="table-fn"} -0.29[\\\\](#t003fn005){ref-type="table-fn"} Other laboratory findings Total cholesterol 0.06 -0.07 -0.15[\\](#t003fn003){ref-type="table-fn"} -0.02 -0.21[\\\\](#t003fn005){ref-type="table-fn"} -0.20[\\\\](#t003fn005){ref-type="table-fn"} LDL-C 0.03 -0.09 -0.12[\](#t003fn002){ref-type="table-fn"} 0.03 -0.20[\\\\](#t003fn005){ref-type="table-fn"} -0.24[\\\\](#t003fn005){ref-type="table-fn"} Apolipoprotein B 0.03 -0.09 -0.14[\\](#t003fn003){ref-type="table-fn"} -0.09 -0.26[\\\\](#t003fn005){ref-type="table-fn"} -0.26[\\\\](#t003fn005){ref-type="table-fn"} HDL-C 0.04 0.03 0.07 0.04 0.26[\\\\](#t003fn005){ref-type="table-fn"} 0.34[\\\\](#t003fn005){ref-type="table-fn"} Apolipoprotein A1 0.08 0.05 -0.04 0.06 0.11[\](#t003fn002){ref-type="table-fn"} 0.13[\\](#t003fn003){ref-type="table-fn"} HDL \[%\] -0.01 0.06 0.15[\\](#t003fn003){ref-type="table-fn"} 0.02 0.33[\\\\](#t003fn005){ref-type="table-fn"} 0.39[\\\\](#t003fn005){ref-type="table-fn"} Triglycerides 0.01 -0.05 -0.10[\](#t003fn002){ref-type="table-fn"} 0.02 -0.36[\\\\](#t003fn005){ref-type="table-fn"} -0.48[\\\\](#t003fn005){ref-type="table-fn"} Lp(a) 0.01 0.07 -0.05 -0.03 -0.09 -0.09 hsCRP 0.06 -0.01 -0.04 0.02 -0.30[\\\\](#t003fn005){ref-type="table-fn"} -0.29[\\\\](#t003fn005){ref-type="table-fn"} Glucose 0.07 0.05 -0.17[\\\](#t003fn004){ref-type="table-fn"} 0.14[\\](#t003fn003){ref-type="table-fn"} -0.16[\\](#t003fn003){ref-type="table-fn"} -0.15[\\](#t003fn003){ref-type="table-fn"} HbA1c -0.07 -0.07 -0.01 0.24[\\\\](#t003fn005){ref-type="table-fn"} -0.21[\\\\](#t003fn005){ref-type="table-fn"} -0.21[\\\\](#t003fn005){ref-type="table-fn"} Uric acid 0.02 -0.05 -0.01 -0.05 -0.28[\\\\](#t003fn005){ref-type="table-fn"} -0.36 Fibrinogen -0.01 -0.04 -0.06 0.07 -0.22[\\\\](#t003fn005){ref-type="table-fn"} -0.19[\\\\](#t003fn005){ref-type="table-fn"} Albumin -0.02 -0.03 -0.10[\](#t003fn002){ref-type="table-fn"} -0.07 0.02 0.15[\\](#t003fn003){ref-type="table-fn"} ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- p-value \\<0.05 \\* \<0.01 \\\* \<0.001 \\\\\<0.0001. Abbreviations: BMI--body mass index; FSH--follicle-stimulating hormone; HDL-C--high-density lipoprotein cholesterol; hsCRP--high-sensitivity C-reactive protein; LDL-C--low-density lipoprotein cholesterol; LH---luteinizing hormone; Lp(a)--lipoprotein(a); SHBG--sex hormone binding globulin; WC--waist circumference; WHR--waist-to-hip ratio. Multivariable analysis of the association between analyzed parameters and FH of P-CAD {#sec011} ------------------------------------------------------------------------------------- According to the multivariable model that was selected based on the information criteria, the factors independently associated with FH of P-CAD were progesterone, LH, BMI, WHR, HbA1c, and LDL-C. There was also a borderline significant trend regarding the relationship between testosterone and Lp(a) with positive FH of P-CAD ([Fig 2](#pone.0215302.g002){ref-type="fig"}). ![The association between analyzed variables and family history of premature coronary artery disease---multivariable analysis.](pone.0215302.g002){#pone.0215302.g002} Discussion {#sec012} ========== Progesterone in men is synthesized in the testicles and adrenal glands, and its serum level, in contrast to that of women, does not alternate periodically or with age \[[@pone.0215302.ref014]--[@pone.0215302.ref017]\]. Higher levels of progesterone in the offspring of patients with P-CAD detected in the crude comparison, as well as the results of the multivariable analysis, indicate a potential association between progesterone levels and FH of P-CAD. To the best of our knowledge, no previous studies have analyzed the association between endogenous progesterone and cardiovascular disease in humans. The relationship between progesterone and the pathophysiology of CAD was the subject of a few experimental studies on animal models. In 1996, Hanke et al. demonstrated that in the rabbit model of atherosclerosis, progesterone therapy inhibited the beneficial effect of estrogen therapy on plaque size, likely by affecting arterial sex hormone receptors. Moreover, in the same study, therapy with progesterone alone was associated with considerably increased intimal thickening, compared with the castrated control rabbits \[[@pone.0215302.ref014]\]. Recently, Yang et al. demonstrated that progesterone induces the expression of macrophage CD36, a receptor for oxidized LDL that enhances foam cell formation. This effect was shown to be mediated by the transcription factor peroxisome proliferator-activated receptor γ \[[@pone.0215302.ref018]\]. Additionally, Wassmann et al. demonstrated that progesterone antagonizes the vasoprotective effect of estrogen by reducing the expression of extracellular superoxide dismutase and activating NADPH oxidase. Therefore, progesterone increases the production of reactive oxygen species in vascular smooth muscle cells \[[@pone.0215302.ref019]\]. In our study, progesterone levels were significantly correlated with WHR, fasting glucose and HbA1c, which are associated with insulin resistance. The role of progesterone in the regulation of glucose levels has been known since the 1980s \[[@pone.0215302.ref015]\] when it was demonstrated that progesterone increases glucose levels by inhibiting insulin-mediated glucose uptake and metabolism in adipocytes and skeletal muscles \[[@pone.0215302.ref015],[@pone.0215302.ref020]--[@pone.0215302.ref023]\]. Most of the published studies analyzing the role of progesterone in the development of insulin resistance have concerned gestational diabetes; therefore, for obvious reasons, the results of these studies cannot be directly translated to men \[[@pone.0215302.ref023],[@pone.0215302.ref024]\]. Nonetheless, insulin resistance is another potential mechanism through which progesterone levels may affect the risk of cardiovascular disease. Testosterone is the central male sex hormone, which is produced primarily by Leydig cells in the testicles \[[@pone.0215302.ref025]--[@pone.0215302.ref027]\]. In serum, testosterone is primarily transported bound to proteins, and two-thirds of the bound testosterone is bound to SHBG \[[@pone.0215302.ref026],[@pone.0215302.ref028]\]. The production of testosterone by Leydig cells is stimulated by LH, which is synthesized in the anterior pituitary gland \[[@pone.0215302.ref029]\]. Similar to previous studies, in the young, healthy, male subjects included in our analysis, testosterone and SHBG were significantly inversely correlated with fasting glucose, HbA1c, WHR, BMI, lipid parameters (total cholesterol, LDL-C, apolipoprotein B, triglycerides), as well as inflammatory biomarkers \[[@pone.0215302.ref025],[@pone.0215302.ref028]--[@pone.0215302.ref031]\]. In observational studies it has been shown that men with lower testosterone levels are at higher risk for cardiovascular disease, and that subject on treatment with testosterone replacement therapy have a decreased risk of adverse cardiovascular events \[[@pone.0215302.ref028],[@pone.0215302.ref032],[@pone.0215302.ref033]\]. Nonetheless, observational studies have their limitations i.e. they are burdened with reverse causality bias \[[@pone.0215302.ref034]\]. Although some promising findings have originated from small randomized clinical trials (RCTs), which indicated an improvement of lipid parameters, fasting glucose and a decrease in body weight in men with hypogonadism treated with testosterone replacement therapy (TRT) \[[@pone.0215302.ref035],[@pone.0215302.ref036]\], no beneficial effects of TRT in terms of cardiovascular disease have been demonstrated \[[@pone.0215302.ref036]\], with one RCT, discontinued because of the significantly higher incidence rate of adverse cardiovascular events in the TRT group \[[@pone.0215302.ref037]\]. In opposite to observational studies, mendelian randomization analyses demonstrated a potential, causative relationship between higher endogenous testosterone levels and unfavorable lipid profiles, type 2 diabetes, hypertension, and adverse cardiovascular events \[[@pone.0215302.ref034],[@pone.0215302.ref038],[@pone.0215302.ref039]\]. In our study, we have shown for the first time that in young healthy male adults, LH is independently associated with FH of P-CAD. Moreover, there was a borderline significant relationship between higher testosterone levels and FH of P-CAD in the multivariable analysis. Other factors independently associated with FH of P-CAD included a higher BMI and WHR as well as higher LDL and HbA1C levels, which was in agreement with previous studies and may be the result of both hereditary burden and environmental and lifestyle factors \[[@pone.0215302.ref008],[@pone.0215302.ref040],[@pone.0215302.ref041]\]. Study limitations {#sec013} ----------------- Subject with FH of p-CAD, similarly as in other reports, were different form subjects without FH of p-CAD, with regard to BMI and other risk factors of atherosclerosis \[[@pone.0215302.ref008],[@pone.0215302.ref040],[@pone.0215302.ref041]\]. Taking into account the correlation between progesterone and factors associated with insulin resistance and to exclude the potential secondary effects of body weight differences, we addressed this issue in the multivariable analysis, which revealed that BMI was independently from progesterone associated with FH of p-CAD. Additionally, we compared progesterone levels between test group and controls in the following subgroups: normal body weight (BMI 18.5 to 24.9 kg/m^2^), overweight (BMI 25.0 to 29.9 kg/m^2^) and obesity (≥30.0 kg/m^2^). In all analyzed subgroups stratified by BMI, subjects with positive FH of P-CAD had significantly higher progesterone level compared with controls ([S2 Table](#pone.0215302.s002){ref-type="supplementary-material"}---Supporting Information). The primary aim of our study was to compare wide range of biochemical marker levels between subjects with and without FH of P-CAD. Male and female sex hormones were measured in both sexes mainly because of logistic reasons, and we did not aim to test any specific hypothesis regarding sex hormones. Results regarding highly significant association between progesterone and FH of P-CAD, even after adjustment for classical risk factors, although surprising also to our team, are in line with the results of experimental studies. Considering the case-control design of our study, they should be interpreted as hypothesis-generating. Conclusions {#sec014} =========== Progesterone is significantly corelated with markers for insulin resistance in young healthy male adults. Both progesterone and LH levels are significantly associated with FH of P-CAD, independent of traditional risk factors for CAD. Supporting information {#sec015} ====================== ###### List of reagent kits. (DOCX) ###### Click here for additional data file. ###### The comparison of progesterone levels between subjects with and without family history of premature coronary artery disease, across BMI categories. (DOCX) ###### Click here for additional data file. ###### Dataset (raw data). Study dataset. (XLSX) ###### Click here for additional data file. ###### Dataset (imputed data). Dataset with imputed values using random forest algorithm (used for multivariable analysis). (XLSX) ###### Click here for additional data file. [^1]: Competing Interests:The authors have declared that no competing interests exist.	{ "pile_set_name": "PubMed Central" }
Background ========== Genomic imprinting refers to the phenomenon of unequal expression of a heterozygous genotype depending on which parent (father or mother) the imprinted variant is inherited from. It is estimated that approximately 1% of all mammalian genes are imprinted (<http://igc.otago.ac.nz>). Among these imprinted genes, Beckwith-Wiedemann syndrome, Silver-Russell syndrome, Angelman syndrome, and Prader-Willi syndrome are the best known. Numerous methods have been proposed to detect imprinting effects. For a diallelic genetic marker locus, the parental-asymmetry test (PAT) that considers imbalance of parental origins of the variant allele is simple and powerful. A series of generalizations of PAT, such as the pedigree-parental-asymmetry test (PPAT) for general pedigree data, widen its practical range \[[@B1]\]. However, these tests use only information on affected offspring and their parents. Recently, PATu \[[@B2]\] was proposed to take unaffected offspring in a nuclear family into consideration, making fuller use of data to improve power. In this study, we propose a novel parent-of-origin effects test, PPATu, that uses both affected and unaffected offspring in general pedigrees, and apply the method to the Genetic Analysis Workshop 18 (GAW18) data, consisting of 20 large pedigrees, to study the hypertensive phenotype. Previous studies revealed the possible involvement of imprinted genes in hypertension \[[@B3],[@B4]\]. The GAW18 data thus provide us the opportunity to further substantiate such findings. Methods ======= Suppose that the marker of interest has 2 alleles, M~1~and M~2~, and the disease allele is more likely to be associated with marker allele M~1~. Let 0, 1, and 2 represent the marker genotypes M~2~M~2~, M~1~M~2~, and M~1~M~1~, respectively. For a child-parents trio, let F, M, and C denote the marker genotypes of the father, mother, and child, respectively. Throughout this article, mating symmetry is assumed; that is, P(F=f, M=m)=P(F=m, M=f) for all f, m = 0,1,2. We also assume that there is no maternal effect; that is, the maternal genotype does not confer additional risk on the child\'s phenotype. Suppose we have Nindependent pedigrees, and for the i^th^pedigree, we have nu~i~unaffected and na~i~affected offspring. Define $S = \sum\limits_{i = 1}^{N}\left\lbrack {\sum\limits_{j = 1}^{na_{i}}\left( {{I_{F_{ij}}}_{> M_{ij},C_{ij = 1}} - I_{F_{ij} < M_{ij},C_{ij} = 1}} \right) - \sum\limits_{k = 1}^{nu_{i}}\left( {I_{F_{ik} > M_{ik = 1},C_{ik = 1}} - I_{F_{ik} < M_{ik},C_{ik = 1}}} \right)} \right\rbrack$ where Iis the usual indicator function. We can prove that under the null hypothesis of no imprinting effect, E(S) = 0. The unbiased estimator of the variance of Sis $$\hat{V}\left( S \right) = \sum\limits_{i = 1}^{N}\left\lbrack {\sum\limits_{j = 1}^{na_{i}}\left( {{I_{F_{ij}}}_{> M_{ij},C_{ij = 1}} - I_{F_{ij} < M_{ij},C_{ij} = 1}} \right) - \sum\limits_{k = 1}^{nu_{i}}\left( {I_{F_{ik} > M_{ik},C_{ik = 1}} - I_{F_{ik} < M_{ik},C_{ik = 1}}} \right)} \right\rbrack^{2}$$ The standardized test statistic $PPATu = \frac{S}{\sqrt{\hat{V}\left( S \right)}}$ follows the N(0, 1) distribution asymptotically. When there is maternal imprinting effect, PPATu will be positive; when there is paternal imprinting effect, it will be negative. Note that the contributions from trios in a pedigree are not independent, and their correlations are accounted for in the variance. In our simulation study and application below, we compare the performance of PPATu with PPAT, whose statistic is defined without the negative terms in the Sstatistic; that is, without utilizing information on trios with unaffected offspring. More specifically, $PPAT = \frac{\sum\limits_{i = 1}^{N}\sum\limits_{j = 1}^{na_{i}}\left( {I_{F_{ij} > M_{ij},C_{ij} = 1} - I_{F_{ij} < M_{ij},C_{ij} = 1}} \right)}{\sqrt{\sum\limits_{i = 1}^{N}\left\lbrack {\sum\limits_{j = 1}^{na_{i}}\left( {I_{F_{ij} > M_{ij},C_{ij} = 1} - I_{F_{ij} < M_{ij},C_{ij} = 1}} \right)} \right\rbrack^{2}}}$ Results ======= Simulation study ---------------- To evaluate the power of the proposed statistic and to compare with PPAT, we carried out a simulation study under 9 different settings, combinations of 3 sets of haplotype frequencies (H1, H2, H3) and 3 imprinting models (I1, I2, I3) (Table [1](#T1){ref-type="table"}). Our data were simulated based on the general pedigrees from GAW18; their sizes are described below. To gauge the type I error rate, we also considered 9 additional settings, combinations of the same 3 sets of haplotype frequencies and 3 no-imprinting models (N1, N2, N3), also given in Table [1](#T1){ref-type="table"}. We simulated 1000 replicates under each of the settings (a total of 18 combinations). The results are plotted in Figure [1](#F1){ref-type="fig"}, which shows that the empirical type I error rates, at the 0.01 nominal significance level, are all well controlled for both PPAT and PPATu. On the other hand, PPATu is clearly more powerful in all settings, especially when there is a substantial imprinting effect. Conclusions are the same for significance levels 0.05 and 0.005 as well; consequently,the results are not shown for brevity. ###### Combinations of 9 imprinting settings and 9 no-imprinting settings Haplotype frequency^a^ Imprinting model^b^ No-imprinting model^b^ ---- ------------------------ ----- --------------------- ----- ------------------------ ------ ------ ------ ------ ---- ------ ------ ------ H1 0.2 0.0 0.1 0.7 I1 0.26 0.28 0.37 0.39 N1 0.26 0.33 0.39 H2 0.3 0.1 0.0 0.6 I2 0.24 0.26 0.42 0.44 N2 0.24 0.34 0.44 H3 0.3 0.0 0.0 0.7 I3 0.18 0.23 0.53 0.58 N3 0.18 0.38 0.58 ^a^Haplotype frequencies are between disease susceptibility locus(with disease allele D and normal allele d) and marker locus (alleles M1 and M2). ^b^Imprinting models are defined in terms of penetrance probabilities: (ϕ~d/d~, ϕ~d/D~, ϕ~D/d~, ϕ~D/D~). For the no-imprinting models,ϕ~D/d~= ϕ~d/D~. ![Type I error and power for PPATu and PPAT. The solid horizontal line marks the nominal significance level of 0.01. For power, the 9 settings are the 9 combinations of haplotype frequencies and imprinting models (as given in Table 1): 1 = (H1, I1), 2 = (H2, I1), 3 = (H3, I1), 4 = (H1, I2), 5 = (H2, I2), 6 = (H3, I2), 7 = (H1, I3), 8 = (H2, I3), and 9 = (H3, I3). For type I error, the 9 settings are the 9 combinations of haplotype frequencies and the no-imprinting models (as given in Table [1](#T1){ref-type="table"}): 1 = (H1, N1), 2 = (H2, N1), 3 = (H3, N1), 4 = (H1, N2), 5 = (H2, N2), 6 = (H3, N2), 7 = (H1, N3), 8 = (H2, N3), and 9 = (H3, N3).](1753-6561-8-S1-S52-1){#F1} GAW18 data analysis ------------------- We consider GAW18 real genome-wide association studies pedigree data that contain a total of 472,049 single-nucleotide polymorphism (SNP) genotypes on odd chromosomes and phenotype data, including systolic and diastolic blood pressure (SBP and DBP). In our study, we use a hypertensive binary phenotype; an individual is classified as affected if SBP \> 140 mm Hg, or DBP \> 90 mm Hg, or on antihypertensive medication at the first examination. There are 20 pedigrees; the sizes range from 27 to 107 individuals. In total, there are only 157 affected offspring, while there are 709 unaffected ones. Hence, based on the experience gained in our simulation, we expect a substantial gain in information for PPATu that makes use of information from both affected and unaffected individuals when compared to PPAT. To reduce the effect of multiple testing, we first used pedigree disequilibrium test (PDT \[[@B5]\]) to identify SNPs that are associated with hypertension at the 0.05 level, and then performed imprinting effect tests, focusing only on those SNPs. In our analysis, all trios with complete data within each pedigree were included in computing the test statistic. Furthermore, although many tests were performed, we did not attempt to correct for multiple testing given the small sample size (a total of only 20 pedigrees). Table [2](#T2){ref-type="table"} shows the cross-classification of SNPs by different tests. We also provide Figure [2](#F2){ref-type="fig"}, which shows the results for a combination of significant levels. As one can see from Figure [2](#F2){ref-type="fig"} and Table [2](#T2){ref-type="table"}, PPATu detected more SNPs with smaller pvalues than PPAT. For SNPs detected by both methods, the results for PPATu are more significant. Table [3](#T3){ref-type="table"} shows the 2 SNPs (rs12947636 and rs1674137) that are significant at the 0.01 level for both PDT and PPATu. It appears that these 2 SNPs are novel, as they were not previously identified as associated with hypertension, thus further study is warranted to substantiate the finding. ###### Cross-classification of results based on pvalues fromassociation(PDT) andimprinting (PPAT and PPATu) tests PPAT PPATu -------- -------- -------- -------- -- -------- -------- -------- PDT \<0.05 \<0.01 \>0.05 \<0.05 \<0.01 \>0.05 \<0.05 245 1 12,752 556 59 12,441 \<0.01 17 0 798 30 2 785 \>0.05 7008 35 14,681 15,007 1328 6682 ![*Significant imprinting results (pvalue \< 0.05) for SNPs having small pvalues for association test (815 SNPs with pvalue \< 0.01) across all odd chromosomes. (Top) Based on PPAT that uses only affected off spring and their parent data (17 SNPs identified). (Bottom) Based on PPATu that uses both affected and unaffected off spring (30 SNPs identified).](1753-6561-8-S1-S52-2){#F2} ###### SNPs having pvalues \<0.05 for both association and imprinting tests SNP Chr Gene Position P\_~PDT~ P\_~PPAT~ P\_~PPATu~ Imprinting ------------ ----- ------------ ---------- ---------- ----------- ------------ ------------ rs12947636 17 SLC39A11* 70992565 0.0090 0.6171 0.0080 maternal rs1674137 19 TSKS 50258030 0.0089 0.6547 0.0038 maternal rs11606492 11 PLEKHA7 17000241 0.0488 0.1797 0.0176 paternal The first 2 SNPs are significant at the 0.01 level for PDT and PPATu. The third SNP is significant at the 0.05 level for PDT and PPATu, and is within the PLEKHA7gene previously identified as associated with hypertension. To gain a more global view of the extent of the role of imprinting in hypertension, we also carried out genome-wide testing for imprinting effects without restricting to SNPs with small pvalues from PDT. The analysis has resulted in larger sets with pvalues smaller than 0.05, which we call \"significant\"for easy reference. A search in the NCBI database found 49 genes implicated to be associated with hypertension in previous studies. Figure [3](#F3){ref-type="fig"} gives the numbers of SNPs with significant imprinting or association effects within these genes. We can see that more SNPs identified by PPATu are within these genes than those identified by PPAT. Specifically, 47 and 15 SNPs within these genes are identified as having significant imprinting effects by PPATu and PPAT, respectively. In fact, the proportion of SNPs identified by PPATu that are in known genes is higher (1.5 times) than that identified by PPAT, although this difference is not substantial. Other than 1 SNP, the rest did not reach the threshold of significance for PDT, which may be explained by the power loss of association tests like PDT that do not account for the imprinting effect properly. The 1 SNP (rs11606492; Table [3](#T3){ref-type="table"}) that yielded significant results from both PDT and PPATu (but not PPAT) is within the gene PLEKHA7that has been implicated to be associated with hypertension in previous studies \[[@B6],[@B7]\]. ![*Number of SNPs with significant imprinting or association effects (pvalue \< 0.05) that are within genes previously identified as associated with hypertension. The pink, yellow, and blue circles denote results from PDT, PPAT, and PPATu, respectively.](1753-6561-8-S1-S52-3){#F3} Conclusions =========== In this article, we propose an imprinting test that utilizes both affected and unaffected individuals from general pedigrees, the type of data provided by GAW18. We expect PPATu to be more powerful than the existing test PPAT \[[@B1]\] because the former makes full use of information by taking unaffected offspring into consideration. Indeed, our simulation study shows that PPATu has higher power than PPAT without an elevated type I error rate based on the GAW18 pedigrees. Our results from analysis of the GAW18 data using PPATu leads to the identification of a number of SNPs that are within genomic regions previously implicated for the hypertensive phenotype. Nevertheless, further investigation is warranted especially to evaluate the performance of the methods under different study designs and ascertainment criteria. Competing interests =================== The authors declare that they have no competing interests. Authors\' contributions ======================= LS designed the overall study; FZ conducted statistical analyses and drafted the manuscript. Both authors wrote, read, and approved the final manuscript. Acknowledgements ================ The authors would like to acknowledge the NIH grant that supports GAWs and the GAW18 data providers. This work was supported in part by NSF grant DMS 1208928. The GAW18 whole genome sequence data were provided by the T2D-GENES Consortium, which is supported by NIH grants U01 DK085524, U01 DK085584, U01 DK085501, U01 DK085526, and U01 DK085545. The other genetic and phenotypic data for GAW18 were provided by the San Antonio Family Heart Study and San Antonio Family Diabetes/Gallbladder Study, which are supported by NIH grants P01 HL045222, R01 DK047482, and R01 DK053889. The Genetic Analysis Workshop is supported by NIH grant R01 GM031575. This article has been published as part of BMC Proceedings*Volume 8 Supplement 1, 2014: Genetic Analysis Workshop 18. The full contents of the supplement are available online at <http://www.biomedcentral.com/bmcproc/supplements/8/S1>. Publication charges for this supplement were funded by the Texas Biomedical Research Institute.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-ijms-18-02486} =============== Gut-brain orexigenic peptide ghrelin \[[@B1-ijms-18-02486]\], a natural ligand of the growth hormone secretagogue receptor (GHS-R1A), has been recently shown to play a critical role in food reward \[[@B2-ijms-18-02486]\] as well as reward, motivation and intake of alcohol and reward of several stimulants (for review, see \[[@B3-ijms-18-02486],[@B4-ijms-18-02486]\]). In addition to hypothalamus, the central GHS-R1As are expressed in important reward related areas including striatum, nucleus accumbens (NAC), amygdala, prefrontal cortex, hippocampus, and ventral tegmental area (VTA) \[[@B5-ijms-18-02486],[@B6-ijms-18-02486],[@B7-ijms-18-02486],[@B8-ijms-18-02486],[@B9-ijms-18-02486],[@B10-ijms-18-02486],[@B11-ijms-18-02486],[@B12-ijms-18-02486]\]. Available literature involving ghrelin in opioid abuse and addiction is still limited and inconclusive \[[@B13-ijms-18-02486],[@B14-ijms-18-02486],[@B15-ijms-18-02486],[@B16-ijms-18-02486],[@B17-ijms-18-02486]\]. The self-administration study of Maric et al. \[[@B13-ijms-18-02486]\] showed that ghrelin administered intracerebroventricularly (i.c.v.) could increase heroin intake, however, pre-treatment with a peptide GHS-R1A antagonist (i.c.v.) did not affect heroin self-administration. However, we have shown in our earlier study \[[@B16-ijms-18-02486]\] that premedication with the GHS-R1A antagonist, a triazole non-peptidic derivative JMV2959 \[[@B18-ijms-18-02486]\], significantly and dose-dependently reduced morphine-induced dopamine efflux in the nucleus accumbens shell (NACSh), a brain structure that is crucially important for drug reward mediation \[[@B19-ijms-18-02486],[@B20-ijms-18-02486]\], and attenuated behavioural stimulation, particularly stereotypical behaviours induced by morphine. This was subsequently confirmed in mice \[[@B15-ijms-18-02486]\]. Engel et al. \[[@B15-ijms-18-02486]\] also described that JMV2959 significantly reduced expression of morphine-induced conditioned place preference (CPP) in mice and we have recently documented the same in rats \[[@B21-ijms-18-02486]\]. Conclusively, this suggests a significant involvement of central ghrelin system in changes induced by morphine/opioids in the mesolimbic dopaminergic system, changes which are associated with processing of neural reward. The opioid/µ-receptor agonists' rewarding reinforcing properties are traditionally associated with opioid-induced accumbens dopamine efflux \[[@B22-ijms-18-02486],[@B23-ijms-18-02486],[@B24-ijms-18-02486],[@B25-ijms-18-02486]\] caused by supressing γ-aminobutyric acid (GABA) release from VTA interneurons, which tonically inhibit mesolimbic dopamine neurons \[[@B26-ijms-18-02486],[@B27-ijms-18-02486],[@B28-ijms-18-02486]\]. However, intra-accumbens opioid administration also induced dopamine increase \[[@B29-ijms-18-02486],[@B30-ijms-18-02486],[@B31-ijms-18-02486]\] and inhibited accumbens fast amino-acid-mediated synaptic transmission \[[@B32-ijms-18-02486],[@B33-ijms-18-02486],[@B34-ijms-18-02486]\]. Opioids are self-administered into the NACSh as well as VTA \[[@B28-ijms-18-02486],[@B35-ijms-18-02486]\]. Thus, at least part of the opioid rewarding effects may be due to their direct or indirect effect on synaptic transmission in the NAC. Both dopaminergic and non-dopaminergic circuits can contribute to VTA and NAC opioid reward \[[@B28-ijms-18-02486],[@B36-ijms-18-02486]\]. The endogenous cannabinoid system provably significantly participates in reinforcing processes of opioids \[[@B37-ijms-18-02486],[@B38-ijms-18-02486],[@B39-ijms-18-02486],[@B40-ijms-18-02486],[@B41-ijms-18-02486],[@B42-ijms-18-02486]\]. Cannabinoid CB1 receptor (CB1) antagonists such as SR141716A (rimonabant) reduced the opioid rewarding effects in both conditioned place preference models \[[@B43-ijms-18-02486],[@B44-ijms-18-02486],[@B45-ijms-18-02486]\] and intravenous self-administration \[[@B24-ijms-18-02486],[@B44-ijms-18-02486],[@B46-ijms-18-02486],[@B47-ijms-18-02486]\], whereas δ-9-tetrahydrocannabinol (THC), a CB1 agonist, increased the reinforcing effects of the intravenously self-injected heroin \[[@B48-ijms-18-02486]\]. Caille and Parsons \[[@B49-ijms-18-02486]\] documented that intravenous self-administration of heroin was significantly attenuated by CB1 antagonist SR141716A infused into the NACSh, thence CB1 receptors possibly modulate reward of opioids through the ventral striatopallidal projection. There is also evidence of cannabinoid--opioid interaction mediated by activation of presynaptic CB1 and µ-receptors (colocalization or heterodimerization) of the same or linked accumbens neurons, particularly NACSh, similar to the neurons within the VTA \[[@B37-ijms-18-02486],[@B50-ijms-18-02486]\]. However, systemic SR141716A pre-treatment had no significant effect on morphine-induced dopamine increases in the NACSh \[[@B46-ijms-18-02486],[@B51-ijms-18-02486]\]. It follows that CB1 receptors in the NACSh significantly modulate opioid reinforcement through dopamine-independent mechanisms \[[@B52-ijms-18-02486]\]. The contemplated role of the CB2 recently discovered in the VTA in reward is not yet clear \[[@B53-ijms-18-02486]\]. It has also been described \[[@B52-ijms-18-02486]\] that during the heroin self-administration N-arachidonoylethanolamine (anandamide, AEA) levels were significantly enhanced and simultaneously 2-arachidonoylglycerol (2-AG) levels significantly decreased in the dialysates from the NACSh. This is in accordance with our previous microdialysis study in rats \[[@B17-ijms-18-02486]\] where we observed a significant AEA increase and significant 2-AG decrease in the NACSh dialysates after acute morphine dose as well as when morphine was administered during prolonged abstinence from repeated morphine. In addition, Vigano et al. \[[@B54-ijms-18-02486]\] measurements of accumbens 2-AG and AEA contents post mortem after sub-chronic morphine administration in rats showed AEA increase and 2-AG decrease. Thus, the endocannabinoids input in the opioid motivational properties was supported, possibly particularly AEA increase induced by opioids in the NACSh contributes to opioid reward \[[@B52-ijms-18-02486]\]. The most-characterized endocannabinoids are anandamide (AEA) \[[@B55-ijms-18-02486]\] and 2-AG \[[@B56-ijms-18-02486],[@B57-ijms-18-02486]\]. AEA and 2-AG differ in many aspects, they are formed in various brain structures under different conditions and are uniquely affected by different stimuli, including pharmacological interference \[[@B58-ijms-18-02486],[@B59-ijms-18-02486],[@B60-ijms-18-02486],[@B61-ijms-18-02486]\]. Endocannabinoids are formed "on demand" and they act as retrograde messengers in the central nervous system (CNS) through activation of presynaptic CB1 receptors on both excitatory and inhibitory synapses \[[@B59-ijms-18-02486],[@B62-ijms-18-02486],[@B63-ijms-18-02486],[@B64-ijms-18-02486]\]. Endocannabinoids, which are released after depolarization in the NAC and from VTA dopaminergic neurons, possibly modulate glutamatergic and GABAergic afferents as retrograde messengers on different neuronal receptors \[[@B65-ijms-18-02486]\]. As explained earlier, opioids trigger the release of AEA in the NACSh, which is possibly participating in opioid reward in a dopamine-nondependent manner \[[@B52-ijms-18-02486]\], however, compounds that increased brain concentrations of AEA and prolonged AEA's effects did not influence heroin self-administration in rats, thus the importance of this release is still rather unclear \[[@B48-ijms-18-02486]\]. Thus far, we have even more limited knowledge about the role of 2-AG in opioid reward processes in the NACSh. In our previous study, we have documented a significant interaction between ghrelin and endocannabinoids in the morphine-induced changes in the NACSh \[[@B17-ijms-18-02486]\]. Premedication with JMV2959 before morphine, which was administered in acute doses or in a challenge dose during prolonged abstinence from chronic morphine, dose-dependently reversed the morphine-induced AEA increase in the NACSh leading to a significant AEA drop. JMV2959 intensified significantly the acute morphine-induced decrease in 2-AG concentrations and reduced challenge morphine induced 2-AG decrease in the NACSh. Besides our results, we have found several studies substantiating relevant interactions between ghrelin and cannabinoid systems in the food intake regulation by the brain-gut axis \[[@B66-ijms-18-02486],[@B67-ijms-18-02486],[@B68-ijms-18-02486],[@B69-ijms-18-02486],[@B70-ijms-18-02486],[@B71-ijms-18-02486],[@B72-ijms-18-02486]\]. Fentanyl, a 4-anilidopiperidin derivate synthetized in 1959 \[[@B73-ijms-18-02486]\], is generally considered as a µ-receptor-selective agonist about 100-fold more potent than morphine \[[@B74-ijms-18-02486]\], using active transportation through the blood--brain barrier \[[@B75-ijms-18-02486],[@B76-ijms-18-02486]\]. Fentanyl and new opioid synthetic derivatives have been recently increasingly abused in the USA, Canada and Europe \[[@B77-ijms-18-02486]\]. The goal of the present study was to establish whether the GHS-R1A antagonist, the substance JMV2959 could influence the fentanyl-induced effects on anandamide and 2-AG in the NACSh and specify the involvement of GHS-R1A located in the VTA and NAC. Opioid/morphine administered systemically or into the NAC can also stimulate accumbens GABA efflux \[[@B78-ijms-18-02486],[@B79-ijms-18-02486]\] and there is evidence about its contribution to the opioid reinforcing properties \[[@B29-ijms-18-02486],[@B80-ijms-18-02486]\]. It has been suggested that the simultaneous activation of µ and GABA-A receptors which are co-expressed on GABAergic interneurons significantly supress GABA efflux onto dopamine nerve endings, disinhibits dopamine neurons and enhances dopamine efflux \[[@B29-ijms-18-02486]\], thus elevated GABA concentrations potentiate opioid-induced dopamine release in the NAC. However, the VTA/NAC dopamine neurons can co-release other modulators such as endocannabinoids, GABA, glutamate, and variety of neuropeptides, which may contribute to the behavioural consequences of inhibiting or stimulating "dopamine" neurons. Various dopamine non-dependent mechanisms, where GABA is supposed to play an important role, also participate in opioid reinforcing processes \[[@B28-ijms-18-02486],[@B81-ijms-18-02486]\]. CB1 and µ receptors are located on GABA neurons in several brain areas including VTA and NAC \[[@B42-ijms-18-02486],[@B82-ijms-18-02486]\]. Recently, it has been suggested how important is the role of ghrelin system in regulation of GABAergic transmission in the central nucleus of amygdala (CeA) together with a complex interaction of ghrelin and ethanol at CeA GABAergic synapses \[[@B83-ijms-18-02486]\]. GHS-R1A activation also attenuated hypothalamic GABA release \[[@B84-ijms-18-02486]\]. Therefore, another goal of our present study was to test the influence of ghrelin antagonism on the fentanyl-induced accumbens GABA efflux and specify the involvement of GHS-R1A located in the VTA and NAC. To get a more complex picture, we have monitored also behavioural changes in rats during the microdialysis experiment, when ghrelin antagonist was administered intraperitoneally before fentanyl. 2. Results {#sec2-ijms-18-02486} ========== 2.1. The Effects of Growth Hormone Secretagogue Receptor (GHS-R1A) Receptor Antagonist on Fentanyl-Induced Accumbens Anandamide (AEA) Extracellular Concentration Increase {#sec2dot1-ijms-18-02486} -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ### 2.1.1. Pre-Treatment with Intraperitoneal JMV2959 {#sec2dot1dot1-ijms-18-02486} The influence of intraperitoneally administered ghrelin antagonist on fentanyl-induced increase of extracellular AEA in the NACSh is illustrated in [Figure 1](#ijms-18-02486-f001){ref-type="fig"}. Baseline levels of AEA did not significantly differ between animals in all presented experiments. As expected, acute systemic fentanyl (30 µg/kg subcutaneous---s.c.) administration evoked a statistically significant efflux of AEA in the NACSh. The two-way ANOVA for repeated measures (RM) followed by Bonferroni's multiple comparisons procedure has shown a significant group effect: saline + fentanyl 30 µg/kg vs. saline + saline group (F1,10 = 813.8, p \< 0.001) and time effect (F10,100 = 81.9, p \< 0.001); time course of AEA changes in the NACSh after saline/fentanyl injection differed significantly between the two groups of rats (time × group interaction, F10,100 = 86.8, p \< 0.001). The fentanyl-induced AEA increase reached the maximum effect 220% of baseline mean level 60 min after fentanyl administration (p \< 0.001). Pre-treatment with the GHS-R1A antagonist JMV2959 administered intraperitoneally (i.p.) turned the fentanyl-induced accumbens AEA increase and induced a significant decrease with the maximum drop 50% of baseline mean level. Thus, the JMV2959 pre-treatment effect was highly statistically significant: JMV2959 3 min/kg + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: effect of group F1,10 = 217.3, p \< 0.001; effect of time F10,100 = 7.9, p \< 0.001; time × group interaction F10,100 = 78.0, p \< 0.001. Observed changes within the JMV2959 pre-treatment group in comparison to baseline were also significant (p \< 0.001) (JMV2959 3 mg/kg + saline vs. saline + saline: effect of group F1,10 = 18.7, p \< 0.01; effect of time F10,100 = 14.9, p \< 0.001; time × group interaction F10,100 = 13.7, p \< 0.001). The JMV2959 pre-treatment induced decrease/reversal of accumbens AEA was observed within about 20--150 min after fentanyl administration, then the AEA levels crossed the baseline levels and reached a significant AEA increase with maximum 117% of baseline. A single dose of JMV2959 3 mg/kg i.p. had no effect on accumbens AEA and neither for saline i.p. ### 2.1.2. Pre-Treatment with JMV2959 Administered into the Ventral Tegmental Area (VTA) {#sec2dot1dot2-ijms-18-02486} [Figure 2](#ijms-18-02486-f002){ref-type="fig"}a illustrates the observed influence of ghrelin antagonist, administered into the VTA, on changes in accumbens AEA induced by 30 µg/kg s.c. fentanyl. The 30 µg/kg dose of fentanyl together with intra-VTA Ringer's solution induced practically the same AEA increase as the above described fentanyl with systemic saline: Ringer's solution/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + saline: effect of group F1,10 = 243.0, p \< 0.001; effect of time F9,90 = 55.1, p \< 0.001; time × group interaction F9,90 = 57.8, p \< 0.001; with maximum effect 217% of baseline level. JMV2959 pre-treatment into the VTA 5 min before fentanyl significantly and dose-dependently reduced the opioid-induced AEA increase. The lower dose (2 µg) pre-treatment caused a drop of AEA accumbens levels to the baseline concentration and the higher dose (10 µg) even induced significant AEA decrease with the maximum drop 78% of baseline level (p \< 0.05). For the lower JMV2959 dose: JMV2959 2 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + fentanyl 30 µg/kg: effect of group F1,10 = 168.7, p \< 0.001; effect of time F9,90 = 40.5, p \< 0.01; time × group interaction F9,90 = 42.9, p \< 0.001. The effects of JMV2959 2 µg/VTA with fentanyl 30 µg/kg on the accumbens AEA did not significantly differ from Ringer's solution/VTA with saline. For the higher JMV2959 dose: JMV2959 10 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + fentanyl 30 µg/kg: effect of group F1,10 = 295.2, p \< 0.001; effect of time F9,90 = 22.9, p \< 0.001; time × group interaction F9,90 = 54.5, p \< 0.001. JMV2959 10 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + saline: effect of group F1,10 = 5.1, p \< 0.05; effect of time F9,90 = 2.6, p \< 0.05; time × group interaction F9,90 = 2.2, p \< 0.05. A single dose of JMV2959 2 µg as well as 10 µg/VTA, similarly to Ringer's solution/VTA had no significant effect on accumbens AEA. ### 2.1.3. Pre-Treatment with JMV2959 Administered into the Nucleus Accumbens (NAC) {#sec2dot1dot3-ijms-18-02486} [Figure 2](#ijms-18-02486-f002){ref-type="fig"}b illustrates the observed influence of ghrelin antagonist administered into the NACSh on accumbens fentanyl-induced AEA changes. The 30 µg/kg fentanyl effects on accumbens AEA were practically the same with both pre-treatment with saline i.p. and Ringers's solution into the VTA. Thus, for ethical reasons, we did not create a new group with fentanyl 30 µg/kg without any pre-treatment, but we have used the group with systemic saline pre-treatment instead (see [Section 2.1.1](#sec2dot1dot1-ijms-18-02486){ref-type="sec"}) for testing the JMV2959 effects when administered into the NACSh; for statistical evaluation we have used mean of three baselines and results after fentanyl administration (saline + fentanyl 30 µg/kg vs. saline + saline: effect of group F1,10 = 823.5, p \< 0.001; effect of time F9,90 = 76.0, p \< 0.001; time × group interaction F9,90 = 81.1, p \< 0.001; maximum increase 220% of baseline level). We have used the dialysis probe for administration of JMV2959 into the NAC. After baseline samples were collected, perfusion with Ringer's solution (2 µL/min) was switched to perfusion with 8 mM or 40 mM JMV2959, respectively, for 15 min, starting 5 min before fentanyl administration; thereafter the inlet tube was switched back to Ringer's solution. Pre-treatment with both JMV2959 doses into the NACSh dose dependently reversed the fentanyl-induced accumbens AEA increase to a significant decrease with maximum drop 84% of baseline level 60 min after fentanyl (lower JMV2959 dose) and 61% of baseline level 40 min (higher JMV2959 dose) after fentanyl administration, respectively. For the lower JMV2959 dose: JMV2959 8 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: effect of group F1,10 = 120.2, p \< 0.001; effect of time F9,90 = 46.5, p \< 0.01; time × group interaction F9,90 = 60.4, p \< 0.001. For the higher JMV2959 dose: JMV2959 40 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: effect of group F1,10 = 158.8, p \< 0.001; effect of time F9,90 = 27.0, p \< 0.01; time × group interaction F9,90 = 70.0, p \< 0.001. The JMV2959/NAC pre-treatment induced decrease/reversal of accumbens AEA after fentanyl administration were observed only during first intervals, the AEA levels returned to baseline levels at about 90 min (lower JMV2959 dose) and 120 min (higher JMV2959 dose) after fentanyl administration. The lower JMV2959 dose even reached significant accumbens AEA increase within the last one and half hour with maximum 120% of baseline level. (JMV2959 8 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + saline: effect of group not significant---n.s.; effect of time F9,90 = 20.8, p \< 0.001; time × group interaction F9,90 = 21.5, p \< 0.001) (JMV2959 40 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + saline: effect of group F1,10 = 7.8, p \< 0.05; effect of time F9,90 = 16.0, p \< 0.001; time × group interaction F9,90 = 16.2, p \< 0.001). Administration of single lower 8 mM JMV2959 dose into the NAC and saline i.p. did not significantly influence the accumbens AEA. Administration of single higher 40 mM JMV2959 dose induced slight but significant AEA decrease with maximum 94% of baseline (JMV2959 40 mM/15 min/NAC + saline vs. saline + saline: effect of group F1,10 = 13.5, p \< 0.01; effect of time F9,90 = 2.4, p \< 0.05; time × group interaction F9,90 = 2.2, p \< 0.05. 2.2. The Effects of GHS-R1A Receptor Antagonist on Fentanyl-Induced Accumbens 2-Arachidonoylglycerol (2-AG) Extracellular Concentration Decrease {#sec2dot2-ijms-18-02486} ------------------------------------------------------------------------------------------------------------------------------------------------ ### 2.2.1. Pre-Treatment with Intraperitoneal JMV2959 Administration {#sec2dot2dot1-ijms-18-02486} The influence of intraperitoneally administered ghrelin antagonist on fentanyl-induced 2-AG decrease in the NACSh is illustrated in [Figure 3](#ijms-18-02486-f003){ref-type="fig"}. Baseline extracellular concentrations of 2-AG did not significantly differ between animals in all presented experiments. Fentanyl administration induced a statistically significant decrease of accumbens shell 2-AG with maximum drop 81% of baseline level 1 h after fentanyl administration: saline + fentanyl 30 µg/kg vs. saline + saline: effect of group F1,10 = 197.6, p \< 0.001; effect of time F10,100 = 18.4, p \< 0.001; time × group interaction F10,100 = 16.0, p \< 0.001. Pre-treatment with JMV2959 intensified the fentanyl-induced accumbens 2-AG decrease. The 3 mg/kg i.p. JMV2959 pre-treatment significantly deepened the fentanyl-induced 2-AG drop into maximum 59% of baseline level: JMV2959 3 mg/kg + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: effect of group F1,10 = 566.1, p \< 0.001; effect of time F10,100 = 246.0, p \< 0.001; time × group interaction F10,100 = 33.1, p \< 0.001. (JMV2959 3 mg/kg + fentanyl 30 µg/kg vs. saline + saline: effect of group F1,10 = 391.4, p \< 0.001; effect of time F10,100 = 321.5, p \< 0.001; time × group interaction F10,100 = 290.7, p \< 0.001). JMV2959 3 mg/kg i.p. administered in a single dose did not significantly influence the accumbens 2-AG, also saline i.p. had no effect on 2-AG in the NACSh. ### 2.2.2. Pre-Treatment with JMV2959 Administered into the VTA {#sec2dot2dot2-ijms-18-02486} [Figure 4](#ijms-18-02486-f004){ref-type="fig"}a illustrates the observed influence of ghrelin antagonist administered into the VTA on the fentanyl--induced accumbens 2-AG decrease. The 30 µg/kg dose of fentanyl together with intra-VTA Ringer's solution induced practically the same 2-AG decrease as the above described fentanyl with systemic saline, with maximum of 82% of baseline level: Ringer's solution/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + saline: effect of group F1,10 = 80.0, p \< 0.001; effect of time F9,90 = 26.7, p \< 0.001; time × group interaction F9,90 = 28.0, p \< 0.001. Pre-treatment with JMV2959 influenced the fentanyl-induced accumbens 2-AG decrease differently depending on the given dose. The 2 µg JMV2959/VTA dose slightly but significantly attenuated and simultaneously also prolonged the fentanyl-induced 2-AG decrease, with maximum 87% of baseline. On the contrary, the 10 µg JMV2959/VTA dose significantly deepened the accumbens fentanyl-induced 2-AG decrease to maximal drop 77% of baseline. For the lower dose: JMV2959 2 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + fentanyl 30 µg/kg: effect of group n.s.; effect of time F9,90 = 68.3, p \< 0.001; time × group interaction F9,90 = 13.9, p \< 0.001. For the higher dose: JMV2959 10 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + fentanyl 30 µg/kg: effect of group F1,10 = 39.8; effect of time F9,90 = 169.5, p \< 0.001; time × group interaction F9,90 = 5.1, p \< 0.001. In comparison to the control group the lower 2 µg/kg JMV2959 dose with fentanyl significantly decreased the accumbens 2-AG: JMV2959 2 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + saline: effect of group F1,10 = 605.3, p \< 0.001; effect of time F9,90 = 23.9, p \< 0.001; time × group interaction F9,90 = 30.0, p \< 0.001. The higher 10 µg JMV2959 dose with fentanyl, also induced significant 2-AG decrease in the NACSh: JMV2959 10 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + saline: effect of group F1,10 = 985.9, p \< 0.001; effect of time F9,90 = 129.1, p \< 0.001; time × group interaction F9,90 = 142.7, p \< 0.001. A single dose of JMV2959 2 µg as well as 10 µg/VTA did not significantly influence the accumbens 2-AG levels, the same as Ringer's solution/VTA with saline s.c. ### 2.2.3. Pre-Treatment with JMV2959 Administered into the NAC {#sec2dot2dot3-ijms-18-02486} [Figure 4](#ijms-18-02486-f004){ref-type="fig"}b illustrates the observed influence of ghrelin antagonist administered into the NACSh on the accumbens fentanyl-induced 2-AG decrease. In addition, the 30 µg/kg fentanyl effects on accumbens 2-AG were practically the same with both, pre-treatment with saline i.p. as well as Ringers's solution into the VTA. Thus, again, we have used the group with systemic saline pre-treatment of fentanyl (see [Section 2.2.1](#sec2dot2dot1-ijms-18-02486){ref-type="sec"}) for testing the JMV2959 effects when administered into the NAC (for statistical evaluation we have used mean of three baselines and fentanyl effects) (saline + fentanyl 30 µg/kg vs. saline + saline: effect of group F1,10 = 196.6, p \< 0.001; effect of time F9,90 = 17.3, p \< 0.001; time × group interaction F9,90 = 15.9, p \< 0.001; maximum decrease 81% of baseline level). Pre-treatment with both JMV2959 doses into the NAC significantly and dose dependently deepened the fentanyl-induced accumbens 2-AG extracellular concentrations with maximum drop to 74% (lower JMV2959 dose) and 63% of baseline (higher JMV2959 dose), respectively. For lower JMV2959 dose: JMV2959 lower dose 8 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: effect of group F1,10 = 160.9, p \< 0.001; effect of time F9,90 = 52.6, p \< 0.001; time × group interaction F9,90 = 2.3, p \< 0.05. For higher JMV2959 dose: JMV2959 higher dose 40 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: effect of group F1,10 = 79.9, p \< 0.001; effect of time F9,90 = 80.6, p \< 0.001; time × group interaction F9,90 = 10.7, p \< 0.001. Both JMV2959 pre-treatments with fentanyl induced significant 2-AG decrease in comparison to saline + saline: for lower dose: JMV2959 lower dose 8 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + saline: effect of group F1,10 = 656.3, p \< 0.001; effect of time F9,90 = 22.2, p \< 0.001; time × group interaction F9,90 = 23.1, p \< 0.001; for higher dose: JMV2959 higher dose 40 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + saline: effect of group F1,10 = 309.4, p \< 0.001; effect of time F9,90 = 67.7, p \< 0.001; time × group interaction F9,90 = 68.9, p \< 0.001. Single lower JMV2959 dose administered into the NAC and saline i.p. did not significantly influence the accumbens 2-AG. Administration of single higher JMV2959 dose 40 mM/15 min/NAC induced slight but significant 2-AG decrease with maximum 97% of baseline (JMV2959 40 mM/15 min/NAC + saline vs. saline + saline: effect of group F1,10 = 6.6, p \< 0.05; effect of time F9,90 = 3.1, p \< 0.01; time × group interaction F9,90 = 3.3, p \< 0.05. 2.3. The Effects of GHS-R1A Receptor Antagonist on Fentanyl-Induced Accumbens γ-Aminobutyric (GABA) Extracellular Concentration Increase {#sec2dot3-ijms-18-02486} ---------------------------------------------------------------------------------------------------------------------------------------- ### 2.3.1. Pre-Treatment with Intraperitoneal JMV2959 Administration {#sec2dot3dot1-ijms-18-02486} The influence of intraperitoneally administered ghrelin antagonist on fentanyl-induced increase of accumbens shell extracellular GABA is illustrated in [Figure 5](#ijms-18-02486-f005){ref-type="fig"}. GABA baseline levels did not significantly differ between rats in all presented experiments. As expected, acute systemic fentanyl (30 µg/kg s.c.) administration induced a statistically significant increase of GABA in the NACSh (saline + fentanyl 30 µg/kg vs. saline + saline: effect of group F1,10 = 105.0, p \< 0.001; effect of time F10,100 = 57.8, p \< 0.001; time × group interaction F10,100 = 56.4, p \< 0.001; maximum increase 192% of baseline level). The GHS-R1A antagonist, JMV2959 i.p. administration 20 min before fentanyl prevented the fentanyl-induced accumbens GABA increase maintaining its concentration almost on the baseline level: JMV2959 3 min/kg + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: effect of group F1,10 = 158.1, p \< 0.001; effect of time F10,100 = 65.2, p \< 0.001; time × group interaction F10,100 = 55.9, p \< 0.001. Accumbens GABA levels within the JMV2959 + fentanyl group were significantly above the saline levels during the first four intervals after fentanyl (JMV2959 3 mg/kg + saline vs. saline + saline: effect of group F1,10 = 50.1, p \< 0.05; effect of time F10,100 = 4.6, p \< 0.001; time × group interaction F10,100 = 4.9, p \< 0.001; maximum increase 112% of baseline). A single dose of JMV2959 3 mg/kg i.p. had no effect on accumbens GABA and the same was true for saline. ### 2.3.2. Pre-Treatment with JMV2959 Administration into the VTA {#sec2dot3dot2-ijms-18-02486} [Figure 6](#ijms-18-02486-f006){ref-type="fig"}a illustrates the observed influence of ghrelin antagonist administered into the VTA on changes in accumbens GABA induced by 30 µg/kg s.c. fentanyl. The 30 µg/kg dose of fentanyl together with intra-VTA Ringer's solution again induced practical the same GABA increase as the above described fentanyl with systemic saline: Ringer's solution/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + saline: effect of group F1,10 = 65.3, p \< 0.001; effect of time F9,90 = 11.0, p \< 0.001; time × group interaction F9,90 = 11.0, p \< 0.001; with maximum effect 188% of baseline level. JMV2959 pre-treatment into the VTA 5 min before fentanyl using both doses significantly and comparably reduced the opioid-induced GABA increase to the baseline levels. During the two last hours of microdialysis the accumbens GABA levels were oscillating around border significant increase in comparison to baseline concentrations (around 110% maximum). For the lower JMV2959 dose: JMV2959 2 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + fentanyl 30 µg/kg: effect of group F1,10 = 63.9, p \< 0.001; effect of time F9,90 = 18.3, p \< 0.01; time × group interaction F9,90 = 10.1, p \< 0.001. For the higher dose: JMV2959 10 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + fentanyl 30 µg/kg: effect of group F1,10 = 101.5, p \< 0.001; effect of time F9,90 = 20.0, p \< 0.01; time × group interaction F9,90 = 16.1, p \< 0.001. Both doses of JMV2959/VTA with fentanyl 30 µg/kg showed very low but significant GABA increase only at the end of the third hour after fentanyl administration: JMV2959 2 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + saline: effect of group n.s.; effect of time F9,90 = 3.4, p \< 0.01; time × group interaction F9,90 = 3.0, p \< 0.01; maximum effect 109% of baseline; JMV2959 10 µg/VTA + fentanyl 30 µg/kg vs. Ringer's solution/VTA + saline: effect of group n.s.; effect of time F9,90 = 3.4, p \< 0.01; time × group interaction F9,90 = 2.8, p \< 0.01; maximum effect 109% of baseline. A single dose of JMV2959 2 µg/VTA similarly to saline had no effect on accumbens GABA. A single 10 µg/VTA dose of JMV2959 induced slight but significant increase of accumbens GABA: JMV2959 10 µg/VTA + saline vs. Ringer's solution/VTA + saline: effect of group n.s.; effect of time n.s.; time × group interaction F9,90 = 2.3, p \< 0.05; with maximum effect of 104% of baseline. ### 2.3.3. Pre-Treatment with JMV2959 Administration into the NAC {#sec2dot3dot3-ijms-18-02486} [Figure 6](#ijms-18-02486-f006){ref-type="fig"}b illustrates the observed influence of ghrelin antagonist administered into the NACSh on the accumbens fentanyl-induced GABA increase. Again, the 30 µg/kg fentanyl effects on accumbens GABA were practically the same with both, pre-treatment with saline i.p. as well as Ringer's solution into the VTA and the group with systemic saline pre-treatment of fentanyl (see [Section 2.2.1](#sec2dot2dot1-ijms-18-02486){ref-type="sec"}) was used for testing the JMV2959 effects when administered into the NAC (mean of three baselines + effects of fentanyl) (saline + fentanyl 30 µg/kg vs. saline + saline: effect of group F1,10 = 105.2, p \< 0.001; effect of time F9,90 = 46.4, p \< 0.001; time × group interaction F9,90 = 45.0, p \< 0.001; maximum increase 191% of baseline level). Pre-treatment with both JMV2959 doses into the NAC significantly reduced the fentanyl-induced accumbens GABA increase to the baseline levels. Since 100 min after fentanyl administration the accumbens GABA levels stayed slightly significant above the baseline concentrations (112--116% maximum). For lower JMV2959 dose: JMV2959 lower dose 8 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: effect of group F1,10 = 76.1, p \< 0.001; effect of time F9,90 = 33.6, p \< 0.001; time × group interaction F9,90 = 53.7, p \< 0.001. For higher JMV2959 dose: JMV2959 higher dose 40 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: effect of group F1,10 = 74.2, p \< 0.001; effect of time F9,90 = 22.7, p \< 0.001; time × group interaction F9,90 = 35.1, p \< 0.001. Both JMV2959 pre-treatments with fentanyl induced slight but significant GABA increase only during the last/third hour after fentanyl administration in comparison to saline + saline: for lower dose: JMV2959 lower dose 8 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + saline: effect of group n.s.; effect of time F9,90 = 8.9, p \< 0.001; time × group interaction F9,90 = 8.0, p \< 0.001; for higher dose: JMV2959 higher dose 40 mM/15 min/NAC + fentanyl 30 µg/kg vs. saline + saline: effect of group n.s.; effect of time F9,90 = 3.8, p \< 0.001; time × group interaction F9,90 = 3.3, p \< 0.05. Single lower as well as higher JMV2959 dose administered into the NACSh did not significantly influence the accumbens GABA and also administration of saline had no effect on accumbens GABA. 2.4. Additional Behavioural Assay {#sec2dot4-ijms-18-02486} --------------------------------- As expected, fentanyl induced typical significant biphasic (inhibition-stimulation) behavioural changes in the rats, as illustrated in [Figure 7](#ijms-18-02486-f007){ref-type="fig"}, in accordance with our previous studies with morphine \[[@B16-ijms-18-02486],[@B17-ijms-18-02486],[@B21-ijms-18-02486],[@B85-ijms-18-02486]\]. The fentanyl-induced changes were more dynamic in comparison with morphine and the behaviour returned to almost normal control parameters at the end of the experiment. We were interested in the ghrelin antagonist effects. JMV2959 pre-treatment 20 min before fentanyl induced moderate but significant changes within the fentanyl-evoked stimulation. In comparison to saline + fentanyl, in the JMV2959 pre-treated rats, we especially observed decreased locomotion, the fentanyl-induced behavioural stimulation was cut with JMV2959, the parameters of locomotion reached control levels 30 min earlier (between intervals 140 and 160 min) (see [Figure 7](#ijms-18-02486-f007){ref-type="fig"}a), although the fentanyl + saline locomotion was still significantly higher within 140 and 160 intervals in comparison to the control group (locomotion: saline + saline vs. fentanyl + saline time × group interaction F10,100 = 11.7; p \< 0.05). The incidence of stereotypical behaviours ([Figure 7](#ijms-18-02486-f007){ref-type="fig"}b) was present within shorter period, it was lowered, but apparent (less confined gnawing and licking, almost no stereotyped sniffing). The rats started with gnawing and licking afterwards during still remaining stupor positions earlier than within the fentanyl + saline group, which shifted the beginning of occurrence of stereotypies. In comparison to fentanyl + saline, the pre-treatment with JMV2959 significantly increased the immobility scores ([Figure 7](#ijms-18-02486-f007){ref-type="fig"}c) during the last hour of the experiment (increased sedation) and during the last interval we observed sedation with eyes closed, similarly to the saline + saline group and correspondingly to the decrease in locomotion changes. In the fentanyl + saline rats, the immobility scores remained significantly lower within 140 and 160 intervals in comparison to control (immobility: saline + saline vs. fentanyl + saline time × group interaction F10,100 = 8.6; p\< 0.05), but the JMV2959 pre-treated rats reached the scores of control group already before the interval 160 min. Catalepsy ([Figure 7](#ijms-18-02486-f007){ref-type="fig"}d) remained practically unchanged (JMV 3 mg/kg + fentanyl 30 µg/kg vs. saline + fentanyl 30 µg/kg: locomotion---effect of group n.s.; effect of time F10,100 = 44.0, p \< 0.001; time × group interaction F10,100 = 3.03, p \< 0.05.; stereotypies---effect of group F1,10 = 19.4, p \< 0.05; effect of time F10,100 = 50.0, p \< 0.001; time × group interaction F10,100 = 4.3, p \< 0.001; immobility---effect of group F1,10 = 15.0, p \< 0.01; effect of time F10,100 = 26.9, p \< 0.001; time × group interaction F10,100 = 2.2, p \< 0.05; catalepsy---effect of group n.s.; effect of time F10,100 = 47.0, p \< 0.001; time × group interaction n.s.). During the last interval of the total 5 h microdialysis experiment, the control rats exhibited less locomotion parameters, increased immobility, sedation and closed eyes in comparison to the baseline intervals. The effects of single 3 mg/kg JMV2959 dose on the rat behaviour did not significantly differ from saline. 3. Discussion {#sec3-ijms-18-02486} ============= Our results for the first time indicate significant involvement of GHS-R1As within the VTA and NAC in opioid/fentanyl-induced changes in the NACSh endocannabinoid AEA and 2-AG as well as GABA extracellular concentrations. Concerning endocannabinoids, the present results extent and particularize our previous study with morphine and systemic JMV2959 pre-treatment \[[@B17-ijms-18-02486]\]. Fentanyl is considered as a µ-receptor-selective agonist, 100-fold more potent than morphine, mediating its reinforcing properties through both VTA as well as NAC µ-receptors \[[@B28-ijms-18-02486],[@B31-ijms-18-02486],[@B74-ijms-18-02486]\]. The 30 µg/kg subcutaneous fentanyl used dose is about 300-fold lower than the morphine 10 mg/kg s.c. dose in our previous study and yet fentanyl induced much higher AEA increase in the NACSh (220%) in comparison to morphine (142% of baseline mean). However, the induced accumbens 2-AG decrease was practically comparable with fentanyl (81%) and morphine (85% of baseline). This indicates association of opioids-induced accumbens AEA efflux with their µ-receptor affinity. The observed opioid/fentanyl-induced accumbens AEA increase and 2-AG decrease are also in accordance with Vigano \[[@B54-ijms-18-02486]\] and Caille \[[@B52-ijms-18-02486]\]. Pre-treatment with JMV2959 in allF doses and types of administration (i.p., into the VTA or NAC) significantly changed the fentanyl-induced accumbens AEA increase and 2-AG decrease. The fentanyl-evoked AEA increase was reversed by pre-treatment with intraperitoneal 3 mg/kg dose of JMV2959, inducing a significant decrease. The fentanyl-induced 2-AG decrease was significantly deepened by the 3 mg/kg i.p. JMV2959. This is fully in accordance with our previous study with morphine and intraperitoneal JMV2959 pre-treatment \[[@B17-ijms-18-02486]\]. In one experiment of our previous study \[[@B17-ijms-18-02486]\], we have confirmed participation of central ghrelin system in the observed changes of AEA and 2-AG concentrations induced by opioid/morphine in the NACSh, when we co-administered ghrelin (40 µg/kg) with JMV2959 3 mg/kg i.p. before morphine, and ghrelin had abolished all the observed JMV2959 effects. In the present study, we also observed a reverse of fentanyl-induced accumbens AEA increase when fentanyl s.c. was administered together with JMV2959 into the NACSh (perfusion with 8 or 40 mM for 15 min); however, the observed significant dose-dependent AEA decrease had sharper drop and shorter duration in comparison with the intraperitoneal JMV2959 effect, which was possibly due to the local (NAC) type of JMV2959 administration. The intra-VTA pre-treatment with JMV2959 (2 or 10 µg) prevented the fentanyl-induced accumbens AEA increase, but only the higher dose reversed the AEA levels to significant decrease. It is difficult to compare effects of differently administered various doses, but considering the effects of the pre-treatment JMV2959 doses, which per se/alone did not significantly influence the accumbens AEA, it can be suggested that GHS-R1As of both VTA and NAC brain structures participate in the significant JMV2959 reversal effects on the fentanyl/opioid-induced AEA increase in the NACSh, with major involvement of the NAC structure. It has been described that predominant (but not selective) µ-receptor antagonist naltrexone did not attenuated ghrelin-induced food intake, locomotor stimulation and accumbens dopamine release \[[@B10-ijms-18-02486],[@B86-ijms-18-02486]\], thus it has been suggested, that capability of reinforcement reduction through GHS-R1A antagonism does not include µ-receptors. It has also been described that subchronic JMV2959 (but not ghrelin) treatment significantly increased opioid peptide enkephalins/δ-agonists levels within the VTA and striatum and β-endorphin/κ-agonist levels within hippocampus and these changes are considered to contribute to the JMV2959/GHS-R1A antagonist-induced attenuation of opioid/morphine reinforcement \[[@B15-ijms-18-02486]\]. To our knowledge, the influence of subchronic JMV2959 on µ-receptor endogenous ligands/β-endorphin has not yet been tested. However, our present results documented that GHS-R1A antagonism significantly affected the selective µ-opioid fentanyl-induced accumbens AEA changes which are believed to contribute to opioid reinforcement. Furthermore, in the present study surprisingly, all pre-treated JMV2959 doses and types of administration affected the fentanyl-induced AEA increase in a similar manner (mainly) without showing significant influence on the AEA levels "per se". The JMV2959 3 mg/kg i.p. dose, doses administered into the VTA and lower dose into the NAC did not induce significant changes in accumbens AEA, although the higher 40 nM/NAC dose produced slight but significant AEA decrease (94% of baseline mean). Which corresponds with our previous study \[[@B17-ijms-18-02486]\], when 6 mg/kg i.p. JMV2959 dose induced significant AEA drop to 92% of baseline and 3 mg/kg i.p. dose did not have significant influence. The above-summarized findings indicate presumable complexity of possibly several ghrelin involving neural pathways participating in the observed antagonism of fentanyl-induced accumbens AEA increase (and probably also in other here further presented changes). Possible cooperation of two or more neural systems and/or indirect effects might be considered. After all, such pattern is known from ghrelin influence of accumbens dopamine \[[@B87-ijms-18-02486]\] and it also has been documented, that ghrelin orexigenic effect is dependent on several central networks, such as dopamine, cannabinoid, opioid and serotonine systems (see review \[[@B69-ijms-18-02486]\]). In addition, the high constitutive activity of the GHS-R1A might possibly play some role \[[@B88-ijms-18-02486]\]. Blocking of GHS-R1A by JMV2959 pre-treatments seem to prevent development of following fentanyl-induced ghrelin involving pathways and changes, which might at least partly contribute in massive pre-treatment effects using "per se" not effective doses. Further investigation is necessary to clarify the appropriate mechanisms. As mentioned earlier, it is thought that opioid-induced AEA increase in the NACSh possibly contribute to the opioid reinforcement through CB1 receptor-mediated process independent of dopamine \[[@B46-ijms-18-02486],[@B49-ijms-18-02486],[@B52-ijms-18-02486]\]. Anandamide had been intravenously self-administered by squirrel monkeys, which proves its reinforcing properties \[[@B89-ijms-18-02486]\]. Thus, the observed dose-dependent reversal of AEA increase evoked by opioids in the NACSh, which was caused by ghrelin antagonist, suggests an important contribution of central ghrelin system in the assumed anandamide impact to the opioid reinforcement. Our results indicate, that the GHS-R1A receptors within the NACSh as well as VTA both participate significantly in the opioid-induced accumbens anandamide increase, with possibly emphasized impact of the NACSh ghrelin signalling. Similar to the 3 mg/kg JMV2959 i.p. administered effects, intra-accumbens administration of JMV2959 (perfusion with 8 or 40 mM for 15 min) also deepened significantly and dose-dependently the fentanyl-induced 2-AG accumbens decrease. When JMV2959 was administered into the VTA (2 or 10 µg), only the higher JMV2959 dose significantly deepened the fentanyl-induced 2-AG decrease. The lower JMV2959/VTA dose attenuated but prolonged the fentanyl-induced accumbens 2-AG decrease. Thus, both types of JMV2959 administration significantly changed the fentanyl-induced accumbens 2-AG decrease, but the pre-treatment effects seemed more expressed with the administration into the NACSh, similarly to the AEA. Analogously, the noticeable JMV2959 pre-treatment effects on fentanyl-evoked accumbens 2-AG decrease were induced using mainly per se not effective doses, with exception of the 40 mM/NAC JMV2959 dose, which induced significant 2-AG decrease to 97% of baseline mean, similarly to our previous study, when 6 mg/kg i.p. JMV2959 induced 2-AG decrease to 93% of baseline \[[@B17-ijms-18-02486]\]. Again, possible complex ghrelin involving mechanisms/pathways should be considered participating in the predominant intensification of fentanyl-evoked accumbens 2-AG decrease induced by GHS-R1A antagonist pre-treatment. Further investigation is necessary. However, the significance of 2-AG decrease induced by acute opioid administration is so far difficult to interpret, as well as the observed intensified decrease with ghrelin antagonist. Recently considered and documented important role of 2-AG in reinforcement and addiction processes within the mesolimbic system has been mainly associated with chronic drug use (including opioids), changed motivational states, cue-evoked reward seeking, withdrawal and synaptic plasticity and also 2-AG increase \[[@B90-ijms-18-02486],[@B91-ijms-18-02486]\]. GHS-R1A, CB1 and µ-opioid receptors are expressed within the NACSh as well as the VTA, thus interaction among the appropriate signalling systems within these brain structures could be considered above \[[@B5-ijms-18-02486],[@B37-ijms-18-02486],[@B38-ijms-18-02486],[@B42-ijms-18-02486],[@B50-ijms-18-02486],[@B92-ijms-18-02486]\]. However, further investigation is necessary to elucidate the actual purpose and consequences of their possible relationships, including our findings mentioned The present data together with our previous results \[[@B16-ijms-18-02486],[@B17-ijms-18-02486]\] associated with endocannabinoids and ghrelin signalling involvement in the opioid reinforcement, suggest that ghrelin signalling is possibly significantly involved in both dopamine-dependent as well as dopamine-independent opioid reinforcing processes in the NACSh. Midbrain GHS-R1As, co-localized with dopaminergic and cholinergic receptors \[[@B5-ijms-18-02486],[@B93-ijms-18-02486]\], functionally interact in amplification of the dopaminergic signalling in the VTA neurons and stimulate accumbens dopamine efflux \[[@B86-ijms-18-02486],[@B94-ijms-18-02486]\]. Indeed, the ability of ghrelin antagonism to decrease opioid-induced accumbens shell dopamine increase has been documented previously \[[@B15-ijms-18-02486],[@B16-ijms-18-02486]\]. The brain endocannabinoid system is important for regulation of dopamine signalling during reinforcement processes \[[@B61-ijms-18-02486],[@B95-ijms-18-02486]\]. Anandamide, when administered intravenously, also increased dopamine in the NACSh \[[@B61-ijms-18-02486]\] as a neurochemical effect common to rewarding stimuli. The CB1 antagonist SR141716A infused into the NACSh significantly attenuated intravenous self-administration of heroin \[[@B49-ijms-18-02486]\], but systemic SR141716A pre-treatment had no effect on morphine-induced dopamine increase in the NACSh \[[@B46-ijms-18-02486],[@B51-ijms-18-02486]\]. Thus, accumbens CB1 receptors possibly significantly modulate opioid reinforcing properties through dopamine-independent mechanisms \[[@B52-ijms-18-02486]\]. Thus, the significant reduction/reversal of opioid/fentanyl-induced accumbens AEA increase caused by JMV2959 might indicate an important participation of ghrelin signalling in the presumed non-dopaminergic opioid reinforcing mechanisms. Considering possible interactions between endocannabinoids and central ghrelin signalling, functional cooperation of CB1/endocannabinoids and GHS-R1A/ghrelin within hypothalamus has been described, which possibly contribute to ghrelin orexigenic effects, but similar link within NACSh to our knowledge has not been evidenced \[[@B66-ijms-18-02486],[@B67-ijms-18-02486],[@B68-ijms-18-02486],[@B69-ijms-18-02486],[@B72-ijms-18-02486]\]. Further research is required to elucidate the appropriate involved mechanisms. It has been suggested, that the opioid-induced GABA efflux in the NACSh \[[@B78-ijms-18-02486],[@B79-ijms-18-02486]\] contributes to the opioid reinforcing properties \[[@B29-ijms-18-02486],[@B80-ijms-18-02486]\]. The µ-opiod receptors are strategically located to modulate GABA release in the NAC, because they are expressed on medium spiny projection neurons (from VTA) and spiny interneurons in the NAC \[[@B27-ijms-18-02486],[@B96-ijms-18-02486],[@B97-ijms-18-02486]\]. The majority of neurons within NAC are GABAergic medium spiny neurons (MSNs) \[[@B98-ijms-18-02486]\] that send their output to several brain structures including the VTA \[[@B99-ijms-18-02486]\] and ventral pallidum \[[@B100-ijms-18-02486]\]. These GABA neurons receive inputs from the VTA (dopaminergic) and from hippocampus, amygdala and medial prefrontal cortex (glutamatergic) \[[@B101-ijms-18-02486],[@B102-ijms-18-02486],[@B103-ijms-18-02486]\]. The VTA GABA neurons, which project to the NAC synapse largely onto cholinergic interneurons \[[@B104-ijms-18-02486]\]. The simultaneous activation of µ and GABA-A receptors, which are co-expressed on GABAergic interneurons in the NAC markedly suppresses GABA release onto dopamine nerve endings, thus disinhibiting/enhancing dopamine efflux \[[@B29-ijms-18-02486]\]. Alternatively, dopamine release may be evoked indirectly through activation of GABA-A and µ receptors on GABAergic medium spiny projection neurons in the VTA \[[@B27-ijms-18-02486],[@B28-ijms-18-02486]\]. In either case, GABA elevated concentrations potentiate opioid-induced accumbens dopamine release. In the present study, we have demonstrated for the first time that pre-treatment with ghrelin antagonist in all given doses and types of administration prevented the opioid/fentanyl-induced accumbens GABA efflux in a very similar way. These findings suggest that ghrelin antagonism may inhibit opioid-induced dopamine release, which has been described previously \[[@B15-ijms-18-02486],[@B17-ijms-18-02486]\], at least in part, by attenuating opioid-evoked GABA release in the NACSh. Since we have found comparable effects of JMV2959 administered into the VTA, the NAC and i.p. onto the fentanyl-induced accumbens GABA increase, possibly both modulatory mechanisms within the NAC and VTA participate on these ghrelin antagonist effects. These new findings further support the presumed complexity of possibly several ghrelin involving pathways which contribute to opioid reward and reinforcement. It has been also established, that opioids' activation of opioid receptors in the NAC decreases GABA release in their major projection area the ventral pallidum, generating dopamine-independent biological effects. It has been well documented, that CB1, µ as well as GHS-R1As are present on GABA presynaptic terminals in various brain structures \[[@B28-ijms-18-02486],[@B42-ijms-18-02486],[@B81-ijms-18-02486],[@B84-ijms-18-02486]\]. CB1 receptors are located on inhibitory inputs to GABAergic medium spiny projection neurons in the NAC and it has been suggested that endocannabinoids release evoked by depolarization in the NAC and from the VTA dopaminergic neurons may act as retrograde messengers on reachable receptors of GABAergic afferents \[[@B65-ijms-18-02486]\]. It was also described that VTA GABA-A/GABAergic system importantly contributes to the dopamine-independent opioid reinforcement \[[@B81-ijms-18-02486],[@B105-ijms-18-02486]\]. Thus, the observed opioid-induced accumbens GABA increase might also contribute to dopamine-independent opioid reinforcement mechanisms, possibly involving endocannabinoids or ghrelin. Recently, important interactions between central ghrelin and GABAergic systems have been implicated within the CeA \[[@B83-ijms-18-02486]\] and hypothalamus \[[@B84-ijms-18-02486]\], hereby we have observed significant ghrelin and accumbens GABA interaction. Assuredly, further investigation of the particular involved mechanisms is necessary. The additional behavioural assay confirmed that ghrelin antagonism moderately but significantly attenuated the fentanyl-induced behavioural stimulation (locomotion, stereotypical sniffing, partly also confined gnawing), which are considered as a sign of activation of nigrostriatal pathway \[[@B36-ijms-18-02486],[@B106-ijms-18-02486]\]. It seemed that JMV2959 i.p. pre-treatment slightly speeded up onset of stimulation phase of fentanyl-induced biphasic behavioural changes, reduced especially fentanyl-induced increased incidence of walking, rearing and stereotypical sniffing and generally accelerated return of the fentanyl changed behaviour (locomotion and immobility) to the normal/control condition. In our previous study, JMV2959 in 1, 3 and 6 mg/kg i.p. single doses did not significantly influence rat locomotor activity 25 min after the JMV2959 administration, when monitored for 20 min using Ethovision program (Noldus, Wageningen, The Netherlands) \[[@B21-ijms-18-02486]\]. Our results signify a strong participation of accumbens endocannabinoids, particularly anandamide, but also GABA in the neural opioid/fentanyl reinforcing processes and suggest that ghrelin antagonism may play an important role in the NACSh endocannabinoid/AEA and GABA changes possibly related to opioid/µ-receptor agonist reinforcement. Although GHS-R1A receptors within both NACSh as well as VTA have been found to contribute significantly to these effects, administration of ghrelin antagonist into the NACSh seemed to have stronger impact on the accumbens endocannabinoid opioid-induced changes. Collectively, our results indicate significant engagement of central ghrelin in GABA and endocannabinoid mechanisms in opioid/fentanyl reinforcement and encourage further investigation to assess, if ghrelin antagonism or substances affecting GABA or endocannabinoid concentrations and actions, such as GHS-R1A antagonists, can be used to prevent opioid/fentanyl-seeking behaviour. 