The Impact of the Serotonin Transporter Gene Promoter DNA Methylation on Anorexia Nervosa(cid:0)A Preliminary Longitudinal Study

The serotonin system has been reported to be involved in the pathogenesis of anorexia nervosa (AN). The promoter region of serotonin transporter gene (SLC6A4), also called 5-HTTLPR, responsible for serotonin reuptake, has received much attention, especially 5-HTTLPR methylation, which was associated with abnormal eating behaviors. The study was aimed to explore the association between 5-HTTLPR methylation and AN, as well as its predictive effect on therapeutic response. Methods and were recruited. follow-up. 5-HTTLPR methylation levels and clinical symptoms were assessed at for participants, at methylation was measured by MassArray methods and clinical were mainly by the EDE-Q6.0 scale. abnormal eating behaviors in AN. The promoter region of serotonin transporter gene (SLC6A4), also called 5-HTTLPR, has received much attention. In our study, it found untreated AN patients had higher methylation at 5-HTTLPR than healthy controls, indicating its vital roles in AN pathogenesis. Further research needs to investigate the effect of 5-HTTLPR methylation on therapeutic response and explore its treatment role in AN patienrs.


Introduction
Anorexia nervosa (AN) is one of the most severe and complex eating disorders (EDs) characterized by strict restrictions of energy intake, a strong desire to stay extremely thin, a pathological fear of weight gain combined with a distorted body perception, which mainly affects young women [1]. AN patients are divided into restricting type (AN-R) and binge/purging type (AN-BP) based on whether they had binge eating and vomiting behaviors in the diagnostic and Statistical Manual of Mental Disorders (DSM-V) [2].
These symptoms make patients unable to comprehend the seriousness of the illness, as a result, they stay a pathological state persistently, which leads to serious distress and burden for their family and society. Especially, the prevalence of the disease is rising continuously in China and has become comparable to that worldwide [3]. Growing interest has focused on the pathogenesis and treatment of AN.
AN is believed to be the product of Gene x environment (GxE) interaction, which plays a vital role in the psychology of EDs [4] . The research is still relatively novel, among which DNA methylation is the most widely studied. DNA methylation is one of the most crucial epigenetic modi cations, mediated by environmental effects on gene expression without changing DNA sequences [5]. It is a process in which a methyl group is added to the nucleotide base pair cytosine next to a guanine, that happens at the speci c dinucleotide units called CpG islands (CpGs), leading to reduced gene expression [6].
The serotonin system has been reported to be involved in the pathogenesis of anorexia nervosa [7,8].The serotonin transporter gene (SLC6A4), also known as serotonin transporter (5-HTT), located on human chromosome 17q11.1-q12 [9], coding for the 5-HTT serotonin transporter protein, was supposed to be implicated in EDs [10]. It also acts a key role in regulating eating behaviors by controlling hunger and promoting satiety and being highly in uenced by nutritional intake [11]. Several animal studies remained the role of 5-HT in the inhibition of feeding behaviors, which could promote satiety and produce an anorexic effect [12,13].
Serotonin Transporter Linked-Polymorphic Region (5-HTTLPR), located at the promoter region of the SLC6A4 gene, is one of the most investigated polymorphisms [14], which was discovered to impact the transcriptional activity of serotonin and regulate its expression [15]. There is some evidence that the 5-HTTLPR genotype affects vulnerability to a broad range of mental disorders including depression [16], panic disorder (PD) [17], post-traumatic stress disorder (PTSD) [18], and so forth. However, scarce studies have been performed in ED patients and longitudinal studies in AN have not yet been investigated. SLC6A4 DNA methylation could result in reducing 5-HTT gene expression, which was reported to be associated with abnormal eating behaviors such as food restriction and food avoidance in the previous study [19]. Thus, we make a high hypothesis that SLC6A4 DNA methylation acts a critical part in restrictive eating and AN pathogenesis.
This study provides a novel perspective to help us understand the epigenetic mechanism of 5-HTTLPR methylation for AN. This is the rst investigation to explore whether there is association between 5-HTTLPR methylation level and AN, as well as clinical symptoms before and after treatment in AN patients. We expect that DNA methylation will be associated with AN, together with clinical symptoms and therapeutic response of AN.