4. Materials and Methods {#sec4-ijms-18-02486} ======================== 4.1. Animals {#sec4dot1-ijms-18-02486} ------------ Male Wistar rats (adult 8 weeks old; 200--250 g; Velaz, Praha-Lysolaje, Czech Republic) in groups of 6 were used. The animals were housed in polycarbonate cages and given free access to food and water with room temperature (22--24 °C), constant humidity (50--60%), and a 12-h light/dark reversed cycle for at least 7 days before the experiments, which were performed from 8 am to 3 pm during the dark period. Procedures involving animals and animal care were conducted in compliance with international laws; protocols respected EU Directive (2010/63/EU, 22 September 2010) and the Guidelines of the European Union Council (86/609/EU, 24 November 1986) and followed the instructions of the National Committee for the Care and Use of Laboratory Animals. Experiments were performed in accordance with the Animal Protection Act of the Czech Republic (No. 246/1992 Sb, 15 April 1992) and were authorized by the Expert Committee for Protection of Experimental Animals of the Third Faculty of Medicine, Charles University in Prague. 4.2. Drugs and Chemicals {#sec4dot2-ijms-18-02486} ------------------------ Fentanyl citrate was purchased from Sigma-Aldrich (St. Louis, MO, USA). JMV2959 (1,2,4-triazole derivate), which has been proved to be an GHS-R1A antagonist \[[@B18-ijms-18-02486]\], was provided by Anton Bespalov (AbbVie, Heidelberg, Germany). All reagents were analytical grade. Fentanyl (30 µg/kg) was always dissolved in saline and administered subcutaneously (s.c.) 0.1 mL/100 g of body weight and saline was used as placebo; the chosen dose was selected following the literature as reliable analgesic and discriminative dose increasing accumbens dopamine \[[@B19-ijms-18-02486],[@B107-ijms-18-02486],[@B108-ijms-18-02486]\]. JMV2959 was dissolved in saline, when administered intraperitoneally (i.p.) 20 min before fentanyl; the selected dose 3 mg/kg JMV2959 s.c. was determined based on our previous studies in Wistar rats \[[@B16-ijms-18-02486],[@B17-ijms-18-02486]\] and the literature \[[@B109-ijms-18-02486],[@B110-ijms-18-02486]\]. The dose 3 mg/kg JMV2959 had no effect on the rat behaviour. The intracerebral JMV2959 doses were in accordance with the literature \[[@B111-ijms-18-02486],[@B112-ijms-18-02486]\]. When JMV2959 was administered intra-cerebrally, JMV2959 was dissolved in the Ringer's solution (adjusted to pH = 7.0) and Ringer's solution was used as a placebo. Doses 2 or 10 µg of JMV2959 were administered into the VTA at a volume of 0.5 µL for 1 min; the cannula stayed in place for another minute and after was retracted (5 µL microsyringe; Innovative Labor System, Stutzerbach, Germany). The administration sites were verified following the end of the experiment ([Figure 8](#ijms-18-02486-f008){ref-type="fig"}b), and only animals with correct injection sites were included in the statistical analysis. We have used the dialysis probe for administration of JMV2959 into the NAC. During perfusion with Ringer's solution (always 2 µL/min) the inlet tube was switched to tube filled with 8 mM or 40 mM solution of JMV2959 in the Ringer's solution for 15 min, starting 5 min before fentanyl administration; thereafter, the inlet tube was switched back to Ringer's solution. The position of each dialysis probe was histologically verified ([Figure 8](#ijms-18-02486-f008){ref-type="fig"}a) after the completion of each microdialysis experiment and only animals with correct probe positions were included into the statistical evaluations. 4.3. In Vivo Microdialysis {#sec4dot3-ijms-18-02486} -------------------------- The acute effects of fentanyl in rats were monitored after pre-treatment with JMV2959 or saline/Ringer's solution, using the in vivo microdialysis model in the nucleus accumbens shell (NACSh). In separate groups, JMV2959 (3 mg/kg) was administered intraperitoneally or into the VTA (2 or 10 µg) or into the NAC (8 or 40 mM/15 min perfusion). After three baseline samples collecting, JMV2959 was administered intraperitoneally 20 min before fentanyl (s.c.). After four baseline samples collecting, JMV2959 was administered into the VTA 5 min before fentanyl or the NAC was perfused with JMV2959 for 15 min, starting 5 min before fentanyl. The dialysis samples were collected for a total of 260 min at 20 min intervals. Dialysate levels of AEA, 2--AG and GABA were analysed using high-sensitivity liquid chromatography combined with mass spectrometry. ### 4.3.1. Surgery {#sec4dot3dot1-ijms-18-02486} The method is described in detail in Sustkova-Fiserova et al. \[[@B16-ijms-18-02486],[@B17-ijms-18-02486]\]. Under ketamine---xylazine surgical anaesthesia (ketamine 100 mg/kg i.p., Narketan, Vetoquinol; xylazine 10 mg/kg i.p., Xylapan, Vetoquinol), using a stereotaxic instrument (Stoelting Co., Wood Dale, IL, USA), a disposable dialysis guide cannula (MAB4 probes, Agnthos, Lidingo, Sweden) was implanted in rats into the nucleus accumbens shell (NACSh coordinates---anterior (A): +2.0 mm and lateral (L): ±1.2 mm from bregma and vertical (V): 6.2 mm from occipital bone) \[[@B113-ijms-18-02486]\]. The cannula was then secured to the skull with dental cement and an anchoring screw. The guide was randomly assigned to the left or right side. In experiments, were JMV2959/Ringer's solution was administered into the VTA, two guide cannulas were implanted together on the same site, one into the NACSh (coordinates see above) and one into the VTA (VTA: A: −5.3 mm and L: ±0.8 mm from bregma and V: 8.2 mm from the skull) (unilaterally). The used coordinates to target the VTA district linked with ghrelin-involving food/drug motivation are chosen following literature \[[@B7-ijms-18-02486],[@B12-ijms-18-02486]\]. Postoperative, the rats were housed in their own individual cages. After the end of the microdialysis experiments, the placements of the dialysis probe (NACSh) as well as placement of the infusion cannula (VTA) were verified histologically ([Figure 8](#ijms-18-02486-f008){ref-type="fig"}). Only animals with correct probe/cannula placement were used for subsequent statistical analysis. ### 4.3.2. Microdialysis and Chemical Analysis Assay {#sec4dot3dot2-ijms-18-02486} In accordance with Sustkova-Fiserova \[[@B16-ijms-18-02486],[@B17-ijms-18-02486]\], 48 h after implantation, a probe (MAB4, 2 mm active cuprophane membrane, Agnthos, Sweden) was inserted into the guide cannula and the probe was flushed with artificial cerebrospinal fluid (Ringer's solution; 147 mM NaCl, 2.2 mM CaCl~2~ and 4.0 mM KCl; pH 7.0) at a constant rate of 2.0 μL/min (Univentor 864 Syringe Pump, Agnthos); we used dual swivels (Agnthos). After 60 min of habituation (when the dialysate was discarded), 20 μL samples were collected in small polyethylene tubes at 20-min intervals; the other 20-μL part of each interval dialysate served for other neurotransmitters detection. After three consecutive baseline samples, rats were injected with saline or JMV2959 (i.p.), which was followed (20 min later) by fentanyl or saline (s.c.) injection (in separate experiments). In experiments with JMV2959 administration into the VTA/NAC we have collected four baseline samples and 5 min before fentanyl, JMV2959 was administered into the VTA or NAC perfusion started and samples were further collected starting with fentanyl administration. Sampling continued for 3 h following injection of fentanyl or saline. Immediately after sampling, the samples were frozen at −70 °C. The amount of AEA, 2-AG and GABA in the dialysate were quantified using high-performance liquid chromatography combined with mass spectrometry (HPLC-MS). The appropriate HPLC-MS determination methods were described in detail earlier \[[@B17-ijms-18-02486],[@B114-ijms-18-02486]\]. Thus, here only brief explanation: determination of GABA and endocannabinoids in the dialysate consisted of lyophilization in freeze dryer (Labconco Free Zone, Kansas City, MO, USA) to concentrate the substances from the dialysates, and detection using liquid chromatography combined with electrospray ionization tandem mass spectrometry (LC--ESI-MS/MS) which consisted of a chromatograph Accela 1250 (Thermo Scientific, Waltham, MA, USA), autosampler Accela (Thermo Scientific) and a TSQ Vantage mass spectrometer (Thermo Scientific). The data were acquired and processed using Xcalibur 2.1.0 software (Thermo Scientific). The in vitro recovery (probe MAB4, 2 mm, Agnthos) of anandamide (AEA) and 2-AG has been determined in our previous study \[[@B17-ijms-18-02486]\]; the average recovery of AEA was 51 ± 4% and for 2-AG 53 ± 5%. However, detected extracellular concentrations from the NACSh dialysates oscillated around 0.9--3.0 ng/mL of anandamide and 0.1--0.7 ng/mL of 2-AG. The limit of quantification (LOQ) for AEA was 240 pg/mL and the LOQ for 2-AG was 280 pg/mL. In our previous study \[[@B115-ijms-18-02486]\] has been described, that GABA efflux in the NACSh in our microdialysis experiments is Ca^2+^ dependent which indicates, that GABA extracellular concentration in the dialysates reflects the overflow of neuronally/from the synapses released neurotransmitter. 4.4. Additional Behavioural Assay {#sec4dot4-ijms-18-02486} --------------------------------- During the experiment, when 3 mg/kg JMV2959 was administered intraperitoneally 20 min before fentanyl, similarly to our previous studies with morphine \[[@B16-ijms-18-02486],[@B17-ijms-18-02486],[@B21-ijms-18-02486],[@B85-ijms-18-02486]\], rats' behavioural changes were monitored simultaneously in the course of microdialysis measurements. The following behavioural categories were monitored: immobility (eyes closed, akinesia, reduced responsiveness to environmental cues, sedation), catalepsy (exophthalmos, trunk rigidity frozen postures), locomotion (non-stereotyped activity, walking, rearing, grooming, sniffing), stereotypical behaviours (licking, stereotypical sniffing, confined gnawing) as was described previously in Sustkova-Fiserova et al. \[[@B17-ijms-18-02486]\] (and in accordance with \[[@B116-ijms-18-02486],[@B117-ijms-18-02486]\]. An observer, who was blinded to the treatment, scored all the behavioural categories at each microdialysis interval (every 20-min). The intensity or incidence of any behavioural changes which occurred during each/the whole 20 min interval were assessed using predefined anchor points on a 4 point scale: 0 = absent/no incidents; 1 = mild/1--5 incidents; 2 = moderate/medium/6--10 incidents; 3 = marked/maximum/more than 11 incidents. Behavioural changes were scored during the entire microdialysis period: 60 min baseline + 20 min pre-treatment and 3 h following fentanyl or saline administration. 4.5. Statistical Analysis {#sec4dot5-ijms-18-02486} ------------------------- Raw data for endocannabinoids and GABA, not corrected for probe recovery, expressed as ng/mL/sample, were transformed into a percentage of baseline levels (mean of three or four intervals prior to pre-treatment). In addition, changes in behavioural parameters, within the 20-min intervals, were analysed. We used Sigma Plot 13 (Systat Software, Inc., San Jose, CA, USA) for statistical evaluation of the time course neurochemical and behavioural data. For statistical differences between the treatment groups (JMV2959 + fentanyl), (saline/Ringer's solution + fentanyl), and (saline/Ringer's solution + saline) relative to time-related changes during the in vivo microdialysis experiment, we have used two-way analysis of variance for repeated measures (ANOVA RM analysis) followed by Bonferroni multiple comparisons procedure. The group of animals was entered as the between-group factor and the time-points as repeated within-subject measures, we compared all treatments to baseline mean; 20-min intervals over 200--180 min of post-treatment. All statistical tests were performed as two-sided at a significance level of 0.05 (the p values of \<0.05, \<0.01 and \<0.001 defined statistical significance). Groups of 6 animals were used; results are presented as the mean ± SEM. We thank Vera Mendlova, Nada Hemberova and Eva Sulcova from the Department of Pharmacology; Petr Zach, from the Department of Anatomy, Third Faculty of Medicine, Charles University in Prague; and statistician Marek Maly, from The National Institute for Public Health for their excellent technical assistance. The GHS-R1A antagonist JMV2959 was kindly provided by Anton Bespalov, AbbVie, Heidelberg, Germany. This study was supported by the Grant Agency of the Charles University GAUK 742214 and GAUK 748216, Project PROGRES Q35 and Project 260388/SVV/2017. Magdalena Sustkova-Fiserova was responsible for the conception, design and coordination of the study, acquisition and interpretation of the statistical data and drafting of the manuscript. Chrysostomos Charalambous was responsible for the main statistical evaluation of the data, graph creation, revision of the manuscript, and participation in the experiments. Kamila Syslova analysed the microdialysates. Tereza Havlickova, Marek Lapka, Pavel Jerabek and Nina Puskina participated evenly in the experiments. The authors declare no conflicts of interest. 2-AG 2-arachydonoylglycerol AEA N -arachidonoylethanolamine CeA central nucleus of amygdala GABA γ-aminobutyric acid GHS-R1A growth hormone secretagogue receptor A1 MSNs medium spiny neurons NAC nucleus accumbens NACSh nucleus accumbens shell VTA ventral tegmental area ![Effects of ghrelin receptor antagonist JMV2959 3 mg/kg administered intraperitoneally (i.p.) on the fentanyl-induced accumbens anandamide (AEA) levels. JMV2959 was administered following three 20 min baselines and 20 min before fentanyl/saline (intervals: baseline = −60 to −20 min; JMV2959 pre-treatment = 0 min; fentanyl = 20--180 min) (means ± SEM). The effects are illustrated as follows: saline + fentanyl (filled circle), 3 mg/kg JMV2959 + fentanyl (open circle), 3 mg/kg JMV2959 + saline (open triangle), saline + saline (dotting). Differences to saline + saline group are expressed as \\\* p \< 0.001. Differences between fentanyl + saline and 3 mg/kg JMV2959 + fentanyl effects are expressed as ^§§§^ p \< 0.001. The horizontal arrow shows intervals with appropriate significant changes (^§§§^); the oblique arrows show the time of administration of J/S = JMV2959/saline and F/S = fentanyl/saline.](ijms-18-02486-g001){#ijms-18-02486-f001} ###### Effects of growth hormone secretagogue receptor (GHS-R1A) antagonist JMV2959 administered following four 20 min baselines into the ventral tegmental area (VTA) or into the nucleus accumbens (NAC) always 5 min before fentanyl/saline on the accumbens AEA levels (means ± SEM). The effects are illustrated as follows: (a) Ringer's solution/VTA + fentanyl (filled circle), 2 µg JMV2959/VTA + fentanyl (open circle), 10 µg JMV2959/VTA + fentanyl (open square), 2 µg JMV2959/VTA + saline (continuous line), 10 µg JMV2959/VTA + saline (open triangle), and Ringer's solution + saline (dotting); and (b) saline + fentanyl (filled circle), JMV2959 lower dose 8 mM/15 min/NAC + fentanyl (open circle), JMV2959 higher dose 40 mM/15 min/NAC + fentanyl (open square), JMV2959 lower dose 8 mM/15 min/NAC + saline (continuous line), JMV2959 higher dose 40 mM/15 min/NAC + saline (open triangle), and Ringer's solution + saline (dotting). Differences between treatments and the control group (Ringer's/VTA + saline or saline + saline) are expressed as \\\* p \< 0.001, \\ p \< 0.01, \* p \< 0.05. Differences between fentanyl and fentanyl in combinations with the higher JMV2959 dose (10 µg/VTA or 40 mM/15 min/NAC, respectively) are expressed as ^\#\#\#^ p \< 0.001, ^\#\#^ p \< 0.01. Differences between fentanyl and fentanyl in combinations with the lower JMV2959 dose (2 µg/VTA or 8 mM/15 min/NAC, respectively) are expressed as ^§§§^ p \< 0.001, ^§§^ p \< 0.01. The horizontal arrows show intervals with appropriate significant changes (^§§§^ or ^\#\#\#^); the oblique arrows show the time of administration of J/S = JMV2959/saline or J/R = JMV2959/Ringer´s solution and F/S = fentanyl/saline. ![](ijms-18-02486-g002a) ![](ijms-18-02486-g002b) ![Effects of GHS-R1A antagonist JMV2959 3 mg/kg i.p. on the fentanyl-induced accumbens 2-AG concentration. JMV2959 was administered following three 20 min baselines and 20 min before fentanyl/saline (means ± SEM). The effects are illustrated as follows: saline + fentanyl (filled circle), 3 mg/kg JMV2959 + fentanyl (open circle), 3 mg/kg JMV2959 + saline (open triangle), saline + saline (dotting). Differences to saline + saline group are expressed as \\\* p \< 0.001, \\ p \< 0.01, \* p \< 0.05. Differences between fentanyl + saline and 3 mg/kg JMV2959 + fentanyl effects are expressed as ^§§§^ p \< 0.001. The horizontal arrow shows intervals with appropriate significant changes (^§§§^); the oblique arrows show the time of administration of J/S = JMV2959/saline and F/S = fentanyl/saline.](ijms-18-02486-g003){#ijms-18-02486-f003} ![Effects of ghrelin receptor antagonist JMV2959 administered following four 20 min baselines into the VTA or into the NAC 5 min before fentanyl/saline on the accumbens 2-AG levels (means ± SEM). The effects are illustrated as follows: (a) Ringer's solution/VTA + fentanyl (filled circle), 2 µg JMV2959/VTA + fentanyl (open circle), 10 µg JMV2959/VTA + fentanyl (open square), 2 µg JMV2959/VTA + saline (continuous line), 10 µg JMV2959/VTA + saline (open triangle), and Ringer's solution + saline (dotting); and (b) saline + fentanyl (filled circle), JMV2959 lower dose 8 mM/15 min/NAC + fentanyl (open circle), JMV2959 higher dose 40 mM/15 min/NAC + fentanyl (open square), JMV2959 lower dose 8 mM/15 min/NAC + saline (continuous line), JMV2959 higher dose 40 mM/15 min/NAC + saline (open triangle), and Ringer's solution + saline (dotting). Differences between treatments and the control group (Ringer's/VTA + saline or saline + saline) are expressed as \\\* p \< 0.001, \\ p \< 0.01, \* p \< 0.05. Differences between fentanyl and fentanyl in combinations with the higher JMV2959 dose (10 µg/VTA or 40 mM/15 min /NAC, respectively) are expressed as ^\#\#\#^ p \< 0.001, ^\#\#^ p \< 0.01, ^\#^ p \< 0.05. Differences between fentanyl and fentanyl in combinations with the lower JMV2959 dose (2 µg/VTA or 8 mM/15 min, respectively) are expressed as ^§§§^ p \< 0.001, ^§§^ p \< 0.01, ^§^ p \< 0.05. The horizontal arrows show intervals with appropriate significant changes (^§§§^ or ^§§^ or ^\#\#\#^); the oblique arrows show the time of administration of J/S = JMV2959/saline or J/R = JMV2959/Ringer´s solution and F/S = fentanyl/saline.](ijms-18-02486-g004){#ijms-18-02486-f004} ![Effects of GHS-R1A antagonist JMV2959 3 mg/kg i.p. on the fentanyl-induced accumbens γ-aminobutyric (GABA) concentration. JMV2959 was administered following three 20 min baselines and 20 min before fentanyl/saline (means ± SEM). The effects are illustrated as follows: saline + fentanyl (filled circle), 3 mg/kg JMV2959 + fentanyl (open circle), 3 mg/kg JMV2959 + saline (open triangle), saline + saline (dotting). Differences to saline + saline group are expressed as \\\* p \< 0.001. Differences between fentanyl + saline and 3 mg/kg JMV2959 + fentanyl effects are expressed as ^§§§^ p \< 0.001. The horizontal arrow shows intervals with appropriate significant changes (^§§§^); the oblique arrows show the time of administration of J/S= JMV2959/saline and F/S = fentanyl/saline.](ijms-18-02486-g005){#ijms-18-02486-f005} ###### Effects of ghrelin receptor antagonist JMV2959 administered following four 20 min baselines into the VTA and into the NAC 5 min before fentanyl/saline on the accumbens GABA levels (means ± SEM). The effects are illustrated as follows: (a) Ringer's solution/VTA + fentanyl (filled circle), 2 µg JMV2959/VTA + fentanyl (open circle), 10 µg JMV2959/VTA + fentanyl (open square), 2 µg JMV2959/VTA + saline (continuous line), 10 µg JMV2959/VTA + saline (open triangle), and Ringer's solution + saline (dotting); and (b) saline + fentanyl (filled circle), JMV2959 lower dose 8 mM/15 min/NAC + fentanyl (open circle), JMV2959 higher dose 40 mM/15 min/NAC + fentanyl (open square), JMV2959 lower dose 8 mM/15 min/NAC + saline (continuous line), JMV2959 higher dose 40 mM/15 min/NAC + saline (open triangle), and Ringer's solution + saline (dotting). Differences between treatments and the control group (Ringer's/VTA + saline or saline/saline) are expressed as \\\* p \< 0.001, \\ p \< 0.01, \* p \< 0.05. Differences between fentanyl and fentanyl in combinations with the higher JMV2959 dose (10 µg/VTA or 40 mM/15 min/NAC, respectively) are expressed as ^\#\#\#^ p \< 0.001, ^\#\#^ p \< 0.01, ^\#^ p \< 0.05. Differences between fentanyl and fentanyl in combinations with the lower JMV2959 dose (2 µg/VTA or 8 mM/15 min/NAC, respectively) are expressed as ^§§§^ p \< 0.001, ^§§^ p \< 0.01, ^§^ p \< 0.05. The horizontal arrows show intervals with appropriate significant changes (^§§§^ or ^\#\#^ or ^\#\#\#^); the oblique arrows show the time of administration of J/S = JMV2959/saline or J/R = JMV2959/Ringer´s solution and F/S = fentanyl/saline. ![](ijms-18-02486-g006a) ![](ijms-18-02486-g006b) ![Effects of ghrelin receptor antagonist JMV2959 (3 mg/kg i.p.) on behavioural changes induced by fentanyl are illustrated in four monitored categories as means of behavioural scores (±SEM) separately: (a) locomotion; (b) stereotypies; (c) immobility; and (d) catalepsy. Behavioural changes during baseline period (intervals −60, −40, −20) and the pre-treatment with JMV2959/saline (0), are followed by 20--180 min of fentanyl/saline effects (20--180 min). The behavioural effects are illustrated as follows: saline + fentanyl (filled circle), JMV2959 + fentanyl (open circle), saline + saline practically identical with JMV 3 mg/kg + saline (cross with dotting). Differences between fentanyl and JMV2959 + fentanyl effects are expressed as ^§§^ p \< 0.01, ^§^ p \< 0.05. The horizontal arrows show intervals with appropriate significant changes (^§§^); the oblique arrows show the time of administration of J/S = JMV2959/saline or J/R = JMV2959/Ringer´s solution and F/S = fentanyl/saline.](ijms-18-02486-g007){#ijms-18-02486-f007} ![Schematic locations of: dialysis probes in the nucleus accumbens shell (a); and sites of infusions into the ventral tegmental area (b). Schematic locations of probe tips in animals which were involved in analyses of accumbens neurotransmitter concentrations (the bold lines indicate the dialyzing positions on the [Figure 7](#ijms-18-02486-f007){ref-type="fig"}a) and locations of JMV2959/Ringer's solution administrations into the VTA (dark dots in the lower part of slices on the [Figure 8](#ijms-18-02486-f008){ref-type="fig"}b) as described in the atlas of Paxinos and Watson \[[@B113-ijms-18-02486]\]. The distance from bregma (in mm) is indicated on the left of each schematic.](ijms-18-02486-g008){#ijms-18-02486-f008}	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1} =============== Magnesium valproate is chemically known as magnesium 2-propylpentanoate. Magnesium valproate is an anticonvulsant used in the treatment of epilepsy and bipolar disorder, as well as other psychiatric conditions requiring the administration of a mood stabilizer \[[@B1]--[@B6]\]. Chemical structures of magnesium valproate and four of its process related impurities are given in [Figure 1](#fig1){ref-type="fig"}. To our knowledge, there is no paper describing an ultraperformance liquid chromatographic (UPLC) method that allows the separation of magnesium valproate and its impurities in bulk drugs. Some articles exist on isocratic liquid chromatographic methods for the determination of magnesium valproate \[[@B7], [@B8]\]. However, these isocratic methods use short columns and are suitable for assay only since they focus on the main peak. The gas chromatographic methods and colorimetric methods are also in literature \[[@B9]--[@B11]\]. The methods for determination of valproic acid and sodium valproate are also useful for achieving the best results in this work \[[@B12]--[@B14]\]. Some of the bioanalytical methods are also available in literature \[[@B15]--[@B21]\]. Applying the above methods for the separation of impurities in a bulk magnesium valproate sample gives poor separation. Therefore, an attempt was made to develop a new, rapid, and sensitive method for the determination of magnesium valproate and its process related impurities. To access the reproducibility and wide applicability of the developed method, it was validated as per international code of harmonization norm, which is also mandatory \[[@B22]\]. This paper also deals with the validation of the developed UPLC method for the assay of magnesium valproate from its bulk and pharmaceutical dosage form. 2. Material and Methods {#sec2} ======================= 2.1. Chemicals and Reagents {#sec2.1} --------------------------- Parth Laboratories Pvt. Ltd. (Rajkot, India) has provided magnesium valproate and its process related impurities working standards and samples. Ammonium dihydrogen orthophosphate for HPLC, orthophosphoric acid (HPLC grade), and HPLC grade acetonitrile were purchased from Spectrochem Pvt. Ltd. HPLC grade water used was purified by Milli-Q Elix-3 water purification system. 2.2. Instrumentation {#sec2.2} -------------------- The Waters Acquity UPLC chromatographic system was used to perform development and validation. This system consists of a binary solvent manager, multiple wavelength ultraviolet detector, sample manager, and column oven connected to a multi-instrument data acquisition and processing system Empower 2.1 version. Sartorius microbalance and Equiptronics branded balance and heating oven was used for the weighing and heating purpose while Spinco ultrasonic bath was used for degassing purpose. 2.3. Method Development {#sec2.3} ----------------------- Analytical method development consists of the following steps which can be bound up by the literature survey, previous experience, and chemical nature of the reagents used in the development. 2.4. Mobile Phase Selection {#sec2.4} --------------------------- On the basis of literature survey, several exploratory runs have been performed but initially proper selectivity and resolution between the drug substance and its impurities were not achieved. After furnishing more importance to the literature, it was concluded that since one impurity specifically 2-(1-ethyl-methyl)pentanoic acid is a structural isomer of the drug component (valproic acid), the 5 mM ammonium dihydrogen orthophosphate with pH = 3.0 has given the maximum resolution with acetonitrile as organic component at isocratic elution, 55 : 45, v/v. 2.5. Column Selection {#sec2.5} --------------------- Column selection is the most important part in the method development. In this case most suitable column chemistry was bridge ethyl hybrid (BEH) C~18~ among the column chemistry available with us. Acquity BEH C~18~ (100 mm × 2.1 mm i.d., 1.7 μ particle size) has given the best outcomes. The resolution between the 2-(1-ethyl-methyl)pentanoic acid and the valproate salt was not achieved by the other columns. 2.6. Detection Wavelength Selection {#sec2.6} ----------------------------------- The standard solution was screened over 190 nm to 400 nm using the advantage of photodiode array detector. On the basis of peak absorption maxima and peak purity index, the 215 nm was decided as the detection wavelength which has provided the maximum chromatographic compatibility to the method. 2.7. Mobile Phase Preparation {#sec2.7} ----------------------------- The mobile phase consists of acetonitrile: 5 mM ammonium dihydrogen orthophosphate (pH = 3.0) (45 : 55) was prepared by dissolving 575 milligram of ammonium dihydrogen orthophosphate for HPLC in 1000 mL ultrapure (HPLC grade) water; then the pH of the buffer solution was adjusted up to 3.0 using 0.1% orthophosphoric acid solution. This was further mixed with acetonitrile by mentioned ratio and filtered through 0.22 μn filter followed by degassing in ultrasonic bath for 20 min. 2.8. Standard Solution Preparation {#sec2.8} ---------------------------------- Magnesium valproate and its related impurities working standard solution containing 200 μg/mL were prepared in a 100 mL volumetric flask by dissolving 20.0 mg each in 25 mL acetonitrile : water (50 : 50) and then diluting to volume with the same diluent. Further this was filtered through 0.22 μn filter followed by degassing in ultrasonic bath for 20 min. The chromatogram of working standard is given in [Figure 2](#fig2){ref-type="fig"}. 2.9. Sample Solution Preparation {#sec2.9} -------------------------------- Magnesium valproate and its related impurities sample (different batch) solution containing 200 μg/mL were prepared in a 100 mL volumetric flask by dissolving 20 mg each in 25 mL acetonitrile : water (50 : 50) and then diluting to volume with the same diluent. Commercially, magnesium valproate is available in 200 mg, 300 mg, and 500 mg pharmaceutical dosage form. Average weight of the tablet was found by weighing 20 tablets and then it was dissolved and diluted to achieve the 200 μg/mL concentration. Further this was filtered through 0.22 μn filter followed by degassing in ultrasonic bath for 20 min. The chromatogram of sample solution and commercial is shown in Figures [2](#fig2){ref-type="fig"} and [3](#fig3){ref-type="fig"}, respectively. 3. Results and Discussion {#sec3} ========================= With reference to the method development, chromatographic parameters used for method validation experiments are given in [Table 1](#tab1){ref-type="table"}. Method for the determination of magnesium valproate and its related impurities in bulk drug is further validated as per ICH Q2(R1) guideline \[[@B21]\]. Validation of analytical method was performed using magnesium valproate and its process related impurities working standard and sample batch drug substance. The accuracy experiment was performed by recovery study at three levels, 150%, 100%, and 50%, of the standard concentration. The working standard of magnesium valproate and its related impurities were added to the samples and the recovery was calculated which was between 98 and 102%; these were well within the acceptance criteria. The method precision was assessed using multiple preparations of a single sample. Six different preparations of the same magnesium valproate and its related impurities sample, each 0.20 mg/mL, were analyzed in triplicate on the same day. New solutions were prepared and analyzed on each of two successive days. The %RSD values obtained for the peak areas of magnesium valproate, valeronitrile, pentanoic acid, 2-ethyl pentanoic acid, and 2-(1-methyl, ethyl)pentanoic acid were not more than 2.0. The intermediate precision study was performed using another Acquity BEH C~18~ column (100 mm × 2.1 mm i.d., 1.7 μ particle size). The %RSD values were of the same magnitude as above. The linearity of the method was evaluated by analyzing eight solutions in the concentration ranges 50--350 μg/mL for each solution of magnesium valproate and its related impurities. The peak areas obtained from different concentrations of the drugs were used to calculate linear regression equations. These were y = 10.15x − 4.714, y = 10.06x + 21, y = 10.12x + 5.285, and y = 10.37x − 21.85 with correlation coefficients of R ^2^ = 0.999, R ^2^ = 0.998, R ^2^ = 0.999, R ^2^ = 0.997, and R² = 0.998 for pentanoic acid, 2-ethyl pentanoic acid, 2-(1-methyl, ethyl)pentanoic acid, valeronitrile, and magnesium valproate, respectively. The high values of the correlation coefficients were indicative of linear relationships between analyte concentration and peak area. The limits of detection (LOD) and quantification (LOQ) were established by evaluating the minimum level at which the analytes could be readily detected and quantified with accuracy, respectively. The LOD and LOQ for each component were 1.0 μg/mL and 2.0 μg/mL. The signal to noise ratio was more than 3 and 10 for LOD and LOQ, respectively. The selectivity of the method was evaluated by injecting blank matrix, each individual impurity, magnesium valproate standard solution, and spiked solution to check the interference of the diluent as well as the standard solution on each other. The method was proved as highly selective that there was no interference on any component to others. The separation factor (α) was investigated for all the impurities, which found 0.33, 0.64, 0.88, and 0.90 for pentanoic acid, 2-ethyl pentanoic acid, 2-(1-methyl, ethyl)pentanoic acid, and valeronitrile, respectively. The influence of five chromatographic parameters (k) on the separation was investigated. The parameters examined were the amount of acetonitrile in mobile phase, the pH of the ammonium dihydrogen orthophosphate solution, and the amount of ammonium dihydrogen orthophosphate solution in the mobile phase. No such impact of the small changes on the above parameters was observed which suggest that the method is highly robust. The developed UPLC method shows a good separation of magnesium valproate to its impurities. The robustness study indicated that mainly the pH of the ammonium dihydrogen orthophosphate solution should be monitored carefully to ensure the best separation as this has a significant effect on the separation. The method is found to be selective, precise, sensitive, and linear, which is also proved from the summary of method validation ([Table 2](#tab2){ref-type="table"}). The method can be used for the determination of magnesium valproate and the identification of the impurities present in the pharmaceutical dosage form. The authors are grateful to the National Facility for Drug Discovery through NCEs Development & Instrumentation Support to SMPEs, Department of Chemistry at Saurashtra University (Rajkot, India), for providing the state-of-the-art analytical facilities. They are thankful to UGC and DST (DPRP Programme), Government of India, for providing financial assistance in terms of meritorious research fellowship and research assistantship. They are also thankful to the M/s Parth Laboratories Pvt. Ltd. (Rajkot, India) for providing working standards and samples. Conflict of Interests ===================== The authors declare that there is no conflict of interests regarding the publication of this paper. ![Chemical structure of magnesium valproate and its process related impurities.](ISRN2014-412704.001){#fig1} ![Chromatograms of working standard and sample solution.](ISRN2014-412704.002){#fig2} ![Chromatogram of commercial sample of magnesium valproate.](ISRN2014-412704.003){#fig3} ###### Chromatographic conditions of proposed analytical method. Parameters Optimum condition ---------------------- ----------------------------------------------------------------- Mobile phase A 5 mM ammonium dihydrogen orthophosphate (PH = 3.0) Mobile phase B acetonitrile (HPLC grade) Column Acquity BEH C~18~ (100 mm × 2.1 mm i.d., 1.7 μ particle size) Flow rate 0.3 mL/min Isocratic elution at 55 : 45, v/v Detection 215 nm UV Diluent Acetonitrile : water (1 : 1) Injection volume 5 μL ###### Summary of validation study. Validation experiment PA 2-EPA 2-1-MEPA VA MgVA ------------------------------- ------------------------ ----------------------- ------------------------ ------------------------ ------------------------ Specificity No interference No interference No interference No interference No interference LOQ (μg/mL) 2.0 2.0 2.0 2.0 2.0 LOD (μg/mL) 1.0 1.0 1.0 1.0 1.0 Linearity $I$ Correlation coefficient 0.999 0.998 0.999 0.997 0.998 $II$ Regression equation y = 10.15x − 4.714 y = 10.06x + 21 y = 10.12x + 5.285 y = 10.37x − 21.85 y = 10.29x − 13.28 Method precision (n = 6) %RSD = 0.2 %RSD = 0.2 %RSD = 0.2 %RSD = 1.2 %RSD = 0.1 Int. precision (n = 6) %RSD = 1.2 %RSD = 0.8 %RSD = 0.8 %RSD = 1.8 %RSD = 0.7 Accuracy (% recovery) 99--101 98--100 99-100 98--102 99-100 Robustness No significant change No significant change No significant change No significant change No significant change [^1]: Academic Editor: Luigi Janiri	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== Distal radius fractures are common skeletal injuries and occurred at all ages of the general population \[[@CR1], [@CR2]\]. These types of fractures are reported as having one of the highest incidences accounting for over 15% of bone fractures \[[@CR3], [@CR4]\]. Distal radius fracture management usually includes a plaster cast, splint, and a moulded synthetic material cast to immobilise the injured upper extremity \[[@CR5]--[@CR9]\]. A normal course of the treatment includes the application of a cast and several follow-up clinical visits lasting four to six weeks \[[@CR7], [@CR10], [@CR11]\]. The traditional casts are described as having both poor ventilation and an improper fit, while also causing discomfort. These morbidities associated with conventional casts may result in cast complications such as cutaneous diseases, bone and joint injuries, or malunion \[[@CR7], [@CR12], [@CR13]\]. The rate of cast-related complications published is high with up to 31% being reported in published studies \[[@CR14]\]. The 3D printing technology is rapidly advancing in medical applications \[[@CR15]\]. In the manufacture of rehabilitation tools, 3D printing technology is being applied to orthopaedic cast fabrication to create patient-specific features with an appropriate fit and a ventilated structure \[[@CR16]\]. Jake Evill and Deniz Karasahin proposed a novel design of casts with a web-like structure and fabricated it by using a 3D printing technique. The cast models were built from 3D--scanned images of subjects' limbs and created by using computer-aided design software, which can generate a Stereolithography (STL) file, a standard file format widely used for 3D printing. Kim et al. developed a hybrid model of wrist orthosis composed of a 3D--printed frame and an injection moulding shell \[[@CR16], [@CR17]\]. Although 3D printing has made advances at a rapid pace in the development of casting techniques, all published technologies for the design and fabrication of 3D--printed casts are still in the concept stage without clinical application \[[@CR16]\]. To date, there is no clinical study investigating the application of 3D--printed casts \[[@CR18]\]. Our previous published study has developed a rapid and intelligent technique to create a patient-specific orthopaedic cast fabricated by 3D printing \[[@CR16]\]. Orthopaedic surgeons with little knowledge of engineering can design a high-quality cast in a short amount of time by utilising the technique. The novel cast also has additional features such as being ventilated and hygienic advantages that potentially lower the risk of complications. To our knowledge, there is no published clinical trial using the 3D--printed cast for the treatment of distal radius fractures. This study is to perform a pioneer clinical trial for the casting technology utilising 3D--printed casts for the treatment of distal radius fractures. Methods {#Sec2} ======= Ten patients (age range from 5 to 78 years old) including four males and six females are involved in the clinical trial. There were six patients who suffered distal radius and ulnar styloid fractures. Distal radius fractures were present in three patients. One patient sustained fractures of both the distal radius and ulna. Exclusion criteria included pathological/open fractures, fractures requiring internal fixation and patients who were not available for local follow-up. All clinical trials were performed in southern China where the weather was humid and warm, normally above 30 degrees Celsius during the clinical testing period. Patients first underwent closed reduction using traditional plaster cast fixation due to tissue swelling. The first treatment stage lasted for one week after swelling subsided. 3D--printed orthopaedic casts developed by our published techniques \[[@CR16]\] were applied to these patients after one week fixation. Two follow-up examinations and investigations were performed about the second and sixth week after the application of the 3D--printed cast \[[@CR11], [@CR19]\]. Patients' limb injury data are obtained from computerised tomography (CT) scanning or magnetic resonance imaging. An injured patient is first examined using radiography to identify the fracture type and locate the region of the injury. Physicians then perform the closed reduction followed by plaster casting in the initial phase. To obtain workable data for later cast design, both arms are scanned by a CT imaging system (Aquillion CX 64, Toshiba, Japan) or MR (Achieva 1.5 T, Philips) imaging equipment. The 3D scanning system using a 3D scanner, which was employed in our previous study \[[@CR16]\], is not utilised in the clinical trial due to the difficulty of the scanning procedure. To use a 3D scanner, patients must keep their arms at a position without movement for several minutes. This procedure is difficult to conduct. In China, the use of CT or MR imaging is recommended for further diagnosis and treatment of the injury. Before scanning, a technician makes marks on the patient's arm to indicate the scanning scope as shown in Fig. [1(a)](#Fig1){ref-type="fig"}. The technician helps the patient to bend his wrist at an angle (Fig. [1](#Fig1){ref-type="fig"} (b)), which is termed a casting angle, where the casting arm keeps the wrist at the same angle during the treatment period. The value of the casting angle is determined by an orthopaedic surgeon according to the injury details. The patient lies on the patient table of the scanning system, and then raises and bends his wrist at the casting angle (Fig. [1(c)](#Fig1){ref-type="fig"}). Both hands with symmetric postures are scanned to obtain raw data (Fig. [1(d)](#Fig1){ref-type="fig"}). Data from the other forearm without injury can be the alternative due to the swelling of the injured forearm during the initial inflammation stage.Fig. 1Scanning method: (a) determine the scanning scope by making marks on the patient's arm; (b) bend the wrist at the casting angle; (c) the patient lies on the table and bends both wrists at the casting angle; (d) scans are taken of both hands To evaluate the clinical trial, we designed questionnaires concerning clinical efficacy and patient satisfaction \[[@CR6], [@CR7], [@CR20]\]. Orthopaedic surgeons and patients worked together to complete the questionnaire after seven weeks of treatment. The design of these questionnaires is based on the existing questionnaire used in the hospital performing the clinical test and published studies \[[@CR7]\]. Two questionnaires have been designed including two groups of survey questions associated with treatment-related issues and patient satisfaction assessments \[[@CR7]\]. The first group lists four survey questions related to clinical effect based on expert opinions and previous investigation carried by Chinese orthopaedic surgeons (Table [1](#Tab1){ref-type="table"}). Pressure sore is an important assessment item that is a common complication caused by traditional casting \[[@CR14]\]. Other common complications, such as the stability of immobilisation, the severity of compromised blood flow, and pressure-related patient discomfort, are also included in the assessment. The first questionnaire is completed by the surgeon who exams patients at the second and sixth week after the application of a cast. Patient satisfaction or patient sensation and behaviour is another assessment including comfort, patient compliance, preference between 3D--printed and plaster cast, cast odour, and skin itchiness (Table [2](#Tab2){ref-type="table"}). As a pilot study, the assessment of the patients' preference addresses patients' opinions about the physical structure and wearer comfort. The inflammation-induced uncomfortable experience, especially during the initial stage of treatment, should be ruled out from the assessment. With the assistance from doctors, patients complete the second assessment questionnaire after the second and seventh weeks of casting. Doctors only explain the details of the questionnaire to let the patient fully understand questions without any personal recommendations to affect selection.Table 1Assessment of clinical effectiveness of a 3D--printed castAssessment ItemAssessment contents and grading standardexcellent-3good-2acceptable-1poor-0Stability of immobilisationNo loss of reductionSlight shift but no need for re-manipulationReinforced same castLoss of reduction requiring further procedureBlood circulationGood terminal circulation with a florid complexionVenous obstruction relief after physical movement or arm liftingPale skin, low temperature of the armSignificant ischaemia of involved limb, compartment syndromeWear-pressure-related painNo painSlight pain with a minor influence on sleepMild pain caused poor-quality sleepSevere pain caused difficulty falling asleepPressure soresNo abnormality of the skinNon-blanchable erythema of the intact skinSkin breakdown or bleeding blisterFull thickness skin loss Table 2Patient satisfaction questionnaireItemAssessment contents and grading standardExcellent-3Good-2Acceptable-1Poor-0Patient comfortVery comfortableOccasional irritationMedium irritationBad experience wearing castPatient complianceStrong willingness to try new castMinor doubtDubious but compliedAccepted reluctantlyPatient preference between 3D--printed cast and conventional plaster castOpted for 3D--printed cast without hesitationPreferred a 3D--printed cast to a plaster castEither cast is acceptableInsisted on using conventional castCast odour and smellNoneSlight cast odourSmelly cast after heavy sweatingStinky castSkin itchinessNo itchRarely itchyFrequent itch but tolerableSeverely itchy Patients' raw models are input into the intelligent system developed by our previously published study to perform patient-specific design as per clinical requirements \[[@CR16]\]. Short arm casts are designed for the treatment of distal radius fractures. A cast model is designed as a holed surface pattern with a flare on the lower edge near the elbow for the consideration of ventilation and wearer-friendly features (Fig. [2](#Fig2){ref-type="fig"} ). Fig. 2The cast model designed by our developed techniques Medically compatible materials such as polypropylene (PP) and polyamide (PA2200) are employed in the 3D printing fabrication of an orthopaedic cast. These materials are China Food and Drug Administration (CFDA) approved as Class I materials for rehabilitation devices. We utilise selective laser sintering (SLS) 3D printer EOS P395 (Germany) or Stereolithography (SLA) printer RS4500 (UnionTech, China) to produce the cast. Technically, post-processing for a 3D--printed cast is necessary for producing a final physical model. It includes padding a 3D--printed cast, mechanical grinding or rolling sharp edges, and adding fixation components. As shown in Fig. [3](#Fig3){ref-type="fig"}, fixation straps are mounted on the cast to adjust the assembly and create a cast that is tailor-fitted to an injured limb. In particular, cushion pads are glued on some specific anatomical regions, such as the wrist and ulnar styloid process, to avoid local high pressure and scratching of the skin.Fig. 3Post-processing includes adding the fixation strap and padding on the specific anatomical regions close to the wrist and ulnar head Results {#Sec3} ======= A personalised and 3D--printed cast is fabricated as a split structure with two half parts but still keeps the circumferential structure when applied to an injured extremity. A short arm cast extends from the mid-forearm to the distal, proximal crease \[[@CR8]\] (Fig. [4](#Fig4){ref-type="fig"}). In some mild cases with slight injuries of a forearm, a short arm cast can extend from the mid-forearm to the middle area between the wrist and distal crease (Fig. [4](#Fig4){ref-type="fig"}).Fig. 4A short arm cast extends from the mid-forearm to the distal or proximal crease. To consider the mobilisation of the wrist for a slight injury, a short arm cast can also extend from the mid-forearm to just above the wrist crease All patients prefer the custom-fit feature that establishes comfortable contact on the injured arm. The surrounding Velcro straps allow the adjustment of the assembly of the cast to accommodate the swelling forearm in the initial inflammatory phase of a fracture (Fig. [5](#Fig5){ref-type="fig"}). The 3D--printed cast satisfies the orthopaedic requirement for the treatment of a fracture in terms of the seven-week follow-up. The novel cast maintains fracture bone alignment and immobilises the forearm during the healing process. No loss of reduction is found in all participating patients. Moreover, no breakage occurred in any cast during the treatment period.Fig. 5The application procedure of a 3D--printed cast with a split structure. The split design and Velcro strap allow the cast to be adjusted and have custom-fitted features Pressure sores, common casting complications occurring in a traditional cast, were not present in any patients. However, one female patient (age 56, height 5′1″, weight 41 kg) had a blister with around 5 mm diameter on the bony prominence near the head of the ulna (Fig. [6](#Fig6){ref-type="fig"}). The patient adjusted the cast by herself without consulting with a physician. It then gave rise to a tight fit. No pain and no complaints were reported from the patient. The blister has disappeared one day after a physician adjusted the cast. A skinny arm can be easily bruised from the cast with relatively tight contact. This complication, skin breakdown, was not present in any patients even though thin patients were involved in the clinical trial.Fig. 6A blister caused by high local pressure near the region of the ulnar head Patients were taught how to check for compartment syndrome by themselves within one week of the application of the 3D--printed cast. As one of the most serious complications of traditional casting techniques, compartment syndrome was not found in any of the patients examined in the second and sixth week after casting. Orthopaedic surgeons inspected the skin appearance under the cast, and no visual signs of pressure damage were found in these patients. Complications associated with compromised blood flow commonly occurring when using circumferential casts and splints were not present during the healing period. The first group of assessment (Table [1](#Tab1){ref-type="table"}), clinical efficacy, was scored as 9.8 out of 12 and 30% of patients scored this group lower than 9 of the score. The lowest score was 8 out of 12 in the first group evaluation (Table [3](#Tab3){ref-type="table"}). The second group of the assessment (Table [2](#Tab2){ref-type="table"}), patient satisfaction to the application of the novel cast, was scored as 11.5 out of 15 (Table [3](#Tab3){ref-type="table"}). In this questionnaire, 80% percent of patients scored the new casting technique over 11 out of 15, and no patients scored fewer than 8. Both assessments of patient comfort and comparison between a plaster and a 3D--printed cast were assigned full marks. However, patient compliance was assigned as 2.1 out of 3 where the lowest score was rated.Table 3Assessment resultsClinical efficacyPatient satisfactionMean score9.8/1211.5/15Highest score11/1214/15Lowest score8/129/15 Discussion {#Sec4} ========== An appropriate casting technique not only holds the fracture reduction at a proper anatomic position but also minimises the risks of complications related to distal fracture. Complications including cutaneous diseases, compartment syndrome and vascular comprise, have been reported in conventional cast application due to the unbalanced pressures and high stiffness \[[@CR3], [@CR12]\]. Traditional casting creating mould utilising plaster and thermoplastic has rigid structure without flexibility and poor ventilation. In addition, the swelling of soft tissue occurring in an injured forearm at the initial stage makes it difficult to create patient-specific features. 3D--printed casts are featured as patient-specific and fully ventilated as well as lightweight structures \[[@CR16]\]. 3D--printing technology is an image-based technology combined with rapid prototyping that can create a patient-specific cast in terms of injury regions and severity. An orthopaedic surgeon performs the closed reduction for a displaced fracture followed by casting fixation. Patient-specific features of casting play an important role to maintain the alignment and avoid loss of closed reduction. Moreover, the custom-fit structure ensures the matching surface geometry between the cast and arm and thus disperses pressure. The ventilated structure featured in the novel cast confers the benefits of improved patient comfort and reduced risk of cutaneous complications. Acquisition of a patient's image is an important process for performing cast design. In the imaging process, fracture patients are required to hold a special posture to keep the forearm or lower leg in a natural position. Participants with fractures have swollen arms where the original shapes are difficult to be discerned. This study proposes the mirror technique, which scans the counterpart of an injured limb. Eight out of 10 testing casts used the contralateral arm. Bilateral symmetry exists in natural biological structures. For example, surgeons assume the bilateral symmetry of the human skeletal system for surgery purposes. Also, Islam et al. demonstrated the symmetric morphology of some human bones \[[@CR21]\]. Thus, the mirror technique offers relatively accurate patient data and minimises imaging difficulties for the patient. All patients participating in these clinical trials have completed the entire therapeutic course without negative clinical consequences. Rather, the patients in these trials had superior clinical outcomes (Fig. [7](#Fig7){ref-type="fig"}). No clinical trial using 3D--printed cast has been reported. The 3D--printed cast offers custom-fitted immobilisation during the entire treatment process. A cast with a correct fit can apply appropriate orthopaedic pressure on the injured arm to maintain bone alignment even after significant deformities occurred in the soft tissue. No patients found the loss of reduction in these clinical trials due to the technique that ensured they were fitted correctly.Fig. 7a A fractured forearm fitted with a cast; (b) the recovery of the broken arm with a superior clinical outcome Pressure-related complications are reported with traditional casts, especially circular casts. The cast utilised for this study are designed as splitting and circular structures with padding components, which can split the cast to accommodate the injured limb and relieve pressure. Slightly slow circulation is found after two weeks of cast application due to the wearing pressure. Participated patients can relieve the symptom through raising the injured arms. Slight pain is developed from wearing pressure, but no complaint is reported from patients. Wearing pressures are necessary for any casting or splint technology to maintain the reduction of the fracture and perform orthopaedic corrections effectively \[[@CR22]\]. Patient-specific and 3D--printed casts develop proper fit by the shape of the injured arm and thus reduce the risk of stress concentration since large contact areas are created. Concentrated stress is considered as an important factor associated with compartment syndrome. These technical advantages are present in the clinical trial so that no cases have compartment syndrome. Both clinicians and patients are highly concerned about pressure sores developing beneath the cast. To a patient with a distal radius or Colles' fracture, pressure sores frequently develop in some regions of a forearm, such as on the heads of the radius or ulna, due to their protruding shapes. High local stresses generated from wearing pressure occur in such regions. In the early two weeks of cast application, pink but not broken skin arises on the regions of heads of the radius or ulna accounting for 20% of participating patients using a 3D--printed cast without a padding layer. In addition, even though custom-fit features are created, a slight motion between the injured arm surface and the cast is common regardless of the type of the cast application. The pink skin on those regions results from high local pressure and motion-related scratch. Two approaches to improve design have been proposed in the 3D--printed cast: creating bump shape on those regions and padding the regions as shown in Fig. [8](#Fig8){ref-type="fig"}. Eight out of 10 patients were applied those improved casts. Moreover, the material of cast with high stiffness potentially developed a discomfort contact to the skin. Those two improvements potentially reduce the risk of high local pressure and gain clinical benefits to solve pressure-related skin problems. The improvements bring sound clinical outcomes that no reddened or pink skin arises from the casting arm.Fig. 8Projecting areas like the head of ulna or radius bone have an increased risk of developing high local wearing pressure under a cast. An improved design creates a bump shape to avoid direct contact between the cast and the underlying skin As a novel cast patients and clinicians never used, a 3D--printed cast gets high credit from patients with its ventilated, comfortable and fashionable design. All patients assign the highest score on the assessment of comfort. The patient-specific structure creates moderate contact like a fitted sleeve covering the injured arm and prevents injuries from external impacts and scratches. The ventilated structure keeps the microenvironment dry between the skin and the cast and reduces the risk of cutaneous complications. It should be noted that the lightweight structure still maintains a high level of strength. Engineering analysis performed in our previous study demonstrated that the structure of the printed cast was able to resist either normally imposed loads or accidentally impact loads \[[@CR16]\]. In this study, breakage due to poor strength of the structure was not found in any of the cases. The wearer-friendly design minimises the interference with daily activities of the patient. Patient compliance or treatment adherence is a challenge for a physician to carry out this research. The details of treatment technologies and potential side effects were informed by the physician. In the early stage of the clinical trial, most patients who show curiosity rather than suspicion on the application of the 3D--printed cast were mostly attracted by the novel 3D printing technology with non-interventional treatment. A few patients were sceptical since no such technology had been used in clinical applications. After two weeks of application of the novel cast, 100% of patients opted to utilise a 3D--printed cast instead of a plaster cast (Fig. [9](#Fig9){ref-type="fig"}).Fig. 9The cumbersome structure of a conventional plaster cast compared with a 3D--printed cast. Patients express a strong preference to utilise a 3D printed cast instead of a conventional cast There are some limitations of this study concerning casting technology and clinical testing. First, acquisition of a patient's data is a challenge for a patient with a fracture. A patient is asked to keep a required position during scanning for the purpose of data integrity. Additionally, we use mirror technology to acquire the counterpart of the fractured arm. Variation between arms is shown in the normal anatomy. Further investigations should be required to evaluate the bilateral symmetry of human bones digitally. There is no complaint about the fitting issue by using the mirror technique. The availability of biocompatible materials for 3D printing fabrication is limited. We used polypropylene (PP) and polyamide (PA2200) with Young's modulus 1300Mpa. As our experiences, a material with stiffness higher than 2000Mpa of Young's modulus, like acrylonitrile butadiene styrene (ABS), may cause uncomfortable contact to the skin. In addition, although a ventilated structure has been created, physicians receive a few of complaints about the smelly cast from participating patients who sweat a lot in hot weather. It should be noted that all clinical trials were carried out in a southern Chinese city with hot and humid weather during the testing period. The fabrication cost of a 3D printing cast is relatively high, about \$150 US Dollars, compared with conventional alternatives less than \$50 US Dollars. The rapid growth of 3D printing technologies would reduce the fabrication cost in the near future. As a pilot study, we are mainly concerned with the clinical feasibility of the application of the 3D--printed cast. Due to the limited amount of patients, the initial study investigated a small-sized group without performing statistical analysis for comparison among conventional and 3D--printed casts. As a new casting technology, we conducted the clinical trials with the focus on mild distal fracture. Cast design and 3D--printing fabrication would require a longer time. In emergency situations requiring a cast for the case of an acute fracture, as per our initial experiences, the 3D--printed cast would not be suggested currently. Its personalised structure and ventilated as well as lightweight design bring about quality patient experiences as confirmed by positive assessments of patient comfort and acceptance. Those two assessments would be helpful for the improvement of the current casting technology. This pilot study provides an initial experience on how to apply the 3D printing technique in the treatment of fractures. The future study would perform clinical trials on a larger scale sample size, and as a comparison study among conventional and 3D--printed casts to assess whether clinical differences were statistically significant. Conclusions {#Sec5} =========== This study performs a pilot study with the focus on the clinical trial for the treatment of distal radius fracture using patient-specific and 3D printing cast developed by our published techniques. The novel cast performs circumferential support for the fractured forearm and prevents the injury from external impact. The patient-specific design maintains the alignment of fracture bones and creates the custom-fit and moderate wearing pressure to avoid compartment syndrome and pressure sores, which is considered as the casting complication to challenge conventional casting technology. The ventilated structure and fashionable design of the novel cast combined with the 3D printing fabrication increase patient comfort and satisfaction. Superior clinical outcomes have been obtained from clinical trials. The authors wish to acknowledge Diya Ma, Matthew-Lun Wong, Ka-Long Ko, Ka-Hei Ko and Jin-Peng Lee for their important contributions to the software development. Funding {#FPar1} ======= The work described in this paper was supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No.: CUHK 14113214), grants from the Innovation and Technology Commission (Project No: ITS/149/14FP, GHP/028/14SZ, ITS/293/14FP), grants from CUHK Technology and Business Development Fund (Project No.: TBF16MED002, TBF16MED004), a grant from The Science, Technology and Innovation Commission of Shenzhen Municipality (Project No.: CXZZ20140606164105361), and a grant from The Scientific Research Project of Guangdong Province (Project No.: 2014B090901055). YC performed the clinical trial, designed the questionnaires and edited the manuscript. HL conceived the cast design, carried out the clinical trial, and both drafted and edited the manuscript. LS carried out data collection and analysis. XZ carried out data collection and provided clinical advice. WH conducted the clinical trial and critically revised the manuscript. DW carried out the method design and development and critically revised the manuscript. All authors read and approved the final manuscript. Ethics approval and consent to participate {#FPar2} ========================================== The study concerning the client's right to privacy. The study was approved by the Ethical Committee of The Third Affiliated Hospital of Southern Medical University (reference number: 201,603,006) and was conducted according to the principles of the Declaration of Helsinki. The experimental methods used in the study were in accordance with medical practice, and the subjects were protected. Competing interests {#FPar3} =================== The authors declare that they have no competing interests. Publisher's Note {#FPar4} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.	{ "pile_set_name": "PubMed Central" }