Methods
Study design AN patients in this study came from two different projects. The methylation of 5-HTTLPR and clinical symptoms were assessed at the baseline for all patients, however, only thirty patients from one project completed the subsequent follow-up. The methylation level of 5-HTTLPR were measured at the twelfth week and clinical symptoms were evaluated at the fourth and twelfth week for follow-up patients. Meanwhile, they received routine outpatient treatment, mainly including physical and mental examinations, psychological counseling, nutritional evaluation, and drug treatment based on their situations. There is no required treatment in our study. The study protocol and process were approved by the ethics committee of the Shanghai Mental Health Center.

Participants
A total of ninety-one AN patients aging 13-60 years were recruited from the department of psychological outpatient and inpatient in Shanghai Mental Health Eating Disorder Treatment Center. All patients were assessed and diagnosed by two senior psychiatrists according to DSM-5 (American Psychiatry Association). Each of them met the following inclusion and exclusion criteria: (a) female, Chinese Han, Junior high school or above; (b) BMI from 13.0 kg/m 2 to18 kg/m 2 , no serious or chronic somatic disease; (c) no history of another psychiatric disorder; (d) have not taken psychiatric and hormone drugs since the illness onset or at least for twelve weeks; (e) not currently pregnant or breastfeeding; (f) capable of performing all tests and signed informed consent. These patients included AN-R subtypes (n = 44) and AN-BP subtypes (n = 47). Eighty-seven healthy controls (HC) were recruited among colleges and health care workers through advertisements. The inclusion and exclusion criteria were in accordance with those of AN patients, except that they were not diagnosed with anorexia nervosa on the basis of DSM-V.

Clinical assessment
At the baseline, demographic information such as age, gender, education level, as well as family history, psychological characteristics and clinical features of all participants were collected. The follow-up patients needed to be assessed the psychological characteristics and clinical features at the fourth and twelfth week, which were measured using the following instruments:

Eating Attitudes Test-26 (EAT-26)
The scale is a 26-item, self-screening questionnaire used to assess the symptoms and attitudes of ED patients. It mainly consists of three dimensions: diet, gluttony, food preoccupation concept and oral control. Its standards have been evaluated in Hong Kong [20]. Our previous results preliminarily con rmed that the reliability and validity of the scale in mainland China are good with internal consistency of 0.910, test-retest reliability of 0.817, convergent validity of 0.744, speci city of 0.85 and approximate sensitivity of 0.66 [21].
Questionnaire Version of the Eating Disorders Examination (EDE-Q 6.0) The scale is a 28-item, self-report questionnaire widely used to assess the severity of ED, that consists of four subscales: restricted eating, attention to eating, attention to body shape, and attention to weight. The Chinese version of the scale was provided by Professor Li of Hong Kong University, whose reliability and validity were good with internal consistency of 0.95, test-retest reliability of 0.79, and convergent validity of 0.72 [22]. In addition, our previous study preliminarily con rmed that the reliability and validity of the scale is good in mainland China [23].

Other questionnaires
In addition to the above scales, some other rating scales were also included. Hamilton Depression Rating Scale (HAMD), Hamilton Anxiety Scale, (HAMA) and Yale Brown Obsessive Compulsive Scale (YB-OCS) were adopted to evaluate the depression, anxiety and compulsive symptoms respectively in AN patients.