Memory is a ubiquitous part of our daily experience. It is not just the occasional act of remembering an event or an appointment, or being able to recognize an individual as someone seen in the past, the benefits of previous experience facilitate all aspects of thinking, taking action, and perceiving. Therefore, it is not surprising that the organization of memory, the manner in which it is expressed or used, and its representation in the brain have been of substantial interest to philosophers, scientists, and laypeople alike. The idea that memory can take multiple forms was advocated in the early 1800s by the philosopher Maine de Biran, who proposed that representative, mechanical, and sensitive memory were differentially associated with conscious awareness, rigid/inflexible behavior, and emotional lability (Maine de Biran, [@B65]; see also Eichenbaum and Cohen, [@B32]). Empirical investigations conducted by modern-day cognitive psychologists and cognitive neuroscientists have extended these views, identifying different classes of memory and characterizing their permanence over time, how they might be expressed in thought and action, and whether or not they depend on conscious apprehension of encoded and/or retrieved content. Indeed, it seems that early philosophical ideas were not far from the mark, as much of the driving force behind contemporary research involves identification and differentiation of memory systems and the neural substrates that support them. Initial evidence for the memory systems view came from now-classic reports (e.g., Milner, [@B70]) of spared and impaired memory performances following damage to structures located in the medial aspect of the temporal lobes (MTL; i.e., the hippocampus and adjacent cortical structures in the parahippocampal region). These studies, reviewed briefly below, invariably indicated that performances on tests of memory that proceed relatively automatically (i.e., in the absence of effortful retrieval attempts) and without conscious awareness were largely spared, while those that required amnesic patients to consciously recover (e.g., via recognition or recall) information learned in the very same sessions was devastatingly impaired. Based on reports of such dissociations, it was proposed that one of the fundamental distinctions between memory systems is whether or not they support conscious access to remembered content. This view that memory systems divide on consciousness has now become so ensconced in the field as to be considered essentially a first principle from which new findings can be anchored. Indeed, it has been proposed that one could infer whether or not performance on a particular task was supported by explicit memory processes based on the presence or absence of MTL activity in neuroimaging data (cf. Henson, [@B57]). Only recently have we come to realize that the putative systemic division of labor between conscious and unconscious memory is not so clean. Therefore, a primary goal of this review is to examine recent evidence that has been advanced against the view that the hippocampus is selectively critical for conscious memory. As part of this objective, we review evidence for hippocampal involvement in successful memory expression (e.g., in eye movement behavior) when explicit responses are incorrect, hippocampal contributions to learning when participants are unaware of the contingencies that support increasingly proficient responding, and hippocampal recruitment during encoding when materials are presented subliminally (i.e., masked from conscious awareness). Along the way, consideration is given to criticisms that have been levied against these findings (provided that they exist), and potential explanations for differences in the reported results are proposed. In addition, some of the methodological and interpretive pitfalls that can undermine claims for or against hippocampal contributions to memory without awareness are considered; these are described in the context of the associative priming literature, which has a relatively long history in the investigation of questions about memory, awareness, and the MTL. It is important to note that findings are reported from several neuropsychological investigations and that many of these were conducted with patients of mixed etiology (e.g., Alzheimer\'s disease, Korsakoff\'s syndrome) who were selected based on their documented memory impairment. When findings can be more confidently attributed to the hippocampus (e.g., because recruitment was limited to patients with relatively circumscribed hippocampal lesions) we say so in the text. Ultimately, it is concluded that we have indeed reached a tipping point, and that the role of the hippocampus in memory has more to do with the processing and representational demands of the task at hand than with conscious awareness. The implications of this conclusion for some theoretical perspectives of hippocampal function are briefly considered, and some suggestions are made as to how best to proceed in future work. Historical context: dissociation of conscious and unconscious memory expression following amnesia ================================================================================================= The nature of the memory impairment following amnesia, and by extension, the role of the hippocampus and adjacent MTL cortical structures in learning and memory has been a source of considerable interest and intense theoretical debate. As described briefly above, one question that has been subject to much disagreement in the literature concerns the proposed selective role of the hippocampus in conscious (i.e., aware or explicit) expressions of memory (cf., Postma et al., [@B85]; Reder et al., [@B88]; Henke, [@B54]). While it is generally agreed that conscious recollection, which involves calling to mind the components of past events and experiences, depends critically on hippocampal integrity, the role of this structure in unconscious (i.e., unaware or implicit) expressions of memory has generally been regarded as far more speculative. Such speculation is well-founded, as several early neuropsychological findings documented dissociations between certain forms of learning, which were intact in amnesia, and the ability to consciously recognize or report information encountered in the context of the associated learning experience. As we shall see, this work inspired theoretical frameworks of hippocampal function that hinged on conscious access to remembered content (e.g., Moscovitch, [@B73]; Squire, [@B115]; see also Graf and Schacter, [@B41]), though not all theoretical perspectives share this view (e.g., Cohen and Eichenbaum, [@B20]; Henke, [@B54]). The critical role played by the hippocampus in conscious retrieval of new memories was evident in the seminal description of the famous amnesic patient H.M., which contains anecdotal descriptions of behavior based on informal interactions and reports of impaired performances on standardized neuropsychological tests of memory (Scoville and Milner, [@B106]; see also Milner et al., [@B71]). In each case, the ability of this individual to lay down lasting long-term memory representations amenable to explicit report was severely impaired. For example, when queried a half-hour after taking lunch, it was indicated that H.M. could not recall having gone to lunch, let alone exactly what had been enjoyed or the company he had been in. And in formal testing it was noted that "once he had turned to a new task the nature of the preceding one could no longer be recalled, nor the test recognized if repeated" (Scoville and Milner, [@B106]; p. 108). Not long after these initial impairments were documented, what has since been described as an astonishing demonstration of intact learning and long-term retention of novel information was reported by Brenda Milner (Milner, [@B70]; see also Milner et al., [@B72]). In this work, H.M. learned, over the course of several trials, to trace within the lines of an object using only the image reflected from a mirror (i.e., the mirror tracing task; see Figure [1A](#F1){ref-type="fig"}). In keeping with prior observations, this example of preserved motor skill learning was accompanied by an inability to accurately describe the circumstances in which the new skill was acquired and severely impaired recognition of the apparatus that had been used during testing. Such results have since been replicated with the additional striking report that H.M.\'s retention of the acquired skill was still evident as many as 359 days after initial exposure to the task (Gabrieli et al., [@B36]). Importantly, it was also demonstrated that this finding was not limited to mirror tracing as follow-up investigations indicated that H.M. could successfully acquire, and subsequently retain, new motor skills in the context of several other experimental tasks (e.g., rotary pursuit, bimanual tracking, tapping; cf. Corkin, [@B23]). ![Classic evidence for intact unconscious memory following amnesia. (A) Left: Illustration of the mirror tracing apparatus. The paper and the participant\'s hand are blocked from view, and tracing (within the lines of the star) is accomplished via the reflection of the image in the mirror. Right: Performance of patient H.M. on the mirror tracing task as reported by Milner ([@B70]; see also Milner et al., [@B72]). The number of tracing errors (i.e., tracing outside of the boundaries) declined across trials and sessions illustrating learning and retention of this motor skill. (B) Left: Illustration of the materials used to examine memory for primed items. Participants are presented with increasingly complete pictures of objects across trials and are asked to name the object, even in its degraded form. Right: Performance of amnesic patients and controls on the picture naming task as reported by Warrington and Weiskrantz ([@B130]). Both groups made fewer errors (i.e., provided fewer incorrect names, even for degraded images) when the picture set was presented repeatedly across several trials and sessions illustrating a reprocessing or priming effect. (C) Left: Illustration of word triplets used in the mirror-reading task. Some of these triplets were seen just once, others were repeated several times during testing. Participants were to read each triplet as quickly as possible when it was presented. Right: Performance of amnesics and controls on the mirror reading task reported by Cohen and Squire ([@B22]). Mean reading times of patients and controls decreased consistently across several days of testing providing evidence for intact perceptual skill learning in amnesia. Patients did not benefit as much as controls from repeated presentations of the same triplets. \[Figure [1A](#F1){ref-type="fig"}, right, reprinted with permission from Elsevier; Figure [1B](#F1){ref-type="fig"}, right, reprinted by permission from Macmillan Publishers Ltd; <http://www.nature.com/>; Figure [1C](#F1){ref-type="fig"}, right, reprinted with permission from AAAS; <http://www.aaas.org/>\].](fnhum-06-00080-g0001){#F1} The above results provided initial empirical evidence for the multiple memory systems perspective and they also inspired investigators to question whether or not intact long-term learning went beyond the domain of motor skills. That this was indeed the case, was initially reported by Warrington and Weiskrantz ([@B130]; see also, Milner et al., [@B71]) who showed that amnesic patients, like matched controls, could more readily identify pictures of fragmented words and objects (i.e., identify them in an increasingly degraded form) with repeated exposure to the stimulus set (see Figure [1B](#F1){ref-type="fig"}). Facilitated identification of degraded pictures was retained from day to day during testing, and savings was documented for one patient even after a 3 month retention interval. This type of learning is an example of repetition priming, which refers to a benefit in performance that accrues with repeated exposure to particular items when memory is tested indirectly---that is, without reference to a memory test or to the fact that some of the items had been seen previously (cf. Schacter, [@B99]; Richardson-Klavehn and Bjork, [@B92]). Such effects have also been documented when participants are instructed to complete three letter stems (e.g., POS\_\_\_\_\_) with "the first word that comes to mind." Unbeknownst to the participant, a subset of these stems correspond to words encountered during an earlier study phase (e.g., POSTURE), and priming is evident to the extent that participants complete the stems with studied words more often than would be expected by chance (i.e., without the associated study exposure). With these instructions, which simply orient participants to the stem-completion requirement, the task is "treated more as a 'guessing game' than a formal test of memory" (Warrington and Weiskrantz, [@B130]; p. 974), and amnesic patients perform much like controls. In contrast, when instructions reference prior experience, as in cued recall or recognition, learning effects observed in amnesia are much reduced (e.g., Warrington and Weiskrantz, [@B130]; Squire et al., [@B117]). Following on the heels of these reports, intact performances on cognitive (e.g., Tower of Hanoi; see Cohen, [@B19]) and perceptual skill learning tasks were also observed (Cohen and Squire, [@B22]). In one of these experiments (Cohen and Squire, [@B22]), amnesic patients acquired a pattern analyzing skill (i.e. reading mirror reversed text) as quickly as a control group, but failed to show the normal amount of additional facilitation to a subset of word triplets that were presented 20 times over the course of the experiment (see Figure [1C](#F1){ref-type="fig"}). The small amount of repetition-based facilitation that was evident in the performances of amnesic patients was subsequently considered akin to effects of perceptual priming, and controls were said to have shown greater benefit of repetition because unlike patients, they had explicit knowledge of the repeated items and could likely retrieve the remaining two words after having read the first (Cohen, [@B19]). Indeed, when participants were asked about the occurrence of repetitions, patients indicated that there were none, whereas controls reported them spontaneously. A more objective assessment revealed that explicit recognition was severely impaired among amnesic patients, even for the subset of words seen 20 times previously. Such results provide strong evidence for dissociation of skilled performance on the one hand and conscious recollection of the events surrounding learning on the other. Models of hippocampal function: a range of perspectives on consciousness ======================================================================== It was based on reports like those described briefly above that the initial distinction between declarative ("knowing that") and procedural ("knowing how") memory systems was drawn (Cohen and Squire, [@B22]). According to this perspective, the hippocampal system (i.e., the hippocampus along with the entorhinal, perirhinal, and parahippocampal cortices) is critically involved in the representation of a particular type of memory, declarative memory, which is selectively compromised in amnesia. Because memory for the procedures or operations that support normal acquisition of skilled performance was largely preserved, it was proposed that a different brain system must support procedural memory. Borrowing from the nomenclature of Bruner ([@B8]), knowledge garnered by the procedural memory system was said to be without record (inaccessible to verbal reports; absent any explicit knowledge of events surrounding learning), and to reflect the tuning and modification of dedicated cortical and subcortical processors that are engaged during the performance of a particular task. Declarative memory was said to be with record (consciously accessible; amenable to verbal report) and to reflect the combined outcomes of having engaged said processors (Cohen, [@B19]; Squire and Cohen, [@B115a]). The resulting declarative memory database "provided the basis for access to facts acquired during the course of experience and for conscious recollection of the learning experiences themselves" (Cohen, [@B19]; p. 97). Most important for the purpose of this review, contemporary versions of this theory continue to share in common with a handful of other models (e.g., Moscovitch, [@B73]) the proposed dichotomy between aware and unaware expressions of memory, and their dependence or not on structures in the hippocampal system (cf. Squire, [@B115]; Squire et al., [@B116]). The proposed link between conscious awareness and hippocampal system function has perhaps been articulated most strongly in the component process model developed by Moscovitch and colleagues (e.g., Moscovitch and Umilta, [@B75]; Moscovitch, [@B73]). According to the model only information that has been consciously apprehended is encoded by the hippocampal system, which then integrates the elements of that event, and the conscious experience of it, into a memory trace. An index of this memory trace is laid down in the hippocampus and can subsequently be used to recover the components of previously experienced events when associatively bound cues are available. The components themselves (e.g., faces, objects, words) are represented in domain-specific posterior neocortical (and other) modules that process and store perceptual and conceptual records of stimulus input. As an aside, processing mediated by these domain-specific modules is said to be sufficient for item-specific (e.g., repetition priming) or skill-based (e.g., mirror tracing) implicit memory expression, a view that is generally consistent with what has been proposed in the context of other models described here (e.g., Cohen and Eichenbaum, [@B20]; Squire, [@B115]; Henke, [@B54]). Returning to the hippocampal system, explicit recovery of memories for past events and experiences (i.e., episodic memory) depends on an effective cue, which triggers the hippocampal index and elicits automatic and obligatory recovery of associated elements as well as the conscious experience of the original event. The retrieval process is said to be automatic because events are often remembered without intention---they "pop into mind much as preattentive perceptual stimuli pop out of their background" (Moscovitch, [@B73]; p. 260)[^1^](#FN1){ref-type="fn"}. Important for our purposes, a distinguishing feature of this model as it was originally articulated is that consciousness is built into newly established hippocampus-dependent memory representations, and is automatically recovered during retrieval. Notably, a recent extension of the component process model seems to exercise a bit more flexibility with respect to consciousness of hippocampus-supported memory expression (Moscovitch, [@B74]; see also Sheldon and Moscovitch, [@B108]). Specifically, it is now suggested that conscious recollection involves (at least) two separate, but interacting stages of processing. During stage one, studied content is retrieved automatically and obligatorily without concomitant awareness in a process mediated by the hippocampal system. Successful associative retrieval at this stage of processing may be revealed when indirect testing methods are used (e.g., speeded lexical decision, eye tracking). During the second, slower stage of processing, retrieved content becomes consciously accessible and can influence explicit responses (e.g., on recall or recognition tests) in a process said to be mediated by interaction of the hippocampal system with prefrontal and perhaps parietal lobe structures. Based on this proposal, one might infer that when the second stage of processing is disrupted or deficient indirect testing measures will reveal evidence of hippocampus-dependent memory in the absence of associated awareness (cf. Hannula and Ranganath, [@B48]). While there is an undeniable link between the MTL memory system and conscious memory expression, there is sharply declining acceptance of the hypothesis that conscious memory is the linchpin function of the hippocampus. For example, in a direct extension of the declarative memory theory, Cohen and Eichenbaum ([@B20]; see also Eichenbaum et al., [@B33]) proposed that structures in the parahippocampal region (i.e., the perirhinal, parahippocampal, and entorhinal cortices) support memory for individual items and that the hippocampus itself is critically involved in binding together arbitrary (i.e., non-derivable) relationships among items encountered in the context of some scene or event. A key property of the resulting relational memory representations is compositionality---individual items retain their representation in memory independent of, and in addition to, the larger relational structure in which they are bound. Compositionality permits flexible memory expression which means that relational memory retrieval is not rigidly tied to the particulars of the original learning experience or the environment associated with memory acquisition; hippocampus-dependent relational memories can be expressed in variety of contexts and in response to any number of relationally bound cues. In contrast, rigidly bound (or fused) associations, those that lack flexibility and compositionality, are said to be the province of structures in the parahippocampal region (e.g., Eichenbaum and Bunsey, [@B31]; Pascalis et al., [@B82]). On this view, "the hippocampal system plays only an indirect role in consciousness---it organizes the database, so to speak, on which other brain systems may operate and, in so doing, determines the structure and range of conscious recollections" (Cohen et al., [@B21]; p. 148; see also Ryan and Cohen, [@B94]; Moses and Ryan, [@B77]). The modified version of the component process model (Moscovitch, [@B74]), which postulates a two-stage retrieval process, appears to share at least certain aspects of this view---namely, that processing mediated by the hippocampal system during retrieval provides the raw materials (i.e., reinstates previously bound elements) that may then be communicated with other brain regions to support conscious memory performance. Importantly though, neither the component process model nor the declarative memory theory (cf. Squire et al., [@B116]) distinguish between types of representation mediated by the hippocampus and adjacent MTL cortical structures as is done in the relational memory theory. One final model to be considered here, which diverges most noticeably from views of hippocampal function in which conscious awareness plays a prominent role, was proposed recently by Henke ([@B54]). According to the model, memory systems are best distinguished not on the basis of conscious apprehension, but rather by the type of processing they support. In this case, conscious awareness of encoded and retrieved content does not influence which processing mode is engaged, and therefore does not figure into the model. Instead, processing modes and the memory systems on which they depend are differentiated based on the speed of memory acquisition (i.e., rapid, single trial learning vs. slow, incremental learning), the number of elements to be encoded (i.e., single items vs. associations), and the flexibility of the resulting memory representations. Rapidly encoded novel associations that lend themselves to flexible memory expression are said to depend upon interactions between the hippocampus and specialized neocortical processing sites. When encoding is temporally extended (i.e., involves more than one study exposure) the hippocampus might augment performance early in learning, but would not be required for successful memory acquisition. Instead, temporally extended learning is said to depend critically on interaction of neocortical structures with the basal ganglia, the cerebellum, or the parahippocampal gyrus depending upon task demands. The end result of having engaged in repeated processing of the same information is an inflexible, rigidly bound associative memory representation. A final processing mode, which involves rapid encoding of single (or unitized) items, is said to depend critically on the parahippocampal gyrus. While this model clearly shares several features in common with the relational memory theory, no strong claims have been made in the context of the relational memory theory as to whether or not hippocampus-mediated encoding can proceed subliminally. More generally, the proposal that the hippocampus is absolutely unnecessary for learning when repeated study exposures are provided represents a departure from the relational memory theory, which emphasizes representational characteristics of hippocampus-dependent memory above all else. As can be seen, the models of hippocampal (and hippocampal system) function described here, which represent a subset of those proposed in the literature, provide a range of perspectives on the relationship between the hippocampus and conscious apprehension of retrieved and/or encoded content (for a summary see Table [1](#T1){ref-type="table"}). In the paragraphs that follow, evidence for and against the proposed role of the hippocampus in memory without awareness is evaluated. For ease of comparison to the models described above, these studies are considered separately as a function of whether they were designed to investigate hippocampal involvement in implicit memory retrieval, implicit learning that develops over the course of several experimental trials, or implicit (and in the examples provided, subliminal) encoding. First, however, findings from the associative priming literature are used to illustrate some of the investigative pitfalls that can complicate interpretation of results reported in studies of unconscious learning and memory. As part of this objective, we highlight two important issues: (1) the potential for explicit memory contamination on indirect tests of memory, and (2) the need to characterize the nature of the underlying memory representations that support performance on these tests. Following discussion of these issues, we shall see that the associative priming literature provides some initial evidence in favor of the view that the hippocampus does indeed play a critical role in unconscious memory expression even when strict criteria are in place to minimize explicit contributions to performance. ###### Summary of hippocampal models and associated perspectives on requirements for consciousness of encoding and retrieval. Model Representative citation MTL brain regions Type of memory Consciousness of Encoding Consciousness of Retrieval --------------------------------- ----------------------------- ----------------------------------------- ------------------------------------------------------------------- ------------------------------- -------------------------------- Declarative memory theory Squire et al. ([@B116]) Hippocampus and MTL cortical structures Conscious memory for facts and events Yes Yes Component process model Moscovitch ([@B73]) Hippocampus and MTL cortical structures Components of consciously apprehended events Yes Yes Two-stage model of recollection Moscovitch ([@B74]) Hippocampus (PFC and parietal cortex) Consciously recollected events N/A Stage 1---No Stage 2---Yes Relational memory theory Eichenbaum et al. ([@B33]) Hippocampus Flexible, compositional inter-item and item-context relationships -- Not required MTL cortical structures Single items and inflexible (fused/unitized) associations -- Not required Processing model Henke ([@B54]) Hippocampus Rapidly encoded, flexible associations Not required Not required Parahipp. Gyrus (BG and cerebellum) Incrementally learned, inflexible associations Not required Not required Parahipp. gyrus Single items and inflexible (fused/unitized) associations Not required Not required MTL cortical structures include the perirhinal, parahippocampal, and entorhinal cortex; BG, basal ganglia; PFC, prefrontal cortex; --, not explicitly described in the model. Note that the 2-stage model of recollection is an extension (and subset) of the component process model. Associative priming: identifying investigative pitfalls in studies of "unconscious" memory ========================================================================================== The initial collection of investigations designed to identify spared and impaired memory performances in amnesia, described briefly in the opening section, provides strong support for theoretical positions that postulate a divide in memory systems based on whether or not conscious awareness accompanies demonstrations of successful retrieval. The priming literature, for example, demonstrated normal or near normal retrieval of studied content, but only when indirect testing methods were used (cf. Schacter, [@B99]). A limitation of these studies was the selective assessment of memory for individual items. Whether or not the same pattern of performance would be observed for tests of memory that required participants to form new associations among items that did not have pre-existing relationships had yet to be examined. In the event that they could be documented, demonstrations of intact priming for novel associations in amnesia would further substantiate claims that the hippocampus was not required for, and does not support, unaware expressions of memory. In the mid-1980s, two independent research groups provided initial reports of intact associative priming in memory impaired individuals (Graf and Schacter, [@B41]; Moscovitch et al., [@B76]). In both studies, implicit memory for unrelated word pairs was spared despite gross impairment of explicit memory for the same materials. This outcome seemed not to depend upon the exact testing methods that had been used, as different paradigmatic approaches were adopted by the two groups. In one case (Graf and Schacter, [@B41]), the dependent measure was successful cued stem-completion. Following a study phase in which participants generated sentences that incorporated both words from unrelated cue-target pairs (e.g., KINDLY-STICK → the kindly woman picked up the stick) an indirect stem-completion test was administered. Results showed that both patients and controls were more likely to complete stems with studied targets when cue-target pairs were intact (e.g., KINDLY-STI\_\_\_) than when they were not (e.g., IRRATIONAL-STI\_\_\_). In the remaining case (Moscovitch et al., [@B76]), the dependent measure was reading speed. During a learning phase participants read several unrelated word pairs (e.g., MERCHANT-TRIBUTE, DAMPNESS-SILENCE, JOURNEY-TREMBLE) and were instructed to study them. Shortly thereafter, with speeded reading instructions, patients and controls read lists of intact pairs (e.g., MERCHANT-TRIBUTE) more quickly than lists of re-arranged pairs (e.g., DAMPNESS-TREMBLE). Because the individual words had been studied in both experimental conditions, the resulting facilitation was attributed to implicit memory for previously presented associations. Based on these outcomes, it was concluded that the hippocampus and related limbic structures were not critical for the initial acquisition and subsequent retention of novel associations; rather, their role was to "make newly acquired information available to consciousness" (Moscovitch et al., [@B76]; p. 345). At face value these results seem difficult to reconcile with the relational memory theory (Cohen and Eichenbaum, [@B20]) and related models (Henke, [@B54]), which propose a critical role for the hippocampus in relational memory processing and representation with or without awareness. But, as described in the sections that follow, the results reported in these early experiments were not always successfully replicated. In the wake of these failed replications, researchers became increasingly convinced that explicit memory processes were contributing to cued stem-completion performance (e.g., Shimamura and Squire, [@B110]; Bowers and Schacter, [@B6]; Reingold and Goshen-Gottstein, [@B91]; McKone and Slee, [@B69]; Gooding et al., [@B38]), and that speeded reading was often mediated by inflexible perceptual or production-based memory representations (e.g., Poldrack and Cohen, [@B84]). According to the relational memory theory (Cohen and Eichenbaum, [@B20]; Eichenbaum et al., [@B33]; Eichenbaum and Bunsey, [@B31]; Cohen et al., [@B21]) and the processing model proposed by Henke ([@B54]), this type of representation should not depend critically on hippocampal integrity. Because issues of awareness and representation require careful consideration in any claim for or against a role for the hippocampus in unconscious memory, relevant findings from the associative priming literature are considered in the sections that follow. Confounding effects of explicit memory contamination ---------------------------------------------------- One of the greatest challenges posed by investigations of implicit memory is the development of a paradigm that can be used to index past experience absent any confounding effects of conscious awareness. Without rigorous tests, explicit contributions to performance can go undetected and the use of incidental instruction alone does not eliminate this confound (e.g., Bowers and Schacter, [@B6]). Following the initial report of intact cued stem-completion in amnesia (Graf and Schacter, [@B41]), reanalysis of the original data, combined with a new follow-up investigation, revealed that only those patients characterized as having "mild" memory deficits showed intact associative priming (Schacter and Graf, [@B102]). Indeed, with just one exception (Cermak et al., [@B12]) work conducted with amnesic patients by other labs indicated that any benefit of context reinstatement in the cued stem-completion paradigm was negligible (Cermak et al., [@B13]; Mayes and Gooding, [@B67]; Shimamura and Squire, [@B110]). Insight into what might be driving this idiosyncratic pattern of results was provided by experiments conducted with neurologically intact college-age participants who completed post-test awareness questionnaires. Based on their responses, it was found that only those participants classified as "test-aware" showed associative priming on the stem-completion test (Bowers and Schacter, [@B6]; McKone and Slee, [@B69]; Gooding et al., [@B38]). Even more problematic, a subset of participants confided that they had deliberately completed the stems with a word that had not been studied (McKone and Slee, [@B69]; Gooding et al., [@B38]). Based on this evidence, there is general agreement that associative priming elicited by cued stem-completion is influenced by explicit memory contamination (cf. Schacter et al., [@B101]), and under these circumstances impaired performance among amnesic patients is not surprising. This line of work is instructive because it illustrates how explicit retrieval processes can contaminate performance on indirect tests of memory. However, it is important to note that the reported outcomes do not completely preclude the possibility that associative priming was mediated by unconscious processes on a subset of the trials, or the possibility that performance on a given trial might reflect both unconscious and conscious influences. Consistent with the first possibility, Bowers and Schacter ([@B6]) suggested that test-awareness may have developed consequent to several successful completions when intact cue-target pairs were presented. This proposal was challenged by McKone and Slee ([@B69]) who had "aware" participants identify the pair from the stem-completion task that they felt had triggered their awareness. Based on these subjective reports it was determined that awareness developed early in testing, and summary statistics limited to trials preceding awareness provided no evidence for associative priming. Importantly, caution was advised in the interpretation of this result because it was based on so few trials. Of additional concern, because the awareness questionnaire was administered after completion of the indirect test, participants may have misestimated when awareness emerged or may not have even had access to this information (Tversky and Kahneman, [@B121]). Under these circumstances, they may have simply selected the first pair from the list that they recognized now, after having been informed of the link between study and test. Even if we assume that associative priming in the cued stem-completion paradigm was never strictly implicit, performance on individual trials may have been influenced by both unconscious and conscious memory processes (cf. Gooding et al., [@B38]). This possibility is consistent with the recently proposed modifications of the component process model described above, in which it is indicated that conscious recollection is a two-stage process (Moscovitch, [@B74]; Sheldon and Moscovitch, [@B108]). Specifically, because cued stem-completion tests are not speeded, and therefore provide ample time for the second stage of processing to proceed, awareness of successful completions may emerge even if the initial retrieval process was not consciously mediated. To the extent that both processes are implicated in cued stem-completion performance, the relative absence of associative priming in amnesia may indeed reflect genuine implicit associative memory impairment. This possibility cannot be dismissed on the basis of post-test awareness questionnaires, and would likely benefit from additional investigation with methods that can more precisely estimate the time-course of memory processes that emerge over the course of individual experimental trials (e.g., event-related potentials, magnetoencephalography). Characterizing memory representations that support performance -------------------------------------------------------------- Consistent with findings that conscious memory processes are relatively slow to develop (e.g., Yonelinas and Jacoby, [@B136]), alternative methods that emphasized rapid responding were also used in associative priming investigations to more effectively minimize explicit memory contamination. As indicated earlier, intact associative priming in memory impaired individuals was initially documented not only with stem-completion, but also in a study that used speeded reading as the dependent measure (Moscovitch et al., [@B76]). The first attempted replication of this result was unsuccessful (Musen and Squire, [@B78]), but in this case, associative priming was absent from the performances of patients and controls alike, which raised concerns that there may have been insufficient power to detect a statistically significant difference in reading time across lists of intact and recombined word pairs (Light et al., [@B64]; Poldrack and Cohen, [@B84]). In line with this possibility, affirmative evidence for associative priming in speeded reading was reported in a subsequent experiment conducted with college-age participants (Poldrack and Cohen, [@B84]). Having replicated the result, the investigators went on to examine the nature of the newly acquired memory representations. Results showed that reading-based response time facilitation for intact pairs at test was much reduced when the first and second word in a pair had swapped positions (e.g., study: WINDOW-REASON; speeded reading: REASON-WINDOW) and was completely absent when a word was interposed between the paired items (e.g., study: WINDOW-REASON; speeded reading WINDOW and REASON). Similar results were reported in a speeded lexical decision task when paired words were presented sequentially, rather than simultaneously, disrupting previously established perceptual gestalt (Goshen-Gottstein and Moscovitch, [@B39]). Such findings imply that new learning was due to the acquisition of an inflexible or rigidly bound memory representation that could only support associative priming when word pairs were presented exactly as they had been seen or spoken during encoding. Under these circumstances performance may reflect the operation of the same specialized neocortical processors that have been implicated in single item priming and procedural learning (Henson, [@B57]), or the operation of binding mechanisms in the parahippocampal region that can support memory for unitized word pairs (e.g., Quamme et al., [@B87]; Haskins et al., [@B51]; see also Erikson and Desimone, [@B34]). Because specialized neocortical processors are not typically compromised in amnesia, and because there may be some sparing in the parahippocampal region, amnesic patients might be expected to show normal levels of associative priming when basic perceptual properties encountered during study are reinstated on the indirect test. Results largely confirm these expectations (Gabrieli et al., [@B37]; Goshen-Gottstein et al., [@B40]; Verfaellie et al., [@B122]). However, when this constraint was not met---for example, when paired words were presented sequentially rather than simultaneously---amnesic performance has typically been impaired (Paller and Mayes, [@B80]; Savage et al., [@B97]; Carlesimo et al., [@B11]); this is a finding to which we will return. Notably, at least one set of results is not easily accommodated by the above dichotomy. Despite the use of methods that retained the perceptual gestalt of intact pairs when memory was tested indirectly, Yang et al. ([@B134]) reported that associative priming was impaired, and indeed absent, among amnesic patients who completed their experiment. This finding is even more remarkable because participant recruitment was limited to patients who had relatively selective MTL lesions, and several procedural details were in place to preclude explicit contamination. Critically, a follow-up fMRI investigation conducted with neurologically intact participants localized the associative priming effect to the parahippocampal cortex (Yang et al., [@B135]). To the extent that performance on this task was supported by a fused representation of studied associations, such results would be predicted by the relational memory theory and by Henke\'s processing model; in contrast, the reported outcome presents a challenge for the declarative memory theory, which proposes that memory processes mediated by MTL structures are necessarily available for conscious report (e.g., Squire, [@B115]). So as not to leave any loose ends, and because one of these investigations is especially relevant to questions about a potential role for the hippocampus in implicit memory, we now return to findings of associative priming impairment with sequential testing procedures noted briefly above (Paller and Mayes, [@B80]; Savage et al., [@B97]; Carlesimo et al., [@B11]). As you may recall, effects of associative priming were abolished when such procedures were used with college-age participants, an effect that was attributed to disruption of perceptual gestalt (Goshen-Gottstein and Moscovitch, [@B39]). So why then was associative priming successfully documented among controls recruited to participate in the neuropsychological investigations? It seems likely that combined use of multiple encoding exposures (i.e., between two and six across experiments) and requirements to perform at least two different elaborative encoding tasks during the study phase may have been contributing factors. Such manipulations may have minimized the acquisition of fused perceptual representations, supporting instead, the acquisition, and subsequent use of higher-order relational memory representations. These representations, which by some accounts (e.g., the relational memory theory) are said to depend critically on hippocampal integrity, may then have been better suited to support successful associative priming among controls when sequential testing procedures were used. One of these investigations (Carlesimo et al., [@B11]) is particularly notable because associative priming was absent only from the performances of amnesic patients who had selective hippocampal damage; those with amnesia consequent to thalamic or basal forebrain damage, in whom hippocampal volumes were within normal limits, had associative priming scores that were fully intact, and even numerically larger than the same effect in control data. That explicit contamination was unlikely to have influenced performance is suggested by savvy methodological choices---intact pairs were rarely encountered on the indirect test (i.e., the number of trials was large, but few of these trials were intact pairs) and the first item from each pair was presented very quickly (i.e., 300 ms) and immediately replaced by the second item, which remained in view until participants indicated whether it was a word or a non-word. This approach provides little incentive to engage in effortful retrieval attempts and little time for strategic processing of the first word from a given pair. Consequently, this result provides strong evidence for hippocampus-mediated associative priming, even in the absence of awareness. Converging evidence for hippocampus-supported unconscious memory ================================================================ The number of investigations that point to hippocampal involvement in learning and memory without awareness has increased substantially in recent years, making a strict division of labor across memory systems based solely on subjective memory outcome less tenable (cf., Reder et al., [@B88]; Henke, [@B54]; Dew and Cabeza, [@B28]). While some of the reported findings have been subject to debate, the evidence does indeed suggest that we have reached a tipping point. In what follows, recent evidence for (and in some cases against) hippocampal involvement in unaware expressions of memory is presented. As indicated earlier, these studies are considered separately based on whether they were designed to investigate hippocampal involvement in implicit memory retrieval, implicit learning, or implicit encoding. As will be seen, some of the models described earlier can accommodate the reported findings more easily than others. Evidence for hippocampus-supported retrieval without awareness -------------------------------------------------------------- We have already seen at least one example from the associative priming literature that provides compelling support for the proposal that hippocampal function extends beyond conscious memory expression (i.e., Carlesimo et al., [@B11]). The evidence is convincing not only because the task was carefully designed to minimize explicit contributions to performance, but also because impairments were evident only among amnesic patients with selective hippocampal damage. These results are complemented by converging reports of impaired associative priming from a handful of other investigations that also used sequential testing procedures (Paller and Mayes, [@B80]; Savage et al., [@B97]) and by reports of impaired associative priming when information from different modalities (e.g., voice-word associative priming) must be bound to support performance (Schacter et al., [@B100]; see also Curran and Schacter, [@B26]). However, because patients in the latter studies were of mixed etiology, impaired performances cannot be attributed unambiguously to the hippocampus. A converging methods approach could help disambiguate these findings, but unfortunately most neuroimaging investigations have focused on identification and characterization of the neural substrates of single item priming (Henson, [@B57]). Among the associative priming studies utilizing neuroimaging methods (Mayes et al., [@B68]; Badgaiyan et al., [@B4]) concerns have sometimes been raised that performance may have been subject to explicit memory contamination, which then precludes conclusions about a role for the hippocampus in unconscious memory. Another approach used to examine hippocampus-dependent implicit memory expression takes advantage of the fact that eye movement behavior is influenced by prior exposure to visual materials (see Hannula et al., [@B46] for review). In several experiments, it has been demonstrated that eye movements distinguish studied from novel items (e.g., Althoff and Cohen, [@B2]), and that patterns of viewing can provide evidence of memory for inter-item and item-context relationships (e.g., Ryan et al., [@B93]; Hannula et al., [@B49]). These eye-movement-based memory effects are expressed rapidly and obligatorily in patterns of viewing (e.g., Ryan et al., [@B96]), occur well in advance of explicit recognition responses (e.g., Hannula et al., [@B49], [@B47]), and have even been documented in the absence of conscious access to retrieved content (e.g., Ryan et al., [@B93]). When combined with neuropsychological or neuroimaging methods, eye movement investigations provide insight into hippocampal function. For instance, it has been demonstrated that like neurologically intact controls, amnesic patients with hippocampal damage make fewer fixations and sample fewer distinct regions of previously studied as compared to novel items (e.g., faces, buildings, scenes; Althoff and Cohen, [@B2]; Ryan et al., [@B93]) and they do so despite poor recognition of the same materials (Althoff et al., [@B3]; Althoff, [@B1]). This intact eye-movement-based reprocessing, or repetition, effect resembles findings of intact repetition priming described earlier. In contrast, amnesic patients fail to benefit from prior exposure when relational memory is tested. In one of these investigations (Ryan et al., [@B93]), participants were presented with pictures during a test phase while their eye movements were recorded. Some of these pictures were seen previously and remained unchanged, others had also been seen previously, but were now systematically manipulated such that relationships among scene elements had changed (e.g., soap to the left of the sink was now to the right of the sink). Under these circumstances, neurologically intact individuals looked disproportionately at the part(s) of the scene that had been manipulated, and did so even when they failed to report the difference explicitly. The same eye-movement-based relational memory effect was completely absent from the viewing behavior of amnesic patients, including those with damage limited to the hippocampus (Ryan and Cohen, [@B95]). Such results suggest that the hippocampus is critical for relational memory, irrespective of whether or not said memory is accessible to awareness, and are predicted by the relational memory theory (Cohen and Eichenbaum, [@B20]) and by Henke\'s processing model (2010). Recently, it was reported that disproportionate viewing of scene changes was only evident when participants were aware that the change had been made (Smith et al., [@B112]). Based on this null result, it was said that patients in the study conducted by Ryan et al. ([@B93]) failed to show the viewing effect because it depends on awareness. However, changes in viewing behavior absent awareness have been successfully documented in several other experiments (Hayhoe et al., [@B52]; Hollingworth et al., [@B59], [@B60]; Hollingworth and Henderson, [@B61]; Henderson and Hollingworth, [@B53]; Beck et al., [@B5]). As such, the result reported by Ryan et al. ([@B93]) can arguably be said to provide a second strong demonstration of hippocampal involvement in implicit expressions of memory. Complementing these results, recent work indicates that eye-movement-based memory effects may precede conscious awareness (e.g., Hannula et al., [@B49]; see also Hannula et al., [@B47]), and that activity in the hippocampus predicts these viewing effects even when explicit recognition has failed (Hannula and Ranganath, [@B48]). In these experiments (Hannula et al., [@B49]; Hannula and Ranganath, [@B48]), participants studied scene-face pairs and were tested with three-face displays superimposed on studied scenes. All three faces in each test display were studied, and on a subset of the trials, one of the faces had been paired with the scene during the encoding phase. When a preview of the scene was provided, disproportionate viewing of the associated face was evident approximately 500--750 ms after three-face display onset and preceded explicit recognition of that face by as much as 1000 ms[^2^](#FN2){ref-type="fn"}. This preferential viewing effect was evident within the same time frame regardless of the instructions provided at test, and even when it was counterproductive to view the associate. When amnesic patients with hippocampal damage were tested, viewing did not distinguish the associated face from the other two faces in the display, which provides evidence for relational memory impairment in these individuals. A follow-up investigation conducted with neurologically intact participants, which combined fMRI and eye tracking methods, showed that hippocampal activity differences during presentation of the scene cue predicted the viewing effect, even when analyses were limited to trials in which participants failed to identify the associated face in their explicit recognition responses (Hannula and Ranganath, [@B48] see Figure [2](#F2){ref-type="fig"}). Activity differences associated with recognition accuracy were localized to lateral prefrontal cortex, and these differences, greater for correct than for incorrect responses, were correlated with those in the hippocampus during presentation of the three-face test displays. Like the findings reported by Ryan et al. ([@B93]), such results implicate the hippocampus in relational memory retrieval, even without awareness. Consistent with the two-stage model of conscious recollection, the results further suggest that successful explicit retrieval may depend upon interaction of the hippocampus with prefrontal regions. ![Illustration of the basic paradigm and results from Hannula et al. ([@B49]) and Hannula and Ranganath ([@B48]). (A) Examples of scene-face pairs presented during the study trials, along with a single, associated test display in which the face on the left was the match (i.e., was the associate of the scene). Each test trial began with the presentation of a scene cue meant to prompt retrieval of the associated face. (B) Eye movements from a representative participant superimposed on the test display shown above. Fixations are indicated by white circles and the size of the circle was proportionate to the amount of viewing time directed to the fixated region. Transitions from one fixation to the next are indicated by red lines. (C) Proportion of total viewing time directed to correctly identified matching faces vs. faces that were merely selected from displays that did not contain a match broken down into 250 ms time bins following the onset of the three face test display. Neurologically intact control participants showed disproportionate viewing of the matching face just 500--750 ms after the faces were presented; no evidence of relational memory was evident in the eye movement behavior of amnesic patients. (D) Bilateral regions of the hippocampus for which BOLD signal was greater for incorrect trials during presentation of the scene cue when subsequent viewing of the match was high vs. when subsequent viewing of the match was low in college-age participants. Trial-averaged time courses extracted from the left and right hippocampal regions, respectively, show differences in BOLD signal between high and low viewing trials during presentation of the scene cue. This result illustrates hippocampal recruitment, even when explicit memory has failed.](fnhum-06-00080-g0002){#F2} Evidence for hippocampus-supported learning without awareness ------------------------------------------------------------- Hippocampal contributions to implicit memory have also been documented in investigations that give rise to learning that unfolds gradually over the course of several experimental trials or blocks. As was the case for the retrieval paradigms, evidence demonstrating hippocampus-supported implicit learning has been obtained in investigations using multiple research methods (e.g., fMRI, neuropsychological investigation) and across a variety of experimental tasks (i.e., contextual cueing, visual statistical learning, sequence learning, transitive inference), each of these are described in turn below. Perhaps more than any other investigative approach, the contextual cueing paradigm, developed by Chun and Jiang ([@B15]), has generated considerable debate about the existence of hippocampus-dependent implicit memory expression (e.g., Chun and Phelps, [@B17]; Manns and Squire, [@B66]; Greene et al., [@B44]; Preston and Gabrieli, [@B86]; Westerberg et al., [@B131]). Critically, this paradigm has also been subject to rigorous empirical investigation that has provided important insights into the nature of the memory representations that support performance (e.g., Jiang and Wagner, [@B62]; Brady and Chun, [@B7]). Although the proposal that learning in this task is truly implicit (e.g., Chun and Jiang, [@B16]) has been challenged, even skeptics have confirmed that contextual cueing is evident in the performances of "unaware" participants (Smyth and Shanks, [@B114]). Therefore, this extensive literature provides a strong context for evaluating conflicting evidence for and against hippocampal contributions to task performance absent concomitant awareness of the contingencies that support learning. Participants in contextual cueing experiments search for a target (a rotated T) among an array of distractors (rotated L\'s) and upon locating the target, make a button press response specifying whether it was rotated left or right. Over the course of several experimental blocks, a subset of the search arrays is presented repeatedly so that the distractors provide a stable spatial configuration predictive of target location; the remaining arrays are novel (i.e., presented just once) and have no predictive value (see Figure [3A](#F3){ref-type="fig"}). Results consistently show that the latency to locate the target decreases over the course of the experiment for both novel and repeated arrays (i.e., a general effect of skill learning), and that there is additional response time facilitation for repeated (vs. novel) arrays, which has been dubbed the contextual cueing effect. Based on early speculation that response time facilitation to repeated contexts depended ostensibly on the binding of target location to the global distractor configuration in which it was repeatedly embedded, it was proposed that contextual cueing should depend critically on structures in the MTL, particularly the hippocampus (Chun and Phelps, [@B17]). This proposal was notable because the seminal set of experiments documented the contextual cueing effect despite at-chance explicit recognition of repeated displays (Chun and Jiang, [@B15]). ![The contextual cueing task. (A) Illustration of the basic contextual cueing paradigm. Participants attempt to locate a target (i.e., a rotated T) from among an array of distractors (i.e., rotated L\'s) as quickly as possible upon display presentation. A subset of the displays are repeated several times over the course of the experiment. (B) Illustration of the results from Chun and Phelps ([@B17]). Both amnesic patients and controls show improved skill learning (i.e., faster target identification) over the course of the experiment, but patients show no evidence for contextual cueing. (C) Illustration of conditions used in the modified version of the contextual cueing task that was used to examine transfer effects, and was developed by Jiang and Wagner ([@B62]). Trained displays were seen during learning, and were also used to create recombined displays. Color coding (shades of blue) was not used in the experiment, but is meant to illustrate how the recombined display was created. Half of the elements in the display were from one trained display, the remainder were from the other. The lines connecting the display elements are meant to illustrate the differences in overall gestalt of the display configuration, and were not present during testing. Notice that the shape of the global configuration is different for the recombined display than for both trained displays. Novel displays controlled for any facilitation in responding that might be due to the target location alone. The standard skill learning effect is illustrated in the graph on the left. Effects of transfer are illustrated in the graph on the right. The time required to locate and respond to a target was not different for recombined and old displays, and both of these display types elicited faster responses than novel displays. Such results provide evidence for flexible transfer of learning from trained displays to recombined displays. \[Figure [3B](#F3){ref-type="fig"} reprinted by permission from Macmillan Publishers Ltd; <http://www.nature.com/>; Figure [3C](#F3){ref-type="fig"}, graphs, reprinted with kind permission from Springer Science+Business Media.\].](fnhum-06-00080-g0003){#F3} In an effort to identify the neural substrates of contextual cueing Chun and Phelps ([@B17]) conducted a neuropsychological experiment with MTL amnesic patients and a matched comparison group. Despite normal skill learning (i.e., decreased latency for novel and repeated arrays across experimental blocks), amnesic patients failed to demonstrate any additional facilitation for displays presented repeatedly over the course of the experiment (see Figure [3B](#F3){ref-type="fig"}). At first blush, this result seems strikingly similar to the one reported by Cohen and Squire ([@B22])---amnesic patients show normal skill learning, but fail to benefit from repeated presentation of the same displays. The critical difference is that participants in the comparison group were rarely aware that some of the displays had been presented repeatedly. And more importantly, whether they reported having noticed these repetitions or not, they failed to distinguish repeated from novel displays when an explicit recognition test was administered. Based on these findings, it was concluded that "the hippocampal system mediates contextual (relational) encoding in humans, without having to evoke conscious memory processes" (Chun and Phelps, [@B17]; p. 846). The results reported by Chun and Phelps ([@B17]) were soon replicated in healthy participants under the influence of midazolam, a pharmacological agent that can induce a temporary amnesia (Park et al., [@B81]; see also Chun, [@B14]). This effect was particularly striking because the same (neurologically intact) participants who demonstrated intact contextual cueing absent the effects of midazolam were impaired when the drug was administered. One notable short-coming of both investigations was lack of anatomical specificity. While hippocampal atrophy was reported for three of four patients tested by Chun and Phelps ([@B17]), one of whom was subsequently characterized as having damage limited to the hippocampus (Verfaellie et al., [@B123]), the remaining two had additional damage that extended beyond the hippocampus into adjacent MTL cortical structures. Results of a subsequent investigation (Manns and Squire, [@B66]) suggested that contextual cueing was only impaired for a subset of patients with extensive MTL lesions; those with damage limited to the hippocampus performed normally. This outcome challenges the claim that implicit expressions of learning in the contextual cueing task are mediated by the hippocampus. However, as indicated by Manns and Squire ([@B66]), residual hippocampal tissue in patients with limited damage (i.e., ∼32% average tissue loss with none greater than 50%) may have contributed to intact contextual cueing in their investigation, and the patient identified as having limited hippocampal damage in the original work reported by Chun and Phelps ([@B17]) failed to show a contextual cueing effect. Either way, the results are not easily accommodated by the declarative memory theory, as the evidence clearly points to a role for MTL cortical structures in unconscious learning and memory. With the above caveats in mind, it is notable that contextual cueing performance seems to be supported by (at least) two qualitatively distinct memory representations (Jiang and Wagner, [@B62]). Originally, it was assumed that with repeated exposure to the array, the global geometric configuration of the distractors was learned and that contextual cueing was supported by a bound representation of this configuration to the target location in memory (Chun and Jiang, [@B15]). While this does indeed seem to be the case (Jiang and Wagner, [@B62]), evidence suggests that spatial relationships between individual distractors and the target can also support contextual cueing. Specifically, it was found that if a target appears repeatedly in the same location for two separate arrays, each with a different global configuration, contextual cueing was robust even on a transfer test in which half of the distractors from the first array were recombined with half from the second array, a manipulation that changed the global configuration of display elements (see Figure [3C](#F3){ref-type="fig"}). Hence, two representations seem to support contextual cueing: (1) a fused representation of the global array configuration that, by some views (e.g., relational memory theory and Henke\'s processing model), could be supported by structures in the parahippocampal region and (2) a representation of spatial relationships between the target and individual distractors that can give rise to flexible memory expression when novel, recombined displays are presented. One possibility then is that spared performance on the contextual cueing task in hippocampal amnesia reflects preserved representation of global configuration in the parahippocampal region. To the extent that the two documented representations co-exist, one (i.e., the fused associative representation) might compensate for the other (i.e., the flexible relational representation) in cases of hippocampal damage. Whether and when in the course of learning these representations emerge remains to be determined, but the suggestion has been made that early learning might reflect spatial relationships between targets and individual distractors, and that representations of global configuration develop more slowly (Jiang and Wagner, [@B62]). Much like the neuropsychological work, results from recent fMRI investigations attempting to identify the neural substrates of contextual cueing have been mixed (Greene et al., [@B44]; Preston and Gabrieli, [@B86]; Westerberg et al., [@B131]). While one study reported implicit, facilitation-related activity differences in left posterior hippocampus and left parahippocampal cortex (Greene et al., [@B44]), others found that hippocampal activity differences were correlated with explicit recognition of repeated displays (Preston and Gabrieli, [@B86]; Westerberg et al., [@B131]). Notably, those studies failing to show implicit hippocampal differentiation included only trial number (or epoch) and stimulus type (i.e., novel vs. repeated arrays) as parametric factors in the reported fMRI analyses. Greene et al. ([@B44]) likewise did not find implicit hippocampal effects with just these two factors, but recognized that not all repeated arrays showed robust response time facilitation. Including reaction time as a factor in subsequent analyses had two consequences: (1) hippocampal differentiation was revealed as an interaction of stimulus type and reaction time in the left posterior hippocampus, such that the effect was only evident for repeated arrays with greater response time facilitation; and (2) the decrease in unexplained variance allowed the detection of a main effect of stimulus type located bilaterally in the posterior hippocampus. It is, therefore, advised that future investigation of contextual cueing incorporate this approach. As stated above, some neuroimaging studies have revealed hippocampal activity differences associated with explicit recognition of repeated displays (Preston and Gabrieli, [@B86]; Westerberg et al., [@B131]). In one of these investigations 13 of 23 participants performed above chance on the recognition test administered after learning (Preston and Gabrieli, [@B86]); using the same methods, Greene et al. ([@B44]) reported at-chance recognition among their participants. Such differences in behavioral outcome are difficult to explain, but may reflect slight modification of instruction, and/or differences in strategy adopted by the participants to perform the task. Indeed, changes in strategy, based for example on encouragement to guess, can improve recognition performance, but these performance improvements do not always reflect conscious retrieval processes (Voss et al., [@B125]; Voss and Paller, [@B127], [@B128]). The remaining study that reported hippocampal activity differences linked to awareness used a procedure in which participants were instructed about repetitions and told to use this information to their advantage while performing the task (Westerberg et al., [@B131]). Clearly, further investigation is needed to better pin down the variables that drive hippocampal recruitment in contextual cueing. Aside from incorporating the analysis technique described above, future work might also examine whether and how structures in the MTL, and especially the hippocampus, contribute to successful (flexible) transfer (as reported by Jiang and Wagner, [@B62]) when two displays associated with the same target location are used to create a novel configuration. The expectation here would be that activity differences in the hippocampus would support facilitated responding to recombined displays on the transfer test. Response time facilitation in the contextual cueing paradigm derives from the presentation of displays in which target location is predicted by the simultaneous presentation of an invariant context. As such, there is no requirement to integrate display elements across time. An alternative approach that has been used to investigate incidental learning involves manipulation of the transitional probabilities among individual items seen repeatedly over the course of an experiment or an experimental block. A subset of these items are presented in a repeated sequence so that one item has the potential to reliably predict, or cue, the appearance of another (e.g., Face A always precedes Scene A); the remaining items are also repeated, but have no predictive value. Recent work suggests that the hippocampus is engaged by these statistical learning paradigms, and is sensitive to the temporal co-occurrence of items embedded in repeated sequences (Turk-Browne et al., [@B119], [@B120]). In one of these investigations, the hippocampus was the only region that showed reliably enhanced activity during the presentation of the first element in a "paired" set, which suggests that it may be involved in the process of anticipating, or calling to mind, the (temporally) associated element as learning proceeds. This result resembles findings described earlier, in which hippocampal activity differences during presentation of a scene cue predicted viewing of the associated face (Hannula and Ranganath, [@B48]). In both cases, presentation of a cue may elicit retrieval of relationally bound memory representations, perhaps by way of pattern completion processes mediated by the hippocampus (Hannula et al., [@B49]). Importantly, results indicated that these effects reflect implicit learning because participants were unaware of repeated patterns when a post-test awareness questionnaire was administered; however, performance on a post-learning recognition test revealed greater than chance discrimination of sequentially presented pairs vs. foils. While a skeptic might argue that this result implicates conscious processes in learning, recent work (mentioned briefly above) has suggested that the relationship between recognition performance and explicit memory might not be so straightforward. Specifically, it has been found that when participants are encouraged to guess, implicit memory can support successful forced-choice responding (cf. Voss and Paller, [@B128]). Clearly, this is an important issue that will be a source of much future investigation, and is something to consider with respect to the interpretation of results from post-learning recognition tests as these are often administered with instructions that encourage participants to guess. Borrowing from Voss and Paller ([@B128]), researchers could incorporate confidence judgments that include a "guess" response in future work as a first step toward addressing the possibility that correct recognition on awareness tests reflects accurate guessing. As discussed earlier, representational flexibility is considered a key property of hippocampus-dependent memory representations. This flexibility underscores demonstrations of explicit memory, and is made possible because the complex relationships mediated by the hippocampus are necessary for disambiguation (pattern separation) and for the generativity of cued recall (pattern completion; cf. Norman and O\'Reilly, [@B79]). As suggested above, certain implicit tasks may necessarily involve the hippocampus if the associated task demands require pattern completion or pattern separation. One such task is the serial reaction time task (SRTT), which is performed implicitly, depends upon both pattern separation and pattern completion, and requires contributions of the hippocampus and surrounding MTL cortical structures to support performance. In the SRTT, a series of stimuli appear on a screen and each unique stimulus requires a distinct response (e.g., a particular button press) from the participant. Embedded in the series of stimuli are certain repeating sequences, and when these sequences are short, unambiguous, and occur with enough regularity, participants show faster reaction time for items cued by the sequence than for randomly occurring items (cf. Reed and Johnson, [@B89]). Based on the nature of this facilitation, and on findings of performances among amnesic patients that are comparable to those of controls, it has been said that this is a clear example of procedural learning. However, longer sequences may be used wherein the predictive value of multiple items is Markovian. In the simplest case, consider the sequence A-D-C-D-A-B-D-B. In this case, the predictive value of D is ambiguous as it precedes A, B, and C equally often. In order to make correct predictions for items following D, it is necessary to keep track of the item prior to D. That is, A-D unambiguously predicts C, while C-D unambiguously predicts A, and B-D unambiguously predicts B. Critically, in this version of the SRTT performance is not associated with awareness of the repeating sequence, amnesic patients are impaired (Curran, [@B25]), and the impairment is amplified as more items are required for the sequence to be predictive (e.g., Shanks et al., [@B107]). While it may be difficult to draw definitive conclusions about the involvement of the hippocampus based on the data from amnesic patients, several lines of converging evidence make the case stronger. First, normal-memory participants under the influence of systemic diazepam, which induces a temporary amnesia, showed impaired performance on the SRTT comparable to that of organic amnesics (Shanks et al., [@B107]). Second, fMRI reveals that normal memory participants show significantly greater activation in the hippocampus and surrounding MTL cortical structures while performing the SRTT when there are higher-order sequences (like those described above) embedded in the stimulus set (Schendan et al., [@B104],[@B105]). Together, these results provide yet another compelling demonstration of hippocampus-mediated facilitation on a task that is performed without awareness and requires flexible representation of relationships among items. One final well-studied implicit learning task that requires relational flexibility (i.e., pattern separation and pattern completion) is the transitive inference (TI) task. The TI task is of particular interest because it requires both induction and deduction, and allows the inference of unstudied relationships. Thus it provides insights into how we may use past learning to model complex contingencies sufficient for prediction under novelty. Indeed, any discussion of relational flexibility must be predicated on the notion that the fundamental utility of memory is to improve our ability to negotiate future events (cf. Schacter et al., [@B98]). In a typical TI experiment, participants are trained to make certain pairwise discriminations. For instance if A and B are presented simultaneously, participants would learn by trial and error to choose A (hereafter abbreviated A\>B). Such items are referred to as premise pairs, and a series of overlapping premise pairs is acquired by the participant over the course of several trials (e.g., A\>B, B\>C, C\>D, D\> E, E\>F). It is possible to maintain the acquired information as five distinct pieces of information, or alternatively, to integrate it into a more global representation (i.e., A\>B\>C\>D\>E\>F). Note that this global representation requires fewer operations, but also contains information not trained---namely, relationships between non-adjacent pairs. This can be tested by asking participants to make decisions about non-studied pairings such as B?D, B?E, and C?E. Correctly choosing B\>D, B\>E, and C\>E is evidence that participants have ordered the items and have made correct inferences (note that pairings with A or F require no inferences as these are always and never correct, respectively). Interestingly, participants can perform well above chance on inferred items in the absence of explicit strategies (e.g., Greene et al., [@B45]). That is, while participants are aware they have learned the premise pairs, they do not deliberately infer novel relations and are generally unaware that they have made an inference, or that it was possible to do so. In addition, performance on the TI task has been shown to depend critically on the hippocampus from converging evidence as fMRI data shows that success on the TI task depends on activity differences in the hippocampus during both study and test (Greene et al., [@B43]), and hippocampal amnesics are not capable of doing the task (Smith and Squire, [@B113]). There is some concern about how awareness is assessed in this paradigm. Indeed, it has been asserted that successful inference is related to task awareness (Smith and Squire, [@B113]). However, Greene et al. ([@B45]) have shown that a brief training period leads to high inference without awareness, while extended training leads to nearly every participant becoming task-aware. A close examination of the methods used by Smith and Squire ([@B113]) reveals that their conclusion that performance and awareness are related is entirely attributable to overtraining (see Greene, [@B42]). Others are concerned that implicit inference may be an example of pseudo-inference (e.g., Frank et al., [@B35]). According to this argument, the closer an item is to an endpoint of the sequence, the stronger is its reinforcement value. The first item is never wrong, and thus has the highest reinforcement value, the last item is never right, and therefore has the lowest reinforcement value; intermediate items obtain intermediate values according to their proximity to the end items. Thus the B\>D choice can be made without true inference by simply comparing the relatively higher positive weight of B to D. Even if this pseudo-inference account is correct, all of the direct evidence implicates the hippocampus as necessary for success in this task. More importantly, a similar inference task has been devised which has no endpoints, and therefore cannot be solved with a differential weighting strategy (see Figure [4](#F4){ref-type="fig"}; Leo and Greene, [@B63]; adapted from Bunsey and Eichenbaum, [@B9]). As with similar TI studies, participants performed well above chance on all inferences and few exhibited any sign of explicit strategy or knowledge; in fact, most were surprised to learn in debriefing that the items at test had not been studied. ![Illustration of the modified transitive inference paradigm that cannot be solved via pseudo-inference. In this version of the task, three items (e.g., faces, objects) appear on the screen in an upside down triangle on each trial. Participants are instructed to choose one of the top two items via button-press, and are told that the bottom item determines which of the remaining two items is correct. Two sets of these trial types are developed for training. As illustrated in the figure, when "M" is presented on the bottom (in training set 1) participants learn to choose "A", and when "X" is presented on the bottom (in training set 2) participants learn to choose "M". Once criterion learning has been achieved on all of displays presented during learning, the test for transitivity is administered. In this case, participants have not learned (i.e., were not trained on) the relationships among display elements, but must make an inference. Evidence for inference is obtained when, for example, participants choose "A" when "X" is the discriminative stimulus. In this case "X" and "A" are indirectly linked via "M". As indicated in the text, this version of the TI task has no endpoints and cannot be solved using pseudo-inference.](fnhum-06-00080-g0004){#F4} Evidence for hippocampus-supported encoding without awareness ------------------------------------------------------------- As has been discussed, models of hippocampal function diverge with respect to whether or not conscious apprehension of stimulus input is considered necessary during encoding (e.g., Moscovitch, [@B73]; Henke, [@B54]). In line with those who claim that consciousness is critical, the idea that information presented subliminally can support subsequent memory expression seems to go against the grain of conventional findings that document the beneficial influence of deep or elaborative encoding on memory performance (Craik and Lockhart, [@B24]). More generally, in the implicit perception literature, questions about the extent to which subliminal information can be processed perceptually and especially conceptually (i.e., semantically) have incited much debate (e.g., Holender, [@B58]; Dulany, [@B29]; Reingold, [@B90]); such disagreement has led to the development of very strict criteria for demonstrating that stimulus input was in fact completely masked from awareness before claims of implicit processing can be made (see Hannula et al., [@B50] for review). Having been informed by this debate, and the methodological constraints that have been put in place as a result, Henke and colleagues (Henke et al., [@B55],[@B56]; Degonda et al., [@B27]; Duss et al., [@B30]) have developed a program of fMRI research that seems to pose a significant challenge to staunch adherents of the view that the hippocampus cannot or does not contribute to encoding processes when stimulus input is not consciously accessible. In these experiments, subliminal encoding entails presentation of several face-occupation name pairs that are flashed briefly and sandwiched in between a set of visual pattern masks that render the pair invisible (see Figure [5A](#F5){ref-type="fig"}). An incidental memory test that entails supraliminal (i.e., visible) presentation of the faces, absent the associated occupation name, is then administered following a delay (e.g., 5 min)---that any effect of implicit memory can be identified with a delay in place is itself worthy of comment, as effects of implicit perception on performance are thought to be relatively short-lived (Wolfe, [@B133]; see also Simons et al., [@B111]). Nonetheless, evidence for implicit memory has now been successfully documented in several of these experiments, using a wide variety of tasks and measures of performance. For example, when instructed at test to guess the occupational category (i.e., artist or academic) or occupation-related characteristics (e.g., long or short education) of the individual depicted by the face, response times were faster for correct than for incorrect guesses (see Figure [5B](#F5){ref-type="fig"}; Henke et al., [@B56]; see also Duss et al., [@B30]). In addition, results show that participants can classify these faces according to occupation-related characteristics with above-chance accuracy (Duss et al., [@B30]). Based on the processing requirements of these tasks, which require flexible representation of relationships among items, one might expect the hippocampus to be recruited during subliminal encoding and/or incidental retrieval of face-occupation name pairs. The results do not disappoint, as activity differences relative to respective control conditions have been observed in the anterior hippocampus (and perirhinal cortex) not only during the supraliminal retrieval phase[^3^](#FN3){ref-type="fn"}, but also during subliminal encoding (Henke et al., [@B55],[@B56]). Perhaps more compelling, hippocampal (and perirhinal) engagement was correlated with response time differences on the incidental test of memory (see Figure [5C](#F5){ref-type="fig"}). ![Illustration of methods and results from Henke et al. ([@B56]). (A) Illustration of the conditions used during encoding and the masking procedure that renders the stimuli invisible. Two encoding conditions were used in the experiment. In the experimental condition a face was presented along with the name of a particular profession and in the control condition the face was presented alone. During the supraliminal test phase (i.e., no masking), participants made a button-press response indicating whether the person depicted by the face was an academic or an artist. (B) Illustration of subliminal learning effects evident during performance of the incidental memory test. Participants responded more quickly when they correctly guessed the occupational category associated with a face-profession pair seen during subliminal encoding than when they got the categorization response incorrect. No such facilitation was seen in the control condition, which confirmed that this difference was not merely a consequence of certain faces being better matched to one category over the other. (C) Activity differences in the right hippocampus (and perirhinal cortex) sensitive to subliminal encoding of face-occupation pairs (left), and activity differences in the right hippocampus (and perirhinal cortex) that were correlated with the response time differences shown in (B). \[Figure [5](#F5){ref-type="fig"} reprinted with permission from Elsevier\].](fnhum-06-00080-g0005){#F5} The basic paradigm described above was modified in subsequent investigation (Degonda et al., [@B27]) so that the subliminal encoding phase was followed by visible encoding prior to testing. Such manipulations provided more opportunity to investigate influences of implicit exposure on subsequent behavior and brain activity. In three separate experiments, the occupation name paired with a given face during visible encoding was either from the same category (e.g., artist: photographer → guitarist), the other category (e.g., artist: poet → academic: professor), or was exactly the same occupation (i.e., painter → painter) as that associated with that face during subliminal encoding. Under these circumstances, it was found that the ease with which participants could imagine the individual in the associated occupation during the second, visible encoding phase was significantly reduced in the incongruent condition. Moreover, results from the test phase showed that correct identification of the occupational category for a given face was reduced for all three conditions relative to a control in which the face had been paired with a nonsense word during subliminal encoding. As such, there were several converging pieces of evidence for successful acquisition of subliminally encoded content. Consistent with the findings reported in earlier studies by the group, hippocampal activity was evident during subliminal encoding in all three experiments, and greater hippocampal recruitment was positively correlated with the reported deficit in explicit retrieval for both the incongruent and congruent conditions. Such results are perhaps surprising, but as they have been consistently replicated, it seems that models of hippocampal function will need to account for them. Future work might also examine whether or not there are differences in how information is conveyed to the hippocampus when stimuli are presented subliminally. To our knowledge this question has not yet been addressed, but similar issues have been a source of on-going investigation in the emotion literature. The dominant view, which has not gone unopposed (cf. Pessoa and Adolphs, [@B83]), attributes rapid implicit processing of affective visual stimuli to a short latency subcortical pathway that bypasses visual cortex and conveys information directly to the amygdala by way of the superior colliculus and the pulvinar. This research, and the ensuing debate, will likely inform similar questions about hippocampus-supported subliminal memory. At the tipping point: hippocampal function extends beyond conscious memory ========================================================================== As has been conveyed in the preceding section, there is now considerable evidence from experiments conducted with a range of methodological techniques that implicates the hippocampus in memory without awareness. Such findings have been documented in investigations that examined implicit retrieval, implicit learning, and implicit encoding. Perhaps most surprising is the evidence for hippocampus-mediated implicit (and in the examples described above, subliminal) encoding. These experiments, which borrow substantially from lessons learned in the implicit perception literature about how best to eliminate awareness, have been replicated several times, and are not easily accommodated by models of hippocampal function that place a premium on conscious apprehension of encoded content (e.g., Moscovitch, [@B73]). The results are, however, in line with the view that the hippocampus is critically involved in binding together inter-item or item-context relationships that can then give rise to flexible memory expression (e.g., Cohen and Eichenbaum, [@B20]; Henke, [@B54]). Indeed, even a cursory comparison of the experiments reported above reveals quite readily that what these tasks have in common is the requirement for relational memory binding and representation. Additional information as to what exactly this might entail is provided by Olsen and colleagues (this issue) who propose that the specific role of the hippocampus may be to continuously bind the elements of experience and to compare representations retained in memory to current perceptual input. One result of this continuous binding and comparison process is that novel information can be integrated into existing relational memory representations. Consistent with the evidence reviewed here, Olsen and colleagues contend that both of these processes can proceed in the absence of awareness and that the resulting relational memory representation need not be consciously accessible. To date, the bulk of the evidence in support of hippocampus-mediated implicit memory seems to have come from investigations that involve temporally extended (i.e., multiple trial) learning, which according to some views, should not depend upon the hippocampus (Henke, [@B54]). Importantly, Henke ([@B54]) does indicate that there may be exceptions to this rule. Semantic memories, for example, may depend initially on the hippocampus despite repeated exposure, and concessions are made for "rare" episodic memories that retain their flexibility with the passage of time. As there are now several findings that implicate the hippocampus in implicit learning that develops gradually over the course of several experimental trials, further specification of the model describing exactly when and how the hippocampus would be involved in such learning seems warranted. Altogether, the results are perhaps most problematic for those who have retained a strict interpretation of the declarative memory theory (e.g., Squire, [@B115]), in which the hippocampal system is said to mediate only those memories that are amenable to conscious report (i.e., can be "declared"). While the relational memory theory shares the view that MTL structures (including the hippocampus) support declarative memory (cf. Cohen and Eichenbaum, [@B20]), proponents of this view do not take a hard line on the consciousness issue (e.g., Cohen et al., [@B21]; Ryan and Cohen, [@B94]). One possibility then is that declarative memory is perhaps best characterized as a subset of relational memory, and as an especially good example of relational memory when studies are conducted with human subjects. The extant possibility is that the hippocampus is necessary, but not sufficient for conscious recollection. Indeed, it has been proposed that other regions, namely structures in the prefrontal cortex and parietal cortex (Cohen et al., [@B21]; Moscovitch, [@B74]), interact with the hippocampus and in so-doing, contribute to the process of conscious recollection. That this may indeed be the case is suggested by converging evidence implicating the prefrontal cortex in facilitating top-down selection and organization of retrieved content, as well as updating features relevant to the demands of on-going retrieval attempts (see Shimamura, [@B109] for review). The literature proposing a role for the parietal cortex in memory retrieval is relatively recent, but there is now a considerable amount of evidence implicating posterior parietal cortex (PPC) in particular aspects of memory retrieval. Interestingly, activity differences in PPC show that this brain region is responsive not only to correct memory decisions, but also to false alarms, which suggests that it may play an especially important role in subjective experience (e.g., Wheeler and Buckner, [@B132]). Several recent theoretical perspectives (see Shimamura, [@B109] for review) have been advanced in an attempt to explain the role of the PPC in memory. Across perspectives, this region has been implicated in the process of accumulating/monitoring retrieved content (Wagner et al., [@B129]), acting as an episodic buffer (Vilberg and Rugg, [@B124]), and directing attention to retrieved content (Cabeza et al., [@B10]; Ciaramelli et al., [@B18]). Among these perspectives, the "attention to memory model" (Cabeza et al., [@B10]; Ciaramelli et al., [@B18]) has perhaps provided the most detailed account of how the PPC might interact with MTL and PFC regions during retrieval. Specifically, it is proposed that dorsal PPC might interact with (dorsolateral) PFC to mediate top-down, effortful, retrieval processes and that ventral PPC might interact with MTL structures in a bottom-up fashion. In this case, attention is captured by content that has been retrieved by MTL structures, including the hippocampus. More precise specification of how exactly these structures might interact requires further investigation, as does differentiating among competing views (see also Shimamura, [@B109] for another proposal), but these efforts do provide a starting point for identifying when and how other brain regions might contribute to memory retrieval and the associated experience of conscious recollection. Finally, it is worth considering which empirical approaches are most likely to yield success in endeavors to further explore the role of the hippocampus in unconscious encoding and/or subsequent expression of memory. In some cases, it is preferable to design a task that precludes knowledge of the link between task performance and memory. Devising such tasks can be particularly difficult because the complexity of the task demands often favor explicit strategies and because any overtraining can impel awareness of the contingencies that support performance (cf. Greene, [@B42]). Tasks already known to require hippocampal involvement can sometimes be designed to minimize the likelihood of awareness, but in designing such tasks, several factors should be considered. Such factors include, but are not limited to: (1) maintaining the shortest possible task exposure, especially avoiding time for rumination; (2) use of brief stimulus exposure; (3) use of novel stimulus materials; (4) use of more stimulus items than can be deliberately memorized; (5) designing tasks that cannot be supported by verbal strategies; and in this case, (6) avoiding any hint whatsoever that the task involves learning or memory. On the other hand, it is critical to consider that one can lessen task awareness without entirely eliminating its use; as such, one runs the risk that explicit strategies might be used but not detected. The approach adopted by Henke and colleagues (e.g., Duss et al., [@B30]), which makes use of masking procedures to render stimulus input invisible, offers promise in this regard. As an alternative to the above approach, one might opt to combine direct tests (e.g., forced-choice recognition tests) with additional measures that are capable of revealing effects of memory in the absence of awareness. Approaches that have a documented track record along these lines include, but are not limited to, concomitant acquisition of sensitive indirect measures (e.g., eye movements) with behavioral testing (see Hannula et al., [@B46] for review) and incorporation of methods that facilitate, and have the potential to reveal, accurate guessing (cf. Voss and Paller, [@B128]). Use of these methods acknowledges that memory tests are rarely process pure, and may permit investigators to more effectively tease apart the contributions made by explicit and implicit memory to performance. In sum, the existence of implicit, hippocampus-dependent tasks has eluded scientists for decades because such tasks are difficult to design and can readily trigger awareness---consequently, we are near the beginning of this endeavor. However, as reviewed here, we now have a solid starting point as several lines of converging evidence have implicated the hippocampus in unconscious memory. Unraveling the Gordian knot of hippocampal function will undoubtedly require a great deal more empirical instances than are presently available in the literature, making this line of work ripe for further investigation. Conflict of interest statement ------------------------------ The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. ^1^While hippocampus-dependent encoding/retrieval processes are characterized as automatic in the model, this does not preclude the possibility that such processes can be influenced by instruction, task demands, or strategy. Indeed, recent work indicates that memory is improved by encoding strategies that depend critically on hippocampal integrity (e.g., Voss et al., [@B126]). Additional detail about the characterization of hippocampal function as automatic and/or obligatory can be found in Moscovitch ([@B73]). ^2^Results of several experiments have shown that eye-movement-based memory effects precede explicit responses and that the time course of these effects is similar across investigations (cf. Hannula et al., [@B47]). While this consistency is remarkable, we cannot rule out the possibility that retrieved content may be accessible to awareness closer in time to the viewing effects than the findings suggest. To address this issue, we are currently conducting new experiments that will permit us to more precisely examine the temporal relationship between patterns of viewing and explicit recognition. ^3^Note that while the stimuli were visible (i.e., supraliminal) during the retrieval phase, the retrieval process was necessarily implicit because face-occupation pairs had been encoded without awareness (i.e., subliminally). [^1]: Edited by: Joel Voss, Northwestern University Feinberg School of Medicine, USA [^2]: Reviewed by: Joel Voss, Northwestern University Feinberg School of Medicine, USA; Jennifer D. Ryan, University of Toronto, Canada; Rosanna K. Olsen, Rotman Research Institute at Baycrest, Canada	{ "pile_set_name": "PubMed Central" }
All relevant data are within the paper and its Supporting Information files. Introduction {#sec001} ============ The design of superior biologic therapeutics, including monoclonal antibodies, single-domain antibodies, and engineered proteins, involves optimizing their ability to bind to disease-related molecular targets (antigens). Immunization of animals \[[@pone.0181490.ref001]\] and phage display methods \[[@pone.0181490.ref002]\] are the workhorses of antibody discovery and optimization. However, these methods do not always achieve the desired levels of affinity and/or specificity. Computational approaches \[[@pone.0181490.ref003]--[@pone.0181490.ref007]\] provide a complementary strategy for affinity optimization but predicting successful designs reliably has been challenging. One of the most daunting challenges in antibody sequence optimization is prioritizing the sequence space resulting from the combinatorial explosion when mutating multiple sites in complementarity determining region (CDR) loops. For example, for CDR loops containing 60 amino acids there are about 10^9^ different triple mutants possible. This number increases by a factor of 10^3^ for each additional mutation site. Exploring this space computationally through random virtual mutagenesis or other stochastic search methods explores only a small corner of sequence space with no clear stopping criteria. We sought to develop a more deterministic approach for building a focused set of mutations with a high likelihood of improving binding. The ADAPT ([A]{.ul}ssisted [D]{.ul}esign of [A]{.ul}ntibody and [P]{.ul}rotein [T]{.ul}herapeutics) platform was developed to aid in the selection of mutants that improve/modulate the affinity of antibodies and other biologics. ADAPT interleaves computational predictions with experimental validation, significantly enhancing the robustness of the design and selection of mutants. We briefly explain the workflow of the method. The first step is an exhaustive single-point virtual mutagenesis along the entire CDR sequence. Each amino acid is mutated in turn to each of 18 amino acids (mutation to Cys and Pro is excluded) and a binding affinity score is calculated relative to the parent sequence. The affinity score is a median-based composite z-score using three scoring functions (see [Methods](#sec005){ref-type="sec"} section), reflecting how many standard deviations away from the median a particular score is. From the 50 top-scoring mutants approximately 15 mutants are selected based on amino acid and site diversity and manual visual inspection of the virtual mutants. The Fab fragments for these are then produced and their binding affinities measured by SPR. Typically, around 3--4 sites are confirmed experimentally to be favorable sites for mutations. This early experimental validation is critical because it narrows down the subsequent mutations to combinations of just these sites. Thus, we avoid the combinatorial explosion that can occur in constructing double and higher mutants. The double-mutant candidates in the next round are limited to pairs of validated single mutants--up to about 30 distinct double mutants. For a 60-amino acid CDR, this is much less than the theoretical 1.2 x 10^6^ possible double mutants. Similarly, for the construction of triple mutants we restrict our combinations to those of validated double mutants and validated single mutants. Furthermore, the filtering at the single-mutant stage also increases the robustness of the design of higher order mutants. Multiple-site mutants are constructed using only validated single mutants. This strategy was tested on three antibody Fab-antigen complexes: bH1-VEGF-A, bH1-HER2 and Herceptin-HER2 \[[@pone.0181490.ref008],[@pone.0181490.ref009]\]. We chose these three systems because they covered a wide range of initial affinities for the parent sequence, representing increasing challenges for further affinity maturation. All three had crystal structures for the complex, providing a solid starting point for the virtual mutation simulations. Our measured affinities for these by SPR were 44 nM (bH1-VEGF-A), 3.6 nM (bH1-HER2) and 0.058 nM (Herceptin-HER2). Results {#sec002} ======= [Fig 1A](#pone.0181490.g001){ref-type="fig"} shows the results of three rounds of design for bH1-VEGF. In the first round, 14 diverse single mutants were selected from the top 50 z-scores for experimental validation. The selection was driven not just by the z-scores but also by the desire to sample multiple mutation sites and amino acid types (see [Methods](#sec005){ref-type="sec"} section). The selected mutants had z-scores ranging from --8.5 to --1.6 ([S1 Table](#pone.0181490.s001){ref-type="supplementary-material"}). Of these, 11 mutants showed improvements in affinity ranging from 1.1-fold to 9.5-fold as measured by SPR, two had slightly weaker binding and the remaining two had no detectable binding. The mutants with improved affinity were located at four sites (HY33, HD98, HG99, LS30b) in the CDRs. The designations H and L indicate the residue as being in the heavy or light chain, respectively. For the second round, we tested 17 double mutants selected out of 32 possible ones ([S2 Table](#pone.0181490.s002){ref-type="supplementary-material"}) arising from combinations of the validated single mutants. All 17 showed improved affinities over the parent ranging from 2.3- to 40-fold, 15 of which showed affinity improvements greater than either of the component single mutants. Given that the LS30bR mutation had the smallest incremental improvement when combined with other single mutations we restricted the third round to triple combinations of heavy chain mutants used in the second round. All six triple mutants that were made showed further improvements in affinity compared to the double mutants ([Fig 1A](#pone.0181490.g001){ref-type="fig"}), with the best being about 100-fold better than the parent (K~D~ = 44 nM for the parent versus 0.46 nM for the mutant). [Fig 2A and 2D](#pone.0181490.g002){ref-type="fig"} show the sensorgrams for the parent bH1 and best triple mutant, respectively. We see a decreased k~off~ rate with respect to the parent. [Table 1](#pone.0181490.t001){ref-type="table"} lists the k~on~ and k~off~ rates for the parent sequence and triple mutants. For all the triple mutants the improvements in affinities were due primarily to slower k~off~ rates. ![Fold improvements in binding affinity and relative changes in binding free energy (kcal/mol) relative to the parent Fab during three rounds of mutations.\ (a) bH1-VEGF. (b) bH1-HER2. (c) Herceptin-HER2. H and L designate the mutation as being in the heavy or light chain, respectively. Standard deviations of ΔΔG are based on 3 or more replicates, typically. NB = no binding. WT = parent sequence. Residue numbering for bH1 follows that of Bostrom et al. \[[@pone.0181490.ref008]\] Residue numbering for Herceptin follows that in the 1n8z PDB entry.](pone.0181490.g001){#pone.0181490.g001} ![Sensorgrams of the parent Fabs and best triple mutant for each complex.\ The red curves represent the global fits of the data to a 1:1 bimolecular interaction model. The slow off rates of Herceptin and its triple mutant required longer data acquisition times in the dissociation phase to obtain reliable kinetics. The insets in panels (C) and (F) are expanded views of the association phase of these sensorgrams.](pone.0181490.g002){#pone.0181490.g002} 10.1371/journal.pone.0181490.t001 ###### SPR-measured k~on~ and k~off~ rates for the parent sequence and triple mutants. ![](pone.0181490.t001){#pone.0181490.t001g} Mutation k~on~ (10^5^M^-1^s^-1^) k~off~ (10^-4^s^-1^) K~D~ -------------------- ------------------------- ---------------------- -------------- --- ------ ----------- ------------- ---------------- bH1-VEGF Parent 2.2 ± 0.5 90 ± 17 44 ± 17 nM H D98M H Y33W H G99D 1.0 ± 0.1 1.4 ± 0.0 1.5 ± 0.2 nM H D98F H Y33W H G99D 1.4 ± 0.3 1.5 ± 0.0 1.1 ± 0.2 nM H D98W H Y33W H G99D 1.6 ± 0.3 0.7 ± 0.0 0.5 ± 0.1 nM H D98M H Y33W H G99M 2.1 ± 0.7 1.9 ± 0.0 1.0 ± 0.2 nM H D98F H Y33W H G99M 2.2 ± 0.6 2.0 ± 0.0 0.9 ± 0.2 nM H D98W H Y33W H G99M 2.9 ± 0.7 3.4 ± 0.2 1.2 ± 0.2 nM bH1-HER2 Parent 3.7 ± 0.4 13 ± 1.6 3.6 ± 0.6 nM H D98Y H D31K L I29K 22 ± 5 2.8 ± 1.1 0.14 ± 0.09 nM H D98F H D31K L I29K 22 ± 4 2.9 ± 0.6 0.14 ± 0.05 nM H D98K H D31K L I29K 23 ± 4 3.9 ± 1.0 0.15 ± 0.05 nM H D98Y H D31K L I29R 38 ± 4 2.5 ± 0.5 0.07 ± 0.01 nM H D98F H D31K L I29R 27 ± 5 2.9 ± 0.6 0.11 ± 0.02 nM H D98K H D31K L I29R 44 ± 6 4.5 ± 0.3 0.11 ± 0.02 nM Herceptin-HER2 Parent 11 ± 4 0.57 ± 0.09 58 ± 18 pM H D102F H D31K L D28K 61 ± 8 0.11 ± 0.00 1.9 ± 0.2 pM H D102F H D31M L D28K 51 ± 13 0.14 ± 0.03 2.9 ± 0.7 pM H D102F H D31K L D28R 68 ± 4 0.17 ± 0.07 2.5 ± 1.0 pM H D102F H D31M L D28R 50 ± 11 0.19 ± 0.04 4.0 ± 1.4 pM [Fig 3](#pone.0181490.g003){ref-type="fig"} illustrates the structural basis for some of these improvements in affinity. HD98 has unfavorable electrostatic interactions with two nearby carboxylates (D63 and E64) on VEGF ([Fig 3D](#pone.0181490.g003){ref-type="fig"}). Mutating it to Trp eliminates the bad electrostatics and introduces new van der Waals contacts. On the G99 site, mutation to Asp introduces a salt bridge with K48 on VEGF ([Fig 3C](#pone.0181490.g003){ref-type="fig"}). Interestingly, the oppositely charged mutation HG99R also results in improved affinity, albeit less than with HG99D. In this case, HG99R forms a salt bridge with D63 on VEGF ([Fig 3A](#pone.0181490.g003){ref-type="fig"}). Also, the methylene groups of the HG99R and VEGF K48 side chains form nonpolar interactions. Yet another way on improving on HG99 is the HG99M mutation. Instead of creating salt bridges it introduces nonpolar interactions between the Met side chain and side chains of VEGF I83 and K48 ([Fig 3B](#pone.0181490.g003){ref-type="fig"}). We see that even on the same site, there can be quite varied mechanisms of enhancing binding affinity. This also demonstrates the versatility of our method, which is capable of suggesting multiple strategies for improving mutations that do not rely solely on electrostatics. Earlier studies by Lippow et al. suggested that it was difficult to reliably predict affinity enhancements arising from mutations to larger amino acids and they relied more on electrostatics as a predictor of changes in binding free energy \[[@pone.0181490.ref004]\]. More recently, Kiyoshi et al. carried out an exhaustive virtual single-mutant scan of the antibody 11K2 against the chemokine MCP-1 and obtained single mutants that had up to 4.6 fold higher affinity than the wild type as measured by SPR \[[@pone.0181490.ref007]\]. All their successful mutants were also mutations to charged amino acids. Unlike this work, they did not go further and construct double and higher order mutants from their single mutations. In this context, the designed HY33W and HG99M mutants are informative as they do not alter the net charge and rely on an increase of the size of the side chain upon mutation to improve binding affinity. ![Structural basis for increased affinities of the HD98W and HG99X mutations for bH1-VEGF binding.\ The bH1 and VEGF chains are colored grey and cyan, respectively. (a) Modeled binding mode of the HG99R. A salt bridge is formed between HG99R of bH1 and D63 of VEGF. (b) Modeled binding mode of HG99M. Nonpolar interactions are formed between HG99M of bH1 and I83 and K48 of VEGF. (c) Modeled binding mode of the HG99D. A salt bridge is formed between HG99D of bH1 and K48 of VEGF. (d) HD98W removes unfavorable electrostatic interactions of HD98 with D63 and E64 of bH1.](pone.0181490.g003){#pone.0181490.g003} A key aspect of our protocol is the experimental validation of designed single mutants prior to designing and producing double mutants. To illustrate the importance of this step consider the single mutants HD98M and LS30dM. The z-scores for each of these mutants were within the top 5 in the virtual single mutant scan ([S1 Table](#pone.0181490.s001){ref-type="supplementary-material"}). Thus, it would have been tempting to immediately make the double mutant. However, experimental validation of the single mutants showed that LS30dM is in fact destabilizing rather than enhancing, suggesting that making a double mutant incorporating LS30dM would have been a waste of time and resources. To confirm this, the double mutant with HD98M and LS30dM was made and resulted in a greater than 10-fold drop in binding affinity. The second affinity maturation test was carried out on bH1-HER2. [Fig 1B](#pone.0181490.g001){ref-type="fig"} summarizes the results for three rounds of optimization. In the first round, four sites each on the heavy and light chains were identified as candidates for mutation. Thirteen single mutants were produced and tested, with z-scores ranging from --3.1 to --0.9 ([S1 Table](#pone.0181490.s001){ref-type="supplementary-material"}). Nine showed improvements in affinity, albeit somewhat more modest than what was observed for the bH1-VEGF case. In round two, 12 double mutants out of a possible 25 ([S3 Table](#pone.0181490.s003){ref-type="supplementary-material"}) involving four sites were made and tested, all of which showed further gains in affinity. In the third round, 6 triple-mutant combinations of the mutants in round two were made. All the triple mutants showed further increases in affinity, with the best achieving about a 50-fold improvement (parent K~D~ = 3.6 nM versus 0.066 nM for the mutant). [Fig 2B and 2E](#pone.0181490.g002){ref-type="fig"} show the sensorgrams for the parent bH1 and best triple mutant, respectively. The improvement in affinity arises from changes in both the k~on~ and k~off~ rates ([Table 1](#pone.0181490.t001){ref-type="table"}). The increases in k~on~ contribute about a 6- to 10-fold improvement in K~D~ and about a 3- to 6-fold contribution from decreases in k~off~. The change in k~on~ is likely due to the long-range electrostatic interactions \[[@pone.0181490.ref010],[@pone.0181490.ref011]\] between the new Lys and Arg residues in the mutants and the net negative charge of HER2. The third affinity maturation test was carried out on Herceptin-HER2. We considered this a particularly challenging test for affinity maturation since Herceptin is already such a tight binder (K~D~ = 58 pM). [Fig 1C](#pone.0181490.g001){ref-type="fig"} summarizes the results for three rounds of optimization. In the first round, two heavy-chain sites and two light-chain sites were identified as candidates for mutation. Eleven single mutants (z-scores ranging from --3.8 to --1.0, [S1 Table](#pone.0181490.s001){ref-type="supplementary-material"}) were produced, nine of which showed some improvement in affinity. In the second round, seven double mutants were tested and all showed improvements over the single mutant components. The double mutants incorporating the heavy chain HD102F had the best affinity and were used to generate triple mutants with light chain single mutant LD28R and LD28K, respectively. Unfortunately, triple mutants incorporating the most active double mutant (HD102F, HD31R) were heterogeneous upon purification and so were not included in the binding affinity measurements. The best triple mutant of the remaining four had about a 30-fold better affinity than parent (mutant K~D~ = 1.9 pM). [Fig 2C and 2F](#pone.0181490.g002){ref-type="fig"} show the sensorgrams for the parent Herceptin and best triple mutant, respectively. The contributions to the change in affinity break down into approximately 6-fold and 5-fold changes in k~on~ and k~off,~ respectively ([Table 1](#pone.0181490.t001){ref-type="table"}). The four triple mutants also exhibited improvements in apparent binding affinity relative to the parent Herceptin Fab when tested by flow cytometry on two cell lines expressing the HER2 antigen at different cell surface densities ([S1 Fig](#pone.0181490.s006){ref-type="supplementary-material"}, [S5 Table](#pone.0181490.s005){ref-type="supplementary-material"}). Hence, the SPR-based binding affinities to the HER2 ectodomain are qualitatively consistent with the trend observed for binding to the full-length HER2 expressed at the cell surface. Implicit in our approach of stepwise build-up of mutations is the assumption of additivity in the contribution of each mutation to the binding free energy. The data generated in the multiple cycles of maturation allows us to examine this assumption. [Fig 4](#pone.0181490.g004){ref-type="fig"} is a plot of the experimentally measured binding free energy contributions of double/triple mutants versus the sum of the contributions of their component single/double mutants. For the combined data set of all three systems we have a correlation of r^2^ = 0.56 and a slope of 0.79. If we force the intercept to be zero we obtain an r^2^ = 0.48 and a slope of 1.07. We see that the assumption of additivity is quite reasonable for this data set. One possible reason is that the component mutations have all been pre-selected to be favorable mutations. It is likely that the sum of favorable and unfavorable contributions would be less predictable. ![Additivity of contribution of mutations to binding affinity.\ Shown is a scatter plot of the experimentally measured relative binding affinities of double and triple mutants versus the sum of independently measured relative binding affinities of the component single/double mutants. The dashed line is the linear regression line for the entire set.](pone.0181490.g004){#pone.0181490.g004} Discussion {#sec003} ========== The use of a consensus z-score is central to the ADAPT protocol. In a previous study \[[@pone.0181490.ref012]\] we examined the ability of consensus scores for predicting relative binding affinities in antibody-antigen complexes. We found that a consensus z-score using the scoring functions SIE, FOLDEF and Talaris-interface improved the transferability across different systems for enrichment predictions of mutants with improved binding. That study was a retrospective analysis of data in the SiPMAB database \[[@pone.0181490.ref012]\], which contains mutational, structural and binding affinity data for antibody-antigen complexes culled from the literature. However, a true test of the predictive ability of the consensus z-score is to apply it to the design of novel proteins with enhanced binding affinities. In this work, we used the consensus z-score approach for affinity maturation of three antibody-antigen complexes with initial K~D~ values ranging from a modest 44 nM to a tight 56 pM, achieving significant improvement in all three cases. Furthermore, the over 100 mutants across the three systems that were predicted to improve binding affinity were experimentally found to be highly enriched with enhanced binders. This suggests that the consensus z-score used has utility and transferability beyond the retrospective analysis of the SiPMAB database. Further support of the conclusions of the previous study \[[@pone.0181490.ref012]\] about the value of using a consensus z-score is given below. [Table 2](#pone.0181490.t002){ref-type="table"} lists the z-scores and ranks for some selected single mutants for the three scoring functions used in this study--SIE, FOLDEF and Talaris-interface. For example, in the bH1-VEGF system the G99D mutation (a component of the best triple mutant) was identified only by SIE (z-score --6.0, rank 32). FoldEF gave a z-score of 0.56 and rank 522. Talaris-interface gave a z-score of 0.58 and rank of 725. The consensus z-score was --1.6, which brought it within the top 50 average z-scores. Using an average rank would have penalized the low ranks too much and not have given enough weight to the very favourable SIE score relative to the scores of the other single mutants. For the bH1-HER2 system, only FoldEF scored I29R highly (z-score --3.0, rank 19). The values for SIE and Talaris-interface were (z-score --0.88, rank 107) and (z-score 0.0, rank 429), respectively. The consensus z-score of --1.2 for I29R brought it into the top 50 average z-scores. For the herceptin-HER2 system, D102F was in the top 50 only for Talaris-interface (z-score --0.81, rank 27). The values for FOLDEF and SIE are (z-score --0.8, rank 88) and (z-score --1.5, rank 51), respectively. The consensus z-score was --1.0 bringing it within the top 50. The use of the consensus z-score allowed us to retain these single mutants as candidates for experimental validation. They eventually ended up as being part of the best triple mutants and would likely have been missed in the absence of the consensus z-score. Our results are consistent with our earlier findings that the use of consensus z-scores for ranking antibody-antigen binding affinities increases the robustness of the predictions \[[@pone.0181490.ref012]\]. 10.1371/journal.pone.0181490.t002 ###### Individual scoring function z-scores and ranks for selected single mutants. ![](pone.0181490.t002){#pone.0181490.t002g} Mutant[^a^](#t002fn001){ref-type="table-fn"} FoldEF SIE Talaris-interface Consensus ---------------------------------------------- -------- ----- ------------------- ----------- ------- ------------------------------------------- ------- HG99D 0.6 522 --6.0 32 0.6 725 --1.6 LI29R --3.0 19 --0.9 107 0.0 153[^b^](#t002fn002){ref-type="table-fn"} --1.3 HD102F --0.8 88 --1.5 51 --0.8 27 --1.0 ^a^ HG99D, LI29R and HD102F are single mutants for bH1-VEGF, bH1-HER2 and herceptin-HER2, respectively. They are found in the best triple mutants for each of the systems. ^b^ This rank is shared with 404 other mutants with a z-score of 0.0. ^c^ The consensus z-score is the arithmetic average of the z-scores for the three scoring functions. All three consensus z-scores are within the top 50 for their respective systems. In contrast, the average of the ranks would not have placed the mutants within the top 50. An important limitation of ADAPT is the need for a three-dimensional structure of the antibody-antigen complex. Ideally this would be from a crystal structure for the parent sequence or from a high-quality model based on a closely related structure. We expect the reliability of the scoring functions used will be quite sensitive to the accuracy of the structure used as a starting point for modeling the mutations. A related limitation of ADAPT is that it is a very conservative exploration of conformational space. Although we allow some limited movement and repacking of side chains in SIE_SCWRL and FOLDX, the backbone conformation is essentially assumed to remain unchanged by the various mutations. Of course, in general, backbone conformations can be sequence-dependent. However, the spirit of ADAPT is to optimize the amino acid sequence for a given backbone conformation. The expectation is that the side chains that are designed will be highly compatible with the given backbone in the context of the antibody-antigen complex. Another limitation of ADAPT in its present form is that we do not allow for insertions or deletions of amino acids. This certainly precludes whole collections of mutation possibilities. However, in the examples presented here we find that even when the CDR chain lengths are kept fixed the sequence space is rich enough to provide mutants yielding significant improvements in affinity. The application of ADAPT to the three systems described here focused on improving affinity and did not explicitly take developability into account. A possible concern is that the presence of Trp and Met in some of our optimized sequences could introduce potential sites of oxidation and compromise the viability of the mutated antibody as a commercial product. In an actual drug discovery campaign, one could exclude Met and Trp in the mutation scans. In our case, exclusion of those amino acids would have resulted in the selection of other single mutants among the top 50. At this point we don't know whether the resulting mutants would have produced comparable affinity enhancements for bH1 against VEGF. However, we note that the best bH1 and herceptin triple mutants against HER2 contain neither Trp nor Met. Hence, the platform is capable of achieving affinity enhancements without the introduction of Trp or Met mutations. Conclusions {#sec004} =========== We have presented ADAPT, a combined interleaved computational and experimental protocol and workflow for affinity maturation of antibodies. We summarize the main results below. 1. An exhaustive virtual scan of single mutants followed by experimental validation of selected single mutants avoids the combinatorial explosion in exploring multi-site mutations. 2. Validation at the single mutation stage improves the robustness and efficiency of designing double and triple mutations, leading to increased binding affinities in three test systems of about 100-, 50- and 30-fold, respectively. It took less than 40 mutants for each system to achieve these improvements. 100% of double and triple mutants showed affinity improvements versus their parent molecules. The success rate across all mutants was 90%. 3. Reasonable additivity of the contribution to binding free energy of mutations was observed. 4. Electrostatics is a major but not exclusive contributor to enhanced affinity. Examples of mutations to non-polar groups that improve affinity were also observed. 5. Enhanced binding of the designed mutants was achieved by increasing the binding on-rate, decreasing the off-rate or a combination of both. This can arise from a combination of long-range electrostatics with short-range non-polar interactions. 6. The data set of over 100 SPR binding affinity measurements in three congeneric series with mostly affinity-enhancing mutations will be of great value to the development, calibration and validation of computational tools for antibody design. The tight interleaving of virtual mutations and experimental validation gives rise to the success of ADAPT, which promises to be a useful platform for affinity maturation of antibodies. While this strategy has been validated on affinity maturation one can envision expanding the applicability domain of ADAPT towards the optimization of other important biophysical properties along with other biotherapeutic frameworks. Methods {#sec005} ======= Structure preparation {#sec006} --------------------- The initial structures of the bH1-VEGF, bH1-HER2 and Herceptin-HER2 complexes were based on the crystal structures 3bdy, 3be1 and 1n8z, respectively. Only the V~H~ and V~L~ regions of the Fab were used in the simulations. From the HER2 extracellular domain, all residues N-terminal to Cys489 were deleted to reduce the size of the antigen for the calculations. HER2 residues 581--590 are not visible in either 3be1 or 1n8z. The missing loop was reconstructed by grafting in the loop from the crystal structure 1n8y. Crystallographic water and counterions were removed. Hydrogen atoms were added and protonation states at neutral pH were adopted. Polar hydrogen orientations were visually examined and manually adjusted as needed. Structural refinement was then carried out by energy-minimization using the AMBER force-field \[[@pone.0181490.ref013]--[@pone.0181490.ref015]\], with a distance-dependent dielectric (4r~ij~) and infinite cutoff for non-bonded interactions. Non-hydrogen atoms were restrained at their crystallographic positions with a harmonic force constant of 5 kcal/(mol^.^A^2^). Building and scoring mutants {#sec007} ---------------------------- We used three protocols (SIE-SCWRL, FoldX and Rosetta) for modeling mutants and scoring their binding affinities. For each of these protocols, additional refinement of the AMBER-refined parent structure was carried out as prescribed by each of the methods as a starting point for building and scoring the mutants. We describe each of these methods briefly below. A more detailed description of the procedures can be found in Sulea et al.\[[@pone.0181490.ref012]\] ### SIE-SCWRL {#sec008} This is a hybrid method. SCWRL version 4.0 (Fox Chase Cancer Center, Philadelphia, PA; <http://dunbrack.fccc.edu/scwrl4>) \[[@pone.0181490.ref016]\] was used to build and repack the mutant side chains, keeping the rest of the complex fixed. The resulting SCWRL-generated structures of the mutants were then energy-minimized with the AMBER force-field \[[@pone.0181490.ref013]--[@pone.0181490.ref015]\], allowing only a set of residues 6 Å around the mutated residue to be mobile. The Fv-antigen interaction was scored with the SIE \[[@pone.0181490.ref017],[@pone.0181490.ref018]\] function. SIE was originally developed for predicting binding affinities of small molecules with proteins. We use the scoring function here with no alteration of its default parameters. ### FoldX {#sec009} Two types of scores were calculated with FoldX version 3.0b6 (Center for Genomic Regulation, Barcelona, Spain; [http://foldx.crg.es](http://foldx.crg.es/)) \[[@pone.0181490.ref019],[@pone.0181490.ref020]\], Fv-antigen binding affinity scores and Fv stability scores. The RepairPDB routine of FoldX was used to prepare a relaxed structure of the complex prior to mutation. Default parameters were used throughout along with setting the ionic strength to 0.1 M. FoldX was used to repack mutated side chains, keeping the backbone fixed. Side chains in the vicinity of the mutations were also repacked. Binding affinities were calculated using the FOLDEF \[[@pone.0181490.ref019],[@pone.0181490.ref020]\] scoring function. The stability of the isolated mutant Fv relative to the parent was also estimated. Any mutant Fv that was predicted to be more than 100-fold less stable than the parent was excluded from consideration by all affinity scoring protocols. Since the repacking step is a stochastic process and produces non-identical results for each run, ten independent runs were carried out for each mutant and a Boltzmann-weighted average was taken. ### Rosetta {#sec010} We used Rosetta software version 3.5 (University of Washington, Seattle, WA; [http://www.rosettacommons.org](http://www.rosettacommons.org/)) \[[@pone.0181490.ref021],[@pone.0181490.ref022]\] to repack and relax only the mutated side chain in a rigid protein environment. The default weights of the Talaris 2013 scoring function was used for the repacking and relaxation step, followed by rescoring of binding affinity with a modified function, Talaris-interface, whose weights were optimized for alanine scanning mutations (<https://guybrush.ucsf.edu/benchmarks/benchmarks/alanine_scanning>) \[[@pone.0181490.ref022]\]. ### Consensus Z-score {#sec011} The z-score indicates how many standard deviations a particular value in a series of values is from the mean of all the values. $$z_{i} = \frac{x_{i} - x_{\text{mean}}}{\sigma}$$ where σ is the standard deviation of the scores. It provides a convenient way of normalizing the results of various scoring functions and combining them in a useful way. The consensus z-scores reported in [S1 Table](#pone.0181490.s001){ref-type="supplementary-material"} are simply the average z-scores of the mutants across the three binding affinity scoring functions used: SIE, FoldX (FOLDEF) and Rosetta (Talaris-interface). The combination of these three scoring functions was found to give good Spearman rank-order correlations for binding affinity on an antibody-antigen data set of about 200 single-point mutants \[[@pone.0181490.ref012]\]. Consensus z-scores have been used previously in virtual screening of chemical libraries \[[@pone.0181490.ref023]\]. Some scoring function "energies" may be unusually large in magnitude (e.g., when steric clashes are present) and will skew the mean and standard deviations. To alleviate this we use a modified z-score instead that is based on the median and median absolute deviation \[[@pone.0181490.ref024]\]. The median absolute deviation (MAD) is more robust to outliers than the standard deviation. The modified z-score, z~i~, of a particular value x~i~ is defined as $$z_{i} = \frac{x_{i} - x_{\text{med}}}{1.4826 \times \text{MAD}}$$ where x~med~ is the median value. The multiplicative constant 1.4826, which converts the MAD to a standard deviation assuming a normal distribution, is obtained by noting that the MAD is the median of the half-normal distribution and hence can be calculated from the inverse error function $$\text{MAD} = \sigma\sqrt{2}\text{erf}^{- 1}\left( 0.5 \right) = 0.6749\sigma$$ where σ is the standard deviation. With the Rosetta Talaris-interface scoring function used in this study, most of the scores were zero or very close to zero, resulting in a MAD that was zero or very close to zero. This makes the z-score undefined or ill-behaved. In this case we used the mean absolute deviation (MeanAD) instead of the MAD. The MeanAD, which is larger than the MAD, is always nonzero unless all values are identical. The z-score is then calculated as $$z_{i} = \frac{x_{i} - x_{\text{mean}}}{1.2533 \times \text{MeanAD}}$$ The multiplicative constant 1.2533 for the MeanAD is obtained by noting that the MeanAD is the mean of the half-normal distribution. ![](pone.0181490.e005.jpg){#pone.0181490.e005g} MeanAD = σ 2 / π The ADAPT software was set to automatically switch to using MeanAD if MAD \< 0.03. Selection of mutants {#sec012} -------------------- Lists of the single-point mutants corresponding to the top 50 consensus z-scores for each of the three complexes are shown in the [S1 Table](#pone.0181490.s001){ref-type="supplementary-material"}. A subset of the top 50 was selected for experimental validation. The reduction of the top-50 list to a smaller subset was driven primarily by cost considerations. In an actual drug discovery setting, it would not be unreasonable to test all 50 single mutants at this stage. The selection of the subset was based not solely on the values of the composite z-scores. A previous study showed that while a collection of mutants with favorable z-scores will be enriched in true positives, the z-scores are only modestly accurate in ranking binding affinities within the set \[[@pone.0181490.ref012]\]. Rather than choosing strictly by z-scores, we also sought to have diversity in the mutation sites explored and in the residue types introduced. In addition, visual inspection of the modeled structures was used to further evaluate the proposed mutants. For example, for herceptin the LS50Y mutation had a slightly more negative consensus z-score than the LS50F and LS50W mutations largely due to the very favorable score given by FoldX. Examination of the FoldX predicted structure for LS50Y showed that the score was based on a structure that disrupted an internal salt bridge between D82 and K105 in HER2. Being internal to HER2, this loss of a salt bridge was not reflected in the FoldX interaction energy. We thus opted to go with the LS50F and LS50W mutations for which the aforementioned salt bridge was preserved in the predicted structures. The validated single mutants that exhibit improved binding affinities were then used to construct virtual double mutants that were then scored using the three scoring functions. The small number of virtual double mutants makes it difficult to interpret z-scores calculated using the double-mutant score distribution in a statistically meaningful way. Instead, we calculated z-scores using the MAD (or MeanAD, as the case may be) and median of the corresponding single-mutant scan for that scoring function. This allowed a direct comparison of the double-mutant z-scores with the single-mutant ones to see if the double mutants showed further predicted improvements in affinity. The consensus z-scores were calculated as before, i.e., as arithmetic averages of individual z-scores. The double-mutant z-scores were primarily used to weed out any possible incompatible combinations of single mutants. For example, it is conceivable that two bulky mutations on adjacent sites may not be simultaneously viable. With the single mutants obtained in this work, such incompatibilities were not observed. In selecting the subset of double mutants for experimental validation we generally chose double mutants whose single-mutant components showed the best binding affinities experimentally. For example, for the bH1-VEGF system, we restricted the choice of double mutants involving HD98 to use only HD98M, HD98F or HD98W --the best single mutants at that site. Again, as in the single-mutant stage, it would be feasible in a drug discovery setting to skip the subset selection and test the entire double-mutant list. The validated double and single mutants were then used to generate the triple mutants. [S2](#pone.0181490.s002){ref-type="supplementary-material"}--[S4](#pone.0181490.s004){ref-type="supplementary-material"} Tables list the consensus z-scores for the double and triple mutants. The general workflow is shown schematically in [S2 Fig](#pone.0181490.s007){ref-type="supplementary-material"}. Calculation of K^WT^/K {#sec013} ---------------------- Each batch of SPR measurements of mutant Fabs always included a determination of K^WT^ for the parent Fab as well. It is this measured K^WT^ for the specific batch that was used in calculating the K^WT^/K ratios (fold improvements in binding affinity) for the mutants. These ratios were averaged over replicate (three or more) measurements to provide the values tabulated in [Fig 1](#pone.0181490.g001){ref-type="fig"}. We felt that using a reference measured at the same time as the mutant would help reduce the effect of systematic errors and yield less noisy values for the relative improvements in affinity. In [Fig 1](#pone.0181490.g001){ref-type="fig"} we also report the parent dissociation constant, $K_{\text{D}}^{\text{WT}}$, which is the overall average K^WT^ using multiple batches of SPR measurements across different mutants. One could use this overall average K^WT^ in calculating K^WT^/K, leading to slightly different numbers. For example, the improvements for the best triple mutants reported in [Fig 1](#pone.0181490.g001){ref-type="fig"} are 104, 46 and 32, respectively for bH1-VEGF, bH1-HER2 and Herceptin-HER2 using our chosen approach. If we use the overall average K^WT^ instead, we obtain 96, 55 and 31 as the corresponding improvements. There is not a large discrepancy between the two approaches and we opted for the former one over the latter. ### Fab production {#sec014} cDNA for the designed Fab heavy and light chains were ordered from commercial vendors (GeneArt, Life Technologies). These contained signal peptide sequences and heavy-chain C-terminal His~8~ tags for purification. The Fabs were produced by co-transfection of CHO-3E7 cells. 2 x 10^6^ CHO-3E7 cells were plated in 2 ml of F17 medium. The cells were transfected with 2 μg total DNA (containing 500 ng each of the heavy chain and light chain constructs) using PEI. The cells were maintained at 37^°^C for 24 h after which they were transferred to 32^°^C for 6 days. The cell-culture supernatant was harvested and analyzed by SDS-PAGE for expression. His-tag purification was carried out and followed by UPLC-SEC, for verification of homogeneity as required. SPR measurements {#sec015} ---------------- GLC sensorchips, the Biorad ProteOn amine coupling kit (EDC, sNHS and ethanolamine), and 10mM sodium acetate buffers were purchased from Bio-Rad Laboratories Ltd. (Mississauga, ON). Recombinant soluble human HER2 extracellular domain (termed HER2 here) was purchased from eBioscience (San Diego, CA). Full-length isoform 165 of human VEGF-A (termed VEGF here) was produced recombinantly and purified. PBS running buffer with 0.05% Tween20 (PBST) was purchased from Teknova, Inc. (Hollister, CA). All surface plasmon resonance assays were carried out using a BioRad ProteOn XPR36 instrument (Bio-Rad Laboratories Ltd., Mississauga, ON) at a temperature of 25C using PBS running buffer containing 0.05% Tween 20 (Teknova, Hollister, CA) with the addition of 3.4 mM EDTA. Three GLC chip channels were prepared for antigen immobilization by injecting a 1:10 dilution of the standard BioRad sNHS/EDC solutions for 140 s at 100 μL/min in the ligand direction. The first channel was used as a blank control; the remaining two channels were used for HER2 and VEGF immobilization. Immediately after the activation, 2.5 ug/mL solutions of HER2 and VEGF in 10 mM NaOAc pH 4.5 were injected in the ligand (vertical) direction at a flow rate of 25 μL/min until approximately 150 resonance units (RUs) were immobilized. Remaining active groups in all three channels were quenched by a 140 s injection of 1M ethanolamine at 100 μL/min in the ligand direction. After immobilization of the antigens, each Fab variant was injected in the analyte (horizontal) direction for binding kinetics determination to the HER2 and VEGF surfaces. In each injection cycle, two buffer injections for 30 s at 100 μL/min in the ligand direction were used to stabilize the baseline. Next, 3-fold serial dilutions of each Fab variant (top nominal concentrations of 30, 60 or 120 nM) and a buffer blank were simultaneously injected over the blank, HER2 and VEGF surfaces at 50 μL/min for 120 s with a 900 s (or longer as required) dissociation phase. The SPR surfaces were regenerated by two 18-second pulses of 0.85% phosphoric acid at 100 μL/min to prepare for the next injection cycle. Sensorgrams were aligned and double-referenced using the buffer blank injection and the blank interspots, and the resulting sensorgrams were analyzed using ProteOn Manager software v3.1. The double-referenced sensorgrams were fit to the 1:1 binding model to determine the k~on~ (s^-1^M^-1^) and k~off~ (s^-1^). The binding affinity (K~D~) was determined from the ratio of k~off~/k~on~. Each Fab variant was injected in triplicate. Abbreviations. EDC: 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride; sNHS: N-hydroxysulfosuccinimide; SPR: surface plasmon resonance; EDTA: ethylenediaminetetraacetic acid; PBS: phosphate buffered saline; HER2: Receptor-tyrosine kinase ErbB2 soluble ectodomain; VEGF165: 165 amino acid isoform of vascular endothelial growth factor. Cell-binding assay {#sec016} ------------------ ### Cell lines {#sec017} The human breast cancer cell line MCF-7 (HTB-22) and the human ovarian SKOV-3 cells (HTB-77) were obtained from American Type Culture Collection (ATCC; Manassas, VA, USA) and maintained in recommended culture conditions. The cell line identities were authenticated with short tandem repeat (STR) profiling. Absence of mycoplasma contamination was tested by PCR. ### Binding affinity measurement by flow cytometry analysis {#sec018} Following treatment with cell dissociation solution (Sigma), cells were collected, washed and incubated for 2 hours on ice with Fabs at final concentrations ranging from 0.1 to 300 nM. A standard IgG isotype (human IgG; Jackson Immunochemicals) was used as negative control. Cells were then washed twice and incubated for 1 h on ice with 1/100 dilution of conjugated secondary antibody AF488-anti-human IgG (H+L) (Jackson Immunochemicals). Cells were washed and resuspended in medium containing 1% propidium iodide (Invitrogen). Each sample was filtered through a Nitex membrane and analyzed on a BD™LSRII flow cytometer (Beckton-Dickinson). Two thousand alive/single-cell events are acquired per sample. Fluorescence was evaluated using the BD FACSdiva™ Software. ### Data analysis {#sec019} Specific mean fluorescence intensity (MFI) is calculated for each sample point by subtracting the MFI value of negative control (background). Binding curves (specific MFI vs. linear or log antibody concentration) were fitted with GraphPad Prism 6 software using a One-site specific with Hill slope four-parameter nonlinear regression curve fitting model to determine apparent K~D~ values for each Fab. Supporting information {#sec020} ====================== ###### Top 50 consensus Z-scores for single mutants. The first column in S1 Table lists the mutation sites involved in the top 50 consensus z-scores. The first letter (H or L) refers to the heavy or light chain, respectively. Scores in red correspond to mutants that were selected for production and experimental validation. (PDF) ###### Click here for additional data file. ###### Consensus z-scores for double and triple mutants of bH1-VEGF. The z-scores shown are relative to the distribution of scores in the exhaustive scan of single mutants, i.e., the median (or mean) absolute deviations of the single-mutant scores were used in computing the z-scores. In red are the mutants that were produced and validated experimentally. Z-scores were calculated for all double mutants arising from combinations of validated single mutants that showed an improvement in affinity ([Fig 1](#pone.0181490.g001){ref-type="fig"}, main text). For example, for bH1-VEGF experimental validation of double mutants involving the D98 site was restricted to those containing HD98M, HD98F or HD98W because those were found to be the best single mutants for that site ([Fig 1](#pone.0181490.g001){ref-type="fig"}, main text). The first letter (H or L) refers to the heavy or light chain, respectively. Scores in red correspond to mutants that were selected for production and experimental validation. The selection of double mutants to test took into account the improvement in affinity afforded by the single mutants together with the calculated z-score for the double-mutant combination. (PDF) ###### Click here for additional data file. ###### Consensus z-scores for double and triple mutants of bH1-HER2. The z-scores shown are relative to the distribution of scores in the exhaustive scan of single mutants, i.e., the median (or mean) absolute deviations of the single-mutant scores were used in computing the z-scores. In red are the mutants that were produced and validated experimentally. (PDF) ###### Click here for additional data file. ###### Consensus z-scores for double and triple mutants of Herceptin-HER2. The z-scores shown are relative to the distribution of scores in the exhaustive scan of single mutants, i.e., the median (or mean) absolute deviations of the single-mutant scores were used in computing the z-scores. In red are the mutants that were produced and validated experimentally. No double mutants involving light chain mutants LD28R or LD28K were made since the plan was to incorporate both of these in triple mutants anyway. In selecting which triple mutants to make, we decided to focus on the heavy chain double mutants with the best affinities as measured by SPR ([Fig 1](#pone.0181490.g001){ref-type="fig"}, main text). (PDF) ###### Click here for additional data file. ###### Apparent K~D~ values of Herceptin Fab and its triple mutants by flow cytometry analysis (n = 3). (PDF) ###### Click here for additional data file. ###### Binding titration curves to determine dissociation constants, K~D~. Mean fluorescence intensity (MFI) from FACS data is plotted versus the concentration of different anti-HER2 Fabs on MCF-7 (A) or SKOV-3 (B) cell lines. Data are fit using One-site specific with Hill slope four-parameter nonlinear regression curve fitting model (see [Methods](#sec005){ref-type="sec"} in main text). Independent triplicate titrations were performed on different days, with similar results. The parent Herceptin Fab is indicated with blue filled circles and blue curve, and its triple mutants from [S5 Table](#pone.0181490.s005){ref-type="supplementary-material"} with the other colors and symbols. Panels A and B of S1 Fig show dose-response binding curves to MCF-7 and SKOV-3 cells for the Herceptin Fab (blue symbols) and its four affinity-matured triple mutants listed in [Fig 1C](#pone.0181490.g001){ref-type="fig"}. The data indicate that the affinity-matured variants have similar apparent binding among themselves and improved binding relative to the parental Herceptin Fab. This ranking is in qualitative agreement with the SPR data (see [Fig 1C](#pone.0181490.g001){ref-type="fig"} and [Table 1](#pone.0181490.t001){ref-type="table"}). Apparent K~D~ values based on this cell-based assay are listed in [S5 Table](#pone.0181490.s005){ref-type="supplementary-material"}. One can note the smaller magnitudes of binding affinity improvements obtained with the cell-based experiments relative to the improvements obtained with the SPR assay, a consequence of the washing steps used in the cellular assay. (PDF) ###### Click here for additional data file. ###### Affinity maturation workflow. (PDF) ###### Click here for additional data file. ###### Fold improvements in binding affinity and relative changes in binding free energy (kcal/mol) relative to the parent Fab during three rounds of mutations (Excel spreadsheet of data in [Fig 1](#pone.0181490.g001){ref-type="fig"}). (XLSX) ###### Click here for additional data file. We are grateful for the technical assistance of Cassio Baptista, Karine Brault, Alina Burlacu, Richard Gingras, Suzanne Grothé, Denis L'Abbé, Paule Lachance, Sonia Lamontagne, Linda Lamoureux, Félix Malenfant, Yuliya Martinova, Marie-Josée Massariol, Roger Maurice, Marie Parat, Josée Plamondon and Hongtao Qi. We thank Hervé Hogues for useful discussions. [^1]: Competing Interests:The authors have declared that no competing interests exist. [^2]: Current address: Weill Cornell Medical College, New York, NY, United States of America [^3]: Current address: Formation Biologics, Montreal, Quebec, Canada	{ "pile_set_name": "PubMed Central" }