SLC6A4 promoter methylation
Venous blood samples were obtained from all participants and collected by EDTA tubes. DNA was extracted using a DNA extraction kit (Tiangen, China) and then puri ed and converted by bisul te using a EZ DNA Methylation Kit™ protocol (Zymo Research, USA). The promoter-associated CpG island was targeted in our assay, primers of which were used to amply bisul te treated DNA. The primers for SLC6A4 were designed by the online software EpiDesigner webunit (www.epidesigner.com, Agena Bioscience, USA), that were as follows: forward 5'-aggaagagagTTGTTAGGTTTTAGGAAGAAAGAGAGA-3' and reverse 5'-cagtaatacgactcactatagggagaaggctCCCTCACATAATCTAATCTCTAAATAACC-3', a T7-promoter tag was attached to the reverse primer for the subsequent PCR ampli cation. The reaction conditions for the PCR were as follows: 94∘C for 4min, followed by 45 cycles at 94∘C for 20sec, 56∘C for 30sec, 72∘C for 1min, then 72∘C for 3min. The PCR products were treated with shrimp alkaline phos phatase (SAP), followed by uracil-speci c cleavage reaction using Mass CLEAVE™ Reagent Kit (Agena, USA). The Agena Mass ARRAY platform was used to analyze the methylation levels of each CpG island in the SLC6A4 gene promoter. Methylation data were generated by the EpiTYPER software v1.2(Agena, USA). The speci c principle was shown in the gure 1.

Statistical analysis
All statistical analysis were conducted by SPSS statistical analysis software (version 22.0; IBM SPSS, Chicago, IL, USA) and GraphPad Prism 8.0 (https://www.graphpad.com/guides/prism/8/userguide/index.html, accessed on 25 April 2021). The data were expressed as numbers and frequencies for categorical variables and analyzed by χ2-test. The continuous variables that were normally distributed are presented as mean ± standard and differences between them were determined by T test and one-way analysis of variance (ANOVA). While the variables that were not normally distributed are presented as the median (maximum and minimum), and differences between them were determined by Mann-whitney U test and Kruskal-Wallis ANOVA test. Apart from the methylation level of CpG4 unit in 5-HTTLPR, others were all normally distributed. Pearson correlation analysis (normal distribution data) and Spearman correlation analysis (non-normal distribution data) were adopted to analyse the association between DNA methylation level in 5-HTTLPR and clinical features. Repeated-measures ANOVAs were used to compare the changes of DNA methylation levels in 5-HTTLPR before and after treatment. All tests were two-tailed and p-values < 0.05 were considered to be statistically signi cant.

Demographic and clinical characteristics
Ninety-one AN patients without treatment and eighty-seven HCs completed the study. All participants were female with mean age of 18.96±4.14 years old and mean education level of 12.18±2.94 years. The mean age of AN onset was 16.5±2.64 years old and the mean course of AN was 15 years. As shown in Table 1, there were no signi cant differences in age (P=0.068) and education level (P=0.062) between the two groups. The current BMI index, ideal body weight and ideal BMI index of AN patients were signi cantly lower than those of HC group (P<0.05). But no signi cant differences were found in the maximum body weight (P=0.820) and maximum BMI (P=0.531) between the two groups. Clinical characteristics have signi cantly higher scores in EAT-26 and EDE-Q6.0 scales (P<0.05), the abnormal eating symptoms of AN patients were more serious than that of HC groups.

Differences of DNA methylation levels in 5-HTTLPR in AN patients and healthy controls
All CpG units of the 5-HTTLPR methylation level were analyzed by MassArray Epityper mass spectrometer and quanti ed as percentage. The ampli ed DNA fragment of each sample contained 7 CpG units, among which more than 50% of GpG3 units failed to be detected successfully, so they were not analyzed, six units were nally included in the statistics. Through intergroup comparison, the methylation level of the CpG11, CpG24.25, CpG31.32 and CpG_mean in AN group was found to be signi cantly higher than those in HC group (P=0.039, 0.042, 0.018, 0.001 respectively). See Figure 2.
Differences of DNA methylation levels in 5-HTTLPR in different subtypes of AN patients and healthy controls AN patients were further divided into restricting (n=44) and binge-eating/purging subtypes (n=47). Through Kruskal-wallis ANOVA test, it showed signi cant differences in CpG24.25 and mean methylation level among AN-R, AN-BP and HC groups (P=0.008). Further, the methylation level of CpG24. 25 and CpG_mean in AN-BP group was signi cantly higher than that of HC groups (P=0.027, 0.031 respectively). However, there was no signi cant differences between AN-R group and AN-BP group (P>0.05), AN-R group and HC group (P>0.05). See Figure 3.