Introduction {#Sec1} ============ Human rhinoviruses (HRVs), members within genus Enterovirus in the family Picornaviridae, have a positive-strand RNA genome of 7,200 nucleotides covalently linked at the 5′ end to the viral protein 3B (VPg) and is translated cap-independently by internal ribosomal entry into a polyprotein (VP4-VP2-VP3-VP1-2A-2B-2C-3A-3B-3C-3D), which yields the 11 proteins through various independently functioning intermediates, upon cleavage by viral proteases \[[@CR1]\]. HRVs were originally classified into two species, A and B, and approximately 100 serotypes \[[@CR2]--[@CR4]\]. Advances in molecular methods led to the characterization of the third HRV group, a genetically heterogeneous third species, HRV species C(HRV-C) \[[@CR5], [@CR6]\]. Identification of the specific strain in HRV disease has been difficult because the traditional serological method is insensitive, labor intensive, and cumbersome \[[@CR7]\], which requires the isolation of HRV in susceptible cell cultures and neutralization tests against all 101 serotype-specific antisera \[[@CR3]\]. To improve the technique for defining the links between illnesses and specific strains of HRV, more sensitive and faster molecular methods have been developed. PCR, RT-PCR, and real-time RT-PCR represent the latest diagnostic methods that provide robust, reproducible results. We present here an RT-PCR for identification of HRVs and a multiplex RT-PCR that simultaneously identifies 12 respiratory viruses; these assays are routinely used in our laboratory for diagnosis of HRV infection in clinical samples and have been optimized to enable simultaneous handling of large numbers of samples. Materials {#Sec2} ========= Sample Preparation {#Sec3} ------------------ Sterile collection tubes with nasopharyngeal aspirates.Flocked tipped swabs with nasal secretions.Sterile saline (0.9 % solution of sodium chloride).Microcentrifuge.Eagle's Minimum Essential Medium (E-MEM) OR Virus Transport Medium (VTM) (see Note [1](#Sec15){ref-type="sec"}). Isolation of Total RNA {#Sec4} ---------------------- Viral RNA isolation kit.Ethanol (96--100 %).Microcentrifuge tubes (1.5 ml).Sterile, RNase-free pipette tips with aerosol barrierMicrocentrifuge (with rotor for 2-ml tubes). RT-PCR {#Sec5} ------ Ethanol (96--100 %).Reverse Transcriptase Kit (e.g., SuperScriptTM III, Invitrogen).DNA Polymerase PCR Kit (e.g., GoTaq Green Master Mix, Promega or MangoTaq, Bioline).Seeplex^®^ RV15 ACE Detection Kit (Seegene).Primers (Table [1](#Tab1){ref-type="table"}).Table 1Primers for HRV nested-PCRPrimer namePrimer sequence (5′--3′)Target geneAmplicon lengthFirst run PCRRV-F1CTCCGGCCCCTGAATRYGGCTAARV-R1TCIGGIARYTTCCASYACCAICCSecond run PCRRV-F2ACCRASTACTTTGGGTRWCCGTG5′ NCR-VP4/VP2110 bpRV-R2CTGTGTTGAWACYTGAGCICCCABases are shown in single-letter symbols; A Adenine, C Cytosine, T Thymine, G Guanine, W A/T, S C/G, R A/G, Y C/T, I InosineRNase-free tips (10, 100, and 1,000 μl).96-Well plates for PCR with membranes for sealing.GeneAmp PCR system 9700 (Applied Biosystems, see Note [3](#Sec15){ref-type="sec"}). Gel Electrophoresis {#Sec6} ------------------- Agarose.DNA gel-loading dye.DNA molecular weight marker.Ethidium bromide.Electrophoretic buffers.Electrophoresis apparatus.UV transilluminator.QIAxcel^®^ capillary electrophoresis system. Methods {#Sec7} ======= Sample preparation (SeeNotes[4](#Sec15){ref-type="sec"}--[6](#Sec15){ref-type="sec"}) {#Sec8} --------------------------------------------------------------------------------------------------- All samples are collected in sterile tubes in the clinic and sent to microbiology laboratory for testing within 3 h of collection. Samples should be stored at 4 °C at this time (see Note [4](#Sec15){ref-type="sec"}).For respiratory virus diagnostics two kinds of samples are collected, nasopharyngeal aspirates (NPA) and nasal swabs. NPAs require specialist equipment that must be handled by trained personnel; this procedure should be performed according to local safety and clinical guidelines. Nasal swabs are collected by inserting flocked tipped swabs into nostrils and must be performed by trained personnel only and according to local safety and clinical guidelines. Swabs must be inserted into tubes with VTM immediately after collection.1.5 ml sterile saline is added into each tube of NPA; tubes are centrifuged at 14,000 rpm (20,000 × g) for 2 min; the supernatant is removed and pellet is resuspended in 200 μl of E-MEM, aliquoted, and kept at −70 °C.Nasal swabs are squeezed ten times against the tube walls in VTM and removed; tubes are centrifuged at 3,000 rpm (200 × g) for 2 min; the supernatant is aliquoted and kept at −70 °C. Isolation of Total RNA {#Sec9} ---------------------- The protocol presented here is based on the QIAamp viral RNA isolation kit; any kit for isolation of total RNA from cell-free body fluids may be used in its stead (see Notes [2](#Sec15){ref-type="sec"} and [7](#Sec15){ref-type="sec"}). Equilibrate samples and Buffer AVE to room temperature (15--25 °C).Check that Buffer AW1, Buffer AW2, and Carrier RNA have been prepared according to the manufacturer's instructions. Redissolve any precipitates in Buffer AVL/Carrier RNA by heating, if necessary, and cool to room temperature before use. All centrifugation steps are carried out at room temperature.Pipette 560 μl of prepared Buffer AVL containing Carrier RNA into a 1.5-ml microcentrifuge tube. Add 140 μl treated NPAs/nasal swab solution to the Buffer AVL/Carrier RNA in the microcentrifuge tube. Mix by pulse-vortexing for 15 s. Incubate at room temperature for 10 min.Briefly centrifuge the 1.5-ml microcentrifuge tube to remove drops from the inside of the lid.Add 560 μl of ethanol (96--100 %) to the sample, and mix by pulse-vortexing for 15 s. After mixing, briefly centrifuge the 1.5-ml microcentrifuge tube to remove drops from inside the lid.Carefully apply 630 μl of the solution from step 5 to the QIAamp spin column (in a 2-ml collection tube) without wetting the rim. Close the cap, and centrifuge at 6,000 × g (8,000 rpm) for 1 min. Discard the filtrate.Carefully open the QIAamp spin column, and repeat step 6.Carefully open the QIAamp spin column, and add 500 μl of Buffer AW1. Close the cap, and centrifuge at 6,000 × g (8,000 rpm) for 1 min. Place the QIAamp spin column in a clean 2-ml collection tube, and discard the tube containing the filtrate.Carefully open the QIAamp spin column, and add 500 μl of Buffer AW2. Close the cap and centrifuge at full speed (14,000 rpm; 20,000 × g) for 3 min. Continue directly with step 10, or to eliminate any chance of possible Buffer AW2 carryover, perform step 9(a), and then continue with step 10. (Optional): Place the QIAamp spin column in a new 2-ml collection tube (not provided), and discard the old collection tube with the filtrate. Centrifuge at full speed for 1 min.Place the QIAamp spin column in a clean 1.5-ml microcentrifuge tube. Discard the old collection tube containing the filtrate. Carefully open the QIAamp spin column and add 60 μl of Buffer AVE equilibrated to room temperature. Close the cap, and incubate at room temperature for 1 min. Centrifuge at 6,000 × g (8,000 rpm) for 1 min. Viral RNA is stable for up to 1 year when stored at −20 °C or −70 °C. RT-PCR {#Sec10} ------ ### First-Strand cDNA Synthesis {#Sec11} We routinely perform reverse transcription using random hexamers and SuperScriptTM III Reverse Transcriptase Kit on GeneAmp PCR system 9700 according to the manufacturer's recommendations. Any other reverse transcriptase system can be used, keeping in mind that some optimization may be required for optimal results (see Note [9](#Sec15){ref-type="sec"}). Briefly centrifuge each reagent provided in the kit.Prepare pre-mix 1 in a sterile, nuclease-free microcentrifuge tube as below, mix by brief vortex, and centrifuge briefly:Pre-mix 1 (for one reaction)5 μg Total RNA 5 μl50 ng/μl random primer 1 μl10 mM dNTP mix 1 μlDNase- and RNase-free water 3 μlTotal volume10 μlCalculate the required total amount of each reagent based on the number of reactions (samples + controls).Label a 96-well plate, as the cDNA plate. Pipette 10 μl of pre-mix into each well of the cDNA plate; cover with the sealing membrane and briefly centrifuge the plate.Place the cDNA plate into GeneAmp PCR system 9700, and incubate at 65 °C for 5 min. Take the cDNA plate out and put it on the ice. Centrifuge briefly before removing the membrane seal.Take another sterile, nuclease-free 1.5 ml microcentrifuge tube labeled as pre-mix 2. Add the following reagents (shown for one reaction); mix by quick vortex, and centrifuge briefly.Pre-mix 2 (for one reaction):10× RT buffer 2 μl25 mM MgCl~2~ 4 μl0.1 M DTT 2 μlRNase OUT(40 U/μl) 1 μlSuperScript III RT(200 U/μl) 1 μlTotal volume of pre-mix 210 μlCalculate the required amount of each reagent based on the number of reactions (samples + controls).Pipet 10 μl of pre-mix 2 into each well of the cDNA plate; cover it with the sealing membrane carefully and briefly centrifuge the plate.Place the cDNA plate into GeneAmp PCR system 9700, run the following cycling parameter: one cycle of (25 °C, 10 min); followed by one cycle of 50 °C, 50 min; then one cycle of (85 °C, 5 min) to stop the reaction. Take the plate out and put it on the ice. Centrifuge briefly before removing the sealing membrane.Pipet 1 μl of RNase H into each well of the cDNA plate; cover carefully with membrane seal and centrifuge briefly.Place the cDNA plate into GeneAmp PCR system 9700, and run one cycle of (37 °C, 20 min).The first-strand cDNA is synthesized and can serve as the template in the nested-PCR or the multiplex PCR or stored at −20 °C for future use. ### Nested-PCR {#Sec12} This procedure consists of two PCR cycles, the first run PCR using the cDNA as template and the second run PCR using the amplimers from the first run PCR as template. First Run PCR Prepare pre-mix 1 in the reagent preparation room (see Note [8](#Sec15){ref-type="sec"}). Completely thaw and thoroughly vortex the buffer prior to use; then briefly centrifuge each reagent.In a sterile, nuclease-free microcentrifuge tube combine the following components to prepare the pre-mix 1 (shown for one reaction); mix by quick vortex and centrifuge briefly.Pre-mix 1 (for one reaction)Nuclease-free water8.25 μl5× colorless GoTaq flexi buffer5 μlMgcl~2~ solution, 25 mM2 μlPCR nucleotide mix, 10 mM each0.5 μlPrimers1 μlGo Taq DNA polymerase (5 u/μl)0.25 μlTotal volume of pre-mix 117 μlCalculate the necessary amount of each reagent based on the number of reactions (samples + controls).Label a 96-well plate as the first run PCR plate. Pipet 17 μl of pre-mix 1 into each well of the first run PCR plate; Transfer the first run PCR plate from the reagent preparation room to the template preparation room. Open the cDNA plate and pipet 3 μl of cDNA template from the cDNA plate into each corresponding well of the first run PCR plate.Cover the 96-well plates with the membrane for PCR plate carefully; briefly centrifuge the plate.Place the first run PCR plate into the GeneAmp PCR system 9700 in the nucleic acid amplification room, and run the following cycling parameter:94 °C 30 s55 °C 1 min45 cycles72 °C 1 min72 °C 10 minSecond run PCR. Prepare pre-mix 2 in the reagent preparation room. Completely thaw and thoroughly vortex the buffer prior to use; then briefly centrifuge each reagent.In a sterile, nuclease-free microcentrifuge tube combine the following components to prepare the pre-mix 2 (shown for one reaction); mix by quick vortex and centrifuge briefly.Pre-mix 2 (for one reaction)Nuclease-free water10.25 μl5× colorless GoTaq flexi buffer5 μlMgcl~2~ solution, 25 mM2 μlPCR nucleotide mix, 10 mM each0.5 μlPrimers1 μlGo Taq DNA polymerase(5u/μl)0.25 μlTotal volume of pre-mix 219 μlCalculate the necessary amount of each reagent based on the number of reactions (samples + controls).Label another 96-well plate as the second run PCR plate. Pipet 19 μl of pre-mix 2 into each well of the second run PCR plate.Transfer the second run PCR plate from the reagent preparation room to the template preparation room. Open the first run PCR plate and pipet 1 μl of DNA template from the first run PCR plate into each corresponding well of the second run PCR plate.Cover the second run PCR plate with the membrane for PCR plate carefully; briefly centrifuge the plate.Place the second run PCR plate into the GeneAmp PCR system 9700 in the nucleic acid amplification room, and run the following cycling parameter:94 °C 30 s55 °C 1 min35 cycles72 °C 1 min72 °C 10 minAfter analysis of the PCR products by electrophoresis, irradiate the PCR products with UV light (365 nm) for 20 min to prevent carryover contamination. ### Multiplex PCR for 15 Respiratory Viruses {#Sec13} We usually use the Seeplex RV15 ACE Detection Kit to simultaneously detect 15 common respiratory viruses. The Seeplex kit is very widely used in clinical microbiology laboratories around the world. The viruses included influenza (Flu) A/B, respiratory syncytial viruses (RSV) A/B, human metapneumovirus (hMPV), human parainfluenza viruses (PIV) 1/2/3/4, human adenovirus (AdV), human rhinovirus (HRV), human coronavirus 229E/NL63 and OC43/HKU1, human bocavirus (HBoV) and human enterovirus. Prepare master mix in the reagent preparation room. Completely thaw and thoroughly vortex the buffer prior to use, and then briefly centrifuge each reagent.Add the following reagents (shown for one reaction) to a sterile, nuclease-free microcentrifuge tube, mix by quick vortex, and centrifuge briefly:5× RV 15 ACE PM 4 μl2× Multiple master mix 2 μl8-MOP solution 3 μlWater 8 μlTotal volume of master mix17 μlCalculate the necessary amount of each reagent based on the number of reactions (samples + controls).Label a 96-well plate as the RV15 PCR plate. Aliquot 17 μl of master mix into each well the RV15 PCR plate.Transfer the RV15 PCR plate from the Reagent preparation room to the template preparation room. Pipet 3 μl of cDNA template from the cDNA plate into each corresponding well of the RV 15 PCR plate. Use 3 μl of ACE NC for the negative control PCR. Use 3 μl of ACE PC for the positive control PCR.Cover the RV 15 PCR plate with sealing membrane and briefly centrifuge the plate.Place the RV 15 PCR plate into the preheated 94 °C GeneAmp PCR system 9700 in the nucleic acid amplification room, and immediately run the PCR reaction using the following program:94 °C 0.5 min60 °C 1.5 min40 cycles72 °C 1.5 min72 °C 10 min ### Analysis of PCR Products {#Sec14} PCR products may be analyzed by agarose gel electrophoresis or by capillary electrophoresis. Figure [1](#Fig1){ref-type="fig"} depicts a typical pattern obtained by gel electrophoresis, while Fig. [2](#Fig2){ref-type="fig"} depicts a typical pattern obtained by capillary electrophoresis. Fig. 1Gel image of rhinovirus nested-RT-PCR products. The nasopharyngeal aspirate (NPA) samples (S) 1, 3, 5, 9, 10, 21, and 32 were collected from seven children with lower respiratory tract infection. Respiratory epithelial cells were collected from NPAs and processed for RNA isolation followed by nested-RT-PCR as described in Subheadings [2](#Sec2){ref-type="sec"} and [3](#Sec7){ref-type="sec"}. PCR products were separated on 2 % agarose gel electrophoresis and visualized by ethidium bromide staining; the relevant DNA marker is indicated Fig. 2Image of Seeplex RV 15 Multiplex products using QIAxcel capillary electrophoresis. NPA samples (S) 1--9 were collected from nine children with lower respiratory tract infection. NPA samples were processed for cell isolation, RNA extraction, and reverse transcription as in Fig. [1](#Fig1){ref-type="fig"}. Seeplex RV 15 Multiplex PCR was performed as described in the text and PCR products separated by the QIAxcel^®^ Novel 12-channel capillary electrophoresis system. The reference table for RV15 ACE Detection (B set) marker is indicated Separate the PCR products from the second run PCR by 2 % agarose gel electrophoresis and visualize by UV transillumination (see Chapter 10.1007/978-1-4939-1571-2_3 for detailed method). For reactions containing the 5× Green GoTaq™ Reaction Buffer, load samples onto the gel directly after amplification.For reactions not containing any indicator dyes, add gel-loading dye (e.g., bromophenol blue) before loading on the gel.Please see Fig. [1](#Fig1){ref-type="fig"} for a picture of a sample gel of nested-PCR products from nasal swabs collected from children with suspected lower respiratory tract infection.There are single bright bands at the 110 bp size location of the positive control (PC), lanes labeled S5, S10, S21, and S32, but no bands around the same location in the negative control (NC), S1, S3, and S9 lanes.Capillary electrophoresis using QIAxcel system (see Note [10](#Sec15){ref-type="sec"}). The 96-well plates can be run directly in the QIAxcel^®^ capillary electrophoresis system. Separation is performed in a capillary of a precast gel cartridge. Each sample is automatically loaded (according to voltage and time parameters) into an individual capillary and voltage is applied. As the molecules migrate through the capillary, they pass a detector that measures the fluorescent signal. A photomultiplier detector converts the emission signal into electronic data, which are then transferred to the computer for further processing using BioCalculator software. After processing, the data are displayed as an electropherogram and a gel image. The 0.2-ml 12-tube strips containing QX Alignment Marker and QX Intensity Calibration Marker (if required) should fit loosely in the MARKER1 and MARKER2 position.QX Alignment Markers should be replaced every 15--20 runs or as needed.When not in use, the 12-tube strip containing QX Alignment Marker should be stored at −20 °C. It should be equilibrated to operating temperature (20--25 ºC) and centrifuged briefly before use.If the QIAxcel gel cartridge is being used for the first time, intensity calibration should be performed (see Note [11](#Sec15){ref-type="sec"}).Preparation of the gel cartridge:Add 10 ml QX Wash Buffer to both reservoirs of the QX Cartridge Stand (provided with the QIAxcel instrument) and cover with 3 ml mineral oil (supplied).Remove the QIAxcel gel cartridge from its packaging and carefully wipe off any soft gel debris from the capillary tips using a soft tissue.Remove the purge cap seal from the back of the QIAxcel gel cartridge and place it in the QX Cartridge Stand.Incubate new cartridges in the QX Cartridge Stand for 20 min prior to use.Preparing the buffer tray:Allow all reagents to equilibrate to room temperature (15--25 °C) before use.Wash the buffer tray with hot water and rinse thoroughly with deionized water.Fill the WP and WI positions of the buffer tray with 8 ml QX Wash Buffer.Fill the BUF position of the buffer tray with 18 ml QX Separation Buffer.Carefully add mineral oil to all three positions to prevent evaporation: add 2 ml mineral oil to positions WP and WI and 4 ml mineral oil to position BUF. All three positions should be covered with mineral oil.Insert the buffer tray into the buffer tray holder so that the slots for the 12-tube strips face the front of the instrument.Preparing QX Alignment Markers:Load 15 μl QX Alignment Marker into each well of a QX 0.2 ml 12-tube strip.Add one drop of mineral oil to each well, and insert the strip into the MARKER1 position of the buffer tray. Important: The 12-tube strip should fit loosely in the MARKER1 position on the buffer tray.Installing a QIAxcel gel cartridge and smart key:Remove the QIAxcel gel cartridge from the QX Cartridge Stand; open the cartridge door and insert the QIAxcel gel cartridge into the QIAxcel system. The cartridge description label should face the front and the purge hole should face the back of the system.Insert the smart key into the smart key socket. The smart key can be inserted in either direction.Close the cartridge door. The cartridge ID, number of runs remaining, and cartridge type will be displayed automatically in the "Instrument Control" window once the smart key is latched. The system will not recognize the cartridge and will not operate if the smart key is not inserted.Add the DNA size marker (supplied with the kit) into the black well (see Note [12](#Sec15){ref-type="sec"}).Load the 96-well plate containing the samples. The minimum sample volume required for analysis is 10 μl. Less than 0.1 μl of the sample will be loaded onto the QIAxcel gel cartridge for analysis. The remaining DNA can be kept for reanalysis or downstream processing.Results for nine nasal swab samples assayed by the Seeplex RV15 multiplex PCR products are shown in Fig. [2](#Fig2){ref-type="fig"}. According to the reference table provided by the manufacturer, samples 2 and 3 are probably positive for influenza A (a second specific PCR and/or sequencing of the PCR product needs to be performed to confirm), sample 5 is positive for HRV, and sample 9 is positive for CoV-OC43. The size of samples 2 and 3 is equal to influenza A, the size of sample 5 is equal to human rhinovirus A/B/C, and the size of sample 9 is equal to human coronavirus OC43. Samples 4, 6, 7, and 8 are negative for all the respiratory viruses detected by Seeplex assay. Notes {#Sec15} ===== E-MEM or VTM can be used to collect samples for virus testing. If desired, commercial VTM preparations can be used.The protocol described here has been optimized for the Roche MagnaPure LC 2.0 system; however, any comparable system for isolation of total RNA from clinical samples may be used, but may require modification of the protocol to optimize results.Any PCR equipment that is able to handle multiple samples in a 96-well format will be suitable.The clinical samples collected from patients with lower respiratory tract infection should be kept in 4 °C refrigerator and sent to microbiological laboratory within 8 h post-collection; If samples cannot be sent to laboratory within 8 h post-collection, they should be frozen at −70 °C.Appropriate personal protective equipment (laboratory coat, mask, gloves) should be worn at all times when handling human specimens.All clinical samples collected from patients with lower respiratory tract infection should be handled in a biological safety cabinet.When the number of samples to be analyzed is less than 10, it is quicker and more convenient to isolate the total RNA from samples manually by using the QIAamp^®^ Viral RNA Mini Kit. When the number of samples is more than 10, it is laborious and time consuming to isolate the total RNA from samples manually and an automated system, e.g., the Roche MagnaPure LC 2.0, may be used.Prepare the PCR pre-mix only in the pre-mix preparation room. Assay the first-strand cDNA synthesis and perform template DNA/RNA or cDNA pipetting in the template room. Do not apply the first-strand cDNA synthesis in the pre-mix preparation room to avoid contamination of the following PCR test. Ensure that the plates are covered and sealed with a membrane before they are transfered to PCR machine in the PCR amplification room.The 8-channel micropipette is preferred when using 96-well plates. It is less laborious and more accurate to aliquot reagent and pipet template DNA/RNA or cDNA into each well. Using multichannel micropipette also helps to reduce the contamination during PCR assay process.The QIAxcel system offers a number of advantages over traditional slab-gel electrophoresis, including higher detection sensitivity, less sample wastage (minimal sample input volumes), fast analysis of up to 96 samples, and automated loading and analysis, making it the preferred method for analysis of DNA products when large numbers of samples are processed.This step is not necessary if the QIAxcel gel cartridge has already been calibrated, unless it is being used on a different QIAxcel instrument or a different computer is used to operate the instrument. If a different computer is used, the calibration log file must be transferred from the computer used to operate the instrument to the new computer so that it is not necessary to run the calibration again.Once created, a single DNA reference marker table can be used for the entire life of the cartridge. However, variations in separation temperature of the cartridge and buffer can introduce variations in DNA size determination. For optimal results, we recommend creating a new DNA reference marker table every 8 runs or after each 96-well plate. This work was supported by the Public Health Academic Leader Project (grant number GWDTR201201) from Shanghai Municipal Health Bureau.	{ "pile_set_name": "PubMed Central" }
1. Introduction =============== Overweight and obesity occur when energy intake exceeds energy expenditure over time \[[@B1-nutrients-07-04336],[@B2-nutrients-07-04336]\]. The majority of short-term weight loss strategies reduce energy intake through calorie-controlled diets that can require substantial changes to the routine of individuals \[[@B3-nutrients-07-04336],[@B4-nutrients-07-04336]\]. However, long-term dieting adherence can be difficult to achieve \[[@B4-nutrients-07-04336],[@B5-nutrients-07-04336],[@B6-nutrients-07-04336]\]. The Dual Process model of information processing \[[@B7-nutrients-07-04336],[@B8-nutrients-07-04336]\] may explain some of the difficulty people have in maintaining healthy dietary changes. The Dual Process model proposes that information processing comprises rational and intuitive components. Formation of new healthy behaviours is an intentional behaviour change, and is proposed to be governed by the rational system which is usually slow, logical and deliberate. However, intentional change may be compromised by stress, fatigue, emotions, multi-tasking, cognitive load, and other competing intentions, such as taste preferences, satisfying a craving for unhealthy food, saving time and money, or consuming unhealthy food and beverages at social events \[[@B7-nutrients-07-04336],[@B8-nutrients-07-04336]\]. This is when the intuitive system becomes dominant, turning decision-making into an automatic, subconscious process, controlled by habits \[[@B7-nutrients-07-04336]\]. Consumers' inability to identify and recall more than 90% of food-related decisions made throughout the day \[[@B9-nutrients-07-04336]\] highlights the mindless, or habitual, nature of dietary behaviours that are driven by the intuitive system \[[@B9-nutrients-07-04336],[@B10-nutrients-07-04336]\]. The intuitive nature of habits makes them resistant to change. One of the ways habitual behaviour may be influenced is by modifying the environment within which decisions are made \[[@B11-nutrients-07-04336],[@B12-nutrients-07-04336],[@B13-nutrients-07-04336]\]. A number of models that have been applied to try and achieve change in health behaviours, such as the Social Cognitive Model and the Ecological and Social Ecological Models \[[@B14-nutrients-07-04336],[@B15-nutrients-07-04336],[@B16-nutrients-07-04336],[@B17-nutrients-07-04336]\] which consider the influence of the environment. However, evaluations of the role of habits and their resistance to change is currently under-researched and underrepresented in the health literature, with most health behaviour change interventions being based on rational decision-making models \[[@B14-nutrients-07-04336],[@B18-nutrients-07-04336],[@B19-nutrients-07-04336],[@B20-nutrients-07-04336],[@B21-nutrients-07-04336],[@B22-nutrients-07-04336]\]. Behavioural economics places particular emphasis on the role of the environment in influencing habitual behaviour. The key tool of influence in behavioural economics is the nudge, which is any addition to, or modification of the environment that is designed to change a consumer's behaviour in a predictable way \[[@B12-nutrients-07-04336]\]. Nudges may increase salience of a particular option or behaviour, making it easier to process, and therefore making that option more appealing. To ensure that freedom of choice is maintained, nudges should be avoidable and the original option should be available so that the choice is voluntary \[[@B12-nutrients-07-04336]\]. Some studies have shown that nudging can influence food choices. Increased visibility, proximity and convenience of food items has been shown to increase selection \[[@B23-nutrients-07-04336]\], while decreased visibility, proximity and convenience reduced selection \[[@B24-nutrients-07-04336],[@B25-nutrients-07-04336]\]. Nudges come in different forms \[[@B26-nutrients-07-04336]\], including salience nudges. Interventions informed by nudging theory have encouraged healthier dietary choices through highlighting nutrition information such as caloric content, traffic light labeling \[[@B27-nutrients-07-04336],[@B28-nutrients-07-04336],[@B29-nutrients-07-04336],[@B30-nutrients-07-04336],[@B31-nutrients-07-04336],[@B32-nutrients-07-04336]\], or descriptive labels that highlight a certain favorable characteristic of the product (for example, "low-fat") \[[@B33-nutrients-07-04336],[@B34-nutrients-07-04336],[@B35-nutrients-07-04336],[@B36-nutrients-07-04336]\]. However these nudging interventions have been conducted predominantly in cafeterias (mainly hospital settings), and only acute effects have been evaluated, limiting their generalizability. Furthermore their efficacy has been variable, possibly due to failure of people to notice/pay attention to the information provided, or a lack of understanding of its meaning \[[@B37-nutrients-07-04336]\]. Seymour et al. \[[@B38-nutrients-07-04336]\] reviewed nutrition information interventions across multiple environments, and concluded they are more effective when conducted in "limited access" sites, such as workplaces. Salience nudges can include simple messages that are easily understood \[[@B26-nutrients-07-04336]\] and therefore may overcome limitations of other nutrition information nudges. Extending this body of knowledge, the current study evaluated salience nudging in a real world environment, and over an extended period of time. The objective of this study was to evaluate the efficacy of using a salience nudge highlighting the low calorie content of low-fat milk, on increasing the selection of low-fat milk in place of full-cream milk selection in a workplace. Milk is a commonly supplied and regularly consumed beverage in workplaces, and an important part of a healthy diet. However, low-fat milk has fewer calories per serve and therefore consumption of low-fat milk may contribute to a reduction in energy intake and obesity. Importantly, the present study evaluated the effect of nudging over the longer-term rather than just a one-off evaluation of an acute effect, thus evaluating its potential to achieve longer-term behavioural change. 2. Experimental Section ======================= Milk selection from the fridge of a university-based research institute was monitored over a 24-week period, from May to December 2013. The kitchen was accessible to approximately 70 staff and research students. The number of people accessing the kitchen each day was not recorded. Ethics approval for the study was obtained from the University of South Australia Human Research Ethics Committee prior to commencement. Two types of milk were used in the study, one was full-cream and the other was low-fat. The full-cream milk (Paul's full-cream milk, Parmalat, Brisbane, Australia) had an energy content of 2860 kJ/L and a fat content of 38g/L and the low-fat milk (Pauls trim milk, Parmalat, Australia) had an energy content of 1932 kJ/L and a fat content of 13g/L. Full-cream and low-fat milk were always available free of charge to all people who used the kitchen and this was not changed during the study. Milk selection was recorded daily at the same time (9.30 am). Baseline observations were collected for a 12-week period, after which a 12-week nudging intervention commenced. The intervention involved placing a salience nudging sign on the low-fat milk which read "Pick me! I am low calorie". Low-fat milk ran out one time during baseline and three times during the intervention period. The selection of both milk types for those days was disregarded and treated as missing data. Spillage was not accounted for, as it was not known whether this occurred or not. However, spillage may have occurred for either milk type. Data were analysed using a repeated measures linear effects mixed model to take account of missing data. Baseline was analysed separately and, because there was no change in milk selection during this period, the average selection was used as the baseline value for analysis of the effect of the intervention. Milk type (i.e., full-cream or low-fat) and time (week) were included as repeated measures within the model as the same kitchen users had access to both milk types and measures of milk selection were repeated daily. Statistical significance was set at p \< 0.05. 3. Results ========== During the baseline period low-fat milk selection was greater than full-cream milk selection (p = 0.001), but did not change from week to week (p = 0.12, milk-type × time). During the intervention period overall milk selection was not different from baseline (p = 0.22), with low-fat milk selection remaining greater than full-cream milk selection (p \< 0.001) and no significant milk-type × time interaction (p = 0.41). However, there appeared to be a transient increase in milk selection during the first part of the intervention period, and sub-analysis of the first two weeks of the intervention period indicated a significant increase in selection of both milk types from baseline (p = 0.03), with a significantly greater increase in low-fat milk selection (p = 0.01, milk-type × interaction; [Figure 1](#nutrients-07-04336-f001){ref-type="fig"}). However, milk selection then returned towards baseline during the rest of the intervention period. ![Selection of low-fat and full-cream milk prior to and during a 12 week nudging intervention. Data analysis was based on weekly selection but has been presented as the mean of 12 weeks of baseline data and means of two- and four-weekly periods during the intervention period for simplicity of graphical representation. Data are mean ± standard error. \Significantly different from baseline (p* = 0.03) and significantly greater increase in low-fat milk selection compared with full-cream milk selection (p = 0.01; first two weeks of the intervention period only).](nutrients-07-04336-g001){#nutrients-07-04336-f001} 4. Discussion ============= The present study indicates that a salience nudge was not effective in changing milk selection from full-cream to low-fat milk. While sub-analysis did indicate a transiently greater increase in low-fat milk selection during the first two weeks of the intervention, the nudge also increased the selection of full-cream milk during this period, thus increasing total milk selection rather than eliciting a shift from selection of full-cream milk to low-fat milk. Therefore, during the first two weeks of the intervention there was an overall increase in milk selection, meaning that there would have been an increase in caloric intake from milk, which may be detrimental when the purpose of the nudge was to increase selection of low-fat milk in preference to full-cream milk to reduce caloric intake. The finding of an overall increase in total milk selection, including full-cream milk, during the first two weeks of the intervention period may be due to a spillover effect \[[@B39-nutrients-07-04336],[@B40-nutrients-07-04336]\] as a result of the nudging sign increasing the salience of milk (i.e., increasing awareness of its availability) resulting in an increase in selection of both full-cream and low-fat milk. There is evidence of spillover effects in the marketing literature \[[@B41-nutrients-07-04336],[@B42-nutrients-07-04336]\], particularly for products that are similar in nature \[[@B40-nutrients-07-04336]\], such as was the case for the low-fat and full-cream milk in the present study. Spillover effects have been evident in other nudging interventions, however in those studies overall selection decreased rather than increased. For example, rearranging the environment to increase visibility of healthier items and reduce visibility of less healthy items in a cafeteria decreased purchases of both healthier and less healthy items \[[@B32-nutrients-07-04336]\]. Thus, while nudging interventions typically aim to increase selection of healthier items, and decrease selection of less healthy items (i.e., substitution of less healthy options with healthier options), this effect does not always occur, but instead changes in overall selection can occur in either direction. The overall lack of efficacy of nudging to change milk selection in the present study is in agreement with the findings of another study by Olstad et al. \[[@B29-nutrients-07-04336]\] who placed descriptive labels (i.e., funky chicken teriyaki wrap) on healthy items, and found no influence on sales. Another study altered placement of snacks across a small set of four shelves to alter visibility of healthy choices and found no influence on snack item selection \[[@B43-nutrients-07-04336]\]. As was the case in the present study, these studies altered the visibility of one item, without changing visibility of the other. In contrast, nudging interventions that increased the visibility, accessibility and/or availability of a healthier items while simultaneously decreasing the visibility, accessibility and/or availability of the less healthy items \[[@B28-nutrients-07-04336],[@B43-nutrients-07-04336],[@B44-nutrients-07-04336],[@B45-nutrients-07-04336],[@B46-nutrients-07-04336]\], have been effective for influencing healthier dietary choices. This suggests that for nudges to be effective in achieving healthier dietary choices, it is not sufficient to only increase availability or awareness of healthier options, but awareness and/or availability of less-healthy options also need to be reduced. Thus, future research could test a message that encourages low-fat milk selection while simultaneously discouraging full-cream milk selection, for example "Pick me (the low-fat option) rather than me (the full-cream milk)!" Another suggestion could be to contrast "Pick me! I am low calorie" placed near the low-fat to "Don't pick me! I am high calorie" placed near the full-cream milk, to see if positive or negative messaging has different influences on milk selection. The transient effect on milk selection that was observed suggests that while nudges may exhibit efficacy acutely, these effects are short-lived and may subside within a relatively short period of time (i.e., within 1--2 weeks), thus future interventions could introduce a new nudging message every two weeks to ensure ongoing salience of the low-fat milk. The present study had a number of limitations. It is possible that the changes in milk selection behaviour that were observed during the initial phase of the intervention period may have been a result of the Hawthorne effect \[[@B47-nutrients-07-04336]\]. This effect occurs when participants are aware that they are being observed, and their behaviour is altered as a result of that knowledge \[[@B48-nutrients-07-04336]\]. Participants in the present study did not have their individual behaviour directly observed, so it was not possible to identify whether particular groups of individuals were influenced more than others by the nudge. Nor were people using the kitchen told that research was being conducted. However, as the present study was conducted in the kitchen of a university research institute, the staff may have been suspicious that milk selection was being observed during the intervention period when the sign appeared on the low-fat milk. The fact that the intervention was conducted in the kitchen of a University-based research institute might limit the generalizability of the findings to other settings. Staff and students who use this kitchen are likely to be from a higher socio-demographic and educational status and numerous studies have reported healthier dietary behaviours among higher socio-demographic groups with better education \[[@B49-nutrients-07-04336],[@B50-nutrients-07-04336]\]. While the health characteristics and dietary intake of the population are unknown, our data indicated that during the baseline period there was greater selection of low-fat milk compared with full-cream milk. This contrasts with figures for overall milk consumption in Australia which show that consumption of full-cream milk is almost double that of low-fat milk (i.e., 1169 million litres of full-cream milk per annum vs. 687 million litres of low-fat milk per annum) \[[@B51-nutrients-07-04336]\]. The greater selection of low-fat milk compared with full-cream milk at baseline in the present study not only suggests that the study sample observed was more health-conscious than the average Australian, and thus not a representative sample, it would also be expected to make it more difficult to further increase low-fat milk selection and decrease full-cream milk selection through nudging. Therefore, it is possible that similar interventions in environments where full-cream milk selection is predominant might result in a different outcome. 5. Conclusions ============== While salience nudging is a potentially feasible intervention that could be easily implemented into workplace environments, the present study showed that over a three-month period a salience nudge was not effective for influencing milk selection in a workplace. Instead, nudging resulted in an overall increase in selection of both low-fat and full-cream milk in the first two weeks, before returning towards baseline milk selection levels. The initial increase in milk selection indicates the presence of a spill-over effect due to the nudge increasing the salience of milk. Caution should be used when nudging dietary choices, to ensure that spillover effects do not result in over consumption. These findings suggest that salience nudging may have limited application for assisting people in changing milk selection. The authors would like to thank Alvin Atlas for statistical advice relating to the data analysis employed in this study. All authors conceived and designed the study. Amy L Wilson and Svetlana Bogomolova were responsible for data collection. Jonathan D Buckley and Amy L Wilson conducted the statistical analyses. All authors contributed to writing of the publication. The authors declare no conflict of interest. [^1]: These authors contributed equally to this work.	{ "pile_set_name": "PubMed Central" }
Background {#Sec1} ========== Over one third (35.8%) of males and 6.6% of females smoke tobacco globally with corresponding proportions of 36.8 and 2.8% in the Eastern Mediterranean Region (EMR) \[[@CR1]\]. The successful smoking cessation or tobacco dependence treatment can restore health status back to normal to a great extent. The risk of premature death and morbidity is reduced by 90.0% if smokers quit before the age of 30 or by 50.0% if they quit by the age of 50 \[[@CR2]\]. Appropriate smoking cessation services including national quit line, nicotine replacement therapy (NRT) and other cessation cost-covered services are now available, mostly in developed countries. But the progress towards tobacco cessation has been very slow since 2010 \[[@CR1]\]. In the six Gulf Cooperation Council (GCC) countries tobacco cessation services are available since 2014 \[[@CR3]\]. The tobacco smoking rates are still high in most EMR countries; including GCC countries, despite serious smoking control efforts \[[@CR4]\]. The tobacco prevalence in the region is about 32.0% for adults (≥15 years) and 36.0% for youth between 13 to15 years and it is on the increase in younger populations. Notably, the age standardized prevalence smoking rates among males are very high in Jordan (43.0%) and Tunisia (45.0%) \[[@CR2]\]. Waterpipe smoking in particular, is increasing in youth of most countries of the region as 39.0% and 31.0% of boys and girls, respectively, smoke it \[[@CR5]\]. Quitting tobacco is often challenging for smokers without support \[[@CR4], [@CR6]\]. However, about 84.0% of smokers can quit if they receive advice from a physician \[[@CR1]\]. In 2013, the age standardized prevalence of current use of any tobacco form was 30.3, 7.1 and 42.7% among adults of sexes combined, males and females, respectively in Bahrain \[[@CR1]\]. Significant proportions of adult male (11.0%) and female Bahraini (6.0%) are current shisha smokers \[[@CR7]\]. Lung cancer is the most common cancer in Bahraini males and second in females with annual age standardized incidence rates of 31.1/100,000 and 10.7/100,000, respectively \[[@CR8]\]. It has also been reported that more than 200 deaths are attributable to tobacco related diseases in Bahrain annually \[[@CR9]\]. Dedicated tobacco cessation clinics are necessary to help tobacco users quit. Bahrain ratified the World Health Organization's Framework Convention on Tobacco Control in 2007 and passed an antismoking law in 1994 with a modified one in 2009 \[[@CR4]\]. Tobacco advertisement, promotion and sponsorship are banned in the country \[[@CR10]\]. QTC were established in three health centers in the Kingdom since 2004 to provide services to residents wishing to quit smoking. The first clinic, Hoora was established in 2004 followed by Hamad Kanoo (2012) and Bank of Bahrain and Kuwait (2014). Although the services including consultation and NRT are given free of cost, the mean consultation cost, incurred by the Ministry of Health (MOH), for advice on smoking cessation is 4.5 Bahraini Dinar (BHD) (12 USD) excluding that of NRT. No efforts were made prior to our study to assess the usefulness of these clinics in smoking cessation. Such studies are essential for health policy makers in their tobacco control efforts; not only in Bahrain but also in other GCC countries. The objectives of the study were to determine the smoking quit rates among males attending the QTC, assess their satisfaction with the treatment services, determine their smoking behavior prior to consultation and describe the sociodemographic determinants of smoking cessation. Methods {#Sec2} ======= This cross sectional study was conducted between 10 August and 30 December 2015. A sample of 354 male patients was estimated and stratified based on the proportion probability to size in the two QTC; Hoora (280) and Hamad Kanoo (74) health centers. All males who attended these clinics in 2013 and 2014 were eligible for inclusion (Fig. [1](#Fig1){ref-type="fig"}). As females made a negligible proportion of the clinic attendees, they were excluded; so was the QTC at the Bank of Bahrain and Kuwait health center as it was recently established, in 2014.Fig. 1Study population and sampling schematic The data were collected through a questionnaire, in Arabic or English, and both versions were pilot tested on 10.0% of the clinic attendees who were excluded from the study. Patients were selected through simple random sampling method from the clinics' registers. The participants were called by phone and data collectors requested them to visit the clinics for interview after informing them about the objectives of the study and that participation was voluntary. The questionnaire had no identity verification and strict confidentiality of the participants was ensured. Interviews were performed by the trained QTC auxiliary staff. Consent forms were signed by the study participants before the interviews. A smoker was considered a quitter if he stopped smoking any tobacco product for at least six months after attending the clinic. Relapse in tobacco smoking was defined as resuming smoking after a complete abstention for at least a month. Descriptive analysis was done to present frequencies of sociodemographic, smoking behavior and quit tobacco related variables. Chi square or t-test was used to draw inferences about categorical and continuous variables, respectively. Results {#Sec3} ======= Data were available for 194 cases from the original sample. Overall, the response rate was 54.8%, with a slightly higher percentage from Hoora Clinic (56.1%) than Hamad Kanoo (50.0%) (Fig. [1](#Fig1){ref-type="fig"}). The majority of the study participants were Bahraini (80.6%), and the rest were other Arab and South Asians. Their mean age was 37.2 years with 37.3% having secondary education or higher and almost one third semi-skilled jobs and 30.4% had ever been unemployed (Table [1](#Tab1){ref-type="table"}). The average monthly income of the 95 who disclosed it was 781.1 ± 535 BHD (2071.91 ± 1419.12 USD).Table 1Sociodemographic characteristics of the patients attending the QTCCharacteristicMean±1SD Age of participant\ (n = 191)37.213.9 Total years of education\ (n = 190)12.73.3 Number of children\ (n = 132)3.51.7 Income (BHD) per month\ (n = 95)781.1535.1 Number of clinic visits2.72.0 Previous unemployment (years)\ (n = 55)2.03.0Nationality\ (n = 175)n% Bahraini14180.6 Other Arab2112.0 Others137.4Educational level\ (n = 193) Primary and below178.8 Intermediate2211.4 Secondary8242.5 Graduate and above7237.3Occupation\ (n = 194) Low professional2010.3 Skilled3116.0 Semi-skilled5628.9 Unskilled84.1 Unemployed42.0 Retired2512.9 Student3116.0 Unspecified199.8Ever unemployed\ (n = 194) Yes5930.4 No13569.6 Sources of information about the QTC and referrals {#Sec4} -------------------------------------------------- The participants' main source of information of the QTC was their friends (36.1%) followed by the primary healthcare establishment (25.1%), physician (7.3%) or other (17.8%). The majority (96.9%) of the patients were referred to QTC by their friends, wives and physicians. Further, 25 were referred due to illness, mainly diabetes mellitus (40.0%) and lung diseases (28.0%). Quit rates and reasons for quitting {#Sec5} ----------------------------------- On an average, the participants made 5.2 ± 11.0 and 3.0 ± 2.2 attempts to quit cigarettes (n = 185) and shisha (n = 41) smoking, respectively. Overall, 56.5% quit all forms of smoking during the study period with 37.6% quitting for six months or longer, after seeking the QTC services. The highest quit rate was among cigar smokers (70.0%) followed by shisha (63.8%), cigarette (55.1%) and pipe (16.7%) smokers (Fig. [2](#Fig2){ref-type="fig"}).Fig. 2Tobacco quit rates for six months or more at the Quit Tobacco Clinics by type of tobacco smoking (n = 194) Health (76.0%) and family (15.7%) were the main reasons, for the participants to quit tobacco smoking. Family (57.6%) and physicians (33.0%) were the main support for those who quit tobacco smoking. Personal problems (34.2%), habit (27%), friends (11.7%), enjoyment (11.7%) and nervousness (8.1%) were identified as the main reasons for not being able to quit more than one form of tobacco. Relapse and withdrawal {#Sec6} ---------------------- Sixty-three participants (35.4%) experienced relapse and started smoking again after successfully quitting. Personal problems (55.0%), friends (15.0%), loneliness (11.3%) and withdrawal symptoms (9.4%) were the reasons for the failure to sustain successful quit attempts. Headache (40.3%) and nervousness (42.9%) were the commonest withdrawal symptoms experienced by the quit attempters. NRT {#Sec7} --- One hundred eighty-one participants received NRT, mostly nicotine chewing gums and patches combined (79%), with only 3.6% given either Bupropion or Champix. One person used traditional herbal treatment to quit tobacco. Only 11.7% purchased NRT from a private pharmacy while the remaining received it for free from the QTC. For those who purchased NRT out of pocket, the cost was unaffordable to 7.2%. When asked if they had to buy NRT, if not provided by the clinic, 51.5% said that they would buy it. Participants' smoking behavior prior to consultation {#Sec8} ---------------------------------------------------- There were 188 cigarette smokers (96.9%) and 6 did not smoke cigarettes at all. 74.5% of the cigarette smokers smoked cigarettes only and the rest smoked it in addition other tobacco products. Forty-seven smoked shisha (23.2%), of whom 5 (10.5%) smoked shisha only and 37 (79.0%) smoked it along with cigarettes and 5 (10.5%) along with cigarettes and other types. Only one smoker smoked other tobacco products only (cigar) while the rest (cigar and pipe smokers) smoked cigarettes or shisha along with that product. The age of starting any type of tobacco smoking was 16.4 ± 7.8 years. Cigarette and shisha smokers {#Sec9} ------------------------------ Participants in our study started smoking cigarettes at an average age of 15.9 ± 4.4 years, smoked 27 ± 16.9 cigarettes per day, for 19.8 ± 12.8 years and spent 1.2 ± 0.7 BHD (3.2 ± 1.9 USD) daily on cigarettes. However, the mean age of starting to smoke, duration of smoking and cost were statistically significantly different when comparing cigarette and shisha smokers (p ≤0.001). The most common reason to start smoking cigarettes was friends (46.8%) or experimentation (29.3%) with LM (33.0%) and Marlboro (32.0%), the commonest brands smoked. Similarly friends (61.0%) and experimentation (13%) were the main reason to start shisha smoking. Fakher and Nakhla tobacco brands were smoked by 55% and 45%, respectively with grape (50.0%), apple (29.5%) and pomegranate (13.6%) the most smoked shisha flavors. Cigarette smokers had stronger craving reflecting very high tobacco dependence than the shisha smokers (p \< 0.001) (Table [2](#Tab2){ref-type="table"}).Table 2Comparison of cigarette and shisha smoking behaviorCigarettes\ (n = 188)\ Mean ± 1SDShisha\ (n = 47)\ Mean ± 1SDAge started smoking in years\ (n = 185)15.9 ± 4.419.7 ± 7.4Number smoked\ (n = 188)\27 ± 16.9 (day)3.4 ± 4.5 (week)Duration of smoking in years\ (n* = 177)\19.8 ± 12.88.2 ± 6.1Smoking cost (BHD)\ (n* = 187)\1.2 ± 0.7 (day)3.1 ± 3.3 (weekly)n* (%)n (%)Time of first cigarette/shisha smoked on a typical day\* Within 30 min of waking up145 (77.6)1 (2.1) \> 30 min of waking up42 (22.4)46 (97.9)Smoking in morning more than rest of the day\* Yes65 (34.9)1 (2.1) No121 (65.1)46 (97.9)Cigarette/shisha most hate to give up\* First in the morning73 (38.6)0 (0.0) Any other116 (61.4)42 (100)Smoked cigarettes during illness\* No94 (50.8)43 (91.5) Yes91 (49.2)4 (8.5)Difficulty avoiding smoking in prohibited areas\* No116 (61.4)45 (95.7) Yes73 (38.6)2 (4.3)\*P \< 0.001 Mean age, total years of education, income and number of children were not statistically significantly different among those who had quit all forms of tobacco compared to those who did not. Similarly, the mean age at starting cigarette smoking, number of cigarettes smoked per day, cigarette smoking quit attempts, duration and cost of cigarettes per week, total years of shisha smoking, number of shisha smoked per week, shisha quit attempts and cost of shisha were not statistically significant among the two groups. We found that those who quit all forms of tobacco were more likely to have been unemployed, smoked cigarettes even when they were sick, were unable to quit other forms of tobacco, would have wanted to seek treatment if they had to pay and were completely satisfied with the QTC (p \< 0.05) (Table [3](#Tab3){ref-type="table"}). The mean QTC visit frequency, counseling sessions, maximum abstinence from tobacco (months), quitting duration (months) and number of quit attempts were higher among those who had quit all forms of tobacco (p \<0.05). Lastly, 68.5% were satisfied with the clinics and 82.5% were completely satisfied with staff behavior and 73.2% with counseling sessions (Table [4](#Tab4){ref-type="table"}).Table 3Quitting all forms of tobacco by selected variablesQuit all forms of tobaccoAll formsNot alln (%)n (%)Ever unemployed\Yes27 (45.8)32 (54.2)No83 (61.5)52 (38.5)Smoke cigarettes while ill\Yes60 (63.8)34 (36.2)No45 (49.5)46 (50.5)Reason not able to quit other product (s)\Friends8 (61.5)5 (38.5)Alone2 (33.3)4 (66.7)Habituation6 (20.0)24 (80.0)Headache and tired0 (0.0)2 (100.0)Enjoy smoking1 (7.7)12 (92.3)Personal Problems15 (39.5)23 (60.5)Nervousness4 (44.4)5 (56.6)Had received treatment if had to pay\Yes64 (64.0)36 (36.0)No46 (48.9)48 (51.1)Overall satisfaction with the clinics\Not satisfied3 (60.0)2 (40.0)Slightly satisfied1 (25.0)3 (75.0)Satisfied11 (39.3)17 (60.7)Very satisfied10 (41.7)14 (58.3)Completely satisfied85 (63.9)48 (36.1\P \< 0.05 Table 4Comparison of quitting all types of tobacco versus not quittingQuit tobaccoNMeanSDMean difference95% CIP valueLowerUpperNicotine gum duration (days)Yes3444.267.817.6−26.4061.720.423No1126.643.4−18.2653.58Nicotine patch duration (days)Yes2930.427.74.3−17.5226.280.688No1326.041.5−22.2430.99Champix treatment duration (days)Yes11097.213.32.5−2.147.250.284No8494.619.8−2.407.51Traditional herbal medicine duration (days)Yes11099.00.02.3−0.505.210.105No8496.615.2−0.945.65Clinic visit frequencyYes813.32.210.351.820.004No562.22.10.351.82Counseling/advice sessionsYes1043.32.11.40.891.98\<0.001No791.91.50.911.96Maximum abstinence from tobacco (months)Yes10518.230.49.3−1.620.280.095No378.824.4−0.6319.26Quitting duration (months)Yes7721.525.6−17.1−33.36−0.930.038No1838.448.6−1257.41228.63Number of quit attemptsYes1066.213.82.9−0.116.090.059No783.22.20.295.68 Discussion {#Sec10} ========== The tobacco-quit rate of all forms of tobacco (56.5%) by QTC attendees is encouraging, indicating that the clinics contribute to the tobacco control efforts in the country and may decrease the health burden of tobacco smoking. It is also worth noting that almost two thirds of the attendees had no relapse after quitting. Study participants smoked for a longer (19.8+ 12.8 years) time than that of shisha smokers (8.2 ± 6.1 years), which might also have contributed to the higher quit rates among the latter group. However, the fact that there were no statistically significant differences between those who completely quit all types of tobacco, cigarettes and shisha versus those who did not by smoking duration, frequency and quit attempts indicates that other factors might play a role in smoking cessation. The higher quit rates among shisha (63.8%) compared to cigarette smokers (55.1%) might suggest that quitting is easier among the former group. Studies on shisha intervention are scarce; however a recent review reported that interventions might help waterpipe smokers successfully quit smoking \[[@CR11]\]. A study on waterpipe smokers in coffee shops in Bahrain reported that 82.0% expressed their ability to quit anytime, but only 40.0% wished to do so \[[@CR12]\]. Cigarette smoking was the commonest form of tobacco smoked (96.9%). The majority smoked cigarettes alone (72.2%) without smoking other types of tobacco. The lower prevalence of shisha smokers attending the QTC can be partly attributed to the fact that shisha smoking may not be considered as addictive and harmful as cigarettes by the Bahraini community \[[@CR13]\]; which warrants further efforts in educating the public about its harmful effects. Age started cigarette smoking (15.9 ± 4.3 years) was lower than that of shisha (19.7 ± 7.4 years). It was also slightly lower than that of cigarette smoking among the general male population (18.4 ± 5.2). Further, the average number of cigarettes smoked daily (27 + 16.9) was higher than that of the general adult population (19.35 ± 12.5) \[[@CR14]\]. Unfortunately, there are no comparative data for shisha smoking with the general population. The majority of the smokers knew about the QTC from their friends. This finding emphasizes the role of friends and peers not only in smoking uptake but also in smoking cessation. Further, the fact that wives asked their husbands to attend the clinic implies that they were worried about the health of their spouses and other members of the family being exposed to second hand smoke. It is disappointing that only 7.3% knew about the clinics from their physicians. This is not surprising as physicians in in the region have low perception of their role in helping smokers quit \[[@CR15], [@CR16]\]. Only 55.0% of the participants reported that their physicians asked them about smoking habits. Desire for quitting was related to physician's advice, family's attitude to shisha smoking, not considering own self being addicted and being a non-Bahraini \[[@CR12]\]. This finding supports the important role of physicians in smoking cessation in Bahrain. It has been recommended in the smoking cessation guidelines for health professionals that they should advise smokers among their patients to quit and refer them to smoking cessation services if needed \[[@CR17]\]. Although 33.0% of the participants reported that they received support from their physicians in quitting, it is disappointing that only 7.3% knew about the QTC from their physicians, in our study, although their advice on quitting tobacco smoking increases quit rates \[[@CR18]\]. In Jordan, 19.9% had received advice from their physicians to stop smoking and only 2.4% had utilized the quitting cessation services \[[@CR19]\]. The fact that 93.3% required NRT implies that the clinic attendees were those who could not successfully quit tobacco smoking through counseling alone, because of tobacco dependence. This also underscores the scale of economic burden on controlling tobacco smoking by the MOH in Bahrain and calls upon more efforts to be focused towards prevention of initiation of tobacco smoking. Our study showed that clients who were able to quit all forms of tobacco had made one additional visit to the QTC (p = 0.004) and 1.4 more counseling sessions (p = \<0.0001) than those who did not quit. We also found that quitters had abstained from smoking for 9.3 months more than the non-quitters (p = 0.038). The lower satisfaction rates related to clinic days and opening hours warrants attention. But the higher satisfaction rates with the staff and the counseling is rewarding. Among the unemployed, there was a statistically significant (p \< 0.05) relationship of successful quitting for those having less number of years of unemployment compared to their counterparts who had been unemployed for longer periods. This is probably due to the fact that the latter could be more stressed and could resort to tobacco dependence as a relief. Over three quarters of the cigarette smokers smoked their first cigarette within the first half hour after waking up, which reflects that they were addicted to cigarette smoking. We found that cigarette smokers had a very high tobacco dependence that they would even smoke during illness (p \< 0.05) and that most of them would not give up a morning cigarette which, along with the fact that they smoked an average of 27 cigarettes per day for the past 20 years. However their desire to quit tobacco was reflected by the fact that successful quitters reported that they could buy NRT even if they had to pay for it (p = 0.05). We did not find any significant difference in the successfully quitting with respect to the type and method of NRT used; however, as most quitters used a combination of a nicotine patch and gum we assume that it was the best method in Bahrain to practice and was found to be highly acceptable and convenient by the quitters. Strengths and limitations {#Sec11} ------------------------- The strength of our study was that it was the first of its kind in assessing a national tobacco intervention from a GCC country. Having QTC management as part of the study team would hopefully improve the quality of QTC services and healthcare policy makers in Bahrain would consider empowering these clinics and increasing their geographical distribution. The main study limitation was the low response rate. Several patients who attended the clinics early 2013 had changed their contact numbers and could not be traced. Others were reluctant to be interviewed. Since our sample was drawn from the patients who came to the health centers for tobacco cessation treatment, the quit rates may not reflect those of the general male population in Bahrain. Conclusions {#Sec12} =========== The fact that more than half of the QTC attendees were able to quit tobacco emphasize the importance of these clinics in tobacco control in Bahrain. Better dissemination of information about the clinics by the healthcare providers is of paramount importance. Public awareness of the QTC clinics through educational establishments, schools, colleges, universities and media is of importance. The MOH may consider increasing QTC accessibility by increasing working hours and open days for the existing QTC. BHD : Bahraini Dinar EMR : Eastern Mediterranean Region GCC : Gulf Cooperation Council MOH : Ministry of Health NCDs : Non-communicable diseases NRT : Nicotine replacement therapy PHC : Primary healthcare QTC : Quit tobacco clinics We are thankful to the QTC team for their support in the study. Funding {#FPar1} ======= The study was funded by World Health Organization/The Eastern Mediterranean Regional Office Special Grant for Research in Priority Areas of Public Health 2014--2015 (RPPH 14--39). Authors' contributions {#FPar2} ====================== RRH designed the study, contributed in the manuscript, JA analyzed and prepared the manuscript, MK and SB contributed to the data collection and management. All authors read and approved the final manuscript. Competing interests {#FPar3} =================== The authors declare that they have no competing interests. Consent for publication {#FPar4} ======================= Not applicable. Ethics approval and consent to participate {#FPar5} ========================================== Ethical approval for the study was obtained from the Research and Ethics Committee at the College of Medicine and Medical Sciences, Arabian Gulf University and from the Research and Technical Support Team, MOH, Bahrain. Data set {#FPar6} ======== The dataset during and/or analyzed during the current study is available from the corresponding author on reasonable request.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION {#s1} ============ Parkinson's disease (PD) is caused by dopamine deficiency, which negatively affects cerebral basal ganglia function[@r1]^)^. The progressive chronic neurodegeneration of the brain associated with PD impairs balance and motor ability. These chronic, negative effects on motor function are accompanied by memory defects and a decreased ability to understand[@r1], [@r2]^)^. In addition, emotional stability deteriorates, which results in difficulty with activities of daily living (ADL) and an increasing feeling of despair and depression; which causes a decline in the quality of life of a patient[@r2], [@r3]^)^. The treatment of PD with medication prescribed by doctors replenishes the deficient dopamine level preventing or delaying nerve cell destruction, thus delaying the progress of the disease[@r4]^)^. However, in cases receiving long-term treatment, disease acceleration, motor disturbance, cognitive impairment, or autonomic nerve malfunction may develop[@r5]^)^. Exercise protects and promotes regeneration of damaged nerves, and enhances balance, physical state, and independency, thereby facilitating ADL. With an appropriate exercise intensity, dopamine levels can be increased and motor disturbances can be improved[@r6], [@r7]^)^. Therefore, treatment needs to integrate prescription medicines and exercise therapy. Exercises in virtual reality provide visual, auditory, and tactile feedback, and the exercise intensity can be modified by adjusting the difficulty of interactions and intensity[@r8], [@r9]^)^. There are no limitations in terms of time or place, thus enabling continuous active exercise[@r8]^)^. Virtual reality dance exercise has been reported to have a positive effect on cerebral palsy, dementia, and the recovery of physical function of stroke patients[@r9],[@r10],[@r11],[@r12]^)^. However, there are insufficient studies of the effects of virtual reality dance exercise on PD patients. In this study, we aimed to examine the effect of virtual reality dance exercise on balance, ADL, and depressive disorder status of PD patients. SUBJECTS AND METHODS {#s2} ==================== Twenty participants with PD were randomly assigned to either the experimental group (5 males, 5 females) or the control group (5 males, 5 females). Patients were able to understand the study requirements and to communicate sufficiently, and could ambulate independently. All participants were informed about the objectives and method of the study and agreed to participate. The study protocol was approved by the Institutional Review Board of Nambu University and was conducted in accordance with the ethical principles of the Declaration of Helsinki. The general characteristics of the participants are listed in [Table 1](#tbl_001){ref-type="table"}Table 1.General characteristics of each groupEG (n=10)CG (n=10)Gender (male/female)5/55/5Age (years)68.4±2.9^a^70.1±3.3Height (cm)165.1±9.1165.2±8.4Weight (kg)63.8±10.263.7±10.1^a^Mean±SD, EG: Experimental Group, CG: Control Group. The age (mean±SD) of the experimental group was 68.4±2.9 years, height was 165.1±9.1 cm, and weight was 63.8±10.2 kg. The age of the control group was 70.1±3.3 years, height was 165.2±8.4 cm, and weight was 63.7±10.1 kg. All participants received 30 minutes of neurodevelopment treatment (NDT) and 15 minutes of functional electrical stimulation (FES) 5 times per week for 6 weeks. The experimental group participated in an additional 30 minutes of dance exercise. The virtual reality dance exercise used the K-Pop Dance Festival (Nintendo Inc., Japan) game for the Wii (Nintendo Inc., Japan) video game system. Songs liked by patients were selected from the various categories of K-Pop music included in the software. A strap was used to fix the remote control to the hands, and the patients tried to mimic the characters on the TV monitor. When subjects properly mimicked the movement, they felt vibrations from the remote control and heard the word "perfect" broadcast by the TV speaker. Balance was assessed using the Berg balance scale (BBS). The BBS consists of 14 items and can be divided into balance for sitting, standing, and position changes. The score given ranges from 0 to 4 points, with a highest possible score of 56 points, and higher scores indicate better balance. ADL was assessed using the Modified Barthel index (MBI). The MBI consists of 10 items: 5 points for personal hygiene, 5 points for taking a bath, 10 points for eating, 10 points for using the toilet, 10 points for ascending stairs, 10 point for putting on clothes, 10 points for defecation, 10 points for urination, 15 points for walking, and 15 points for using the bed, with a highest possible score of 100 points. A score of 0--24 points indicates complete dependence; 25--48, maximum dependence; 50--74, partial dependence; 75--90, slight dependence; 91--99, minimum dependence; and 100, complete independence. The presence and severity of depressive disorder was assessed using the Beck depression inventory (BDI). The BDI consists of 21 items covering emotional and physiological symptoms. The score ranges from 0 to 3 points, with a highest possible total of 63 points. A score of 0--9 points indicates no depression; 10--15, light depression; 16--23, heavy depression; and 24--63, severe depression. The balance scores, ADL score, and depressive disorder status were recorded before and after the 6 weeks of treatment. Data were analyzed using SPPS 12.0 for Windows. Descriptive statistics were used to describe the general characteristics of the participants. The paired t-test was used to compare pre- and post-treatment variables, and the independent t-test was used to compare the treatment groups. The significance level for all tests was α=0.05. RESULTS {#s3} ======= The changes in balance are shown in [Table 2](#tbl_002){ref-type="table"}Table 2.Comparison of BBS, MBI, and BDI values between the experimental and control groupsGroupPrePostD-valueBBSEG46.0±1.3^a^48.1±3.0\2.1±2.3\CG45.0±1.345.4±1.50.4±0.8MBIEG87.9±1.491.1±3.0\3.2±3.0\CG87.4±1.788.2±1.80.8±1.6BDIEG20.4±0.918.2±2.0\−2.2±1.9\CG21.2±1.320.6±1.5−0.6±0.8^a^Mean±SD, \*p\<0.05, EG: Experimental Group, CG: Control Group, D-value: Difference-value, BBS: Berg Balance Scale, MBI: Modified Barthel Index, BDI: Beck Depression Inventory. After 6 weeks of treatment, balance had significantly improved in the experimental group (46.0±1.3 to 48.1±3.0; p\<0.05), while the control group showed no significant improvement (45.0±1.3 to 45.4±1.5; p\>0.05). Compared to the control group, balance of the experimental group was significantly enhanced (p\<0.05). Changes in ADL are shown in [Table 2](#tbl_002){ref-type="table"}. After treatment, ADL had significantly improved in the experimental group (87.9±1.4 to 91.1±3.0; p\<0.05), while ADL of the control group showed no significant improvement (87.4±1.7 to 88.2±1.8; p\>0.05). Compared to the control group, ADL of the experimental group were significantly enhanced (p\<0.05). The changes in depressive disorder status are shown in [Table 2](#tbl_002){ref-type="table"}. After treatment, the depressive disorder status significantly improved in the experimental group (20.4±0.9 to 18.2±2.0; p\<0.05), while the depressive disorder status of the control group showed no significant improvement (21.2±1.3 to 20.6±1.5; p\>0.05). Compared to the control group, the depressive disorder status of the experimental group was significantly improved (p\<0.05). DISCUSSION {#s4} ========== This study aimed to examine the effect of virtual reality dance exercise on the balance, ADL, and depressive disorder status of PD patients. Within-group comparisons demonstrated that balance improved significantly in the experimental group. The group comparison showed balance of the experimental group significantly improved relative to the control group. The present results corroborate those of Esculier et al.[@r13]^)^, Yen et al.[@r14]^)^ and Suarez et al.[@r15]^)^ who showed that virtual reality dance exercise enhanced the balance ability of PD patients. Virtual reality exercise provides visual and auditory feedback while watching the motion of the game character, thus improving balance through integration of feedback from the scala vestibule and proprioceptors. Hence, this method can enhance the balance of PD patients[@r8], [@r9], [@r13]^)^. Virtual reality exercise provides active learning and motivation to patients through an experiential environment[@r16]^)^. Yavuzer et al.[@r17]^)^ showed that virtual reality dance exercise significantly enhances functional independence, and Zhang et al.[@r18]^)^ showed that it resulted in a significant enhancement of ADL. The present study also found there was a significant enhancement of ADL of PD patients following their performance of virtual reality dance exercise. Dance exercise helps to maintain and increase the range and approach of a patient's movement; it also helps with switching the direction of movement, when performing two simultaneous but separate movements in conjunction with music. Moreover, it provides various unpredictable movement patterns, thus helping PD patients who have impaired ADL[@r19],[@r20],[@r21]^)^. A depressive disorder may be caused by the disease itself or may be a side effect of the treatment medication itself. Depressive disorder delays the recovery of physical function and makes it more difficult to perform ADL. Lee et al.[@r22]^)^ and Kim and Lee[@r23]^)^ demonstrated that virtual reality exercise reduces the presence and severity of depressive disorders. The present study corroborates these previous findings since it showed a beneficial effect of virtual reality exercise on the depressive disorder status of PD patients. K-Pop virtual reality dance exercise was performed by the subjects of this study. K-Pop dance utilizes repetitive lyrics and melodies and consists of simple brisk rhythms and beats. Because these features can be used as a targeted treatment and because virtual reality exercise can elicit voluntary participation, the K-Pop virtual reality dance exercise may help improve the depressive disorder status of PD patients[@r23], [@r24]^)^. A limitation of this study was that it was conducted with only 20 PD patients, thus making it difficult to generalize the result to all patients with the disease. Further research with a larger cohort is needed. Another limitation of this study is a lack of specific history regarding the treatments the patients received prior to their participation. There is a need for additional research on the effect of various exercise programs other than virtual reality dance exercise.	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec0005} =============== Polymer electrolytes have been extensively researched due to the possibility of their commercial use in a variety of electrochemical devices such as electrochemical cells, rechargeable batteries, sensors and supercapacitors \[[@bib0005], [@bib0010], [@bib0015], [@bib0020], [@bib0025], [@bib0030], [@bib0035]\]. However, in order polymer electrolytes to be used in lithium ion batteries, further work is still needed to improve their physical and electrical properties. In their efforts to find polymer electrolytes with adequate ionic conductivity and thermal stability at ambient and sub-ambient temperatures, researchers have synthesized and characterized different polymer-salt complexes \[[@bib0040], [@bib0045], [@bib0050], [@bib0055], [@bib0060], [@bib0065], [@bib0070], [@bib0075]\]. Of all the polymer systems investigated, poly(ethylene oxide) (PEO) and poly(propylene oxide) (PPO), complexed with various suitable salts such as LiPF~4~, LiClO~4~ and LiBF~4~ are the most widely employed systems. Other polymer hosts under extensive investigations include poly(methyl methacrylate) (PMMA), poly (acrylonitrile) (PAN), poly(vinyl chloride) (PVC), polyvinyl alcohol (PVA), poly(vinyl acetate) (PVAc) poly(vinylidene fluoride) (PVdF) and, most recently, poly(vinylidene fluoride-co-hexafluoropropylene) (PVdF-HFP). In recent years, polymer blending has attracted a great deal of interest due to its simplicity of preparation and ease of control of its properties [@bib0080]. The possibility of using polymer blend electrolytes for rechargeable lithium batteries was explored by some researchers \[[@bib0085], [@bib0090], [@bib0095], [@bib0100]\]. Reports on lithium ion conducting PVdF-HFP based blend electrolytes have indicated that blending not only results in enhanced conductivity but it also improves the mechanical property of the electrolytes. Among the blend systems studied are PVDF-HFP/PVC/LiClO~4~ [@bib0085], PVDF-HFP/PMMA/LiClO~4~ [@bib0090], PVDF-HFP/hydroxypropyl methyl cellulose (HPMC)/LiMn~2~O~4~ [@bib0095] and PVDF-HFP/PVAc/LiClO~4~ [@bib0105]. In the present study, ion conducting PVdF-HFP:PMMA blends consisting of lithium triflate (LiTf) as dopant salt were prepared to investigate the effect of PMMA concentration on structural, thermal and electrical properties of the polymer electrolyte. Polymer electrolytes composed of PVdF-HFP are expected to have good electrochemical stability and non-combustibility owing to the strong electron-withdrawing functional groups, (-C-F-), in the VdF units [@bib0110] and plasticity due to steric hindrance provided by -CF~3~- pendant group in the HFP monomors [@bib0115]. That is, it has both a crystalline phase due to its VdF units and an amorphous phase due to its HFP monomers. The amorphous phase of PVdF-HFP facilitates the trapping of a large amount of liquid electrolyte providing mobile cations more free volume and hence helping in enhancing ionic conductivity, whereas the crystalline phase acts as a structural support for the formation of free-standing polymer electrolyte film. Also, PVdF-HFP possesses advantageous inherent properties such as high dielectric constant (ε = 8.4) and low glass transition temperature. The high dielectric constant facilitates dissociation of salts, giving rise to a high concentration of charge carriers in the electrolyte and thus higher ionic conductivity. Therefore, PVdF-HFP is chosen as the host polymer in preparing the polymer blend systems in the present study. PMMA is a methacrylic ester polymer which has a very high amorphicity and good transparency. PMMA was chosen as the blending agent by taking into consideration that the presence of amorphous phase is very important for improving conductivity at low temperatures. Its sufficient mechanical strength and compatibility with PVdF-HFP is also another advantage to consider it as a suitable blending agent. The blend films were characterized using XRD, FTIR, SEM and DSC. Conductivity and ionic transference number measurements were carried out for the blend films of varying PVdF-HFP:PMMA compositions. 2. Experimental {#sec0010} =============== PVdF-HFP (Average molecular weight $4 \times 10^{5}$ g/mol, Aldrich, USA) and PMMA (Average molecular weight $1.2 \times 10^{5}$ g/mol, Aldrich, USA) were dried at 70 °C and LiTf (Aldrich, USA) was dried at 100 °C under vacuum for about 10 h before use with the aim of eliminating any moisture. The solvent anhydrous tetrahydrofuran (THF) was used as received. All polymer electrolyte films were prepared by the solution casting technique. The mixture of the appropriate quantities of the polymers and the salt was dissolved in THF by continous magnetic stirring for several hours. The resulting homogenous solution was then cast as film onto a clean petri dish to allow THF to evaporate slowly at room temperature. Then mechanically stable and free-standing films were formed. The films were further dried in a temperature-controlled hot-air oven at 60 °C for 5 h to remove any traces of THF. The weight percentage of the salt was maintained at a fixed amount of 25 wt.%. The PVdF-HFP:PMMA:LiTf compositions used during the film preparation are presented in [Table 1](#tbl0005){ref-type="table"}.Table 1Composition of PVdF-HFP:PMMA:LiTf complexes.Table 1Sample codePVdF-HFP (wt.%)PMMA (wt.%)LiTf (wt.%)BL167.57.525BL2601525BL352.522.525BL4453025BL537.537.525 X-ray diffraction (XRD) studies of the electrolyte films were carried out using PHILIPS X'Pert diffractometer, operating at 40 kV, in the glancing angles ranging from 2θ = 10° to 80°. The FTIR spectra were recorded using IRPrestige-21 (Shimadzu) spectrophotometer over the wavenumber range of 4000--400 cm^−1^. The films were examined under scanning electron microscope (SEM) (ZEISS EVO18) for their morphological properties. Thermal analysis was carried out for the samples using Q20 differential scanning calorimeter at a heating rate of 10 °C/min. Conductivity analysis of the various electrolyte samples was carried out using Hioki 3532 LCR HiTester in the frequency range 42Hz to 1 MHz. The ionic conductivity (σ) of each sample, at various temperatures between 303 and 393 K, was calculated from bulk resistance measurements obtained from impedance plots. Lithium ion transference number of the polymer blend electrolytes was determined by DC polarization technique [@bib0120]. The electrolyte film was coated on both faces with conductive silver paste so that it makes good contact with the stainless steel (SS) electrodes. A small DC potential (∼ 1 V) was applied across the SS/Electrolyte/SS configuration to initiate flowing of current through the system. Consequently, the DC current passing through the electrode-electrolyte system was monitored as a function of time until saturation was reached. Transference number was determined at room temperature (303 K). 3. Results and discussion {#sec0015} ========================= 3.1. XRD analysis {#sec0020} ----------------- In order to investigate the phase nature of the blend films, XRD analysis was performed. The XRD patterns of pure LiTf, PVdF-HFP, PMMA and PVdF-HFP:PMMA:LiTf complexes are presented in [Fig. 1](#fig0005){ref-type="fig"}. The figure shows that high diffraction intensity, sharp peaks of pure LiTf are observed at several values of 2θ. It can be noted from the XRD patterns of the samples in [Fig. 1](#fig0005){ref-type="fig"} (BL1--BL5) that the sharp crystalline peaks corresponding to LiTf are absent in all the polymeric blends, confirming that no separate phase of the lithium salt exists in the blends. This indicates a complete complexation has occurred between the polymer blend and the lithium salt. The diffractogram of pure PVdF-HFP ([Fig. 1](#fig0005){ref-type="fig"}) reveals two well distingushed peaks at 2θ = 18.7° and 20.2°. These peaks indicate the semicrystalline nature of pure PVdF-HFP [@bib0120]. Moreover, the predominantly amorphous nature of PMMA can be observed from the XRD pattern of pure PMMA in [Fig. 1](#fig0005){ref-type="fig"}. However, it is observed that upon blending of both polymers and complexing them with LiTf, the two Bragg peaks of PVdF-HFP disappear and a new broadened peak with lowered intensity is observed in the blend system. This suggests an increase of amorphicity of the polymeric blend films by addition of PMMA and LiTf and the occurrence of complexation between PVdF-HFP, PMMA and LiTf.Fig. 1XRD patterns of LiTf, PVdF-HFP, PMMA and PVdF-HFP:PMMA:LiTf complexes.Fig. 1 3.2. FTIR spectral analysis {#sec0025} --------------------------- [Fig. 2](#fig0010){ref-type="fig"} shows the FTIR spectra of pure PVdF-HFP film, pure LiTf powder, pure PMMA powder and blend polymer electrolyte films containing different concentrations of PMMA in the wave number range 4000 to 400 cm^−1^. The absence of the absorption bands 976, 796, 760 and 530 cm^−1^, which are all associated to the α-phase crystals of the VdF units \[[@bib0125], [@bib0130], [@bib0135], [@bib0140]\], in the PVdF-HFP:PMMA blends indicates that the addition of PMMA to PVdF-HFP results in reduction of crystallinity of PVdF-HFP. The absorption bands of PVdF-HFP detected at 1383 and 490 cm^−1^ are assigned to the wagging vibrations of CH~2~ and CF~2~ groups \[[@bib0145], [@bib0150]\], respectively. These peaks were observed to shift to 1398 and 482 cmcm^−1^, respectively, in the blend polymer electrolytes.Fig. 2FTIR spectra of LiTf, PVdF-HFP, PMMA and PVdF-HFP:PMMA:LiTf complexes.Fig. 2 The strong peaks of PMMA detected at 1445, 1244, 1150 and 989 cm^−1^, which are assigned to O-CH~3~ bending, C-C-O bending, C-O-C asymmetric stretching and CH~2~ wagging [@bib0090], respectively, were all absent in the blends. The IR absorption peaks for PMMA at 2951 (CH symmetric stretching), 1732 (C=O asymmetric stretching [@bib0155]), 843 (C-O-C stretching) and 751 cm^−1^ (CH~2~ rocking [@bib0090]) shifted to 2918, 1718, 873 and 758 cm^−1^, respectively, due to the interaction between PVdF-HFP, PMMA and LiTf in the blend complexes. Moreover, new peaks were identified around 2500, 2305, 2214, 2090 and 1985 cm^−1^ in the blends. The absorption peak of LiTf at 1640 cm^−1^, corresponding to the characteristic peak of Tf^−1^ [@bib0160], got shifted to 1635 cm^−1^ in the polymer blend electrolytes. Most of the other peaks detected for LiTf at various frequencies were absent in the blend systems. The shifting and absence of IR bands of the pure polymers or the salt, as well as the appearance of new peaks in the PVdF-HFP:PMMA:LiTf blend systems suggest that the two polymers are compatibly mixed with each other and strongly interact with the dopant lithium salt. 3.3. Morphological studies {#sec0030} -------------------------- The morphological study of the blend which exhibited maximum conductivity has been carried out using the SEM. [Fig. 3](#fig0015){ref-type="fig"} shows the SEM images of PVdF-HFP, PMMA and PVdF-HFP:PMMA:LiTf (52.5:22.5:25 wt. %). As evident from the images, the surface morphology of PVdF-HFP is significantly modified due to the addition of PMMA and LiTf. Thus, no separate phases exist in the blend system indicating that PMMA can compatibly be blended with PVdF-HFP. It can also be seen from the SEM micrographs in [Fig. 3](#fig0015){ref-type="fig"}(c) that the surface of PVdF-HFP:PMMA:LiTf film has numerous network of interconnecting microstructures as compared to that of pure PVdF-HFP and PMMA films in [Fig. 3](#fig0015){ref-type="fig"}(a) and [Fig. 3](#fig0015){ref-type="fig"}(b). It is, therefore, evident that the blending results in a completely different morphological structure. These interconnecting structures are formed as a result of phase separation between the polymers and the solvent during evaporation. The microstructures can help the lithium ions hop easily in the blend system and the electrolyte absorb more liquid, and hence they are responsible for improving ionic conductivity of the polymer blend electrolyte.Fig. 3SEM micrographs of (a) PVdF-HFP, (b) PMMA and (c) PVdF-HFP:PMMA:LiTf (52.5:22.5:25) (at 2500 × magnification).Fig. 3 3.4. DSC studies {#sec0035} ---------------- Thermal property of the polymer blend electrolyte with the highest conductivity was analyzed using DSC measurements. [Fig. 4](#fig0020){ref-type="fig"} shows the DSC curves of pure PVdF-HFP, pure PMMA, PVdF-HFP:LiTf (75:25 wt.%) and the highest-conducting PVdF-HFP:PMMA:LiTf film (BL3). PVdF-HFP exhibits a sharp endothermic peak at 135 °C \[[Fig. 4](#fig0020){ref-type="fig"}(a)\], which is ascribed to its melting point. The endothermic peak of the curve in [Fig. 4](#fig0020){ref-type="fig"}(b) at 160 °C corresponds to the melting point of PMMA. The melting temperature of PVdF-HFP was observed to shift to a lower temperature (114 °C) \[[Fig. 4](#fig0020){ref-type="fig"}(c)\] by the addition of 25 wt.% LiTf. However, when the PVdF-HFP:LiTf complex is blended with PMMA, the endothermic peak obtained at 114 °C before blending is observed to re-shift towards a higher temperature of 136 °C.This can be observed from the DSC curve for BL3 in [Fig. 4](#fig0020){ref-type="fig"}(d). This implies that the thermal stability of the salted PVdF-HFP electrolyte is improved significantly by blending it with PMMA. In addition, no exothermic peaks were detected for the investigated polymer blend electrolyte, indicating that no decomposition takes place up to at least 350 °C. These results are interesting from device application point of view because wider thermal stability is desirable in polymer electrolytes for battery and fuel applications. The second small endothermic peak observed around 165--174 °C for the blend \[[Fig. 4](#fig0020){ref-type="fig"}(d)\] is attributed to the melting point of the crystalline phase PVdF [@bib0165].Fig. 4DSC curves of PVdF-HFP, PMMA, PVdF-HFP:LiTf (75:25) and PVdF-HFP:PMMA:LiTf (52.5:22.5:25) complex.Fig. 4 3.5. Conductivity studies {#sec0040} ------------------------- The electrical conductivity of blends of PVdF-HFP and PMMA complexed with LiTf was investigated in order to determine the optimal blend composition for maximum conductivity. The conductivity values were calculated using the equation $\sigma = \frac{t}{{AR}_{b}}$, where t is the thickness, A is the area, and R~b~ is the bulk resistance of the polymer electrolyte disc obtained from impedance plots. The electrolyte resistance was determined from the real axis intercept of the low frequency spur in the Nyquist plot for the polymer electrolytes. The impedance plots for the PVdFHFP:PMMA:LiTf blend films containing different PVdF-HFP:PMMA compositions (at room temperature) are shown in [Fig. 5](#fig0025){ref-type="fig"}. The plots show two well-defined regions - semicircular region at high frequency range and linear region at low frequency range. The semicircular region in the high-frequency range is related to conduction process in the bulk of the blend complex and the linear region in the low-frequency range is attributed to the effect of blocking electrodes \[[@bib0170], [@bib0175]\].Fig. 5Impedance spectra of PVdF-HFP:PMMA:LiTf complexes containing different PMMA concentrations at room temperature.Fig. 5 The room temperature conductivity data of the polymer blend electrolytes as a function of PMMA concentration are presented in [Table 2](#tbl0010){ref-type="table"}. It is observed that the ionic conductivity initially increases with PMMA content up to its maximum value at 22.5 wt.% of PMMA. The maximum room temperature (303 K) ionic conductivity achieved for the polymer blend electrolyte comprising 22.5 wt.% of PMMA is $7.4 \times 10^{- 5}$ S cm^−1^. The increase in conductivity with PMMA content could be due to increase in amorphicity of the blend film during complexation (as witnessed by the XRD results). However, upon further increase of the concentration of the PMMA over 22.5 wt.%, the conductivity falls down. The ionic conductivity of PVdF-HFP:PMMA:LiTf blend electrolytes as a function of PMMA concentration at various temperatures is shown in [Fig. 6](#fig0030){ref-type="fig"}.Fig. 6Conductivity of PVdF-HFP:PMMA:LiTf complexes containing different PMMA concentrations at various temperatures.Fig. 6Table 2Ionic conductivity and transference number of PVdF-HFP:PMMA:LiTf complexes at room temperature (303 K).Table 2Sample codeConductivity ($\times 10^{- 5}$Scm^−1^)t~i~BL16.60.84BL26.80.88BL37.40.86BL44.90.86BL54.20.85 The temperature dependence of conductivity was also investigated at various temperatures ranging from 303--393 K in order to understand the conductivity-temperature relationship of the polymer blend. [Fig. 7](#fig0035){ref-type="fig"} shows variation of ionic conductivity (in terms of log σ) versus reciprocal temperature of the polymer blends for 7.5, 15, 22.5, 30 and 37.5 wt.% of PMMA. The conductivity of all the blend systems is observed to increase with temperature. The logarithm of conductivity versus reciprocal temperature plots of the samples studied show nonlinearity, suggesting that the temperature-dependent ionic conductivity of the polymer blend electrolytes obeys the Vogel-Tamman-Fulcher (VTF) rule. Similar behaviour has been observed in other earlier reports on conductivity property of various polymer blend systems \[[@bib0080], [@bib0090], [@bib0115]\]. The observed increase in conductivity with temperature could be due to increased segmental motion of the polymer chains and mobility of ions in amorphous environment. This behaviour may be explained with the help of the free volume theory in polymers \[[@bib0180], [@bib0185]\]. With an increase of temperature, the rate of dissociation of the ionic salt and the thermal movement of the polymer chains increase, leading to higher ionic conductivity in the polymer blend complex.Fig. 7Temperature dependence of ionic conductivity of PVdF-HFP:PMMA:LiTf polymer blend electrolytes containing different PMMA concentrations.Fig. 7 3.6. Ionic transference number {#sec0045} ------------------------------ The variation of DC electric current as a function of time was recorded over a time range of 120 min. [Fig. 8](#fig0040){ref-type="fig"} depicts the current versus time variation for the highest-conducting polymer blend electrolyte at room temperature (303 K). Using the initial total current I and final stabilized current I~e~ after polarization, the lithium ion transference number t~i~ of the various polymer blend electrolyte systems was estimated from the relationship,$$t_{i} = {\frac{I - I_{e}}{I}.}$$Fig. 8Variation of DC current with time for PVdF-HFP:PMMA:LiTf (52.5:22.5:25).Fig. 8 The ionic transference number data for the blend polymer electrolytes are displayed in [Table 2](#tbl0010){ref-type="table"}. The lithium ion transference numbers obtained are high (0.84--0.88) means that the charge contribution comes predominantly from the ions dissociated in the blends. 4. Conclusions {#sec0050} ============== Self-standing polymer blend electrolytes based on PVdF-HFP, PMMA and LiTf were successfully synthesized using the solution casting method. XRD, FTIR, SEM and DSC studies confirm that significant structural and morphological changes take place in the polymer blend electrolyte films upon addition of PMMA and LiTf to PVdF-HFP. The studies also suggest that interaction and complete complexation between the constituents occurred. The conductivity of the polymer blend electrolytes is observed to be dependent on the relative percentage of PMMA. Furthermore, the conductivity-temperature dependence of the blend films seems to exhibit the VTF behaviour. For all compositions, the ionic conductivity of the polymer electrolyte increases with temperature, which can be attributed to the expansion of the polymer matrices to produce free volume for enhanced segmental movement of polymer chains and better mobility of carrier ions. Estimated ionic transference number measurements suggest that the conductivity in the blend systems is predominantly ionic. Declarations {#sec0055} ============ Author contribution statement {#sec0060} ----------------------------- Merhawi A. Gebreyesus: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper. Y. Purushotham: Contributed reagents, materials, analysis tools or data. J. Siva Kumar: Conceived and designed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data. Funding statement {#sec0065} ----------------- This work was supported by the Ministry of Education of the Federal Democratic Republic of Ethiopia and Aksum University under the 'Ethiopian Govt. Scholarship\' scheme. Competing interest statement {#sec0070} ---------------------------- The authors declare no conflict of interest. Additional information {#sec0075} ---------------------- No additional information is available for this paper. The authors are grateful to the Departments of Physics and Chemistry, Osmania University and C-MET Hyderabad for providing research facilities.	{ "pile_set_name": "PubMed Central" }
INTRODUCTION ============ Trypanosoma brucei is a single celled protozoan parasite and the causative agent of human African sleeping sickness and Nagana in cattle. The parasite belongs to the Excavata, an eukaryotic supergroup only distantly related to the mainstream model systems within the Opisthokonta and as such represents a versatile model to study universally conserved proteins and their potential functions in biology. The translationally controlled tumor protein (TCTP) is highly conserved among eukaryotes and is involved in a large variety of processes like cell growth and development, the cell cycle, apoptosis and the protection against cellular stresses, including oxidative stress and heat stress [@B1][@B2][@B3][@B4][@B5][@B6][@B7][@B8]. Moreover, several interacting/binding partners such as the elongation factor eEF-1alpha [@B9], tubulin [@B10], calcium [@B11] or Na^+^/ K^+^-ATPase [@B12] have been identified. We recently showed that T. brucei contains two isoforms TCTP1 and TCTP2 that are exclusively expressed in the procyclic form (PCF) and bloodstream form (BSF) parasites, respectively [@B13]. TCTP1 and TCTP2 have identical 5'UTRs and ten nucleotide changes in the open reading frames (ORFs) that lead to five amino acid changes. The exclusive expression is based on the different nucleotide composition of the 3\'UTRs of the two isoforms, responsible for the different mRNA stabilities [@B13]. TCTP1 is localized in the cytoplasm of PCF cells and loss of the protein causes a growth defect and leads to several phenotypes including a decrease in number and enlargement of acidocalcisomes (ACs) as well as changes in mitochondrial morphology. Furthermore, the cells depleted of TCTP display asymmetrical cell divisions leading to the accumulation of shorter \"tadpole\" like cells [@B13]. While the cytoplasmic localization of TCTP has been described in several systems, the protein is also associated with other structures in the cell. In HeLa cells for example TCTP was mostly found in the nucleus, where it seems to be involved in anti-apoptotic activities since depletion by RNAi led to massive cell death by apoptosis [@B14]. Also, associations with mitochondria have previously been demonstrated, albeit mainly under stress conditions [@B15][@B16][@B17]. TCTP has previously also been described to be secreted and involved in inflammatory response through the release of histamine. How TCTP is released out of the cells remains enigmatic, but interestingly a similar observation was recently reported in trypanosome infected tsetse flies were TCTP might be involved in manipulating the microbiota of the fly [@B18]. As a consequence of harboring only one mitochondrion per cell with a singular mitochondrial genome known as the kinetoplast, trypanosomes display a synchronized mitochondrial and nuclear genome replication and segregation [@B19]. Mitochondrial genome replication initiates prior to nuclear DNA replication and also the segregation of the replicated mitochondrial genome occurs prior to mitosis. A wild type BSF population contains about 80 - 85% cells with one kinetoplast and one nucleus (1K1N), 10 - 15% cells with two kinetoplasts and one nucleus (2K1N) and up to 5% cells with two kinetoplasts and two nuclei (2K2N) stage. Here we provide the first evidence for the localization of TCTP in BSF cells and its requirement for proper cytokinesis and heat stress response in the mammalian infective form of the parasite. RESULTS ======= TCTP2 localization ------------------ In biochemical fractionations using digitonin and differential centrifugation followed by western blotting with the previously described anti-TCTP antibody, the protein is localized to the cytoplasmic fraction (Figure 1A). Since the antibody did not provide specificity in immunofluorescence microscopy, we tagged TCTP2 N- (myc) and C-terminally (triple HA) to evaluate its localization using anti-myc and anti-HA antibodies (Figure 1B). These experiments support the biochemical analysis and show that the ectopically expressed N- or C-terminally tagged TCTP2 is predominantly localized in the cytoplasm with a distinct depletion of the TCTP2 signal in the region of the nucleus (Figure 1B). Furthermore, the protein does not seem to change its localization during the cell cycle (Figure 1B). DAPI was used to stain nuclear and kinetoplast DNA. ![FIGURE 1: Localization of TCTP in bloodstream trypanosomes.\ (A) Western blot of BSF cells extracted with 0.025% digitonin. Total cellular extract (T), supernatant (S) and pellet (P) fractions were analysed by western blot decorated with antibodies against TCTP, ATOM and EF1α.\ (B) Immunofluorescence images of BSF cell expressing N-terminally myc-tagged TCTP2 or C-terminal HA-tagged TCTP2. Myc-TCTP2 was detected with anti-myc antibody (white). The DNA is detected by DAPI (cyan). K, kinetoplast DNA; N = nuclear DNA; Phase, phase contrast microscopy; scale bars: 5 µm.](mic-05-460-g01){#Fig1} TCTP is required for normal cell growth and cell cycle progression ------------------------------------------------------------------ An inducible TCTP RNAi construct targeting the TCTPopen reading frames was stably integrated in a BSF T. brucei cell line. We targeted the TCTP ORF in order to compare the results to the previously published phenotype in the PCF, where we also targeted the TCTP ORF. Depletion of TCTP mRNAs was induced through the addition of tetracycline (tet) and confirmed by western blot decorated with TCTP and EF1α antibodies (Figure 2A; Inset). After 24 hours of RNAi induction, no TCTP could be detected by western blot. Cell growth was monitored for five consecutive days (Figure 2A). Two days post RNAi induction the cells started to grow slower and the growth retardation became more severe on the following days. After five days of TCTP depletion the doubling time was prolonged from six to 24 hours, when compared to the non-induced cells. Phase contrast and DAPI images were used for cell morphology and cell cycle stage analysis (Figure 2B). Over the course of TCTP depletion, we observed an accumulation of cells in the 2K2N stage of the cell cycle. Quantification of kinetoplast-nucleus counts showed that after three days of RNAi induction the number of cells with 2K2N DNA content increases from 4% to 33% (n ≥ 100; Figure 2C). This corresponds to a decrease of cells in G1 of the cell cycle (1K1N) from 80% to 50% (Figure 2C). We did not observe a significant change in the number of 2K1N cells. Phase contrast images did not reveal any obvious change in the cell morphology upon TCTP depletion. ![FIGURE 2: Effect of TCTP depletion, heat and oxidative stress in bloodstream trypanosomes.\ (A) Growth of cells with or without tet was monitored for five consecutive days. Bars represent standard deviation of four independent experiments. Inset: western blot showing TCTP downregulation upon 24 hours of RNAi induction.\ (B) Phase contrast and DAPI images of non-induced (Day 0) and induced (Day 1 and Day 2) bloodstream trypanosomes showing the effect of TCTP RNAi in cell morphology and cell cycle progression.\ (C)K-N counts in percentage following TCTP RNAi (n ≥ 100). Scale bars: 5 µm. (D, E, F) Monitoring growth of T. brucei upon TCTP RNAi followed by heat/oxidative stress.\ (D) BSF trypanosomes were heat shocked (+tet+heat shock, +T+H, red) at 42° C for one hour or (E) incubated with 50 µM sodium hydrogen arsenate (+T+oxidative stress, +T+Ox, green) for three hours, or (F) PCF cells were subjected to 45 minutes heat shock at 41°C. Subsequently the cells were washed in PBS and cultured again in HMI-9+10%FCS. Cell recovery was monitored for five consecutive days. Values represent averages from three independent experiments and error bars indicate standard deviation. Non-induced cells (-tet, -T), 24 hours induced (+tet, +T) cells and non-induced cells but stressed respectively with sodium hydrogen arsenate (-T+Ox) or heat shock (-T+H) were used as controls.](mic-05-460-g02){#Fig2} TCTP is involved in heat-stress response in BSF parasites --------------------------------------------------------- Based on the reports in other model organisms we tested if TCTP in BSF cells is involved in stress response [@B20]. For this we induced TCTP RNAi for 24 hours in bloodstream trypanosomes. At this time the levels of TCTP in BSF are below detection limit and no growth phenotype is observed (see Figure 2A). Following the depletion of TCTP the cells were exposed to heat stress for one hour, then the cells were washed with PBS and cultured in HMI-9 with 10% FCS without tetracycline. The recovery of cell growth was monitored in both cases by following cell growth for five consecutive days (Figure 2D). As controls we used cells that were heat stressed but not induced with tetracyclin as well as cells where RNAi was induced but that were not stressed. Additionally, we also used oxidative stress induced by NaAsO~2~ (Figure 2E) and heat stress in the PCF (Figure 2F). Average cumulative values from three independent experiments were plotted using GraphPrism with error bars indicating the standard deviation. We observed that in the control experiments the BSF cells grew at a wild type rate of approximately four cell divisions per day, while the BSF cells exposed to heat shock after TCTP RNAi showed a delayed recovery by two days. In the first 24 hours post heat shock the cells depleted of TCTP did not grow and a normal growth rate was first observed between 48 and 72 hours post heat shock (Figure 2D). The BSF cells exposed to oxidative stress following depletion of TCTP (Figure 2E) had only one cell division in the first 24 hours of the recovery but regained the normal growth rate (four divisions in 24 hours) two days after the stress was induced. In the PCF cells the heat stress delayed growth for 24 hours, however no significant difference between TCTP depleted cells and the wild type situation was observed. Based on these experiments we suggest that TCTP depletion in BSF cells has only a minor impact on recovery following oxidative stress while it is more pronounced under heat stress. In PCF cells recovery from heat stress does not seem to be influenced by the lack of TCTP. TCTP depletion does not alter mitochondrial or acidocalcisome morphology ------------------------------------------------------------------------ To verify if the depletion of TCTP in BSF trypanosomes would lead to mitochondrial structure abnormalities such as the ones observed in the PCF cells [@B13] we induced TCTP RNAi in BSF cells and visualized the mitochondria using the mitochondrial heat shock protein 70 (mtHSP-70) as a marker. Immunofluorescence images show no aberration in the mitochondrial structure upon two days of TCTP depletion (Figure 3A). Similarly, we asked if TCTP depletion in BSF cells would change the number and/or morphology of acidocalcisomes as we had previously observed in PCF cells. For this we visualized the acidocalcisomes by immunofluorescence microscopy using the vacuolar proton pyrophosphatase (VP1) antibody (Figure 3B, [@B21]. The DNA and cell morphology were visualized by DAPI and phase contrast images. Following three days of TCTP depletion, we did not notice enlarged acidocalcisomes or a significant change in their number (Figure 3B, C). ![FIGURE 3: Effect of TCTP depletion on mitochondria and acidocalcisomes in BSF cells.\ (A) Immunofluorescence images of non-induced (D0) and induced at day two (D2) bloodstream cells stained for mitochondria (HSP-70, green).\ (B) Morphology of acidocalcisomes following downregulation of TCTP in BSF. Non-induced (D0) and induced at day three (D3) BSF trypanosomes were stained for acidocalcisome with the marker VP1 (green). Nuclear DNA and kDNA were detected with DAPI (cyan). Phase contrast images (grey) show cell morphology. Cell morphology is shown by phase contrast images (PH, grey).\ (C) Histograms showing the mean number of acidocalcisomes per cell in D0 and D3 cells (n = 13). Scale bars: 5 µm.](mic-05-460-g03){#Fig3} Identification of possible interactors and binding proteins of TCTP ------------------------------------------------------------------- In order to identify possible targets or interactors of TCTP in BSF trypanosomes we analysed the whole cell proteome before and after TCTP depletion using stable isotope labelling with amino acids in cell culture (SILAC) combined with mass spectrometry (Figure S1A-C, [@B22][@B23][@B24]). Samples form non-induced and induced bloodstream trypanosomes were mixed in 1:1 ratio (cell number) and analyzed by liquid chromatography mass spectrometry. Overall, we could detect the changes of 2557 proteins. Proteome changes upon TCTP depletion for 24, 36 and 48 hours are presented by volcano plots (Figure S1A-C). Each plot represents the proteome comparison between non-induced cells and the respective time point of TCTP RNAi induction. In each plot the x-axes represent the proteins abundance change (log~2~fold change) while the y-axes represent the significance of the observed fold change (log~10~ p-value). Marked in red is TCTP2. Over the course of TCTP RNAi induction, we observed a progressive depletion of TCTP2. We did not detect other proteins whose abundance changed progressively and significantly in all timepoints during the course of TCTP RNAi. Similarly, we did not identify any interaction partners of the PCF TCTP1 (Figure S2) when performing pull-down with the c-terminally HA-tagged version. DISCUSSION ========== We have previously reported that the T. bruceigenome encodes two paralogues of the translationally controlled tumor protein and have named them TCTP1 and TCTP2 [@B13]. The two paralogues are differentially expressed during the parasite life cycle [@B13]. TCTP2 is expressed in BSF cells and is repressed after the transition to the insect form parasites where the second paralogue TCTP1 is almost exclusively expressed. Furthermore, we showed that the different 3\' UTRs of the two paralogues are the basis of the mechanism controlling differential expression and characterized the morphological and organelle phenotypes in the insect form [@B13]. Here, we study the function of TCTP in the BSF cells and compare them to the PCF providing a more comprehensive picture of TCTP in T. brucei (Table 1). ###### Comparison of TCTP1 and TCTP2 in PCF and BSF parasites, respectively. PCF BSF --------------------------------- ---------------------------------------- -------------------------- Expressed paralogue TCTP1 TCTP2 Localization Cytoplasmic Cytoplasmic **TCTP1/2 depletion by RNAi Cell growth Growth inhibition Strong growth inhibition Cell cycle No change Increase in 2K2N cells Cell morphology Tadpole morphology No change Acidocalcisomes Enlarged in size and reduced in number No change Mitochondria Accumulations in the network No change + heat stress No change Decreased growth + oxidative stress NA No change Cell growth** Growth inhibition Strong growth inhibition The localization of TCTP2 in the BSF is very similar to the insect form (Table 1). Based on biochemical and imaging data the protein is mostly distributed throughout the cytoplasm with a distinct depletion in the nucleus. Also, the localization does not change during the BSF cell cycle, similar to what we reported for the PCF trypanosomes [@B13]. A cytoplasmic localization in trypanosomes is consistent with most reports from plant, yeast or mammalian systems [@B16][@B25]. However, some reports have shown nuclear localization in human cells [@B26], as well as transient association with mitochondria and microtubules in yeast cells [@B16][@B25]. In order to test function of TCTP in BSF cells we used RNAi targeting the TCTP ORF. In previous experiments we specifically depleted the BSF TCTP2 by targeting the TCTP2 3'UTR, which led to a growth retardation, and while the overall TCTP levels remained below detection limit in western blotting, the exclusive depletion of TCTP2 was accompanied by a temporary increase of the levels of TCTP1 mRNA [@B13] (Table 1). Interestingly, RNAi experiments targeting the ORF of TCTP, thus also affecting the residual TCTP1 expression, resulted in a much stronger growth inhibition of BSF cells. This supports the previously suggested hypothesis that the residual expression of TCTP1 in BSF cells might partially compensate for the specific loss of TCTP2. In BSF cells we detect a specific increase in the 2K2N cells during TCTP ORF RNA, while specific depletion of TCTP2 leads to a slight increase in 1K1N cells (Figure S3) as was seen for the TCTP depletion in PCF cells. Thus, the specific depletion of the BSF TCTP2 shows a different phenotype than depletion of both paralogues suggesting different functions of the two proteins. Alternatively, the ORF RNAi might be more efficient in depleting TCTP2 since it covers a larger region of the transcript, and this might lead to the different phenotypes. The 2K2N cells have undergone mitochondrial and nuclear DNA replication as well as segregation but seem to be blocked or delayed in cytokinesis. Interestingly, this phenotype was not observed in the BSF specific TCTP2RNAi knockdown (see Figure S3 and [@B13]). Here a small increase in 1K1N cells similar to the PCF TCTP knockdown described previously was observed [@B13].While in BSF parasites cytokinesis is blocked or delayed, the cell division in PCF cells is unequal with one daughter inheriting a not fully developed posterior end leading to a \"tadpole\" like morphology (Table 1). A precytokinesis arrest similar to the one described above was previously observed by Sheader and coworkers upon silencing of VSG in BSF cells. Here the number of 2K2N cells reached up to 60% at 48 hours post induction [@B27]. The authors speculated that this checkpoint which seems absent in the PCF cells is required in vivoto respond to potential changes in VSG levels during cell division [@B27]. For example, TCTP transiently interacts with microtubules and numerous cell cycle proteins including the polo-like kinese-1 [@B28], the checkpoint protein Chfr [@B29], the mitotic regulator Cdc25C [@B8] and nucleolar proteins [@B30]. These proteins interfere in different steps of the cell cycle resulting in a broad number of phenotypes when alternating the levels of TCTP. In this study we aimed to identify TCTP binding partners or effectors using SILAC proteomics after TCTP depletion. We only included timepoints before the growth phenotype appeared in order to avoid the detection of secondary effects. While we could clearly detect the loss of TCTP as a consequence of theTCTP RNAi we did not detect any other significant changes. Thus, TCTP might be involved in maintaining functionality or localization of its interacting partners rather than simply increasing the stability. TCTP has been described as a calcium binding protein for the first time 20 years ago in trypanosomes [@B11] and later confirmed also in other organisms [@B31]. We recently demonstrated that TCTP depletion in PCF parasites causes alterations in the morphology of the major calcium storage organelles, acidocalcisomes and mitochondria [@B13]. In BSF cells both organelles seem not affected by the loss of TCTP, while the growth defect remained. Thus, potentially the effect on the calcium storage organelles in PCF are due to secondary effects. One of the best studied and conserved functions of TCTP is the ability to protect cells against a vast number of cellular stresses. During heat shock, an elevated level of TCTP has been observed in worm parasites such as Schistosoma and Trichinella [@B6][@B7]. This has been hypothesised to act as a stress defence mechanism to protect the parasite during the transition from the cold-blooded vectors (snails) to the worm blooded hosts (vertebrates). On the other hand, silencing of the gene using RNAi led to decreased tolerance to cold and high temperatures in the Brassica oleracea cabbage [@B5]. Here we tested whether the presence of TCTP helps in cell recovery following heat or oxidative stresses. For this we depleted TCTP in BSF cells and subjected them to the heat or oxidative stress, followed by monitoring the cell growth recovery. We found that the cells in which TCTP was depleted prior to the heat stress survive the stress but recover with a delay of two days compared to the controls. Interestingly this result was observed only in bloodstream cells and not in PCF indicating that stress response might be specific for that life cycle stage. Following the mild oxidative stress, both non-induced and induced cells recovered at similar rates. Overall, these data suggest that endogenous levels of TCTP are required for the bloodstream forms recovery following heat stress. TCTP might be chaperoning other proteins and/or mRNAs which help cellular recovery, a model already suggested in the worm parasites [@B6]. MATERIALS AND METHODS ===================== Trypanosome cell lines and culturing ------------------------------------ For RNAi and gene tagging experiments we used transgenic T. brucei bloodstream (New York single marker, NYsm) cell lines co-expressing T7 RNA polymerase and a tet repressor [@B32]. The BSF cells were cultured at 37°C and 5% CO~2~ in HMI-9 medium supplemented with 10% FCS [@B33] in the presence of 2.5 µg/ml geneticin (G418). Plasmid constructs and transfection ----------------------------------- For inducible RNAi against TCTP mRNAs we used a pLEW100 based stem-loop plasmid [@B32][@B34] where an insert of 512 bp targeting the full ORF sequence of the TCTP2 gene was integrated. The constructs were linearized with NotI and 10 µg were transfected in NYsm BSF by electroporation. The positive clones were selected with blasticidin (2.5 g/ml in BSF). Induction of RNAi was done by addition of 1 µg/ml tet. For the C-terminal tagging one allele of TCTP2 in BSF was in situ tagged with a triple Hemagglutinin (HA) epitope [@B35]. For inducible N-terminal c-Myc tagging, the full ORF plus the first 21 nt from the 3'UTR of TCTP2 were amplified by PCR and cloned in pJM-2 vector (a gift from A. Schneider; [@B34]. Upon transfection (as described above) the clones were selected with puromycine. Expression was induced by addition of 1 μg/ml tet. Growth recovery assay --------------------- Downregulation of TCTP in BSF and PCF was induced for 24 hours. The cells were then washed twice in PBS and subjected to either heat or oxidative stress. For the heat shock, the cells were pelleted by centrifugation (2500 rpm/8 min), re-suspended and incubated in pre-warmed media (one hour at 42°C HMI-9 for BSF, 45 minutes at 41°C SDM-79 for PCF). Oxidative stress was induced by incubating BSF for 3 hours in HMI-9 media supplemented with 50 µM sodium arsenate, NaAsO~2~[@B36]. Immediately after each stress the cells were washed in PBS and re-cultured in their normal growth conditions. Their recovery was monitored by counting cell growth in the next four or five consecutive days. Average cumulative values from three independent experiments were plotted using GraphPrism with error bars indicating the standard deviation. Western blot ------------ For western blotting trypanosome pellets were washed in phosphate buffered saline (PBS, pH = 7.2), re-suspended in standard Laemmli buffer (LB) (10^6^ in 15 µl), boiled for 5 min at 95°C and cooled on ice for 5 minutes. For the digitonin fractionation the cells were washed once in PBS, then the pellets were re-suspended in SoTE buffer (0.6 M sorbitol, 2 mM EDTA, 20 mM Tris-HCl, pH 7.5) containing 0.025% digitonin and incubated on ice for 5 min. The cell fractions were then separated by differential centrifugation at 8000 rcf for 5 min at 4°C. The fractions were lysed in LB, boiled for 5 min at 95°C, cooled on ice for 5 minutes and loaded on 10% or 12% SDS-polyacrylamide gels (10^6^ - 10^7^ cells per lane) before subjected to western blotting. The proteins were transferred onto PVDF Immobilon-P membranes (Millipore) using BioRad wet blotting system, blocked for 1 hour at room temperature in 10% skimmed milk or BSA solution in PBST (PBS + 0.1% TWEEN-20) and decorated with the primary antibodies. In this study we used rat-polyclonal anti-TCTP (1:50, Eurogentech), rabbit anti-ATOM (1:10000, [@B37]) and mouse anti-EF1alpha (1:10000, SantaCruz). After washing the primary antibody by incubating the membranes 3 times, 10 minutes each time in PBST, the membranes were incubated for 1 hour at room temperature with the secondary antibody. Secondary antibodies were: rabbit anti-rat HRP-conjugate (1:10000, Dako), swine anti-rabbit HRP-conjugate (1:10000, Dako) and rabbit anti-mouse HRP-conjugate (1:10000, Dako). For the chemiluminescent detection, the SuperSignal system (Pierce) was used and images were acquired with Amersham Imager 600. Immunofluorescence and microscopy --------------------------------- For immunofluorescence, BSF were harvested by slow centrifugation (2000 rpm/5 min), washed once in PBS and then fixed on slides for 4 min with 4% PFA in PBS. The cells were permeabilized for 5 min with 0.2% TritonX-100 in PBS and blocked for 30 min with 4% BSA in PBS. The cells were incubated for 1 hour in a wet chamber with primary antibody diluted in 4% BSA in PBS, washed three times in PBST and incubated again in dark wet chambers with secondary antibodies diluted in 4% BSA in PBS. The cells were mounted with ProLong® Gold Antifade Mountant with or without DAPI (Invitrogen). Images were acquired with Leica DM 5500 fluorescent light microscope and/or Leica SP8 Confocal Microscope System with STED and deconvoluted by Leica LAS AF and Huygens software, respectively. The primary antibodies used in this study were: rabbit anti-myc (1:1000, Sigma), rabbit anti-HA (1:1000, Sigma), mouse anti-mtHSP70 (1:2000, [@B38]) and rabbit anti-VP1 (1:2000, [@B21]). Secondary antibodies were goat anti-rabbit IgG, goat anti-mouse IgG conjugated with fluorophores Alexa Fluor® 488, Alexa Fluor® 594 (1:1000, Invitrogen). SUPPLEMENTAL MATERIAL ===================== ###### Click here for supplemental data file. All supplemental data for this article are also available online at <http://microbialcell.com/researcharticles/the-translationally-controlled-tumor-protein-tctp-is-involved-in-cell-cycle-progression-and-heat-stress-response-in-the-bloodstream-form-of-trypanosoma-brucei/>. [^1]: Conflict of interest: The authors declare no conflict of interest. [^2]: Please cite this article as: Borka Jojic, Simona Amodeo and Torsten Ochsenreiter (2018). The translationally controlled tumor protein TCTP is involved in cell cycle progression and heat stress response in the bloodstream form of Trypanosoma brucei. Microbial Cell: 5(10): 460-468. doi: 10.15698/mic2018.10.652	{ "pile_set_name": "PubMed Central" }