Association analysis between DNA methylation level in 5-HTTLPR and clinical features before treatment in AN patients
According to correlation analysis, it was found that the methylation level of CpG11 unit was positively correlated with the total score of EDE-Q6.0 scale (r=0.216, P=0.047). No signi cant correlation was found between the methylation level of all units in SLC6A4 gene and depression, anxiety and obsessivecompulsive symptoms of AN patients. See Table 2.

Changes in clinical symptoms and methylation level of 5-HTTLPR after treatment in AN patients
Only thirty patients completed the follow-up of twelve weeks. Comparing the clinical symptoms and methylation level before and after treatment by repeated-measures ANOVA, it was found that after fourth and twelfth week treatment, BMI index signi cantly increased (P 0.05), the total score of HAMD, HAMA, YB-OCS scale, both the total score and factor score of EAT-26, and EDE-Q6.0 scale signi cantly decreased in AN patients (P 0.05). In addition, it revealed that the methylation level of CpG11, CpG26.27.28, CpG31.32, CpG33.34.35.36 units and CpG_mean decreased after twelve weeks treatment, but no signi cant differences were found (P 0.05). See Table 3.
Association analysis between the change of DNA methylation level in 5-HTTLPR and the change of clinical features before and after treatment in AN patients.
The DNA methylation level of 5-HTTLPR decreased after 12 weeks treatment compared with basement in AN patients. The changes of each point were described as res_CpG. Similarly, the changes of EDEQ-6.0 total score were described as res_EDEQ-6.0. Correlation Analysis between the change of 5-HTTLPR methylation level and the change of clinical symptoms after treatment showed that the res_CpG1 was signi cantly positively correlated with res_EDEQ-6.0 (r =0.556, P=0.028). However, further considering age, onset age, education level, disease course and other factors as covariables, the in uence of CpG1 unit was no longer signi cant (F=0.908, P=0.409). No correlation was found between the changes of other units and clinical symptoms (P 0.05).