Introduction {#Sec1} ============ Cardiovascular diseases are the main cause of death in patients undergoing haemodialysis^[@CR1]^. Increased arterial stiffness, commonly observed in patients receiving haemodialysis, is a known predictor of the induction of these diseases^[@CR2]--[@CR5]^. It results from the progression of atherosclerosis with vascular intimal and medial calcifications^[@CR6],[@CR7]^. Pulse wave velocity (PWV), the speed at which the arterial pulse propagates through the circulatory system, is a good indicator of the degree of arterial stiffness^[@CR8],[@CR9]^. Generally, carotid-femoral PWV (cfPWV) measurement is considered to be the current gold standard method for assessing central arterial stiffness^[@CR10]^. However, because specialised equipment and specially trained staff are needed to measure the cfPWV, the brachial-ankle PWV (baPWV), which predominantly reflects peripheral arterial stiffness, is widely used owing to its simplicity in East Asian countries^[@CR11],[@CR12]^. Its measurement only requires the wrapping of blood pressure cuffs on the four extremities. However, baPWV measurement may not be appropriate for cardiovascular risk assessment because central arterial stiffness is closely associated with cardiovascular events and mortality. In patients undergoing haemodialysis, maintaining an adequate vascular access is important to reduce morbidity and mortality^[@CR13]^. Therefore, regular monitoring and surveillance must be conducted to check for vascular access failure early or to detect significant vascular access stenosis early. The National Kidney Foundation's Kidney Disease Outcomes Quality Initiative (KDOQI) guidelines suggest conducting regular surveillance of vascular accesses^[@CR14]^. One of the recommended methods of vascular access surveillance is the measurement of the static intra-access pressure ratio (SIAPR), which is the static intra-access pressure normalised to the mean arterial pressure (MAP), using the transducer of haemodialysis machines without additional equipment^[@CR14]^. Meanwhile, vascular access flow (Qa) can be calculated as follows: Qa = MAP/(R~out~ + R~in~), where R~in~ and R~out~ denote the inflow resistance (i.e. resistance of the flow tract upstream of the venous needle), such as the arterial segment of the access, arterial anastomosis, feeding artery, and arterial tree to the heart, and the outflow resistance (i.e. resistance of the flow tract downstream of the venous needle), such as the venous segment of the access, venous anastomosis, and venous outflow to the heart, respectively^[@CR15]^. Conversely, the SIAPR can be calculated as follows: Qa × R~out~/MAP and can be expressed as R~out~/(R~out~ + R~in~)^[@CR12]^. Generally, a high SIAPR is known to be an indirect indicator of vascular access stenosis related to outflow. A low SIAPR in the absence of inflow and outflow stenosis could be attributed to a high resistance to arterial inflow, which is related to increased arterial stiffness. Therefore, prediction of the degree of arterial stiffness using the SIAPR, without additional equipment, might be possible. In this study, we investigated the relationship between the SIAPR and arterial stiffness. In addition, because there is limited knowledge on the clinical usefulness of the SIAPR in predicting cardiovascular events, we evaluated whether the SIAPR had a prognostic value for cardiovascular events compared to known risk factors. Results {#Sec2} ======= Study population {#Sec3} ---------------- A total of 339 patients were on haemodialysis therapy at three dialysis clinics between January 2014 and February 2015. Two hundred ninety-one patients without vascular access dysfunction for 6 months or longer were eligible for inclusion. Eighty-two patients were excluded because of plans for referral to an interventional facility (n = 24), plans for transfer to another haemodialysis centre (n = 12), absence of PWV measurements (n = 4), use of cuffed central catheters (n = 15), and wrist-level fistulae (n = 27). Thus, a total of 209 patients who were on haemodialysis therapy were included in the analysis (Fig. [1](#Fig1){ref-type="fig"}).Figure 1Flow diagram of the study. Baseline characteristics {#Sec4} ------------------------ Table [1](#Tab1){ref-type="table"} shows the baseline demographic and clinical data of all patients. The study population consisted of 209 patients undergoing haemodialysis, 99 (47.4%) of whom were men. Their mean age was 59.8 years, and the underlying cause of end-stage renal disease (ESRD) was diabetes in 114 patients (54.5%). The median dialysis duration before study enrolment was 37.4 months. Of note, the SIAPRs ranged from 0.01 to 0.52, with a median of 0.23 (Fig. [2](#Fig2){ref-type="fig"}). When the patients were divided into two groups according to their median SIAPR, the incidence of previous coronary artery disease, proportion of left ventricular (LV) diastolic dysfunction, E/E′ ratio, and baPWV were significantly higher in the patients with SIAPRs of ≤0.23 than in those with SIAPRs of \>0.23. However, there were no differences between the two groups with respect to age, dry weight, height, MAP, underlying cause of ESRD, vascular access type, medication use, and haemodialysis duration before study enrolment.Table 1Comparison of the baseline characteristics according to the SIAPR.VariablesTotal (n = 209)SIAPR of ≤0.23 (n = 104)SIAPR of \>0.23 (n = 105)P-valueDemographic data Age (years)59.8 ± 11.860.6 ± 11.158.9 ± 12.40.29 Men, n (%)99 (47.4)58 (55.8)41 (39.0)0.02Clinical data Dry weight (kg)60.2 (52.0--68.3)61.3 (53.9--69.4)59.1 (50.0--67.3)0.11 Height (cm)162 (154--168)163 (154--169)160 (153--168)0.26 Systolic blood pressure (mmHg)150.6 ± 25.0152.2 ± 21.2149.0 ± 28.30.35 Diastolic blood pressure (mmHg)77.7 ± 12.177.2 ± 11.878.2 ± 12.40.56 Mean arterial pressure (mmHg)102.0 ± 14.5102.2 ± 12.6101.8 ± 16.20.84Underlying end-stage renal disease cause Diabetes, n (%)114 (54.5)57 (54.8)57 (54.3)0.99 Non-diabetes, n (%)95 (45.5)47 (45.2)48 (45.7)Previous cardiovascular disease Coronary artery disease81 (38.8)56 (53.8)25 (23.8)0.001 Peripheral artery disease12 (5.7)9 (8.7)3 (2.9)0.13 Cerebrovascular disease48 (23)29 (27.9)19 (18.1)0.13Vascular access type grafts, n (%)37 (17.7)13 (12.5)24 (22.9)0.08 Arteriovenous fistulae, n (%)172 (82.2)91 (87.5)81 (77.1)Medication use ß-blockers, n (%)93 (44.5)49 (47.1)44 (41.9)0.45 RAS blockers, n (%)130 (62.2)67 (64.4)63 (60.0)0.51 Vitamin D analogues, n (%)61 (29.2)34 (32.7)27 (25.7)0.34 Calcium-based phosphate binder, n (%)137 (65.6)65 (62.5)72 (68.6)0.44 Non calcium-based phosphate binder, n (%)28 (13.4)13 (12.5)15 (14.3)0.86 Antiplatelet drugs use, n (%)108 (51.7)56 (53.8)52 (49.5)0.63 Duration of haemodialysis^a^ (months)37.4 (12.6--74.0)42.0 (11.4--71.5)30.3 (13.6--74.5)0.50 Single-pool Kt/V1.5 (1.4--1.6)1.5 (1.4--1.6)1.5 (1.4--1.6)0.89 baPWV (cm/s)1935 (1670--2429)2061 (1825--2595)1800 (1568-- 2300)\<0.001Echocardiographic parameters LVEF (%)59.3 (53.1--65.0)56.5 (51.1--63.6)60.5 (55.5--66.7)0.01 LVMI (g/m^2^)135.9 (106.4--164.2)141.1 (127.3--171.3)124.1 (97.1--155.6)0.01 E77.8 (60.9--103.5)79.5 (68.5--104.0)71.1 (58.1--87.8)0.02 A82.0 (65.0--95.8)91.4 (72.7--102.0)88.5 (74.7--104.0)0.90 E/A ratio0.9 (0.7--1.5)0.8 (0.7--1.1)0.8 (0.7--0.9)0.06 DT201.5 (150.1--248.6)203.1 (156.5-- 251.7)210.6 (172.0--255.5)0.28 E/A' ratio0.6 (0.5--0.8)0.6 (0.5--0.7)0.6 (0.5--0.8)0.44 E/E' ratio18.0 (13.6--24.6)20.9 (15.4--25.9)15.4 (12.5--20.5)\<0.001 Left ventricular diastolic dysfunction^b^ (%)134 (64.1)81 (77.9)53 (50.5)\<0.001Note: Values are expressed as medians ± standard deviations, medians (interquartile ranges), or numbers (percentages).Abbreviations: SIAPR, static intra-access pressure ratio; RAS, renin-angiotensin system; baPWV, brachial-ankle pulse wave velocity; LVEF, left ventricular ejection fraction; LVMI, left ventricular mass index; A, peak mitral inflow velocities at late diastole; E, peak mitral inflow velocities at early diastole; DT, deceleration time; A′, late diastolic mitral annular velocities obtained on tissue Doppler imaging; E′, early diastolic annular velocities obtained on tissue Doppler imaging.^a^Before study enrolment.^b^Left ventricular diastolic dysfunction was defined as an E/E′ ratio of \>15.Mean arterial pressure was calculated by diastolic pressure plus a third of the pulse pressure.Figure 2Scattered plots of the SIAPR. Bar and error bar show the median and range, respectively. SIAPR, static intra-access pressure ratio. The SIAPR according to comorbidity {#Sec5} ---------------------------------- Patients with diabetes \[0.22 (0.13--0.32) vs 0.23(0.16--0.37), p = 0.05\], previous peripheral artery disease \[0.07 (0.04--0.26) vs 0.23 (0.15--0.34), p = 0.01\], and previous cerebrovascular disease \[0.18 (0.09--0.25) vs 0.29 (0.18--0.35), p = 0.03\] had significantly a lower SIAPR than patients without diabetes, previous peripheral artery disease, and previous cerebrovascular disease, respectively. Prognostic value of the SIAPR for cardiovascular events {#Sec6} ------------------------------------------------------- During a mean follow-up duration of 48.3 months, 27 patients experienced cardiovascular event. Of these, 18, 4, and 5 patients developed coronary artery disease, cerebrovascular disease, and peripheral vascular disease, respectively. Of note, 4 patients died from cardiovascular event. Additionally, of 4 patients with cerebrovascular disease, 1 patient developed another cerebrovascular disease after the event. The incidence of cardiovascular events was significantly higher in the patients with SIAPRs of ≤0.23 than in those with SIAPRs of \>0.23 (24 patients, 23.1% vs. 3 patients, 2.9%; P \< 0.001). The cumulative probabilities of cardiovascular events were also significantly higher in the patients with SIAPRs of ≤0.23 than in those with SIAPRs of \>0.23 (Fig. [3](#Fig3){ref-type="fig"}). In the Cox regression analysis, when the SIAPR was considered as a continuous variable, an increase in the SIAPR was associated with a reduced risk for cardiovascular events \[hazard ratio (HR): 0.34, 95% confidence interval (CI): 0.21--0.54, P \< 0.001\]. After adjustment for MAP, previous coronary artery disease, AVG, baPWV, and E/E′ ratio, the SIAPR remained to be associated with cardiovascular events (HR: 0.36, 95% CI: 0.21--0.60, P = 0.001) (Table [2](#Tab2){ref-type="table"}).Figure 3Kaplan-Meier analysis of cardiovascular events according to the SIAPR. The incidence of cardiovascular events was significantly higher in the patients with SIAPRs of ≤0.23 than in those with SIAPRs of \>0.23 (P \< 0.001). SIAPR, static intra-access pressure ratio.Table 2Prediction of cardiovascular events using Cox proportional hazards model.VariablesUnivariateMultivariateHR (95% CI)P-valueHR (95% CI)P-valueAge (per 1 y increase)1.03 (0.99--1.07)0.071.01 (0.97--1.04)0.79Men (vs. women)1.34 (0.63--2.87)0.45Mean arterial pressure1.01 (0.98--1.03)0.821.01 (0.98--1.04)0.48Diabetes (vs. non-diabetes)1.09 (0.48--2.45)0.84Previous CAD (vs. previous non-CAD)3.43 (1.49--7.89)0.011.42 (0.58--3.48)0.44AVG (vs. AVF)2.40 (1.09--5.29)0.033.52 (1.39--8.91)0.01ß-blockers use0.57 (0.26--1.27)0.17RAAS blockers use0.91 (0.42--1.99)0.91Antiplatelet drugs use1.52 (0.70--3.33)0.29SIAPR (per 0.1 increase)0.34 (0.21--0.54)\<0.0010.36 (0.21--0.60)0.001baPWV (per 100 cm/s increase)1.08 (1.02--1.16)0.010.97 (0.90--1.05)0.48E/E' ratio1.04 (1.01--1.09)0.051.02 (0.96--1.08)0.53Abbreviations: HR, hazard ratio; CI, confidence interval; CAD, coronary artery disease; AVF, arteriovenous fistula; AVG, arteriovenous graft; SIAPR, static intra-access pressure ratio; baPWV, brachial-ankle pulse wave velocity; E, peak mitral inflow velocities at early diastole; E′, early diastolic annular velocities obtained on tissue Doppler imaging.Mean arterial pressure was calculated by diastolic pressure plus a third of the pulse pressure. Additive prognostic value of the SIAPR for the prediction of cardiovascular events {#Sec7} ---------------------------------------------------------------------------------- To assess the predictive power of the SIAPR, we calculated Harrell's C index to be included in the multivariate Cox regression model (Table [3](#Tab3){ref-type="table"}). Compared with the C-statistic of model 1, which included age, MAP, AVG, previous coronary artery disease, baPWV, and E/E′ ratio, the C-statistic (C-statistic: 0.82, 95% CI: 0.70--0.94, P = 0.02) of model 2, in which the SIAPR was added, significantly increased. In addition, model 2 improved the overall continuous net reclassification index (NRI) (NRI: 63.8%, 95% CI 0.04--0.72).Table 3C-statistics and net reclassification index for the prediction of cardiovascular events.ModelHarrell C statisticContinuous NRIC-statistics (95% CI)P-value^a^Overall (%) with 95% CIModel 1^b^0.77 (0.73--0.86)Ref.Model 2^c^0.82 (0.70--0.94)0.0263.8 (0.04--0.72)^a^P \< 0.05: considered significantly different between model 1 and model 2.^b^Model 1: Age, mean arterial pressure, AVG, previous CAD, baPWV (per 100 cm/s increase), and E/E' ratio.^c^Model 2: Model 1 + SIAPR (per 0.1 increase).Abbreviations: CI, confidence interval; NRI, net reclassification improvement; Ref., reference; AVG, arteriovenous graft; CAD, coronary artery disease; baPWV, brachial-ankle pulse wave velocity; SIAPR, static intra-access pressure ratio.Mean arterial pressure was calculated by diastolic pressure plus a third of the pulse pressure. Discussion {#Sec8} ========== This study is the first to demonstrate that a low SIAPR was significantly associated with cardiovascular events in patients undergoing haemodialysis. Considering that there was the relationship between a low SIAPR and increased arterial stiffness in this study, SIAPR measurement can be a new approach for predicting the development of cardiovascular events in patients undergoing haemodialysis. Increased arterial stiffness occurs frequently in patients with ESRD^[@CR4],[@CR5]^. In addition, many studies have highlighted the role of arterial stiffness in the development of cardiovascular diseases^[@CR3],[@CR10],[@CR16],[@CR17]^. Pathophysiologically, increased central arterial stiffness induces afterload of the LV and poor perfusion of the coronary arteries^[@CR18]^. It results in concentric hypertrophy and fibrosis, which affect LV contraction and relaxation^[@CR18]^. In fact, previous studies have shown that central arterial stiffness is a strong predictor of cardiovascular mortality in patients with ESRD^[@CR17],[@CR19]^. Of note, there is also enough literature evidence demonstrating that the baPWV is an independent predictor of cardiovascular events and mortality^[@CR20]--[@CR25]^. However, theoretically, the baPWV, which reflects peripheral arterial stiffness, is limited in predicting cardiovascular events. In line with these findings, our study showed that although the baPWV was significantly associated with cardiovascular events in the univariate analysis, the HR was too low. In addition, the baPWV was not associated with cardiovascular events in the multivariate analysis. The vascular access approach for haemodialysis is unique in that it creates a low-resistance connection between the arterial and venous circulations^[@CR26]^. The haemodynamics of vascular accesses differ according to the access type. While the arterial pressure in an AVG gradually decreases along the length of the graft, the pressure in AVF dissipates within the first few centimetres of the access^[@CR14]^. Owing to the difference in the pressure profiles of these two access types, the SIAPR in AVF and AVG is known to range from 0.08 to 0.34 and from 0.15 to 0.49, respectively, with reference to the actual pressure at the measurement site divided by the MAP^[@CR14]^. However, elbow-level fistulae do not have collaterals or have lose collaterals, and often haemodynamically behave like grafts. To reduce the difference in the pressure profile between AVF and AVG, patients with wrist-level fistulae that had collaterals were excluded from the present study. Meanwhile, the SIAPR can be calculated as follows: Q~a~ × R~out~/MAP. In addition, because the Qa can be calculated as MAP/total resistance (R~out~ + R~in~), the SIAPR can be expressed as R~out~/(R~out~ + R~in~). Therefore, a low SIAPR in the absence of inflow and outflow stenosis could be attributed to a high resistance to arterial inflow, which is related to increased arterial stiffness. Thereby, patients with abnormal results of vascular access monitoring or surveillance were excluded from our study. In line with this, we observed that the patients with a low SIAPR had a higher proportion of LV diastolic dysfunction related to arterial stiffness and a higher baPWV than those with a high SIAPR. Conversely, we also showed that patients with worse vascular or comorbid status had a lower SIAPR than patients without it. Of note, although patients with AVG had a higher Charlson Comobidity Index^[@CR27]^ than patients with AVF (Supplementary Table [S1](#MOESM1){ref-type="media"}), patients with AVF had a lower SIAPR than patients with AVG due to the difference in the pressure profiles of these two access types \[0.21 (0.13--0.33) vs 0.41 (0.14--0.60), p = 0.01\]. However, the number of patients with AVG was small in this study. On the other hand, the SIAPR is a direct pressure measurement in AVF or AVG, which depend on the MAP. Thus, after excluding the suspicion for local stenosis, the SIAPR can be influenced directly by cardiac and peripheral blood pressure parameters rather than primarily the elastic properties of the arterial wall. To minimize the effect of overhydration on blood pressure parameters in this study, the measurement of SIAPR was performed after the end of haemodialysis therapy. Interestingly, there was a significant association between the SIAPR and cardiovascular events. Another main finding of this study was that the SIAPR had a predictive value for cardiovascular events compared to known risk factors including MAP. This study has several limitations. First, this was a small-scale observational study. Second, the SIAPR is less valuable as a tool for assessing arterial stiffness in fistulae. In fistulae, blood entering the venous system returns through multiple collateral veins. However, only elbow- level fistulae without collaterals were included in this study. In conclusion, because the SIAPR can be measured using the transducer of haemodialysis machines without additional equipment, it may be an attractive approach for predicting cardiovascular events. However, because various clinical conditions need to be considered before this approach can be applied in clinical settings, large-scale observational studies are needed to confirm our findings. Methods {#Sec9} ======= Ethics statement {#Sec10} ---------------- This study was conducted in accordance with the Declaration of Helsinki principles and approved by the institutional review boards of Kangdong Sacred Heart Hospital, Kangnam Sacred Heart Hospital, and Chuncheon Sacred Heart Hospital (reference nos. 2014-01-025, 2014-04-54, and 2014-96). Written informed consent was obtained from all patients before enrollment. Patients {#Sec11} -------- For this prospective observational study, we enrolled patients undergoing haemodialysis at three dialysis clinics (Kangdong Sacred Heart Hospital, Kangnam Sacred Heart Hospital, and Chuncheon Sacred Heart Hospital) between January 2014 and February 2015. Of note, based on these patients' clinical data, we previously showed the associations between vascular calcification and various clinical characteristics^[@CR28],[@CR29]^. All patients regularly underwent vascular access monitoring and surveillance according to the KDOQI guidelines. Vascular access monitoring (i.e. physical examination to detect vascular access dysfunction) was conducted at least once a week by qualified staff. The SIAPR, a special parameter for detecting vascular access dysfunction, was measured every 4 weeks for vascular surveillance. In addition, the patients were screened for abnormal test results suggesting vascular access dysfunction, such as elevated venous pressure, decreased blood flow, unexplained reduction in Kt/V (K = dialyser urea clearance, t = length of the dialysis session, V = urea distribution volume), or elevated negative arterial prepump pressures^[@CR11]^. Patients were eligible if they had undergone haemodialysis without vascular access dysfunction for 6 months or longer. Of note, patients with vascular access dysfunction defined as those who referred to an interventional facility because of abnormal results of vascular access monitoring or surveillance. Patients were excluded if they fulfilled any of the following criteria: 1) plan for referral to an interventional facility because of abnormal results of vascular access monitoring or surveillance, 2) plan for transfer to another haemodialysis centere, 3) absence of PWV measurements, 4) use of cuffed central catheters, and 5) wrist-level fistulae. Definitions {#Sec12} ----------- Cardiovascular disease was defined as a history of coronary artery disease, cerebrovascular disease, or peripheral vascular disease. Coronary artery disease was defined as myocardial infarction or angina; peripheral artery disease was defined as claudication, ischaemic limb loss, and/or ulceration; and cerebrovascular disease was defined as transient ischaemic attack or stroke. Cardiovascular events were defined as coronary artery disease requiring angioplasty or coronary bypass grafting, peripheral artery disease requiring peripheral revascularisation interventions, or cerebrovascular disease diagnosed using imaging studies. PWV assessment {#Sec13} -------------- The baPWV was measured using Vascular Profiler 1000 (VP-1000; Colin Co. Ltd., Komaki, Japan). Brachial and post-tibia arterial pressure waveforms were stored for 10 s using extremity cuffs connected to a plethysmographic sensor and an oscillometric pressure sensor wrapped around the arm and ankle. The baPWV was automatically calculated from the distance between two arterial recording sites divided by the transit time^[@CR28]^. The measurement was performed after the end of haemodialysis session. Echocardiographic measurements {#Sec14} ------------------------------ Comprehensive echocardiographic measurements were performed using an ultrasound machine (Vivid 7; GE Vingmed Ultrasound AS, Horten, Norway) with a 2.5-MHz probe according to the imaging protocol from the American Society of Echocardiography guidelines^[@CR30]^. LV ejection fraction was estimated using the modified biplane Simpson's method in apical two chamber and four-chamber views. LV mass was determined using the method described by Devereux et al.^[@CR31]^, and the LV mass index was calculated by dividing the LV mass by the body surface area. Mitral inflow was assessed using Doppler echocardiography from the apical four-chamber view. The mitral inflow profiles were used to measure the peak mitral inflow velocities at early (E) and late (A) diastole, and their deceleration times. Doppler tissue images of the mitral annulus were also obtained. From the apical four-chamber view, the early (E′) and late (A′) diastolic peak velocities were evaluated. LV diastolic dysfunction was defined as an E/E′ ratio of \>15^[@CR28],[@CR32]^. Echocardiographic measurements were performed after the end of haemodialysis session. Static intra-access pressure surveillance {#Sec15} ----------------------------------------- After assuring that the zero setting on the pressure transducers of the dialysis delivery system has been calibrated for an accuracy within ±5 mmHg, the venous drip chamber pressure and the MAP in the arm contralateral to the access were obtained from the digital pressure display of the dialysis machine 20--30 s after turning off the dialysis blood pump. The static intra-access venous pressure was calculated as follows: P~IA~ = P~DC~ + 0.76 dH, where P~IA~ (mmHg) is the static intra-access pressure, P~DC~ (mmHg) is the venous drip chamber pressure, and dH (cm) is the difference in the height between the venous drip chamber pressure transducer and the venous needle in the access. Subsequently, the P~IA~ was normalised to the MAP^[@CR11]^. The measurement was performed at the end of haemodialysis session. Data collection, exposure, and outcome determination {#Sec16} ---------------------------------------------------- The baseline characteristics, including demographic and clinical data, were obtained from medical records at the time of PWV measurement. The exposure for this study was the SIAPR within two weeks before and after PWV measurement. In addition, the outcome was cardiovascular events that occurred after PWV measurement. Of note, all enrolled patients were followed up until the day of cardiovascular events, loss to follow-up, death, or January 2019, whichever occurred first. Statistical analyses {#Sec17} -------------------- All statistical analyses were performed using the Stata software (version 11.0: StataCorp LP, College Station, TX, U.S.A). Continuous variables were expressed as means ± standard deviations or medians (interquartile ranges) and categorical variables as numbers (percentages). The Kolmogorov-Smirnov test was used to analyse the normality of the distribution of parameters among continuous variables. To compare the differences between the groups, Student's t-test, Mann-Whitney test, the χ² test, or Fisher's exact test was used. The cumulative incidence of cardiovascular events was calculated using the Kaplan-Meier product estimation method. The independent prognostic values for the study outcome were analysed by performing a Cox regression analysis. A multivariate analysis, which included all of the covariates with P values of \<0.1 in the univariate analysis, was performed. Even though a P-value was ≥0.1, potential confounding factor that was known as significant prognostic determinant of cardiovascular events was included in the multivariate analysis. Using Harrell's C index and the continuous NRI, we determined whether exposure had an additive value compared to the risk factors. Supplementary information ========================= {#Sec18} Comorbid conditions of the study population according to vascular access type. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information ========================= is available for this paper at 10.1038/s41598-020-58190-5. This work was supported by a National Research Foundation (NRF) of Korea grant funded by the Korea government (Ministry of Science and ICT) (NRF-2017R1C1B5074168). We thanked the help of a statistician at the institute of clinical epidemiology, hallym university. D.H.S. conceived and designed the study. H.J.J., J.O., Y.K.L., A.C., J.W.Y. and H.K. acquired the data. H.J.J., J.O., A.C. and D.H.S. analysed and interpreted the data. H.J.J. wrote the paper. All authors reviewed the paper for important intellectual content and approved the final version. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. The authors declare no competing interests.	{ "pile_set_name": "PubMed Central" }
![](yjbm00607-0068.tif "scanned-page"){.293}	{ "pile_set_name": "PubMed Central" }
Core tip: Uremic pruritus is a very distressing condition commonly seen in patients with advanced renal failure. Patients respond poorly to the currently available treatment regime. Narrow band ultraviolet (NUV-B) radiation is a treatment option in patients with refractory symptoms. In this study, we selected patients on peritoneal dialysis who had refractory pruritic symptoms, and used NUV-B as an add-on therapy to the standard medical care for a period of 12 wk. We found that using NUV-B improved symptoms in more than 90% of patients. However, the present regime used is not patient-friendly and compliance is poor. INTRODUCTION ============ Uremic pruritus is a very common and troublesome complication seen in patients with advanced chronic kidney disease\[[@B1]-[@B4]\]. The pathophysiology is complex, and many factors have been identified, including skin dryness\[[@B5],[@B6]\], hyperparathyroidism\[[@B7],[@B8]\], calcium phosphate deposition\[[@B9],[@B10]\], imbalances between mu and kappa opioid receptors\[[@B11],[@B12]\] and systemic inflammation. Anemia, inadequate dialysis, elevated serum magnesium and aluminum levels, and hepatitis C infection are also believed to have some contributing effects. However, the causes remain unexplained in many cases. There are strong associations of uremic pruritus with depressive symptoms and poor sleep quality. The pruritus is sometimes severe and refractory to treatment. Narrow band ultraviolet B (NB-UVB) phototherapy is one therapeutic option in these difficult cases. NB-UVB decreases the proinflammatory cytokine levels and induces mast cell apoptosis. In this study, we aim to investigate the role of NB-UVB as an add-on therapy to the standard treatment that is used in treating severe uremic pruritus in peritoneal dialysis patients. MATERIALS AND METHODS ===================== This retrospective study was conducted in the Department of Dermatology at the Jawaharlal Nehru Institute of Medical Sciences Imphal, a tertiary referral center in northeastern India. Adult end-stage renal disease patients on peritoneal dialysis with refractory uremic pruritus were included in the study. The patients were recruited from the nephrology units of three hospitals in Imphal from September 2011 to September 2017. The selected patients were referred to the Dermatology Department at the Jawaharlal Nehru Institute of Medical Sciences Imphal for NB- UVB therapy. Inclusion criteria ------------------ In order to be eligible for inclusion in this study, patients had to be older than 18 years of age, have end-stage renal disease, be on peritoneal dialysis as their treatment modality, and have refractory uremic pruritus. From September 2011 until September 2017, 29 patients satisfied the criteria. Uremic pruritus was defined as pruritus developing in patients with chronic kidney disease in the absence of other systemic, dermatological disorder or psychological factors. Refractory uremic pruritus was defined as uremic pruritus not that was not responsive to any of the two agents known to relieve the symptoms over a 4 wk period. These agents included topical emollients, topical capsaicin, antihistamines, pregabalin, gabapentin and tricyclic antidepressants. Exclusion criteria ------------------ Patients with a prior history of photosensitivity, and other prior dermatological diseases that can cause pruritus, were excluded from this study. Protocol: The patients were administered NB-UVB therapy every other day, three times per week for a total of 12 wk. They were started at a dose of 270 mJ/cm^2^, and then increased by 15% at each visit. If patients had asymptomatic erythema after the session, then treatment was continued at the same dose. The dose was reduced by 15% if the patient developed erythema with minimal pain/itchiness. If the patient developed painful erythema or bullous lesions, treatment was restarted at one-third of the dosage. Phototherapy was administered using "Derma India, Chennai Lightning cubicles PUVA", which is equipped with 24 UVA lamps that emit a radiation spectrum of 320-400 nm with a maximum of 366 nm, and 24 UVB lamps that emit a radiation spectrum of 290-320 nm with a maximum of 300 nm. The patients were allowed to continue with their previous medications/agents for uremic pruritus during the study period. A peritoneal adequacy test was performed in all patients upon entry into the study. Serum calcium, phosphate, intact parathyroid hormone, iron and hemoglobin profiles were evaluated in all patients. A visual analogue scale (VAS) (0 = no pruritus; 10 = most severe pruritus) was used to identify the intensity of itch. VAS was measured at baseline and then weekly until the end of the 12^th^ week. After completion of the treatment protocol, VAS was then measured monthly during the follow-up period. The outcomes were grouped into the following: (1) Complete responders: defined as a VAS score of zero at the end of the study period; (2) Partial responders: defined as a VAS score between one to five at the end of treatment, and with a final VAS score less than the value at baseline; and (3) Non-responders: a VAS score greater than five at the end of the treatment period. Relapse was defined as a VAS score greater than five after previously showing a complete or partial response. After the completion of the treatment protocol, patients were followed-up on a monthly basis for another 6 mo. During the follow-up period, patients were assessed for relapse of the pruritus. VAS scores were recorded during these visits. A feedback form was also provided to the patient. This form allowed them to rate their experience with the treatment protocol and provide suggestions to improve their adherence. Statistical analysis -------------------- Statistical analysis was performed using the SPSS 16 software. Continuous data were described as the mean ± SD, and categorical data by frequency and percentage. Paired t-tests were used to compare the mean VAS scores at baseline with those at the end of the study. A two-sided P score of \< 0.05 was considered significant. RESULTS ======= A total of 29 patients took part in this study. Seven patients dropped out during the treatment period. One patient died during this period. Baseline characteristics of the patients are shown in Table [1](#T1){ref-type="table"}. ###### Baseline characteristics of the patients ----------------------------------------- ----------------- Age (yr) 56.17 ± 15.65 Sex (male/female) 18/11 Hemoglobin (g/L) 9.99 ± 0.99 Corrected calcium (mg/dL) 8.92 ± 1.1025 Phosphate (mg/dL) 4.46 ± 1.35 Intact parathyroid hormone (pg/mL) 132.28 ± 176.63 Mean Kt/V 1.77 ± 0.11 No. of patients with weekly Kt/V \> 1.7 19 Skin phototype IV-20, V-9 Other agents used for pruritus Topical emollient Topical capsaicin Oral anti histaminics: Fexofenadine Pregabalin Gabapentin Amitryptilline ----------------------------------------- ----------------- The mean age of the patients was 56 ± 15 years. The mean duration on peritoneal dialysis of these patients at the time of study was 10 ± 8 mo. The average baseline VAS for pruritus was 7.75 ± 1.02. At the end of the treatment period, the average VAS score was 3.14 ± 1.59, which was a significant drop from the baseline score (P \< 0.05). Twenty-one patients completed the study, and 19 of them (90.4%) showed improvements in pruritus severity. Complete resolution of pruritus was noted in three patients (14.2%). Two patients (9.5%) continued to have persistent pruritus, with VAS scores greater than five (Table [2](#T2){ref-type="table"}). ###### The baseline and weekly visual analogue assessment scores for pruritus Baseline Week 1 Week 2 Week 3 Week 4 Week 5 Week 6 Week 7 Week 8 Week 9 Week 10 Week 11 Week 12 ----------------- -------------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- Mean VAS score 7.75 ± 1.02 7.55 ± 1.02 7.48 ± 0.82 7.03 ± 0.68 6.37 ± 0.49 5.66 ± 0.55 5.53 ± 0.64 5.20 ± 0.70 4.87 ± 0.74 4.60 ± 0.72 3.04 ± 1.32 3.42± 1.63 3.14 ± 1.59 No. of patients 29 29 29 29 27 27 26 25 24 23 22 21 21 VAS: Visual analogue score. Follow-up data were available for 14 patients. The mean VAS score at the end of the follow-up period was 4.14 ± 2.85. Six patients relapsed with pruritus, with VAS scores greater than five. The mean time to relapse was approximately 4.2 ± 2.99 mo. No significant adverse effects attributable to NB-UVB were identified. DISCUSSION ========== Uremic Pruritus is a fairly common entity in patients with advanced renal failure, including patients on hemodialysis as well as peritoneal dialysis. In a recent study in Chinese patients, the prevalence of uremic pruritus in patients on peritoneal dialysis was approximately 62.5%\[[@B13]\]. The usual protocol followed in managing patient with uremic pruritus includes optimization of the dialysis dosage, optimizing treatment of hyperparathyroidism, hyperphosphatemia and anemia. Initially, patients are usually managed with emollients and topical analgesics for symptomatic measures. Many of these patients eventually require systemic medications, such as anti-histamines, pregabalin, gabapentin, and anti-depressants. Hemoperfusion has been used in combination with hemodialysis for hyperparathyroidism and pruritus in hemodialysis patients\[[@B14]\]. A small population of patients continue to have persistent symptoms in spite of all these measures. Phototherapy may be tried as a treatment modality in these cases. In a small open pilot study, Ada et al reported a satisfactory response to NB-UVB in patients with uremic pruritus\[[@B15]\]. The randomized clinical study by Ko et al\[[@B16]\] showed significant improvement in the pruritus intensity, however the beneficial effect was marginal when compared to the control group that received long-wave UVA radiation. In our study, we noted that NB-UVB phototherapy was helpful as an add-on therapy in relieving symptoms of uremic pruritus in patients on peritoneal dialysis. A previous randomized controlled trial by Ko et al\[[@B16]\] failed to show any substantial benefit compared to broadband UVA phototherapy. This lack of benefit was due to improvement in pruritus intensity in the control arm, which they attributed to the placebo effect. However, the population studied in that trial differs from that of our study. In our study, we included only patients with end-stage renal disease who were on peritoneal dialysis. Conversely, the study by Ko et al\[[@B16]\] used a mixed population of patients, including those with chronic kidney disease on conservative treatment. Only three patients in that study were on peritoneal dialysis. Another important difference from that study is the duration of the treatment period. Per our protocol, the total duration of therapy was 12 wk compared to the six week time-span used in the randomized trial. The extended duration of our treatment is based on our preliminary experience with such patients, where we noted a more significant improvement in symptoms when they received a more prolonged course of treatment. The pathophysiology involved in uremic pruritus is very complex and multifactorial. Multiple hypotheses have been proposed, including the potential involvement of anemia, xerosis, hyperparathyroidism, hyperphosphatemia, inadequate dialysis, imbalance of opioid receptors, and inflammation. Some of the factors that contribute to pruritus in non-dialysis patients may not be applicable in patients who are already on peritoneal dialysis. The response of uremic pruritus to phototherapy may differ in the peritoneal dialysis population when compared to patients on hemodialysis. The beneficial effects of NB-UVB is believed to be attributable to the induction of mast cell apoptosis and the reduction in proinflammatory cytokine levels\[[@B17]\]. In our study, the mean hemoglobin levels (9.99 ± 0.9) are very near the target set by KDIGO\[[@B18]\]. The serum phosphate levels (mean value 4.44 mg/dL) and intact parathyroid levels (mean value 132.28 pg/mL) were also reasonable for patients on dialysis. Nineteen out of the 21 patients who completed the study had adequate small solute clearance (weekly kt/v \> 1.7), suggesting that factors other than inadequate dialysis played a significant role in the pathogenesis of uremic pruritus in our patients. We noted that using NB-UVB as an add-on therapy was effective in our patients. The mean VAS score improved from a baseline score of 7.75 ± 1.02 to 3.14 ± 1.59 by the end of therapy. Complete resolution of pruritus was noted in three patients (14.2%). In two patients (9.5%), no significant improvements were noted. Six of the patients who showed an improvement in pruritus ultimately relapsed (31.5%). The beneficial effects of the phototherapy session set in by the 3^rd^ week, and the effect became more pronounced after the end of the 7^th^ week. The mean VAS score at the end of six weeks was 5.53 ± 0.64, which means that many of these patients would have been classified as non-responders if they had received the 6 wk protocol, as in the previous randomized trial. The mean VAS score at the end of the 6 mo follow-up was 4.14 ± 2.85, which was significantly lower than the baseline VAS score (P \< 0.001). Our study is limited by its retrospective nature. Although the population studied here is uniform, it is also small. We noted that it was difficult for the patients to strictly abide by the three times per wk for 12 wk UVB protocol. Upon reviewing the feedback forms, we realized that our patients found the treatment regime to be cumbersome, and many were not willing to enroll in a repeat course of such treatment sessions in the future. This was in spite of the short duration of each session, usually lasting less than two to 3 min. Many of the patients found the frequent visits to the hospital to be very inconvenient. Transportation was also a big hindrance, as many of these patients cannot drive and thus arranging logistical support to bring them to the hospital three times a wk for 12 wk becomes an issue. It is noted that the duration of the overall treatment protocol was longer compared with previous studies. As such, one might consider a slightly shorter duration of NB-UVB treatment (e.g., twice per week for eight to 10 wk, or three times per week for 8 wk). Many questions still remained unanswered at the end of this study. The treatment, even if effective, does not offer long-term protection. Relapse occured in nearly a third of the patients. The optimal treatment duration and frequencies of the session are not known. As such, we need to come up with a large well-designed randomized controlled trial comparing different treatment regimes in order to come up with the most effective yet acceptable regime. In conclusion, using NB-UVB as an add-on therapy to standard agents in refractory uremic pruritus on peritoneal dialysis is effective in reducing the intensity of itching. However, there is a chance of relapse after discontinuation of the phototherapy. The 12 wk treatment regime used in this study is cumbersome and patient acceptance is poor. ARTICLE HIGHLIGHTS ================== Research background ------------------- Uremic pruritus is a common and troublesome entity in patients on peritoneal dialysis. The presence of pruritus affects both sleep quality and overall lifestyle, which can lead to depressive symptoms and mood disorders. In patients with difficult-to-treat pruritus, narrow band ultraviolet B (NB-UVB) can be tried as a treatment option. There is only one randomized controlled trial that has compared the role of NB-UVB in uremic pruritus. There is very limited data regarding the use of NB-UVB in the peritoneal dialysis population. There is therefore an urgent need to identify the effectiveness of such a treatment modality in the peritoneal dialysis population. Research motivation ------------------- With the limited data available, there is no clear-cut consensus regarding the role of NB-UVB in peritoneal dialysis patients who have severe pruritus. The most effective and optimal duration of treatment is also not clear. A previous randomized trial had used a course of three times a week for 6 wk. However, that study mixed in a population of chronic kidney disease patients and also included patients on hemodialysis, peritoneal dialysis, as well as patients who were treated conservatively and had not been initiated on dialysis. There were only three patients in this study who were on peritoneal dialysis, so the results therefore cannot be extrapolated to the peritoneal dialysis population. In this study, we selected a homogenous population of patients with end-stage renal disease who were on peritoneal dialysis with severe uremic pruritus, and used NB-UVB as an add-on therapy to the standard treatment. Research objectives ------------------- The purpose of our study was to assess the effectiveness of NB-UVB as an add-on therapy to standard treatment in peritoneal dialysis patients with refractory uremic pruritus. We included a follow-up 6 mo post-treatment completion to assess for relapse. Patients were also given a feedback form to highlight their experience with the treatment protocol and solicit their suggestions to improve the quality of treatment. Research methods ---------------- This is a retrospective study where peritoneal dialysis patients with refractory uremic pruritus were put on a 12 wk course of NB-UVB, in addition to their standard treatment. We used visual analogue scale (VAS) to record the intensity of pruritus, which was measured during each visit. After the completion of their treatment protocol, patients were followed-up on a monthly basis for six months, and their VAS scores were measured during these visits. The patient feedback forms were also collected during these follow-up visits. Research results ---------------- In this study, we noted that the mean VAS score improved from a baseline of 7.75 ± 1.02 to 3.14 ± 1.59 by the end of treatment. Nineteen out of the 21 patients who completed the study had improvement in symptoms. In three patients (14.2%), complete resolution of pruritus was noted. Two patients (9.5%) continued to have persistent pruritus, with VAS scores greater than five. Six (31.5%) of those patients who showed improvement in pruritus ultimately relapsed. The mean VAS score at the end of the 6 mo follow-up was 4.14 ± 2.85, which was significantly lower than the baseline VAS score (P \< 0.001). Research conclusion ------------------- In this study, we found that NB-UVB therapy is effective as an add-on therapy in difficult-to-treat patients with uremic pruritus in the peritoneal dialysis population. In our study, we used a 12 wk treatment protocol that showed effective results. We noted that the response at 6 wk was suboptimal, and many of our patients would have been classified as non-responders if our treatment was confined to 6 wk period. However, patient compliance was poor, and the frequent visits to the hospital for treatment became an issue when we used the 12 wk regime. We therefore need to come up with an effective treatment regime that will also be acceptable to patients. Research perspectives --------------------- Future studies should try alternative treatment regimes, such as two times per wk for 10 wk, or three times per wk for 8 wk. Institutional review board statement: The study was reviewed and approved by the Jawaharlal Nehru Institute of Medical Sciences Institutional Review Board. Conflict-of-interest statement: All authors have no conflicts of interest to report. Manuscript source: Unsolicited manuscript Peer-review started: April 25, 2018 First decision: May 8, 2018 Article in press: June 28, 2018 Specialty type: Clinical neurology Country of origin: India Peer-review report classification Grade A (Excellent): 0 Grade B (Very good): B Grade C (Good): C Grade D (Fair): D Grade E (Poor): 0 P- Reviewer: Biggar P, Cheungpasitporn W, Watanabe T S- Editor: Ma YJ L- Editor: Filipodia E- Editor: Tan WW [^1]: Author contributions: Sapam R is the main author and contributed to data recording, follow up and analysis; Waikhom R designed the study, did the literature search and contributed to the writing of the manuscript. Correspondence to: Rajesh Waikhom, MD, Associate Professor, Department of Nephrology, Jawaharlal Nehru Institute of Medical Sciences, Imphal East, Manipur, Porompat 795005 India. <rajesh.waikhom@gmail.com> Telephone: +91-897-4007290	{ "pile_set_name": "PubMed Central" }
1. Introduction {#Ch1.S1} =============== Sheep have been raised for milk for thousands of years, which is much longer than cow milk production (Zervas and Tsiplakou, 2011). Sheep milk and its products are widely consumed in some parts of the world, especially in the Mediterranean. Sheep milk has a high degree of similarity to human milk in total fatty acid composition, which makes it a good raw material for infant formula production (Martin et al., 2016). In the sheep industry, prolific sheep usually cannot lactate enough milk for lambs, which could decrease lamb survival rate (Bradford, 1985). A milk replacer is sometimes used as alternative, but it is costly and labor-intensive. Therefore, it is important to identify genes related to sheep milk and then genetically improve sheep milk performance, which will obtain a better profitability in the sheep industry and diversify human milk resources. Specialized strains of livestock have been cultivated by humans in long-standing husbandry practices; artificial and natural selection have imposed detectable selection signatures within genomes. These selection signatures can provide deep insights into selection mechanism and further uncover the causal genes related to relevant phenotypes. In sheep, selection signature analyses of closely related populations with divergent production purposes were successfully implemented in milk traits (Moioli et al., 2013), tail types (Moradi et al., 2012; Moioli et al., 2015; Yuan et al., 2017), gastrointestinal nematode-resistant traits (McRae et al., 2014) and reproductive traits (El-Halawany et al., 2016). In order to find genes associated with ovine milk traits, researchers have conducted a selection signature analysis between five non-milk sheep breeds and five milk sheep breeds, and they have identified some milk-related genes such as ABCG2 and SPP1 (Gutierrez-Gil et al., 2014). However, this previous study still had some limitations in experimental design. First, the sheep breeds that they have chosen have different characteristics not only in milk performance but also in other phenotypic differences, such as in wool traits, which could lead to some false positives of the candidate genes. Meanwhile, this study analyzed milk Lacaune and meat Lacaune genome only using site frequency-based methods, although it has been suggested that the regression-based selection mapping approach is more accurate than that of haplotype-based analysis methods (e.g., extended haplotype homozygosity, EHH; integrated haplotype score, iSH) (Wiener and Pong-Wong, 2011). However, different methods can detect different variation types and complement each other, which could accurately and comprehensively reveal the selection signature that exists within the genome. With meat and milk Lacaune from the Sheep HapMap project (<http://www.sheephapmap.org/hapmap.php>, last access: 6 August 2019), we used a haplotype-based method (hapFLK) and considered population stratification to conduct the selection signatures searching within the sheep genome. We further identified genes related to sheep milk and then provided a potential theoretical basis for sheep breeding. 2. Materials and methods {#Ch1.S2} ======================== 2.1. Experimental data pre-processing {#Ch1.S2.SS1} ------------------------------------- The Illumina Ovine SNP50 BeadChip data of 78 meat-purpose Lacaune sheep and 103 milk-purpose Lacaune sheep were downloaded from the Sheep HapMap project database (<http://www.sheephapmap.org/>, last access: 6 August 2019). The detailed description of these data was well done by Kijas et al. (2012). To facilitate subsequent gene annotation, the PLINK 1.07 software (Purcell et al., 2007) was used to upgrade the map file to match the sheep genome Oar_v3.1 and implement data quality control. Single-nucleotide polymorphisms (SNPs) were excluded from the subsequent analysis: (1) call rate $<$ 90 %, (2) minor allele frequency (MAF) $<$ 0.01 and (3) significantly deviated from Hardy--Weinberg equilibrium ($P < 10^{- 6}$). 2.2. Population structure analysis {#Ch1.S2.SS2} ---------------------------------- The filtered SNPs were pruned using the indep-pairwise option (plink -- file input -- indep-pairwise 50 5 0.1) in PLINK 1.07 software (Purcell et al., 2007) to avoid the strong influence of SNP clusters in principal component analysis (PCA) and relatedness analysis. PCA identifies the principal components representing the population structure based on genetic correlations (shared identity-by-state segments) among individuals. The PCA was implemented using the snpStats R package (<https://www.rdocumentation.org/packages/snpStats>, last access: 6 August 2019). To verify PCA results, ADMIXTURE (Alexander et al., 2009) was implemented. ADMIXTURE estimates ancestry in a model-based manner from large autosomal SNP genotype datasets, and it includes a cross-validation procedure that allows the user to identify the number of presumed ancestral populations ($K$) for which the model has best predictive accuracy. In this study, $K$ was set from 1 to 5, and a 10-fold cross-validation procedure was performed. 2.3. Genome scans for selection signatures using hapFLK {#Ch1.S2.SS3} ------------------------------------------------------- The hapFLK statistics detect selection signatures based on differences of haplotype frequencies between populations (Fariello et al., 2013). Considering the population stratification of samples, the hapFLK method, which is based on haplotype frequency and considers population stratification, was used in this experiment. We used hapFLK software to compute the hapFLK statistic and kinship matrix assuming 10 clusters in the fastPHASE model and used 98 Australia Poll Merino samples as an outgroup (<http://www.sheephapmap.org/hapmap.php>, last access: 6 August 2019). Then the hapFLK statistic was computed as the average across 20 expectation--maximization (EM) iterations to fit the LD model. However, the hapFLK statistic does not strictly follow any of the existing statistical distributions. To investigate the distribution of the hapFLK statistics, we plot a histogram of the hapFLK statistic. Then, we standardized hapFLK following Eq. (1): $${Standardized}\ {hapFLK} = \frac{{Raw}\ {hapFLK} - {mean}\left( {raw}\ {hapFLK} \right)}{{SD}\left( {raw}\ {hapFLK} \right)},$$ where SD (raw hapFLK) is standard deviation of raw hapFLK. Thus, the standardized hapFLK ($Z$ scores) roughly follows a standardized normal distribution. Finally, we computed the $P$ value for each SNP according to this standardized normal distribution. 2.4. Gene annotation {#Ch1.S2.SS4} -------------------- Candidate regions identified by hapFLK were annotated using ovine reference genome (Oar_v3.1). Gene function was annotated using the National Center for Biotechnology Information Gene (<http://www.ncbi.nlm.nih.gov/gene/>, last access: 6 August 2019), which was used for Gene Ontology (GO) analysis. The sheep QTL database (<http://www.animalgenome.org/cgi-bin/QTLdb/OA/index>, last access: 2 May 2019) was used to find whether the known milk-related QTLs are located in selection signature regions. In addition, the online database OMIM (<http://www.ncbi.nlm.nih.gov/omim/>, last access: 6 August 2019) and genomic information from other species, including humans, mice and bovines, were used to predict gene function. 2.5. GO and KEGG enrichment analysis {#Ch1.S2.SS5} ------------------------------------ To extract biological meanings from the list of candidate genes, GO enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed using the OmicShare tools (<https://www.omicshare.com/tools/>, last access: 6 August 2019). For GO enrichment, all candidate genes were mapped to GO terms in the Gene Ontology database. Gene numbers were calculated for every term, and significantly enriched GO terms in candidate genes compared to the genome background were defined by a hypergeometric test. The $P$ value calculated from hypergeometric distribution follows Eq. (2): $$P = 1 - \sum\limits_{i = 0}^{m - i}\frac{\left( \frac{M}{i} \right)\left( \frac{N - M}{n - i} \right)}{\frac{N}{n}},$$ where $N$ is the number of all GO-annotated genes; $n$ is the number of candidate genes in $N$; $M$ is the number of particular GO-term-annotated genes in $N$; $m$ is the number of particular GO-term-annotated genes identified by selection signature in $M$. For pathway enrichment analysis, significantly enriched pathways in candidate genes compared to the genome background were also defined by a hypergeometric test. The calculated $P$ value went through fast discovery rate (FDR) correction following Eq. (3): $${FDR} = P \cdot n/\left( {rank}P \right),$$ where $P$ is the raw $P$ value, $n$ is the number of tests and rank$P$ is the rank for the specific raw $P$ value. Taking FDR $\leq$ 0.05, GO terms and pathways meeting this condition were defined as significantly enriched for candidate genes. 3. Results and discussions {#Ch1.S3} ========================== 3.1. Population genetic structure {#Ch1.S3.SS1} --------------------------------- A total of 46 781 SNPs and 171 individuals were selected for further analysis after quality control. After implementing LD pruning, 13 935 SNPs with low LD were used for PCA and ADMIXTURE analysis. The PCA results showed that all animals can be divided into two groups by the first principal component (PC1): milk and non-milk sheep (Fig. 1a). The first two principal components (PC1 and PC2) can divide these samples into four subgroups (Fig. 1a) and explained 2.8 % and 1.7 % of the variance respectively (Fig. 1b). The results from the ADMIXTURE analysis showed that the least amount of cross-validation error occurred when $K = 4$ (Fig. 1c) indicating that $K = 4$ was the optimal modeling choice. Therefore, these samples could be appropriately divided into four subgroups (Fig. 1d), which was consistent with PCA. ![Population structure of the seven sheep populations. According to PC1, (a) all samples can be divided into four groups, and (b) the first two PCs can explain 2.8 % and 1.7 % variance respectively; (c) when $K = 4$, the least amount of cross-validation error occurred; (d) makes it fairly clear that $K = 4$ was the optimal modeling choice. The blue background represents the meat sheep group; the red background represents the milk group.](aab-62-501-g01){#Ch1.F1} 3.2. Genome scans for selection using hapFLK {#Ch1.S3.SS2} -------------------------------------------- The hapFLK histogram shows that the hapFLK statistic approximately follows a normal distribution (Fig. 2, top right), which is similar to the previous study (Kijas, 2014; Yuan et al., 2017). Therefore, the $P$ values could be calculated from the normal distribution. Negative ${log⁡}_{10}$ $P$ values plotted in genomic order revealed six regions under strong selection (Fig. 2). The genomic location, size, peak SNP and peak genes in the selection regions identified using Oar_v3.1 (Jiang et al., 2014) were summarized in Table S1 in the Supplement. The average size of selective regions was 7.97 Mb ranging from 4.33 to 17.00 Mb. These six selected regions in this study were compared with the six convergence candidate regions (CCRs) identified by Gutierrez Gil et al. (2014). However, only an overlapping region (Chr6: 38.64--43.02 Mb) was found, and the majority of selection regions are not overlapping. A similar situation also appeared in the sheep-tail-type selection signature analyses (Moradi et al., 2012; Moioli et al., 2015). There are several reasons that may explain this result: (1) only one breed was analyzed in the current study, while 10 breeds were incorporated in a previous study (Gutierrez-Gil et al., 2014); (2) the methods between this study and Gutierrez-Gil et al. (2014) were different, which can detect different variants; (3) the low SNP density in both the current study and Gutierrez-Gil et al. (2014) might lead to a low statistical power (Simianer et al., 2014). These results suggested that separate populations selected for similar breeding goals have the low repeatability of selection signature analysis results. ![Manhattan plot of hapFLK statistics. The histogram in the top right shows the hapFLK statistics roughly followed a normal distribution. Black dotted line means the suggestive line; the $- {log⁡}P$ value above this line means significant.](aab-62-501-g02){#Ch1.F2} 3.3. Genes and functional annotations {#Ch1.S3.SS3} ------------------------------------- To understand the function of these selection regions, we mapped them to the sheep genome Oar_v3.1 and sheep QTL database (<http://www.animalgenome.org/cgi-bin/QTLdb/OA/index>, last access: 2 May 2019). As a result, 38 QTLs related to milk traits (Table 1) and 334 candidate genes (Table S1) were identified in selected regions. These 38 milk-related QTLs were enriched in all milk-related QTLs in the sheep QTL database with a hypergeometric test of $P = 2.65506 \times 10^{- 7}$ (significantly different from what is expected by chance). Significantly, the majority of milk-related QTLs (27 out of 38) were located at chromosome 1 (Chr1: 228.88--245.88 Mb). In detail, these 27 QTLs were linked to the percentage of milk fat and milk protein as well as yield of milk protein and milk fat in sheep (Sutera et al., 2019; Hao et al., 2019) suggesting that selection region on chromosome 1 may play an important role in sheep milk performance. In chromosome 13 selection region (Chr13: 43.57--53.55 Mb), two copy number variation regions (CNVRs) (Chr13: 48.83--49.71 Mb, Chr13: 49.01--49.71 Mb) were identified to be significantly ($P = 6.051 \times 10^{- 7}$) associated with milk yield in Valle del Belice sheep (Di Gerlando et al., 2019). However, there was no known milk-related QTL in chromosome 18 selection region (Chr18: 37.95--42.60 Mb). This may be the relatively poor annotation of the current sheep QTL database (release 38), and this will be addressed when more milk-related QTLs are identified in the future. ###### Milk-related QTLs located in selection regions. Region Chromosome Start End Size Sheep QTL -------- ------------ -------- -------- ------- -------------------------------------------------------------------- 1 1 228.88 245.88 17.00 170 224 (Chr1: 233736829--233736869, MF) (Sutera et al., 2019) 170 226 (Chr1: 233736829--233736869, PP) (Sutera et al., 2019) 169 252 (Chr1: 236278074--236278174, PY) (Hao et al., 2019) 169 251 (Chr1: 236278074--236278174, FY) (Hao et al., 2019) 169 400 (Chr1: 236299569--236299669, FY) (Hao et al., 2019) 169 150 (Chr1: 236964320--236964420, FY) (Hao et al., 2019) 169 182 (Chr1: 237198394--237198494, FY) (Hao et al., 2019) 169 181 (Chr1: 237198394--237198494, MY) (Hao et al., 2019) 169 180 (Chr1: 237198394--237198494, PY) (Hao et al., 2019) 169 524 (Chr1: 237368505--237368605, FY) (Hao et al., 2019) 169 208 (Chr1: 237476668--237476768, MY) (Hao et al., 2019) 169 207 (Chr1: 237476668--237476768, PY) (Hao et al., 2019) 169 206 (Chr1: 237476668--237476768, FY) (Hao et al., 2019) 169 442 (Chr1: 237702646--237702746, FY) (Hao et al., 2019) 169 168 (Chr1: 237899096--237899196, MY) (Hao et al., 2019) 169 167 (Chr1: 237899096--237899196, PY) (Hao et al., 2019) 169 166 (Chr1: 237899096--237899196, FY) (Hao et al., 2019) 169 446 (Chr1: 240960592--240960692, MY) (Hao et al., 2019) 169 445 (Chr1: 240960592--240960692, FY) (Hao et al., 2019) 169 280 (Chr1: 241703508--241703608, PY) (Hao et al., 2019) 169 279 (Chr1: 241703508--241703608, FY) (Hao et al., 2019) 169 499 (Chr1: 242397140--242397240, FY) (Hao et al., 2019) 169 388 (Chr1: 242789285--242789385, FY) (Hao et al., 2019) 169 387 (Chr1: 242789285--242789385, PY) (Hao et al., 2019) 169 386 (Chr1: 242789285--242789385, MY) (Hao et al., 2019) 169 551 (Chr1: 243458586--243458686, FY) (Hao et al., 2019) 169 144 (Chr1: 243778419--243778519, FY) (Hao et al., 2019) 2 2 34.95 39.28 4.33 169 594 (Chr2: 37635669--37635769, MY) (Hao et al., 2019) 13 911 (Chr2: 37102076--37260066, PP) (Gutierrez-Gil et al., 2009) 57 738 (Chr2: 32023745--207420807, PP) (Garcia-Gamez et al., 2013) 13 915 (Chr2: 8804882--248905321, MF) (Gutierrez-Gil et al., 2009) 3 3 91.37 97.96 6.59 57 740 (Chr3: 97143203--97187127, MY) (Garcia-Gamez et al., 2013) 4 6 38.64 43.94 5.30 169 477 (Chr6: 41850279--41850379, FY) (Hao et al., 2019) 13 818 (Chr6: 43152047--43302377, MY) (Arnyasi et al., 2009) 13 819 (Chr6: 43152047--43302377, MLACT) (Arnyasi et al., 2009) 13 820 (Chr6: 43152047--43302377, MY) (Arnyasi et al., 2009) 13 821 (Chr6: 43152047--43302377, MLACT) (Arnyasi et al., 2009) 5 13 43.57 53.55 9.98 169 479 (Chr13: 45264465--45264465, FY) (Hao et al., 2019) 6 18 37.95 42.60 4.65 -- Note: milk fat percentage -- MF; milk protein percentage -- PP; milk protein yield -- PY; milk fat yield -- FY; milk yield -- MY; milk lactose yield -- MLACT. Of those candidate genes, succinate receptor 1 (Chr1: 234.11--234.16 Mb, SUCNR1) has been reported as close to the SNP rs417079368 (Chr1: 233.59 Mb), which was significantly ($P = 4.07 \times 10^{- 7}$) associated with milk fat percentage and protein percentage in Valle del Belice sheep (Sutera et al., 2019). Its ligand, succinate, plays an important role not only in adenosine triphosphate generation (Littlewood-Evans et al., 2016) but also in signalling transduction by binding to and activating its specific receptor, SUCNR1 (also known as G-protein-coupled receptor-91, GPR91) (Mu et al., 2017). SUCNR1 is expressed in multi-tissues and organs in sheep, such as the omentum, spleen, liver and mammary gland (Clark et al., 2017). Additionally, the expression of SUCNR1 is related to milk protein trait in sheep (Suarez-Vega et al., 2016). PPARGC1A (Chr6: 43.23--43.33 Mb, PPARG coactivator 1 alpha) located near a multi-effect milk-related QTL region (Chr6: 43.15--43.30 Mb) (Arnyasi et al., 2009). Many scholars have shown that this gene was associated with milk production, milk fat percentage, and other milk-related properties (Khatib et al., 2007; Weikard et al., 2005; Schennink et al., 2009; Cong et al., 2016). ![Second-level GO categories of candidate genes.](aab-62-501-g03){#Ch1.F3} In order to further extract known biological meanings from these 334 candidate genes, GO and KEGG enrichment analyses were implemented using the OmicShare tools (<https://www.omicshare.com/tools/>, last access: 6 August 2019). GO analysis has shown that these candidate genes participated in 47 second-level GO categories (Fig. 3) and significant (FDR $\leq$ 0.05) enrichment in 12 GO terms (Table S1). The highest numbers of genes of second-level GO categories of biological processes (BPs), cellular components (CCs) and molecular function (MFs) are cellular process (GO: 0009987, biological process, 221 genes), cell (GO: 0005623, cellular components, 241 genes), cell part (GO: 0044464, cellular components, 241 genes) and binding (GO: 0005488, molecular function, 189 gene). Previously, genes involved in these second-level GO terms showed strong associations with ruminant milk productivities. For example, it has been reported that genes located in bovine milk yield QTL regions are preferred to enrichment in cellular process (GO: 000987), cell (GO: 0005623), cell part (GO: 0044464) and cellular process (GO: 0009987) (Salih and Adelson, 2009). Further, genes involved in cellular process (GO: 0009987) were found to be associated with fat yield, milk yield, protein yield and fertility index in Nordic red cattle (Iso-Touru et al., 2016). Also, miRNA target genes of goat mammary gland were enriched in cellular processes (GO: 000987, biological process), cell (GO: 0005623, cellular component), cell part (GO: 0044464, cellular component) and cellular process (GO: 0005488, molecular function) (Ji et al., 2012). All these findings support the candidate genes identified in the current study contribute to fundamental physiology of dairy sheep. The most significant GO term of CC is cytoplasm (GO: 0005737, $P =$6.93E10-8, FDR $=$ $2.98 \times 10^{- 5}$). The most significant GO term of MF was the G-protein coupled nucleotide receptor activity (GO: 0001608, $P = 6.37007 \times 10^{- 5}$, FDR $=$ $6.37007 \times 10^{- 5}$) and G-protein coupled purinergic nucleotide receptor activity (GO: 0045028, $P = 6.37007 \times 10^{- 5}$, FDR $=$ $6.37007 \times 10^{- 5}$). The finding that the cytoplasm GO term (GO: 0005737) was enriched in our gene set is interesting. Previous studies have reported that candidate genes associated with milk protein composition traits in a Chinese Holstein population were significantly (FDR $=$ 0.0247) enriched in cytoplasm (GO: 0005737) (Zhou et al., 2019). Apart from GO analysis, KEGG analysis showed that candidate genes could be annotated to 36 KEGG classes (Fig. S1) and could participate in 173 pathways. The highest number of genes of KEGG categories was signal transduction (29 genes). However, no significant pathways were found. 4. Conclusions {#Ch1.S4} ============== Based on haplotype-based methods, the current study has six significant selection regions, which contains 38 known QTLs associated with milk yield. The identified six selection regions harbored 334 candidate genes. Some of the key candidate genes such as SUCNR1 and PPARGC1A may play an important role in sheep milk performance. The findings from this study can be useful to optimize breeding programs to improve the milk-related traits after further functional studies and validation of the association in other independent populations. Supplement ========== 10.5194/aab-62-501-2019-supplement The authors gratefully acknowledge Ruidong Xiang for language editing. Data availability ================= The data of the paper are available from Sheep HapMap project database (, International Sheep Genomics Consortium, 2019). Author contributions ==================== FL and XP designed the experiment and edited the paper. ZY analyzed the data and wrote the paper. WL discussed the results. Competing interests =================== The authors declare that they have no conflict of interest. Financial support ================= This research has been supported by the National Key Technology Research and Development Program of the Ministry of Science and Technology of China (grant no. 2015BAD03B05), the Program for Changjiang Scholars and Innovative Research Team in University (grant no. IRT13019), and the China Agriculture Research System (grant no. CARS-38). Review statement ================ This paper was edited by Steffen Maak and reviewed by Majid Khansefid and one anonymous referee.	{ "pile_set_name": "PubMed Central" }
Introduction {#S1} ============ Glioblastoma (GBM), the most lethal form of primary intrinsic brain tumor in both pediatric and adult populations, has the highest mortality rate among all malignant nervous system neoplasms, with a median survival of only 12--15 months. In 2005, the US Food and Drug Administration (FDA) approved combination temozolomide (TMZ) and radiotherapy treatment in adults with newly diagnosed glioblastoma, as well as the use of TMZ alone as a maintenance treatment ([@B9]). TMZ remains the most widely used and effective first-line chemotherapeutic for GBM patients, with high bioavailability and tolerability. However, over time, the majority of patients with GBM gradually develop resistance to TMZ during treatment, leading to GBM recurrence and treatment failure ([@B13]; [@B33]). Several pathways have been elucidated that regulate TMZ resistance, and expression of the DNA repair protein O^6^-methylguanine-DNA-methyltransferase (MGMT) is considered the predominant cause of TMZ resistance. MGMT expression can be silenced by methylation of the promoter/enhancer region, and polymorphisms in MGMT gene promote MGMT expression and TMZ resistance ([@B35]; [@B3]). Other plausible mechanisms are involved in TMZ resistance as well, including activation of base excision repair (BER), reduced activity of mismatch repair (MMR) genes, histone posttranslational modifications, GBM stem cells and dysregulation of other effectors ([@B5]). Epigenetic variations have been shown to play major roles in mediating the resistance to targeted therapies and conventional cytotoxic agents. Aberrant miRNA expression, such as in mIR-93, noted in clinically relevant tumor subtypes of GBM, influences tumor response to therapy through epigenetic miRNA-based silencing or sensitizing effects ([@B10]). Recent studies suggest that long non-coding RNAs (lncRNAs) are indispensable for the regulation of cellular processes in glioma tumorigenesis and in therapeutic responses ([@B43]). Specifically, several clinically relevant lncRNAs have been correlated with patient outcome in GBM and mediate biological functions, including stemness, immunity, development, regulation of gene expression, and regulation of protein synthesis ([@B26]). Thus, understanding the most relevant mechanisms of TMZ resistance may help identify novel drug targets and more effective therapies. The discovery of lncRNAs has provided insight into the underlying biological mechanisms of glioma phenotypes, which is mediated through their interactions with other cellular macromolecules, including proteins, RNA and DNA ([@B24]). LncRNA nuclear-enriched abundant transcript 1 (NEAT1) contributes to glioma cell growth and invasion through the WNT/β-catenin pathway by scaffolding the EZH2 protein ([@B2]). Findings from Sa's group have shown that lncRNA homeobox transcript antisense intergenic RNA (HOTAIR) promotes drug delivery across the blood tumor barrier (BTB) in glioma treatment by sponging miR-148b-3p ([@B25]). Chen reported that lncRNA AC003092.1 may act as an endogenous "sponge" of miR-195, promoting expression of TFPI-2 and overcoming TMZ resistance in glioma cells ([@B40]). Through chromatin modification, LncPRESS1 disrupts deacetylation of H3K56 by sequestering SIRT6 from chromatin to safeguard pluripotency-specific stem cells ([@B12]). Using RNA expression profiling, Mazor G et al. found that lncRNA TP73-AS1 comprises a clinically relevant lncRNA that influences metabolism-related genes and ALDH1A1, conferring TMZ resistance to GBM stem cells ([@B22]). In addition, studies have shown that lncRNA metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) plays a promising role in TMZ therapeutic response to GBM ([@B1]). Nanocomplex-mediated silencing of MALAT1 effectively sensitizes glioma cells to temozolomide therapy ([@B15]). Furthermore, ML Zhang et al. showed that circRNAs generated from lncRNA LINC-PINT, containing short open reading frames, encode functional peptides that suppress oncogenic transcriptional elongation in glioblastoma ([@B46]). However, it remains largely unknown how specific lncRNAs influence in the mechanical properties of glioma cells in response to TMZ exposure. To address these challenges, we reanalyzed lncRNA profiles in TMZ-resistant glioma cells using four public glioma-associated lncRNAs datasets. Our data identified 12 differentially expressed lncRNAs in TMZ-resistant glioma cells. Among these, lncRNA ADAMTS9-AS2 (ADAM metallopeptidase with thrombospondin type 1 motif 9 antisense RNA 2) was significantly overexpressed. Moreover, alterations in ADAMTS9-AS2 were correlated with TMZ response in glioma patients. Using subsequent functional assays, ADAMTS9-AS2 was found to be involved in fused in sarcoma (FUS)/MDM2 mediated progression in TMZ-resistant GBM. Materials and Methods {#S2} ===================== Cell Culture {#S2.SS1} ------------ T98G and U118 human glioma cell lines, identified by the short tandem repeat (STR) analysis ([Supplementary Data Sheet S1](#DS1){ref-type="supplementary-material"}), were obtained from the Cancer Research Institute, Central South University, China, and have been authenticated by short tandem repeat genotyping (Genesky Biotechnologies Inc., Shanghai). As described in our previous study ([@B4]), we established T98G-R and U118-R TMZ-resistant cell lines with continuous stepwise selection using increasing concentrations of TMZ for greater than 6 months. Next, the half maximal inhibitory concentration (IC~50~) was determined to confirm stable resistance to TMZ. Glioma cells and HEK 293T cells were maintained in DMEM supplemented (C11995500, HyClone) with 10% fetal bovine serum (10099141C, Gibco) and 1% penicillin/streptomycin (10378016, Gibco) at 37°C with 5% CO~2~. Reagents {#S2.SS2} -------- TMZ was purchased from Sigma-Aldrich Corporation Chemicals (PHR1437). Cycloheximide (CHX) was purchased from MedChemExpress (HY-12320). The proteasome inhibitor MG132 was purchased from Selleck Chemicals (S2619). All reagents were dissolved in dimethylsulfoxide (Amresco). Specific siRNAs and smart silencer (smsiRNAs) RNAs were purchased from Ribo (China) with sequences shown in [Supplementary Table S1](#TS1){ref-type="supplementary-material"}. Full-length ADAMTS9-AS2 and MDM2 were PCR-amplified from human cDNA and subcloned into pcDNA3.1(+) to create ADAMTS9-AS2 expression plasmids. A FUS expression plasmid, along with truncation constructs used in this study, were purchased from Addgene (29609, 29610, 29611, 29612) and Vigene Biosciences. The constructs His-Ubiquitin (WT) and His-Ubiquitin (K48R) were kind gifts from Prof. Gang Huang (Cincinnati Children's Hospital, Cincinnati, OH, United States). Overexpressing plasmids (1 μg) or smsiRNAs/siRNAs (100 nM) of indicated genes were transfected into cells using Lipofectamine 3000 (L3000015, Invitrogen) for overexpression and knockdown of indicated genes, respectively, followed by analysis 48--72 h later. Human Tissues {#S2.SS3} ------------- One hundred forty-four glioma tissues were collected between 2015 and 2018 from the Department of Neurosurgery, Xiangya Hospital of Central South University. This project was approved by the Chinese Clinical Trial Register (ChiCTR-RPC-16008569) and the ethics committee of Xiangya Hospital (Changsha, China). Data do not contain any information that could identify patients. Detailed clinical information was collected from patient records and is listed in [Supplementary Table S2](#TS2){ref-type="supplementary-material"}. Samples were obtained from patients during surgery and were immediately snap-frozen in liquid nitrogen until use. Chemotherapy response status was assigned to patients based on progression-free survival (PFS) in GBM patients with TMZ treatment ([@B31], [@B30]) denoted as TMZ response (no recurrence within 4 months after surgical resection) and TMZ non-response (recurrence within 4 months after surgical resection). MTS Assays and IC~50~ {#S2.SS4} --------------------- T98G, U118, and their TMZ-resistant cell lines were seeded in 96-well plates (2 × 10^3^ cells/well) and cultured overnight. Various concentrations of TMZ (0, 0.2, 2, 20, 200, 2,000 μM) or/and indicated transfection reagents were added into the medium for 72 h. Then, MTS assays were performed to determine cell viability following the manufacturer's protocols. Briefly, 20 μl MTS solution (G358C, Promega) was added into each well and incubated for 6 h. Absorbance was detected at 490 nm using a VICTOR X2 microplate reader (PerkinElmer, United States). IC~50~ was determined using sigmoidal concentration-response curve-fitting mode using SPSS software. Variable slope was employed to calculate IC~50~ values \[non-linear regression; dose-response-inhibition; log(inhibitor) vs. response- variable slope (4 parameters)\]. Cell proliferation was determined by treating TMZ-sensitive or TMZ-resistant glioma cells (1 × 10^3^ cells/well) with 100 μM TMZ for 5 days in 96-well plates. Cell proliferation of TMZ-sensitive or TMZ-resistant glioma cells (1 × 10^3^ cells/well) after indicated transfection was determined for 5 days in 96-well plates. Cell Cycle Analysis {#S2.SS5} ------------------- After serum starvation and cell cycle synchronization for 12 h, cells were seeded into T25 flasks at 1 × 10^6^ cells each. As cells adhered to the plate, drug treatments were added to the flask, and cells were incubated for 48 h. For each condition, detached and adherent cells were harvested, fixed in 70% ethanol at −20°C for at least 12 h, and incubated with propidium iodide (20 μg/ml), phosphate-buffered saline (PBS) and RNase A (50 μg/ml) in the dark. Stained cells were detected using flow cytometry (Guava easyCyte 8HT, Millipore, United States). Scratch Assay {#S2.SS6} ------------- The scratch assay was used to measure cell migration in vitro. In brief, cells were seeded into 60 mm dishes (2.5 × 10^5^ cells/well) and cultured until they formed a fused monolayer for 24 h. The smsiRNAs or/and siRNAs treatments were administered for each group according to the experimental design. A P200 pipette tip was used to create a scratch. After 48 h, wound closure was imaged using a microscope with mounted camera. Relative migration distance was measured, using ImageJ software, by determining the fraction of cell coverage across the scratch. Relative migration distance was calculated as follows: (%) = migration area/total area × 100%. Transwell Assay {#S2.SS7} --------------- The transwell assay was used to measure cell invasion in vitro. Matrigel (BD Biosciences, NJ, United States) was mixed with medium in a ratio of 1:8 and placed on the upper surface of each insert in 24-well transwell plates (BD Biosciences, NJ, United States). The chambers were held 6 h in the incubator. In brief, 1 × 10^4^ cells were added to the upper chamber, and 10% FBS was added to the lower chamber. The smsiRNAs or/and siRNAs treatments were administered for each group according