Discussion
Our study adopted preliminarily cross-sectional and longitudinal design to investigate the possible association between 5-HTTLPR methylation level and AN, as well as clinical symptoms before and after treatment in AN patients. We examined seven CpG units of 5-HTTLPR, among which six CpG units were detected except CpG3 unit. AN were signi cantly associated with higher methylation at CpG11, CpG24.25, CpG31.32, suggesting that 5-HTTLPR methylation may contribute to AN pathogenesis. Moreover, the methylation level of CpG24.25 unit and CpG_mean in AN-BP groups, not the AN-R groups were signi cantly higher than that in HC groups. Further, the study failed to discover signi cant difference of all units between AN-R group and AN-BP group, so the 5-HTTLPR methylation might be not a biomarker to distinguish the two subtypes of AN. Meanwhile, positive correlations were reported between CpG11 methylation level and the total score of EDE-Q6.0 scale. Studies have con rmed that 5-HTTLPR methylation level in peripheral blood lymphocyte was associated with that in brain [24], and a majority of epigenetics did research using the peripheral blood lymphocyte samples [25,26], which provide a theoretical basis for our study. Our study showed a signi cant association between SLC6A4 methylation and AN, which was in concordance with the observation in patients with depression [26,27]. Probably because SLC6A4 hypermethylation causes the decrease of mRNA expression, which leads to the decrease of 5-HT transporter protein and 5-HT reuptake, then resulting in the increase of 5-HT concentration in synaptic space and the decrease of 5-HT synthesis by negative feedback, nally it may increase the e cacy of serotonergic antidepressants [28]. While previous study did not nd positive results in AN patients [29], which may be due to the differences in the samples and fragments selected. According to our results, the methylation level of CpG11, CpG24.25 and CpG26.27.28 units may be related to AN. Besides, the methylation level of 5-HTTLPR was connected with eating symptoms but not the accompanying emotional symptoms, indicating it may play a speci c role in AN. Since there are no studies to explore the effect of 5-HTTLPR methylation on the clinical symptoms of AN patients, the conclusion needs to be veri ed from other perspectives, like the brain function [25,30].
Long-term follow-up study on patients with different subtypes of AN showed that most AN started with AN-R, about 88% of AN-R can develop bulimic behaviors, 62% of AN-R eventually develops into AN-BP subtype [31]. To some extent, different subtypes of AN can be seen in different stages of disease development. Our study showed AN-BP patients had higher methylation than healthy controls, probably indicating alterations in trait-related serotonergic function, which might be related with 5-HTT [32]. The differences failed to be found in AN-R patients, may be due to the absence of signi cant changes in biogenetic indicators in the acute phase of AN-R patients.
Considering the possible effect of 5-HTTLPR methylation on long-term AN outcome, some studies have reported the predictive effect of 5-HTTLPR methylation on the therapeutic response in depression patients, the higher the 5-HTTLPR methylation level, the better the antidepressant response [28,33], but there were also contrary results [34], as for how gene methylation affects the therapeutic effect has not been reported. It was the methylation level of CpG11 unit in 5-HTTLPR that was signi cantly associated with therapeutic responses [35]. However, it was not consistent in our study, which may due to the different mechanisms of SLC6A4 gene in the two diseases, in addition, it is hard to detect the association between the SLC6A4 gene methylation and treatment response in a short time for the chronicity and obstinacy of AN. Thus, it may not be suitable as a biomarker to evaluate the outcome of AN.
Few previous studies conducted on 5-HTTLPR in patients with EDs. One of which indicated that the Sallele of the 5-HTTLPR genotype particularly increased the risk susceptibility for AN and was related to its instability and depressive comorbidity [32]. Thus, a study has reported that the effects of 5-HTTLPR genetic variants may contribute to the regulation in modifying its methylation [36], which has been validated in patients with depression [26,28] and anxiety [37].However, Our epigenetic study of 5-HTTLPR did not consider its genotypes, which may make it hard to determine the real consistency of data. Further investigation is needed to examine the underlying mechanisms.
Apart from the shortcomings mentioned above, there were also some other limitations in our study. Firstly, we did not correct for multiple testing due to the exploratory nature, thus increasing the likelihood of receiving a signi cant result. Besides, we have to admit that the weight of HCs was normal and the effects of malnutrition on SLC6A4 methylation should be discussed. On the other hand, we did not consider other CpG units in the region that may be related to AN [30] and the environmental factors that may affect SLC6A4 methylation levels to a large extent [38]. Moreover, only a part of patients nished the follow-up treatment, which may account for the negative associations between 5-HTTLPR methylation and therapeutic responses, or the follow-up time was not enough due to the complexity of AN.
Additionally, we did not detect the mRNA and 5-HT concentration that may be involved in the effect of DNA methylation mechanism. Recently, we have already recruited HCs with low weight, some limitations will be improved in our subsequent research.

Conclusions
Our study detected the methylation level of seven CpG units in 5-HTTLPR gene in patients with AN and found signi cant associations between them for the rst time, supporting 5-HTTLPR methylation may be as a biomarker for AN but not a biomarker to distinguish the two subtypes of AN and predict the therapeutic responses. The discovery provides direction for the epigenetic mechanisms of AN and may offer a new perspective for targeted therapy of AN.

Availability of data and materials
The data and materials supporting the ndings of this study are available from the corresponding author upon reasonable request.
Ethics approval and consent to participate All procedures performed in studies involving human participants were in accordance with the ethical standards of the Ethics Committee of the Shanghai Mental Health Center (2020-32) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. All participants signed the informed consent to the study.

Consent for publication
This article is original and has never been submitted or published elsewhere. All the authors have checked this nal manuscript and approved to submit it. P: t-test  Figure 1 <p>The speci c principle of 5-HTTLPR DNA methylation conducted by the EpiTYPER software.&nbsp; </p> Figure 2 <p> Comparison of 5-HTTLPR methylation rates at each CpG site between AN and healthy controls</p> <p>Note * <em>P</em>&lt;0.05, ** <em>P</em>&lt;0.01</p>