to the experimental design. After 48 h, chambers were fixed and stained with 0.05% crystal violet for 2 h. Cells on the upper surface were gently scraped, and stained cells were imaged and quantified under the microscope. The ImageJ software was used to assist cell counting in cell invasion assays. RNA Extraction and qPCR {#S2.SS8} ----------------------- Total RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. The RNA quality was checked after 1% agarose gel electrophoresis with ChemiDoc XRS system (Bio-Rad, United States) and using Protein nucleic acid spectrophotometer (Beckman Coulter, United States). The A260/A280 ratios of RNA are allowed between 1.8 and 2.2. Total RNA was then reverse-transcribed to cDNA using the PrimeScript^TM^ strand cDNA synthesis kit (6210, Takara). The PARIS Kit (AM1921, Invitrogen) was used to separate nuclear and cytoplasmic RNAs in GBM cells. The qPCR reaction was then performed by CFX96 Touch Real-Time PCR Detection System (Bio-Rad, United States) to determine the expression levels of targets, and performed in triplicate in three independent experiments. The primer sequences of qPCR are shown in [Supplementary Table S1](#TS1){ref-type="supplementary-material"}. Changes in target mRNA levels relative to a reference gene (β-actin) were determined using the 2^--ΔΔct^ method with iTaq Universal SYBR green Supermix as previous reported (1725121, Bio-Rad) ([@B28]). RNA Pull-Down Assay {#S2.SS9} ------------------- LncRNA ADAMTS9-AS2 was transcribed in vitro from the pcDNA3.1(+) vector using the T7 RiboMAX large-scale RNA production system (P1300) and was biotin labeled using the Pierce RNA 3′ end biotinylation kit (20160). Two milligrams of protein extract from T98G-R cells were then mixed with 100 pmol biotinylated RNA, incubated with nucleic-acid-compatible streptavidin magnetic beads and washed (Pierce magnetic RNA-Protein pull-down kit, 20164). Proteins that bound to the streptavidin-coupled dynabeads were resolved using reducing sample buffer and then subjected to western blot. RNA Immunoprecipitation (RIP) {#S2.SS10} ----------------------------- RIP analysis was performed using the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore, Bedford, MA, United States) according to the manufacturer's instructions. Antibodies to FUS and the V5-tag used for RIP were the same as those used for western blot. C-immunoprecipitated RNAs were detected by strand specific qPCR. Western Blot Analysis {#S2.SS11} --------------------- Whole cell protein was isolated using Pierce IP Lysis Buffer (Thermo Fisher Scientific, United States). The PARIS Kit (AM1921, Invitrogen) was used to separate nuclear and cytoplasmic protein fractions in GBM cells. Protein concentrations were quantified using the Micro BCA Protein Assay Kit (23229, Thermo Scientific). Purified proteins were boiled with 4 × loading buffer, and denatured protein samples were separated by SDS-PAGE on 10% polyacrylamide gels. Then, samples were transferred to NC membranes (HATF00010, Millipore). After blocking with 5% non-fat milk for 1 h, membranes were probed with appropriate primary antibodies overnight at 4°C. The next day, membranes were washed with PBS/Tween (PBST) and incubated with appropriate secondary antibodies for approximately 1 h at room temperature. Protein bands were visualized by Immobilon Western chemiluminescent reagents (WBKLS0500, Millipore). Information for utilized primary antibodies is shown in [Supplementary Table S3](#TS3){ref-type="supplementary-material"}. Immunoprecipitation (IP) {#S2.SS12} ------------------------ For ubiquitination assays, cells were transfected according to the experimental requirements followed by treatment with 1 μM MG132 for 6 h. Then, 500 μg protein lysates were incubated with 1 mg/ml specific primary antibody with gentle rocking for 3 h at 4°C. Protein A/G beads (10002D, Invitrogen) were subsequently added to precipitate protein complexes and further incubated with gentle rocking overnight at 4°C. Precipitates were collected, and supernatants were discarded. Pellets were fixed and resuspended in SDS sample loading buffer before boiling. Samples were then subjected to SDS-PAGE for western blot analysis. Data Acquisition and Reanalysis Using Different Bioinformatics Methods {#S2.SS13} ---------------------------------------------------------------------- The bioinformatics analysis of lncRNA profiling in GBM tissues was conducted through several independent bioinformatics databases. LncRNA Modulator Atlas in Pan-cancer (LncMAP) is a user-friendly web platform, providing lncRNA-mediated transcriptional signatures in human cancer tissues ([@B20]). Using the Cancer RNA-Seq Nexus (CRN), we selected two studies (CRN Glioma and CRN Glioblastoma) to explore the lncRNA profiles between the normal and GBM tissues ([@B18]). The Atlas of Non-coding RNAs in Cancer (TANRIC) was used for the further confirmation of the changes of lncRNAs in GBM ([@B17]). Then, the consistently dysregulated genes were identified using a Venn analysis and Heatmap diagram. In addition, to evaluate the roles of lncRNAs in protein complexes, we used the RAID database to version 2.0 (RAID v2.0) ([@B45]), which integrated diverse RNA-associated interactions. Using this tool, we explore the proteins that bind to disregulated lncRNAs. Furthermore, another integrated bioinformatics platform, UbiBrowser ([@B21]), was used to investigate the human E3 ubiquitin-protein ligase-substrate interaction network. Statistical Analysis {#S2.SS14} -------------------- Statistics were calculated using SPSS 20.0 (SPSS Inc., United States). One-way ANOVA (Kruskal--Wallis test) and Chi-square tests were employed to analyze differences in demographic characteristics and clinical data among different groups. Receiver operating characteristic (ROC) curves were established to discriminate GBM response patients from non-response patients. Area under the ROC curve was used as an accuracy index for evaluating the predictive performance of the hypothesized lncRNA. ^∗^p \< 0.05 and ^∗∗^p \< 0.01 were regarded as statistically significant. Results {#S3} ======= Establishment of T98G-R and U118-R TMZ-Resistant GBM Cell Lines {#S3.SS1} --------------------------------------------------------------- To confirm resistant phenotypes, T98G-R and U118-R cancer cells were exposed to various doses of TMZ. After approximately 72 h, we measured cell viability and IC~50~ using MTS assay. Compared to the parental cells, T98G and U118, cell viability rates in T98G-R and U118-R cells were much higher in response to TMZ treatment ([Figure 1A](#F1){ref-type="fig"} and [Supplementary Figure S1A](#FS1){ref-type="supplementary-material"}). IC~50~ values of TMZ-resistant cells were increased more than fourfold compared to parental cells (T98G vs. T98G-R: 163.4 μM vs. 1850.7 μM; U118 vs. U118-R: 87.6 μM vs. 383.6 μM) ([Figure 1A](#F1){ref-type="fig"} and [Supplementary Figure S1A](#FS1){ref-type="supplementary-material"}). To further compare the proliferation capacity of parental and TMZ-resistant GBM cells, the cells were treated with 100 μM TMZ for 3 or 5 days. Significantly different cell cycle distribution was observed between TMZ-resistant cells and their parental cell lines. Upon TMZ treatment, T98G and U118 cells caused marked cell cycle arrest in G2/M phase in a time-dependent manner, while there were no significant changes in T98G-R and U118-R cells ([Figure 1B](#F1){ref-type="fig"} and [Supplementary Figure S1B](#FS1){ref-type="supplementary-material"}). These findings are consistent with the TMZ-resistant phenotype of T98G-R and U118-R cell lines. ![Establishment and characterization of TMZ-resistant GBM cell lines. (A) Cell viability and IC~50~ analysis was performed to evaluate cytotoxicity of TMZ to T98G and T98G-R cells in response to the indicated concentrations of TMZ for 72 h. (B) T98G and T98G-R cells were treated with 100 μM TMZ for 3 and 5 days, and cell cycle was examined by flow cytometry. The above experiments were repeated independently three times with similar results. Each data point represents mean ± SD.](fcell-07-00217-g001){#F1} LncRNA ADAMTS9-AS2 Is Overexpressed in TMZ-Resistant GBM Cell Lines {#S3.SS2} ------------------------------------------------------------------- LncRNAs are essential epigenetic regulators with critical roles in tumor initiation and malignant progression. To examine whether changes in lncRNA are involved in therapeutic response to TMZ in GBM cells, we first performed data mining using CRN Glioma ([@B18]), CRN Glioblastoma ([@B18]), TANRIC-GBM-CHINA ([@B17]), and LncMAP GBM ([@B20]) datasets and evaluated overlapping lncRNAs in GBM samples. The Venn diagram revealed that approximately 68 lncRNAs are common among these four published datasets ([Figure 2A](#F2){ref-type="fig"}). After a preliminary screen through strand-specific qPCR, we identified 12 co-differentially expressed lncRNAs in TMZ-resistant T98G-R and U118-R cells. Among these, nine lncRNAs were frequently upregulated, and three were frequently downregulated ([Figures 2B,C](#F2){ref-type="fig"}). Furthermore, the ADAMTS9-AS2 was the most consistently and markedly overexpressed RNA transcript in TMZ-resistant T98G-R and U118-R cells (p \< 0.01, p \< 0.05, respectively) ([Figures 2D,E](#F2){ref-type="fig"}). Thus, in subsequent experiments, we primarily evaluated the functional roles of ADAMTS9-AS2 in TMZ response in GBM samples. ![ADAMTS9-AS2 is significantly overexpressed in TMZ-resistant GBM cells. (A) Venn diagram of four datasets from different databases. (B) Changes in 68 glioma-related lncRNAs were examined between T98G/T98G-R and U118/U118-R. (C) The Venn diagram indicates the co-upregulated and co-downregulated lncRNAs in TMZ-resistant cells, T98G-R and U118-R. (D) Heatmap showing the 12 co-differentially expressed lncRNAs in TMZ-resistant cells, T98G-R and U118-R. (E) qPCR assay of ADAMTS9-AS2 transcript expression in TMZ-resistant GBM cell lines. Quantitative results are from three independent experiments and are shown as the mean ± SD. ^∗^p \< 0.05, ^∗∗^p \< 0.01.](fcell-07-00217-g002){#F2} Overexpressed ADAMTS9-AS2 Predicts Poor TMZ Response in GBM Patients {#S3.SS3} -------------------------------------------------------------------- To investigate whether ADAMTS9-AS2 expression is associated with TMZ response in GBM patients, we assessed ADAMTS9-AS2 transcriptional levels in TMZ response and TMZ non-response GBM tissues. The results revealed that ADAMTS9-AS2 was significantly upregulated in non-responding tissues compared to responding tissues (7.13 ± 1.07 vs. 1.21 ± 0.12, p \< 0.001) ([Figure 3A](#F3){ref-type="fig"}). A ROC curve was drawn to investigate the potential diagnostic value of ADAMTS9-AS2 expression in differentiating response status in GBM patients. The AUC was 0.84, with the diagnostic sensitivity and specificity reaching 63.87 and 69.99%, respectively ([Figure 3B](#F3){ref-type="fig"}). Moreover, we observed improved PFS among patients with lower ADAMTS9-AS2 expression compared to those with higher ADAMTS9-AS2 expression (6.77 vs. 2.66 months, p \< 0.001) ([Figure 3C](#F3){ref-type="fig"}). Collectively, these clinical data on our patients support the conclusion that ADAMTS9-AS2 represents a significant prognostic marker in GBM patients after TMZ treatment. ![Altered expression of ADAMTS9-AS2 is associated with TMZ response in 144 GBM patients. (A) qPCR assay of ADAMTS9-AS2 transcript expression in patients with differential TMZ response status. (B) ROC curves were established to discriminate TMZ responding patients from non-responding patients. (C) PFS among patients with ADAMTS9-AS2 low (green, median PFS: 6.77 mouths) and ADAMTS9-AS2 high (red, median PFS: 2.66 mouths) groups. ^∗∗^p \< 0.01 were regarded as statistically significant.](fcell-07-00217-g003){#F3} ADAMTS9-AS2 Induces TMZ Resistance by Regulating Metastasis {#S3.SS4} ----------------------------------------------------------- To investigate the effects of ADAMTS9-AS2 on TMZ-resistant behaviors, we determined the Spearman correlation between ADAMTS9-AS2 levels and TMZ sensitivity (IC~50~) in 6 GBM cell lines (T87G, U118, MGR2, U251, U87, C6). IC~50~ values of TMZ were positively correlated with ADAMTS9-AS2 expression levels in these glioma cells (Spearman r = 0.98, p \< 0.001) ([Figure 4A](#F4){ref-type="fig"}). Next, we used a smsiRNAs-mediated knockdown strategy to inhibit ADAMTS9-AS2 expression in TMZ-resistant cells, T98G-R and U118-R ([Figure 4B](#F4){ref-type="fig"} and [Supplementary Figure S2A](#FS2){ref-type="supplementary-material"}). Upon knockdown of ADAMTS9-AS2, T98G-R cells showed enhanced sensitivity to TMZ, which manifested as reduced cell proliferation rates ([Figure 4C](#F4){ref-type="fig"}) and about an 40-fold decrease in IC~50~ ([Figure 4D](#F4){ref-type="fig"}). Similar results were observed in U87-R cells ([Supplementary Figures S2B,C](#FS2){ref-type="supplementary-material"}). In addition, given one of the well-documented mechanisms in TMZ response involves elevated methylation of O6MeG DNA methyltransferase (MGMT) ([@B42]), we further want to evaluate whether the ADAMTS9-AS2 modulated TMZ-resistant behaviors was dependent on the MGMT methylation status in different GBM cells. Interestingly, ectopic expression of ADAMTS9-AS2 significantly upregulated the IC50 values of TMZ in both MGMT-positive cell lines T98G and U118 ([Supplementary Figures S2D,E](#FS2){ref-type="supplementary-material"}) and MGMT-negative cell lines U251 and U87 ([Supplementary Figures S2F,G](#FS2){ref-type="supplementary-material"}). These data collectively support that ADAMTS9-AS2 might represent a predictive marker of TMZ chemosensitivity in GBM cells in MGMT-independent mechanisms. ![ADAMTS9-AS2 supports TMZ resistance in T98G-R GBM cells. (A) The correlation between ADAMTS9-AS2 mRNA expression and IC~50~ values in six GBM cells was quantified by Spearman's rank correlation. (B) qPCR confirmed that smsiRNAs mediate knockdown of ADAMTS9-AS2. Relative cell number (C), TMZ IC50 value (D), invasion ability (E) and migration ability (F) were examined in T98G-R cells after knockdown of ADAMTS9-AS2. Quantitative results of three independent experiments are shown as the mean ± SD. ^∗^p \< 0.05, ^∗∗^p \< 0.01.](fcell-07-00217-g004){#F4} In addition, it has been previously demonstrated that the cancer cells with chemoresistance phenotype exhibit a mesenchymal phenotype with increased migration and invasion capacity compared with the parental cells. Given that metastasis might have promising roles in chemotherapeutic responses ([@B23]; [@B11]), we further assessed whether ADAMTS9-AS2 might influence invasion and migration. As expected, compared to the untreated group, ADAMTS9-AS2 knockdown significantly inhibited invasion in chemoresistant T98G-R and U118-R cells ([Figure 4E](#F4){ref-type="fig"} and [Supplementary Figure S2H](#FS2){ref-type="supplementary-material"}). Scratch analysis showed reduced migratory ability in ADAMTS9-AS2 knock down T98G-R and U87-R cells ([Figure 4F](#F4){ref-type="fig"} and [Supplementary Figure S2I](#FS2){ref-type="supplementary-material"}). Taken together, these finding illustrate that ADAMTS9-AS2 expression induces TMZ resistance via modulation of metastasis, and inhibition of ADAMTS9-AS2 resensitizes TMZ-resistant GBM cells to TMZ. ADAMTS9-AS2 Binds to FUS and Reduces Its K48-Linked Ubiquitination {#S3.SS5} ------------------------------------------------------------------ To screen for interactions between ADAMTS9-AS2 and protein complexes that potentially act as protein scaffolds, we used the RAID algorithm ([@B45]) to identify proteins that bind to ADAMTS9-AS2. FUS, an RNA-binding protein, was identified as the main protein associated with ADAMTS9-AS2 ([Figure 5A](#F5){ref-type="fig"}). First, we applied nuclear and cytoplasmic extraction to visualize the cellular localization and relative abundance of ADAMTS9-AS2 and FUS in TMZ-resistant cells, T98G-R and U118-R. We found that ADAMTS9-AS2 and FUS are primarily distributed in the nuclear region ([Figure 5B](#F5){ref-type="fig"} and [Supplementary Figure S3A](#FS3){ref-type="supplementary-material"}). Furthermore, knockdown of ADAMTS9-AS2 by smsiRNAs downregulated FUS protein levels in T98G-R and U118-R cells ([Figure 5C](#F5){ref-type="fig"} and [Supplementary Figure S3B](#FS3){ref-type="supplementary-material"}). However, no significant changes in ADAMTS9-AS2 were seen in FUS siRNA-treated cells ([Figure 5D](#F5){ref-type="fig"} and [Supplementary Figure S3C](#FS3){ref-type="supplementary-material"}). These data indicate that FUS is a downstream effector of ADAMTS9-AS2. Resistance to TMZ is partly implicated to the low MGMT promoter methylation status and enhanced DNA repair function. To illustrated the mechanism, ADAMTS9-AS2 and FUS expression were analyzed in both MGMT-negative cell lines U251 and U87 and MGMT-positive cell lines U118 and T98G ([Supplementary Figure S3D](#FS3){ref-type="supplementary-material"}). The results showed higher expression of ADAMTS9-AS2 and FUS in relative TMZ-resistant cells than the sensitive GBM cell lines. In addition, we evaluated the effects of ADAMTS9-AS2/FUS on TMZ response in GBM cells. In response to different doses or durations of TMZ treatment, the variation of ADAMTS9-AS2 was very similar to that of FUS in the parent cells, T98G and U118 ([Figures 5E,F](#F5){ref-type="fig"} and [Supplementary Figures S3E,F](#FS3){ref-type="supplementary-material"}), supporting the idea that the ADAMTS9-AS2/FUS axis is involved in TMZ chemotherapy. ![ADAMTS9-AS2 interacts with the co-localized FUS protein. (A) The RAID algorithm was used to predict the binding affinity of ADAMTS9-AS2 to FUS. (B) Subcellular localization of ADAMTS9-AS2 and FUS analyzed from nuclear and cytoplasmic extracts. (C) Protein levels of FUS determined by western blot analyses of lysates from ADAMTS9-AS2 knockdown cells T98G-R. (D) qPCR assay of ADAMTS9-AS2 and FUS transcript expression in ADAMTS9-AS2 or FUS knock down T98G-R cells. Upon different durations (E) or doses (F) of TMZ treatment, the variation tendency of ADAMTS9-AS2 and FUS was analyzed in T98G cells. (G) Proteins isolated from the RNA pull-down assays with biotinylated ADAMTS9-AS2 RNA were identified by western blot analyses using specific anti-FUS antibodies. mRNA isolated from the RIP assays with anti-FUS antibody was identified by qPCR using specific ADAMTS9-AS2 primers in T98G-R and U118-R cells without (H) or with (I) ADAMTS9-AS2 knockdown. (J) Truncated versions of V5-FUS were produced according to the predicted ADAMTS9-AS2/FUS binding domain. mRNA isolated from the RIP assays with anti-V5 tag antibody was identified by qPCR analysis using specific ADAMTS9-AS2 primers in HEK293T cells. All images displayed are representatives of three independent experiments. ^∗^p \< 0.05, ^∗∗^p \< 0.01.](fcell-07-00217-g005){#F5} Next, to confirm the association between ADAMTS9-AS2 and FUS, we used proteins isolated from the RNA pull-down assays and found that ADAMTS9-AS2 directly interacts with both endogenous and exogenous FUS ([Figure 5G](#F5){ref-type="fig"}). Moreover, RIP was performed using a specific FUS antibody to ensure that ADAMTS9-AS2 was specifically immunoprecipitated from cell lysates. The binding capacity in TMZ-resistant cells was much stronger than in parent cells ([Figure 5H](#F5){ref-type="fig"}), and knockdown of ADAMTS9-AS2 significantly weakened the association between ADAMTS9-AS2 and FUS ([Figure 5I](#F5){ref-type="fig"}). In addition, based on the predicted binding sites of ADAMTS9-AS2 in the FUS protein sequence ([Figure 5A](#F5){ref-type="fig"}), we obtained a series of vectors encoding V5-tagged FUS deletion mutants. RIP was performed using a specific V5 antibody to determine that ADAMTS9-AS2 specifically immunoprecipitates FUS through both the RRM and Znf_RanBP2 domains ([Figure 5J](#F5){ref-type="fig"}). Thus, these data confirm that ADAMTS9-AS2 directly binds to FUS in vitro. As FUS transcriptional levels did not change in ADAMTS9-AS2-downregulated T98G-R or U118-R cells ([Figure 5D](#F5){ref-type="fig"} and [Supplementary Figure S3C](#FS3){ref-type="supplementary-material"}), we performed CHX chase assays to determine the protein stability of FUS. While knockdown of ADAMTS9-AS2 surprisingly attenuated the half-life of the FUS protein in TMZ-resistant cells, T98G-R and U118-R ([Figure 6A](#F6){ref-type="fig"}), its overexpression promoted it ([Supplementary Figure S4](#FS4){ref-type="supplementary-material"}). Moreover, global and K48-linked ubiquitination of FUS in ADAMTS9-AS2-downregulated T98G-R and U118-R cells was more augmented than in controls ([Figures 6B,C](#F6){ref-type="fig"}). In HEK293T cells transfected with ADAMTS9-AS2, V5-FUS, and K48-Ubiquitin, ADAMTS9-AS2 knockdown robustly enhanced FUS K48-linked ubiquitination, but this effect was compromised in the presence of the K48R mutant ([Figure 6D](#F6){ref-type="fig"}). These data indicate that ADAMTS9-AS2 inhibits proteasome-dependent degradation of FUS. ![ADAMTS9-AS2 enhances FUS stability by weakening its K48-linked ubiquitination. (A) After treatment with CHX (20 μg/ml) for indicated times, protein levels of FUS were determined by western blot analyses of lysates from ADAMTS9-AS2 knockdown cells T98G-R and U118-R. Western blot analysis of ubiquitin (B) and K48-ubiquitin (C) immunoprecipitated with anti-FUS antibodies in T98G-R and U118-R cells. (D) Western blot analysis of K48-ubiquitin immunoprecipitated with anti-FUS antibodies in ADAMTS9-AS2 downregulated HEK293T cells. Experiments were repeated three times with similar results.](fcell-07-00217-g006){#F6} To identify characteristics of the E3 ligase that interact with FUS, we used a computational predictive system, UbiBrowser ([@B21]). As shown in [Figure 7A](#F7){ref-type="fig"}, FUS interacts most strongly with the murine double minute 2 (MDM2) E3 ligase, with a confidence score of 0.843. Western blot analysis revealed that low MDM2 levels and high FUS levels are clearly seen in T98G-R and U118-R cells ([Figure 7B](#F7){ref-type="fig"}). Furthermore, FUS expression was upregulated in MDM2-downregulated cells, whereas there was no significant change in ADAMTS9-AS2 transcription ([Figure 7C](#F7){ref-type="fig"}). Moreover, the FUS-MDM2 interaction was decreased in T98G-R and U118-R cells ([Figure 7D](#F7){ref-type="fig"}). Knockdown of MDM2 weakened its FUS binding ability, leading to upregulated FUS expression ([Figure 7E](#F7){ref-type="fig"}). We further scanned the interaction between MDM2 and FUS when ADAMTS9-AS2 was knocked down. The FUS/MDM2 interaction was enhanced in ADAMTS9-AS2 knock down T98G-R and U118-R cells ([Figure 7E](#F7){ref-type="fig"}). Then, after full-length and truncated FUS proteins were expressed in HEK293T cells, IP analysis determined that all of FUS fragments containing the RRM or Znf_RanBP2 domains specifically immunoprecipitated MDM2 ([Figure 7F](#F7){ref-type="fig"}). Together, these results demonstrate that ADAMTS9-AS2 might attenuate the interaction between FUS and MDM2, inhibiting MDM2-mediated FUS K48-ubiquitination and degradation. ![ADAMTS9-AS2 attenuates MDM2-mediated FUS ubiquitination and degradation. (A) The UbiBrowser tool identified the E3 ligase that interacts with FUS. (B) Protein levels of FUS and MDM2 were determined by western blot analyses of lysates from T98G-R and U118-R TMZ-resistant cells. (C) Expression levels of ADAMTS9-AS2 and FUS were analyzed in MDM2 knockdown cells T98G-R and U118-R. (D) IP analysis revealed the FUS/MDM2 interaction in T98G-R and U118-R cells. (E) IP analysis revealed the FUS/MDM2 interaction in ADAMTS9-AS2 and MDM2 downregulated cells T98G-R and U118-R. (F) After transfection with the truncated versions of V5-FUS in HEK293T cells, proteins isolated from IP assays using anti-V5 tag antibody were identified by western blot analyses using a specific MDM2 antibody. All data are representative of three independent experiments, and representative images are shown.](fcell-07-00217-g007){#F7} The Effect of Modulating ADAMTS9-AS2/FUS on TMZ Chemosensitivity {#S3.SS6} ---------------------------------------------------------------- To further determine the role of ADAMTS9-AS2/FUS in regulating therapeutic response to TMZ, we used T98G-R and U118-R cell lines stimulated with RNA interference silencing technology to downregulate the ADAMTS9-AS2/FUS signaling pathway. Under these conditions, IC~50~ values and cell proliferation rates were both significantly inhibited in ADAMTS9-AS2 or FUS downregulated cells ([Figures 8A,B](#F8){ref-type="fig"} and [Supplementary Figures S5A,B](#FS5){ref-type="supplementary-material"}), indicating enhanced TMZ chemosensitivity. In addition, transwell and scratch assay showed similar effects for ADAMTS9-AS2 and FUS knockdown with significant inhibition of migration and invasion in both T98G-R and U118-R cells ([Figures 8C,D](#F8){ref-type="fig"} and [Supplementary Figures S5C,D](#FS5){ref-type="supplementary-material"}). Moreover, combined knockdown of ADAMTS9-AS2 and FUS further promoted the sensitivity of TMZ-resistant cells to TMZ ([Figures 8A--D](#F8){ref-type="fig"} and [Supplementary Figures S5A--D](#FS5){ref-type="supplementary-material"}). However, overexpression of FUS could rescue the inhibitory effects of ADAMTS9-AS2 knockdown in both T98G-R and U118-R cells ([Figure 8E](#F8){ref-type="fig"} and [Supplementary Figure S5E](#FS5){ref-type="supplementary-material"}). These data indeed support the conclusion that ADAMTS9-AS2 inhibition enhances the antitumor effect of TMZ in GBM cells by down-regulating FUS expression. ![Knockdown of ADAMTS9-AS2/FUS promotes TMZ chemosensitivity in T98G-R cells. TMZ IC~50~ value (A), relative cell number (B), invasion (C), and migration (D) were examined in T98G-R cells after knockdown of ADAMTS9-AS2 and FUS. (E) FUS overexpression could rescue the inhibitory effects of ADAMTS9-AS2 knockdown in T98G-R cells. The above experiments were repeated independently three times with similar results. ^∗^p \< 0.05 were regarded as statistically significant.](fcell-07-00217-g008){#F8} Discussion and Conclusion {#S4} ========================= Over the last decade, it has been increasingly demonstrated that the majority of the mammalian genome is pervasively transcribed, resulting in the production of numerous lncRNAs ([@B16]). Accumulating evidence indicates that abnormal lncRNAs play multiple roles in maintaining tumor initiation and progression, demonstrating their crucial clinical potential as biomarkers and therapeutic targets ([@B27]; [@B41]). However, the detailed functions of lncRNAs in GBM resistance to TMZ remains to be elucidated in detail. In this present study, we demonstrated that ADAMTS9-AS2 could directly bind to FUS and interfere with its MDM2-mediated K48 polyubiquitination. FUS stabilization by ADAMTS9-AS2 overexpression promotes the cell metastatic behavior, which is required for the TMZ resistance of GBM cells ([Figure 9](#F9){ref-type="fig"}). ![A schematic for FUS stabilization by ADAMTS9-AS2, which promotes TMZ resistance in GBM cells. ADAMTS9-AS2 binds to FUS and interferes with its MDM2-mediated K48 polyubiquitination and degradation. The FUS protein is stabilized by ADAMTS9-AS2 overexpression and promotes cell metastasis, which is required for development of TMZ resistance in GBM cells.](fcell-07-00217-g009){#F9} Growing evidence indicates that lncRNAs regulate the expression of target genes in glioma cells. While dependent upon the cellular and environmental context, interesting conflicting outcomes have been shown regarding lncRNAs in cancer diagnose and prognosis. For example, the lncRNA MALAT1 was categorized as a tumor-suppressive gene in glioma via ERK/MAPK-mediated growth and MMP2-mediated invasiveness, while Xiong et al. found that MALAT1 enhances glioma stem cell viability and promotes gliomagenesis through suppressing miR-129 and facilitating SOX2 expression ([@B7]; [@B39]). ADAMTS9-AS2 was first identified as a novel tumor suppressor that is regulated by DNA methyltransferase-1 (DNMT1). Knockdown of ADAMTS9-AS2 by siRNA inhibited the migration of glioma cells ([@B44]). Recent studies have revealed ADAMTS9-AS2 is a double-edged sword in the initiation and malignant progression of human cancers. Compared to adjacent normal tissue, ADAMTS9-AS2 is downregulated in colorectal cancer and predicts improved prognosis in colorectal cancer patients ([@B19]). In contrast, ADAMTS9-AS2 levels have been found at significantly higher levels in epithelial ovarian cancer than in normal ovaries and benign ovarian cysts shown by lncRNA microarray profiling ([@B34]). Upregulation of ADAMTS9-AS2 facilitates cell migration and invasion via targeting miR-143-3p/integrin α6 signaling in salivary adenoid cystic carcinoma ([@B37]). However, the roles of ADAMTS9-AS2 in TMZ-resistant GBM remain unclear. We performed the first ADAMTS9-AS2/FUS combination analysis in GBM and identified ADAMTS9-AS2 as a proto-oncogene that promotes TMZ resistance through stabilizing the FUS protein. Further clarifying the functions of ADAMTS9-AS2 might uncover the nature of GBM and provide novel targets for treatment. The RNA-binding-protein FUS, also known as translocated in liposarcoma (TLS), is a critical regulator during the characteristic pathological features of amyotrophic lateral sclerosis ([@B32]). Recently, studies have found that FUS mRNA or protein expression is upregulated in liposarcoma ([@B29]), breast cancer ([@B14]), cervical cancer ([@B47]), and FUS promotes malignant progression in non-small cell lung cancer ([@B38]). Functioning as an oncoprotein, FUS has been proven to be essential for the growth of prostate cancer cells by activating androgen receptor signaling ([@B6]). Reducing FUS expression significantly abrogated lncRNA NEAT1 mediated cell survival in breast cancer cells ([@B14]). FUS also regulates the expression of 19 circRNAs by binding to introns in the splice regions, such as for circ_3279 and circ_5306. In glioma, silencing of the FUS gene inhibits the proliferation and migration of neuroblastoma cells and increases their chemical sensitivity to cisplatin by promoting expression of miRNA-141 ([@B36]). In glioma-exposed endothelial cells (GECs), the FUS protein combines with circ_002136, which acts as a miR-138-5p molecular sponge, upregulating SOX13 and SPON2 and increasing angiogenesis of GECs ([@B8]). However, the biological significance of the FUS-mediated therapeutic response is not fully understood. Our study is the first to demonstrate that ADAMTS9-AS2 interacts with FUS in the nucleus to inhibit MDM2-medicated FUS K48-ubiquitination and degradation, which inhibits migration and proliferation in GBM TMZ-resistant cells. Identifying cellular mechanism that drive GBM to be recur and TMZ resistant is critical to improving outcomes in patients. Additional studies are needed to focus on how this resistance developed during the course of GBM progression, especially the development of TMZ resistance at different time interval. Also, the addition of primary GBM cells obtained from patients are needed to determine ADAMTS9-AS2 mediated FUS/MDM2 ubiquitination axis in future studies. This is the first study to provide a detailed characterization of the ADAMTS9-AS2/FUS/MDM2 axis in GBM TMZ resistance. A better understanding of the potential roles of lncRNAs in GBM biology, especially the characteristics of glioma patients, is of great significance for the progression of gene-targeted therapy. Data Availability Statement {#S5} =========================== The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher. Ethics Statement {#S6} ================ The studies involving human participants were reviewed and approved by the Chinese Clinical Trial Register (ChiCTR-RPC-16008569) and the ethics committee of Xiangya Hospital (Changsha, China). The patients/participants provided their written informed consent to participate in this study. Author Contributions {#S7} ==================== YY, ZX, and ZG performed most of the experiments, analyzed the data, and wrote the manuscript. LQ, SZ, XC, and JW helped to perform some of the experiments. ZL, ZZ, XL, and LH provided the study material and supported administrative management. LS conceived and supervised the study, and reviewed and approved the manuscript. All authors read and approved the final manuscript. Conflict of Interest {#conf1} ==================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Funding. This study was supported by grants from the National Natural Science Foundation of China (81572946, 81803035, and 81703036), Natural Science Foundation of Hunan Province, China (2019JJ50932), and Youth Fund of Xiangya Hospital (2017Q17). We thank Elsevier's Language Editing for assistance with English language polishing. Supplementary Material {#S10} ====================== The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fcell.2019.00217/full#supplementary-material> ###### Establishment and characterization of the U118-R TMZ-resistant GBM cell line. (A) Cell viability and IC~50~ analysis was performed to evaluate cytotoxicity of TMZ to U118 and U118-R cells in response to treatment with indicated concentrations of TMZ for 72 h. (B) U118 and U118-R cells were treated with 100 μM TMZ for 3 and 5 days, and cell cycle was examined by flow cytometry. The above experiments were repeated independently three times with similar results. Each data point represents mean ± SD. ###### Click here for additional data file. ###### ADAMTS9-AS2 supports TMZ resistance in GBM cells U118-R. (A) qPCR confirmed smart silencer-mediated knockdown of ADAMTS9-AS2. Relative cell number (B) and IC~50~ values (C) were examined in U118-R cells after knockdown of ADAMTS9-AS2. (D,E) Ectopic expression of ADAMTS9-AS2 significantly upregulated IC50 values of TMZ in MGMT-positive cell lines T98G and U118. (F,G) Ectopic expression of ADAMTS9-AS2 significantly upregulated IC50 values of TMZ in MGMT-negative cell lines U251 and U87. Invasion ability (H) and migration ability (I) were examined in U118-R cells after knockdown of ADAMTS9-AS2. Quantitative results shown are of three independent experiments and represent the mean ± SD. ^∗^p \< 0.05, ^∗∗^p \< 0.01. ###### Click here for additional data file. ###### ADAMTS9-AS2 upregulates the FUS protein, which is involved in TMZ response in U118-R cells. (A) Subcellular localization of ADAMTS9-AS2 and FUS analyzed from nuclear and cytoplasmic extracts in U118-R cells. (B) Protein levels of FUS were determined by western blot analyses of lysates from U118-R ADAMTS9-AS2 downregulated cells. (C) qPCR assay of ADAMTS9-AS2 and FUS transcript expression in U118-R ADAMTS9-AS2 or FUS knock down cells. (D) ADAMTS9-AS2 and FUS Protein levels were evaluated in both MGMT-negative cell lines U251 and U87 and MGMT-positive cell lines U118 and T98G. Upon different durations (E) or doses (F) of TMZ treatment, the variation tendency of ADAMTS9-AS2 and FUS was analyzed in U118 cells. ###### Click here for additional data file. ###### ADAMTS-AS2 regulate the FUS protein stability in parental cells. After treatment with CHX (20 μg/ml) for indicated times, protein levels of FUS were determined by western blot analyses of lysates from ADAMTS9-AS2 over-expressed cells T98G and U118. ###### Click here for additional data file. ###### ADAMTS9-AS2/FUS knockdown promotes TMZ chemosensitivity in U118-R cells. TMZ IC~50~ value (A), relative cell number (B), invasion (C), and migration (D) were examined in U118-R cells after knockdown of ADAMTS9-AS2 and FUS. (E) FUS overexpression could rescue the inhibitory effects of ADAMTS9-AS2 knockdown in U118-R cells. The above experiments were repeated independently three times with similar results. ^∗^p \< 0.05, ^∗∗^p \< 0.01. ###### Click here for additional data file. ###### The sequences for the primers, siRNAs and smsiRNAs. ###### Click here for additional data file. ###### Baseline demographic and clinical characteristics according to ADAMTS9-AS2 expression. ###### Click here for additional data file. ###### The information of indicated primary antibodies. ###### Click here for additional data file. ###### The certificates of cell authenticity by STR analysis. The data sheet contains original STR analysis results, as well as final certification report of glioma cells. ###### Click here for additional data file. ADAMTS9-AS2 : ADAM metallopeptidase with thrombospondin type 1 motif 9 antisense RNA 2 BER : Base excision repair DNMT1 : DNA methyltransferase-1 FUS : Fused in sarcoma GBM : Glioblastoma IC~50~ : Half maximal inhibitory concentration lncRNAs : Long non-coding RNAs IP : Immunoprecipitation MALAT1 : Metastasis-associated lung adenocarcinoma transcript 1 MGMT : O^6^-methylguanine-DNA-methyltransferase MMR : Mismatch repair PFS : Progression-free survival RIP : RNA immunoprecipitation smsiRNAs : Smart silencer TMZ : Temozolomide [^1]: Edited by: Deilson Elgui De Oliveira, São Paulo State University, Brazil [^2]: Reviewed by: Vishwa Jeet Amatya, Hiroshima University, Japan; Sathish Kumar Mungamuri, National Institute of Nutrition, India [^3]: ^†^These authors have contributed equally to this work [^4]: This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Cell and Developmental Biology	{ "pile_set_name": "PubMed Central" }
1. Introduction {#sec1-sensors-15-21500} =============== With the capability of efficiently transforming chemical or biological reaction into measurable signal, biochemical sensors are widely used in chemical industry, biological process, medical diagnostics, healthcare, environmental monitoring, military defense, and scientific research. Compared with other transduction methods, optical sensors possess higher sensitivity and immunity to electromagnetic interference \[[@B1-sensors-15-21500]\]. Among them, sensors based on optical waveguide are considered as the most valuable in real-time in-situ applications due to their advantages of miniaturization and integration. Microring resonator (MRR) \[[@B2-sensors-15-21500],[@B3-sensors-15-21500],[@B4-sensors-15-21500]\] and Mach-Zehnder interferometer (MZI) \[[@B5-sensors-15-21500],[@B6-sensors-15-21500]\] are the two most common configurations employed in waveguide-based sensors for optical signal read-out. MRR biochemical sensors have demonstrated high sensitivity detection of protein, DNA, virus, and bacteria \[[@B7-sensors-15-21500]\]. In order to perform wavelength scanning, a wavelength-tunable laser and an optical spectrum analyzer are required, and the detection limit of the MRR-based sensor is ultimately restricted by the resolutions of this expensive and complex external equipment. By contrast, the MZI read-out configuration is based on the detection of optical intensity and neither tunable laser nor spectrum analyzer are required. Therefore optical waveguide MZI as a sensing platform is more feasible for the miniaturization and integration of the whole sensing system \[[@B8-sensors-15-21500]\]. Silicon-on-insulator (SOI) is a promising material to realize a low-loss optical waveguide, and it has an extremely high refractive index contrast. SOI-based nanowire waveguide \[[@B9-sensors-15-21500],[@B10-sensors-15-21500]\] and slot waveguide \[[@B11-sensors-15-21500],[@B12-sensors-15-21500]\] have been employed to realize high-sensitivity biochemical sensors. The cross-sectional dimensions of these waveguides are generally hundreds of nanometers or even smaller, thus the input/output coupling parts are relatively complex, such as a grating coupler, prism coupler, or large mode converter, which results in high cost and low flexibility of the total sensing system. An SOI rib waveguide with large cross section, i.e. several microns in transversal dimension, possesses high coupling efficiency with standard single-mode glass fibers, and therefore has the potential for integration with optical fiber communication systems and opto-electronic systems \[[@B13-sensors-15-21500]\]. This article proposes an MZI biochemical sensor based on an SOI rib waveguide with a large cross section. Following the introduction to the background of the research, the operation principle of the MZI sensor is presented in [Section 2](#sec2-sensors-15-21500){ref-type="sec"}, then the section dimensions of an SOI rib waveguide with a large cross section are optimized based on the simulations of evanescent field sensing in [Section 3](#sec3-sensors-15-21500){ref-type="sec"}. In [Section 4](#sec4-sensors-15-21500){ref-type="sec"}, the implementations of the MZI structure are analyzed and discussed, including comparison with conventional approaches, numerical simulations of the trench-based waveguide bends and branches, and the performances of the proposed MZI structure. The configuration of the MZI-based sensing platform is proposed in [Section 5](#sec5-sensors-15-21500){ref-type="sec"}, and the discussion and results of this research are given in the last two sections. 2. Principle of Operation {#sec2-sensors-15-21500} ========================= A typical MZI sensor based on optical waveguide consists of a laser source, an optical detection unit and an MZI structure with an evanescent field sensing window, as shown in [Figure 1](#sensors-15-21500-f001){ref-type="fig"}. The MZI structure consists of an input waveguide and an output waveguide, a beam splitter (the left Y-junction), and a beam combiner (the right Y-junction), as well as two straight waveguides between the two Y-junctions as the sensing arm (with a sensitive window) and the reference arm respectively. In the operation, the monochromatic and polarized light from the laser source is coupled into the input waveguide and split equally at the beam splitter. Then the two guiding modes propagate a certain distance along the sensing arm and the reference arm respectively, and recombine at the beam combiner. Based on evanescent field sensing, a phase difference Δφ between the sensing arm and the reference arm occurs when the effective refractive index of the guiding mode in the sensing arm is changed by the biochemical reaction in the sensitive window, resulting in an intensity modulation caused by the interference of the two arms at the waveguide output. Measuring the interference intensity at the output waveguide, the biochemical reaction in the sensitive window is able to be detected. ![The schematic diagram of a typical Mach-Zehnder interferometer (MZI) sensor based on optical waveguide.](sensors-15-21500-g001){#sensors-15-21500-f001} The phase difference Δφ can be expressed as $$\Delta\varphi = \frac{2\pi}{\lambda}\Delta N_{eff}L$$ where, λ represents the operating wavelength, N~eff~ represents the effective refractive index of propagating mode of waveguide, L is the length of the sensitive window on the sensing arm. Assuming the detection target is the refractive index change of bulk solution, the sensitivity of MZI-based sensor can be expressed as: $$S = \frac{\Delta P}{\Delta n}$$ where, Δn represents the refractive index change of the sensitive region, and ΔP is the normalized output power change of the MZI sensor responding to a given Δn. As the partial sensitivity of the interference measurement and evanescent field sensing are defined as ΔP/Δφ and ΔN~eff~/Δn respectively, the Equation (2) can be rewritten as: $$S = \frac{\Delta P}{\Delta\varphi} \cdot \frac{\Delta\varphi}{\Delta N_{eff}} \cdot \frac{\Delta N_{eff}}{\Delta n} = \frac{2\pi}{\lambda} \cdot \frac{\Delta P}{\Delta\varphi} \cdot \frac{\Delta N_{eff}}{\Delta n} \cdot L$$ It can be found that the sensitivity of the MZI-based sensor is determined by the length of the sensitive window (L), and the partial sensitivity of the interference measurement and evanescent field sensing. 3. Section Dimensions of SOI Rib Waveguide {#sec3-sensors-15-21500} ========================================== 3.1. Intensity of Evanescent Field {#sec3dot1-sensors-15-21500} ---------------------------------- The fundamental principle of optical waveguide MZI sensor is based on the evanescent field sensing. According to the Goos-Hanchen effects, the evanescent field is a fraction of optical field that extends to the cladding layer and the substrate layer of the waveguide. In general, there are two types of evanescent field sensing, homogeneous sensing, and surface sensing \[[@B14-sensors-15-21500]\]. The homogeneous sensing and surface sensing for SOI waveguide with large cross section are shown in [Figure 2](#sensors-15-21500-f002){ref-type="fig"}a,b respectively. For the homogeneous sensing, the effective index variation of the propagating mode is produced by the change of the refractive index in the sensitive region. While for the surface sensing, the effective index variation is produced by the change of the thickness of the ultra-thin sensitive layer which is immobilized on the waveguide surface. The thickness of the surface sensitive layer is denoted as t (shown in [Figure 2](#sensors-15-21500-f002){ref-type="fig"}b), which is changed at a small range, about several nanometers. Homogeneous sensing is generally used to detect the concentration change of gas or liquid in the entire sensitive region, and the surface sensing is often applied to detect protein, DNA, virus, and bacteria with the help of immobilized receptor molecules \[[@B15-sensors-15-21500]\]. ![The schematic diagram of sensitive region and sensitive layer at silicon-on-insulator (SOI) rib waveguide with large cross section. (a) Homogeneous sensing, the light olive region is the sensitive region; (b) Surface sensing, the blue region represents the sensitive layer, and its thickness is denoted by t.](sensors-15-21500-g002){#sensors-15-21500-f002} Obviously, the greater the intensity of the evanescent field is, the more sensitive the sensor both in the case of homogeneous sensing and surface sensing will be. The intensity of the evanescent field can be represented by the confinement factor Γ~s~, which is the ratio of the electric field intensity in the sensitive region or the sensitive layer to the entire mode distribution of the guiding mode \[[@B15-sensors-15-21500]\], defined as Equation (4). $$\Gamma_{s} = \frac{\iint\limits_{s}{\left\| {E\left( {x,y} \right)} \right\|^{2}dxdy}}{\iint\limits_{\infty}{\left\| {E\left( {x,y} \right)} \right\|^{2}dxdy}}$$ Generally, in order to achieve low-loss propagation and avoid negative influences due to multimode transmission or cross-polarization interference, the optical waveguides employed in the senor must possess single mode and single polarization. In this article, the SOI rib waveguides with large cross section will be confined by single mode conditions. Once the mode distributions are solved by mode solver programs, such as finite element method (FEM) and finite difference method (FDM), the intensity of the evanescent field can be calculated through area integration. Therefore, taking the maximization of the intensity of the evanescent field as the target, the optimization of waveguide section dimensions can be realized. 3.2. Optimization of Waveguide Section Dimensions {#sec3dot2-sensors-15-21500} ------------------------------------------------- As an example, suppose the sensor based on SOI rib waveguide with large cross section is used to detect the concentration of glucose solution and the refractive index n~c~ of the analyte solution is in the vicinity of 1.33. In order to guarantee high coupling efficiency with the standard single-mode fiber, the total rib height (H) of SOI rib waveguide is set to be 10 μm, and the operating wavelength is selected at 1550 nm so that the refractive index of silicon and SiO~2~ are 3.476 and 1.444, respectively. Considering too small a rib width (w) will reduce the restriction of the rib structure on the guiding mode and too big a rib width (w) will be difficult for system integration (such as the waveguide bends and branches analyzed in next section), the rib width (w) of SOI rib waveguide is set to be in the range of 2.5 μm to 10 μm. The modes of rib waveguide can be denoted as HE~nm~ or EH~nm~ \[[@B16-sensors-15-21500]\], where n = 0, 1, 2..., and m = 0, 1, 2.... HE mode and EH mode are commonly known as quasi-transverse-electric mode and quasi-transverse-magnetic mode respectively. Employing a strict single-mode condition \[[@B17-sensors-15-21500]\], all the SOI rib waveguides with large cross section only support the fundamental guiding modes for each polarization, i.e. HE~00~ and EH~00~. Using the full-vector finite difference method (FDM) \[[@B18-sensors-15-21500]\] to solve the fundamental guiding modes with different dimensions (including the outside rib height h and the rib width w), the intensity of the evanescent field for homogeneous sensing can be calculated, as shown in [Figure 3](#sensors-15-21500-f003){ref-type="fig"}. ![Dependence of the intensity of the evanescent field for homogeneous sensing on rib width (w) with different outside rib height (h). (a) HE polarization; (b) EH polarization.](sensors-15-21500-g003){#sensors-15-21500-f003} It can be seen from [Figure 3](#sensors-15-21500-f003){ref-type="fig"}, for an SOI rib waveguide with the total rib height of 10 μm, the evanescent field intensity of homogeneous sensing exhibits a maximum of 7.436 × 10^−3^% for HE polarization, corresponding to h = 5 μm and w = 4.5 μm, and 9.407 × 10^−3^% for EH polarization, corresponding to h = 5 μm and w = 2.5 μm. Using the same method, dependence of the evanescent field intensity for surface sensing on rib width (w) with different outside rib height (h) also can be calculated. Assuming the refractive index of the sensitive layer n~m~ = 1.45 and the thickness of the sensitive layer t = 10 nm, the evanescent field intensity for surface sensing of the SOI rib waveguide with the total rib height of 10 μm possesses a maximum of 5.089 × 10^−3^% for HE polarization and 7.924 × 10^−3^% for EH polarization, both corresponding to h = 5 μm and w = 2.5 μm. Both for homogeneous sensing and surface sensing, it is evident that EH polarization of the SOI rib waveguide with large cross section is significantly more sensitive than HE polarization due to the larger evanescent field intensity. According to the variational theorem for dielectric waveguides, an analytical method can be used to estimate the partial sensitivity of evanescent field sensing based on the calculated evanescent field intensity \[[@B17-sensors-15-21500],[@B18-sensors-15-21500]\]. Therefore, the partial sensitivity of homogeneous sensing (ΔN~eff~/Δn) possesses a maximum of 2.86 × 10^−3^ for HE polarization (corresponding to h = 5 μm, w = 4.5 μm) and 3.6 × 10^−3^ for EH polarization (corresponding to h = 5 μm, w = 2.5 μm), and the partial sensitivity of surface sensing (ΔN~eff~/Δt) exhibits a maximum of 4.89 × 10^−6^ nm^−1^ for HE polarization and 7.61 × 10^−6^ nm^−1^ for EH polarization (both corresponding to h = 5μm, w = 2.5μm). Therefore, the result of the optimization is that the waveguide mode is the fundamental EH mode, and the cross section dimension of the SOI rib waveguide is H = 10 μm, h = 5 μm, w = 2.5 μm. The simulation of the optimized mode is shown in [Figure 4](#sensors-15-21500-f004){ref-type="fig"}. ![Simulation of the fundamental EH mode of the SOI rib waveguide with H = 10 μm, h = 5 μm, w = 2.5 μm at wavelength of 1550 nm, calculating with full-vector finite difference method (FDM).](sensors-15-21500-g004){#sensors-15-21500-f004} 4. MZI Structure Implementation {#sec4-sensors-15-21500} =============================== 4.1. Conventional Implementations {#sec4dot1-sensors-15-21500} --------------------------------- In the waveguide MZI sensor, the evanescent field sensing is read out by the MZI configuration. The most critical components of the MZI configuration are the beam splitter and the beam combiner, which are waveguide branches and are often identical. For the conventional MZI configurations, there are three structures to realize waveguide branches, as listed in [Table 1](#sensors-15-21500-t001){ref-type="table"}. The data in this table show the technical parameters and the minimum required length of a single branch to fulfill the separation distance of d = 50 μm between the two straight SOI rib waveguides with large cross section. Thus for the SOI rib waveguide with a large cross section, the conventional implementations of waveguide branches lead to an overlong structure which is difficult to be realized. sensors-15-21500-t001_Table 1 ###### The conventional implementations of waveguide branches for SOI rib waveguide with large cross section. Assuming H = 10 μm, h = 5 μm, w = 5 μm. --------------------------------------------------------------------------------------------------------------------------------------------- Structure Type Technology Platform Technical Parameters & Requirements --------------------------------- ------------------------------------------------- --------------------------------------------------------- ![](sensors-15-21500-i001.jpg)\ Mode-matching When d = 50 μm,\ Y-junction 2θ = 0.4°,\ The minimum length of a single branch: L~0~ = 7.2 mm. ![](sensors-15-21500-i002.jpg)\ Waveguide bending When d = 50 μm,\ S-bend splitter R \> 0.26 m,\ The minimum length of a single branch: L~0~ = 5.1 mm. ![](sensors-15-21500-i003.jpg)\ Self-imaging effect \[[@B19-sensors-15-21500]\] When d = 50 μm,\ Multimode interference The minimum length of a single branch: L~0~ \> 12 mm. --------------------------------------------------------------------------------------------------------------------------------------------- 4.2. Trench-Based Bend and Branch {#sec4dot2-sensors-15-21500} --------------------------------- Studies have shown that the bends and branches of waveguides can be realized by using the medium trench, slot, or photonic crystal \[[@B20-sensors-15-21500],[@B21-sensors-15-21500],[@B22-sensors-15-21500],[@B23-sensors-15-21500],[@B24-sensors-15-21500]\]. In this paper, media-filled trenches are used to achieve waveguide bends and branches for an SOI rib waveguide with a large cross section, as shown in [Figure 5](#sensors-15-21500-f005){ref-type="fig"} and [Figure 6](#sensors-15-21500-f006){ref-type="fig"}, respectively. The gray area in these figures represents the filling medium, which can be air, SU8 or other refractive index matching fluid. ![90° trench-based bend geometry of SOI rib waveguide with large cross section.](sensors-15-21500-g005){#sensors-15-21500-f005} ![T-shaped branch geometry of SOI rib waveguide with large cross section.](sensors-15-21500-g006){#sensors-15-21500-f006} Due to the large dimension of the SOI rib waveguide, the computational memory requirement and time-consumption of a three-dimensional finite difference-time domain (3D-FDTD) simulation are huge. The usual way to overcome this issue is to simplify the 3D structure to 2D by using effective refractive index method (EIM), and then simulate the interested structure using two-dimensional finite difference-time domain (2D-FDTD) \[[@B25-sensors-15-21500]\]. In this paper, a 2D-FDTD method with a perfectly matched layer (PML) boundary condition \[[@B26-sensors-15-21500]\] is employed to numerically simulate the above mentioned waveguide bends and branches. The medium trench shown in [Figure 5](#sensors-15-21500-f005){ref-type="fig"} corresponds to a corner mirror, which can change the propagation direction of the waveguide according to total internal reflection. In general, the length and width of the medium trench (L and W as shown in [Figure 5](#sensors-15-21500-f005){ref-type="fig"}) should be large enough to reflect instead of to transmit more mode energy. The parameter D is defined to account for the Goos-Hanchen shift compensation, and it is positive when the trench interface moves away from the bending center (corresponding to the case shown in [Figure 5](#sensors-15-21500-f005){ref-type="fig"}, D \> 0). When the trench interface precisely passes through the bending center, D = 0. Setting the width and length of the air trench at 10 μm and 30 μm respectively, i.e. L = 30 μm, W = 10 μm, the electric intensity map of a 90° air-trench bend is shown in [Figure 7](#sensors-15-21500-f007){ref-type="fig"}. The influence of the parameter D on the bend efficiency of the air-trench bend exhibits as a curve with oscillatory variation in a very small range, as shown in [Figure 8](#sensors-15-21500-f008){ref-type="fig"}. Due to the strong ability of light constraint, there is a high bend efficiency for SOI rib waveguide with large cross section, even when D = ±500 nm. The reason for this oscillatory variation is that there is an angle between the air-trench and the meshing direction in 2D-FDTD simulation to avoid oblique incident light, and the interface of the air-trench and the waveguide is jagged as shown in [Figure 9](#sensors-15-21500-f009){ref-type="fig"}. It can be verified that this effect will be attenuated if finer mesh sizes are adopted. ![The electric intensity map in a plane 4 μm above the SiO~2~ layer for a 90° air-trench bend at a wavelength of 1550 nm. The SOI rib waveguide with a large cross section possesses a total rib height of 10 μm, an outside rib height of 5 μm, and a rib width of 2.5 μm. The guiding mode is the fundamental EH mode. The width and length of the air trench is 10 μm and 30 μm respectively, and the Goos-Hanchen shift compensation D = 0. Bend efficiency calculated through the 2D-FDTD simulation with mesh grid of 5 nm is 0.999746.](sensors-15-21500-g007){#sensors-15-21500-f007} ![The bend efficiency as a function of the Goos-Hanchen shift compensation (parameter D). There is an oscillatory variation, but the amplitude is very small.](sensors-15-21500-g008){#sensors-15-21500-f008} ![The local meshing in the 2D-FDTD simulation at the interface between air-trench and the bending waveguide. The mesh grid is 5 nm.](sensors-15-21500-g009){#sensors-15-21500-f009} Similarly, the electric intensity map of the asymmetric T-shaped air-trench branch with the trench width and length of 97 nm and 30 μm respectively is shown in [Figure 10](#sensors-15-21500-f010){ref-type="fig"}. Based on 2D-FDTD simulation with the mesh grid of 5 nm, the reflection efficiency is 0.497357 and the transmission efficiency is 0.502518. ![The electric intensity map in a plane 4 μm above the SiO~2~ layer for an asymmetric T-shaped air-trench branch at a wavelength of 1550 nm. The SOI rib waveguide with a large cross section possesses total rib height of 10 μm, outside rib height of 5 μm, and rib width of 2.5 μm. The guiding mode is the fundamental EH mode. The width and length of the air trench is 97 nm and 30 μm respectively.](sensors-15-21500-g010){#sensors-15-21500-f010} It is found from the simulations that the splitting efficiencies, including the reflection efficiency, the transmission efficiency and the total transmission efficiency, are functions of trench width (w), as shown in [Figure 11](#sensors-15-21500-f011){ref-type="fig"}. Both for air-filled trench and SU8-filled trench, the transmission decreases and the reflection increases as the trench width increases. With the trench width of 96 nm for air-trench branch and 112 nm for SU8-trench branch, a splitting ratio of 50%:50% is achieved. ![Splitting efficiency as a function of trench width. A 50%:50% splitting ratio is achieved with 96 nm air-trench width or 112 nm SU8-trench width.](sensors-15-21500-g011){#sensors-15-21500-f011} Supposing the trench branch is filled with index matching fluid and has a trench width of 120 nm, the influence of the refractive index of the index matching fluid is shown in [Figure 12](#sensors-15-21500-f012){ref-type="fig"}. As expected, the higher refractive the index of the filled material, the smaller the reflection efficiency and the larger the transmission efficiency. Thus, it can be concluded that the higher refractive index of the filled material results in the larger trench width for a 50%:50% splitting ratio. In addition, using index matching fluid as the filled material for the trench branch can prevent pollution of dust or impurities in the air. ![Splitter efficiency as a function of refractive index of trench. When the trench width is 0.12 μm, a 50%:50% splitting ratio is achieved with the refractive index of 1.76.](sensors-15-21500-g012){#sensors-15-21500-f012} Overall, the trench-based bends and branches can be used for SOI rib waveguide with large cross section, and their performances are obviously superior to the traditional bending and branching structures, such as the higher transmission efficiency and the shorter length of a single branch. Note that the simulation errors mainly come from two aspects, the coupling between modes and the out-of-plane scattering loss, in the process of simplifying the 3D model to 2D \[[@B27-sensors-15-21500]\]. Because the SOI rib waveguides discussed in this paper are confined by the single-mode condition and have strong capability of light constraint, the influences caused by these two factors are very small, and thus the error of the above results obtained from 2D-FDTD simulations is insignificant. 4.3. Proposed MZI Structure {#sec4dot3-sensors-15-21500} --------------------------- A new MZI structure based on an SOI rib waveguide with a large cross section that consists of two trench-based waveguide bends, two trench-based waveguide branches, and two straight waveguides is proposed as shown in [Figure 13](#sensors-15-21500-f013){ref-type="fig"}. These two parallel straight waveguides can act as the reference arm and the sensing arm of the MZI sensor, and their spacing is denoted by d. The parameter S represents the horizontal distance of the two waveguide branches. Compared with the traditional MZI configurations, the proposed configuration possesses two out ports, which can be used solely, or synchronously as mutual reference. ![Schematic of the MZI using media trenches based on SOI rib waveguide with large cross section.](sensors-15-21500-g013){#sensors-15-21500-f013} Using a 2D-FDTD simulation, the optical power map of such a trench-based MZI structure with d = 50 μm, S = 80 μm is shown in [Figure 14](#sensors-15-21500-f014){ref-type="fig"}. The guiding mode is the fundamental EH mode of the SOI rib waveguide with H = 10 μm, h = 5 μm, w =2.5 μm at wavelength of 1550 nm. It shows that only one output port has power output due to the interference effect of the two identical guiding modes in the sensing arm and the reference arm. The transmission of the electric field components in the Z direction E~z~ (out of plane) can be used to clearly explain this interference case, as shown in [Figure 15](#sensors-15-21500-f015){ref-type="fig"}. ![The optical power map of the MZI structure with the fundamental EH-polarized mode of an SOI rib waveguide with H = 10 μm, h = 5 μm, w = 2.5 μm at wavelength of 1550 nm. Because there is no phase difference between the sensing arm and the reference arm, only the port "output 1" has an output signal.](sensors-15-21500-g014){#sensors-15-21500-f014} ![The transmission of the electric field components in Z direction Ez (out of plane) of the MZI structure with EH-polarized mode of SOI rib waveguide with H = 10 μm, h = 5 μm, w = 2.5 μm at wavelength of 1550 nm.](sensors-15-21500-g015){#sensors-15-21500-f015} [Figure 16](#sensors-15-21500-f016){ref-type="fig"}a,b respectively show the normalized power of the SOI rib waveguide mode that propagates along the sensing arm and the reference arm. This normalized power is the ratio of the power on the transmission cross section to the input power of the light source. It can be seen that there are unstable regions at the waveguide bends and branches, which are caused by the interference of the waveguide modes. In particular, the normalized power of the interference enhancement is more than 1. It is found that the length of the unstable transmission segments at the waveguide bends or branches are less than 20 μm, which is a great advantage compared to the conventional implementations of MZI. ![Normalized power of the SOI rib waveguide mode propagating along the sensing arm and the reference arm.](sensors-15-21500-g016){#sensors-15-21500-f016} In the ideal situation (the splitting ratio of every branch in the MZI is 50%:50%), the normalized output power of each output port is a function of the phase difference (ΔΦ) between the sensing arm and the reference arm, as shown in [Figure 17](#sensors-15-21500-f017){ref-type="fig"}. A complementary relationship between the two output ports can be found. The normalized output powers of the output port "output 1" can be expressed as Equation (5). There, ΔΦ~0~ is the initial phase difference caused by machining error or other unbalanced factor. $$P\text{=}\frac{1}{2}\left\lbrack {1 + \cos(\Delta\Phi + \Delta\Phi_{0})} \right\rbrack$$ ![The normalized output power in each output port as a function of the phase difference between the sensing arm and the reference arm.](sensors-15-21500-g017){#sensors-15-21500-f017} 5. MZI Sensing Platform {#sec5-sensors-15-21500} ======================= According to the above analyses and results, the trench-based MZI structure exhibits better performance than traditional configurations. The schematic of the MZI sensing platform based on SOI rib waveguide with large cross section is shown in [Figure 18](#sensors-15-21500-f018){ref-type="fig"}. There is a sensitive window at the sensing arm where selective biochemical sensitive material is used for a specific application. Due to the high coupling efficiency of the SOI rib waveguide with large cross section and the standard single-mode glass fiber, the input and output ports of the MZI sensing platform can be conveniently connected to a laser source and light power detection unit, and remote measurement based on fiber-optic communication can be achieved. With the help of a simple tapered mode converter, the butt coupling of input and output waveguides with standard single mode fiber can be easily realized. A set of simulations show that more than 80% of the coupling efficiency is very easy to achieve. In addition, the output signal can also be detected by integrated photoelectric detectors, as shown in [Figure 18](#sensors-15-21500-f018){ref-type="fig"}. ![Schematic of the MZI sensing platform based on SOI rib waveguide with large cross section.](sensors-15-21500-g018){#sensors-15-21500-f018} As an example, an MZI sensor with the fundamental EH-polarized mode of SOI rib waveguide with H = 10 μm, h = 5 μm, w = 2.5 μm at wavelength of 1550 nm, when L = 10 mm and d = 50 μm, a homogeneous sensitivity of 7296.6%/refractive index unit (RIU) can be obtained according to Equations (3) and (5). Supposing the resolutions of the photoelectric detectors connected to the output ports are 0.2% of the optical power for the intensity measurement, the MZI sensor can achieve a detection limit of 2.74 × 10^−6^ RIU. This detection limit of refractive index is smaller than that obtainable by Si~3~N~4~ rib waveguides (7 × 10^−6^ RIU) \[[@B28-sensors-15-21500]\] and close to that reported in the previous works \[[@B29-sensors-15-21500],[@B30-sensors-15-21500]\]. More importantly, employment of the SOI rib waveguides with large cross sections make the device match well with the communication glass fibers, and the size of the MZI\'s branches are very small, so that the entire sensing platform can be very compact. 6. Discussion {#sec6-sensors-15-21500} ============= Employing with the media-filled trenches, the MZI configuration based on the SOI rib waveguide was realized in micrometer scale, thus a micron-sized and compact biochemical sensing platform based on an SOI rib waveguide with a large cross section was obtained, which is the main advantage of this work. Moreover, high coupling efficiency with standard single-mode glass optical fiber is the most important advantage of SOI rib waveguide with large cross section, which enables the waveguide sensors to integrate with optical fiber communication systems and (opto-) electronic systems, and therefore to realize remote sensing, in situ real-time detecting and possible application in the internet of things. According to the above analysis, the MZI sensor can perform bulk sensing with high sensitivity, which can be used to as chemical sensor. Due to the simple MZI configuration and strong adaptability of evanescent field sensing, the MZI sensing platform can also be used to detect biological reactions as long as suitable receptor molecules are immobilized at the waveguide surface in the sensitive window. To achieve high sensitivity, the trenches of the waveguide branches in the MZI configuration needs to be very narrow, which at the present stage may involve expensive microfabrication processes such as electron beam lithography (EBL) and deep reactive ion etching (DRIE). Therefore a compromise should be made between the sensitivity and the cost of the process. Alternately, employment of filled medium with higher refractive index can lower the difficulty of the process without decreasing the sensitivity. From the analysis of this paper, it can be speculated that the surface roughness of processing and the change of the external environmental temperature in real measurement have effects on the splitting efficiencies of the MZI's branches, but the further effects on the performance of the sensing system need to be further researched in the experiments, which is the next step of this work. In addition, due to the strong anti-interference capability of MZI configuration, these effects might be very small. Fortunately, due to the mass production feature of the SOI process, a very low cost for each sensor chip can be achieved as long as a large market is found. 7. Conclusions {#sec7-sensors-15-21500} ============== An MZI biochemical sensing platform based on an SOI rib waveguide with a large cross section is proposed in this paper. The optimization of the cross section dimensions of the SOI rib waveguide is performed through FDM simulations with the target of maximizing the evanescent field intensity. Medium filled trenches are employed to realize the MZI configuration based on the SOI rib waveguide. The performances of the MZI sensing platform are simulated by using 2D-FDTD method. The optimization of the SOI rib waveguide is that the guiding mode is the fundamental EH mode, and the cross section dimension is that the total rib height H = 10 μm, the outside rib height h = 5 μm and the rib width w = 2.5 μm. When the length of the sensitive window L = 10 mm and the spacing distance between the sensing arm and the reference arm d = 50 μm, the MZI sensor based on SOI rib waveguide with large cross section at an operating wavelength of 1550 nm can achieve a homogeneous sensitivity of 7296.6%/RIU. Supposing the resolutions of the photoelectric detectors connected to the output ports are 0.2%, the MZI sensor can achieve a detection limit of 2.74 × 10^−6^ RIU. This work is supported by Shenzhen Science and Technology Research and Development Funds (No: JCYJ20140709145631545). Dengpeng Yuan, Ying Dong, Yujin Liu and Tianjian Li conceived the platform. Dengpeng Yuan performed all the simulations and wrote the paper. Ying Dong reviewed and edited the manuscript. All authors read and approved the manuscript. The authors declare no conflict of interest.	{ "pile_set_name": "PubMed Central" }