Introduction

Anorexia nervosa (AN) and bulimia nervosa (BN) are severe eating disorders (ED) affecting mainly young females. They are characterized by alterations in eating behavior and weight regulation driven by low self-esteem due to weight preoccupation and perceptions toward body weight and shape. The etiology of these disorders is complex and it is widely accepted that biological, psychosocial and genetic factors are involved [1]. A number of candidate genes with a role in the regulation of eating behavior and body weight have been considered in studies with ED patients [1–3]. One of them is the brain-derived neurotrophic factor (BDNF), which is widely expressed in the brain and is involved in neuronal development and synaptic plasticity. Several lines of evidence indicate that this gene plays a role in energy balance and reward centers of the brain, thus impacting eating behavior, body weight and cognitive impairments in ED [4–7].

Two single-nucleotide polymorphisms (SNPs) in the BDNF gene, namely a valine to methionine substitution in residue 66 (rs6265) and a C-270T transition in the 5′-untranslated (UTR) region (rs56164415), have been the focus of most studies in the ED setting. These works have mostly addressed the role of these variants in the susceptibility to the disorder. For instance, an association between the Met66 allele and the risk for AN, particularly the restricting subtype, has been reported [8, 9], albeit contradictory data have recently been published [10, 11]. In addition, the -270C allele has been shown to modulate the risk for BN and the age at onset of weight loss in ED patients [9].

Previous reports in healthy subjects suggest an association between BDNF variant alleles and personality dimensions such as neuroticism [12], introversion [13] and various anxiety- and depression-related traits [14]. However, the role of BDNF genetic variants as modulators of personality characteristics and psychopathological symptoms in ED patients has been largely overlooked. Individuals with ED display characteristic personality profiles of several traits each of which has been shown to be at least moderately heritable [15]. Studies on personality dimensions have revealed differences between ED patients and healthy females [16, 17], and high levels of traits such as depression, anxiety, impulsiveness, drive for thinness, etc. have been associated with the severity of AN and BN [18, 19].

In the present study, we have aimed to investigate whether BDNF genetic variability may influence personality characteristics and psychopathological symptoms in ED patients and/or modulate the risk for the disorder. For this, in a population of patients with AN or BN and healthy subjects, we have analyzed the presence of the two most commonly studied BDNF SNPs, namely Val66Met and C-270T, as well as that of four other tag-SNPs located in the gene (rs11030102, rs16917237, rs11030119 and rs10835210), which have previously been suggested to impact different psychiatric disorders and/or serum levels of BDNF [20–23].

Methods

Subjects

The study group consisted of 169 unrelated consecutive female patients with AN (n = 106) or BN (n = 63), some of whom had also participated in two other recent studies by our group [24, 25]. The patients attended the collaborating Eating Disorders Unit at the Mental Health Outpatient Clinic in Badajoz (Spain), and were diagnosed by one psychiatrist and one psychologist using the ED section of the Structured Clinical Interview for DSM-IV [26]. Anthropometric (weight, height and body mass index) and psychological parameters (see below) were then collected. Diagnosis was blind to genotype. Exclusion criteria, determined upon screening, included dementia, mental retardation, schizophrenia, Turner’s syndrome, other neurological disorders and underlying endocrine pathologies.

A total of 312 healthy women from the same geographical area as the patients (Health District of Badajoz, Spain) were recruited among University students and staff. Interviews were carried out to guarantee that they had never been diagnosed as having any psychiatric disorder or received any psychiatric treatment. None of the participating control subjects showed anthropometric parameters indicative of present ED.

All the participants were white Spanish individuals who gave written informed consent. The study protocol was approved by the Ethics Committee of the University of Extremadura and was conducted in accordance with the Declaration of Helsinki and its subsequent revisions.

SNPs selection

European population (CEU) SNP data were downloaded from the International Haplotype Mapping Project web site (http://hapmap.ncbi.nlm.nih.gov/cgi-perl/gbrowse/hapmap27_B36). To capture the common SNP variation in the BDNF gene, we analyzed the coding sequence and adjacent 3′ and 5′UTR regions and selected four tag-SNPs (rs11030102, rs16917237, rs11030119 and rs10835210) using the Tagger function of the Haploview 4.0 software (Table 1). The SNPs were selected with a minor allele frequency of at least 0.1, a pair-wise tagging r 2 value of at least 0.80 and were located within a 46.5 kb region that encompasses the BDNF gene (contig NT_009237.18; chromosome positions 27681596 to 27728102). All the SNPs were intronic and captured most of the common variations described for the gene and adjacent regions in public databases.

Table 1 Single-nucleotide polymorphisms genotyped in the BDNF genomic region
Full size table

Genotype analysis

Blood samples from all participants were stored at −80 °C until analysis. Genomic DNA was isolated from peripheral blood leukocytes in 2-ml aliquots of whole-blood samples with a Qiagen blood midi kit (Qiagen Inc., Chatsworth, CA, USA). The purified DNA samples were then stored at 4 °C in sterile plastic vials.

Genotype analyses for rs11030102, rs16917237, rs11030119 and rs10835210 were performed with the single-base extension polymerase chain reaction Sequenom iPLEX-Gold and the mass spectrometry-based platform MassARRAY MALDI-TOF at the Spanish Genotyping National Centre (CEGEN-ISCIII). In brief, the analyses consisted of an initial locus-specific PCR, followed by single-base extension using mass-modified dideoxynucleotide terminators of an oligonucleotide primer which anneals immediately upstream of the polymorphic site of interest. Using MALDI-TOF mass spectrometry, the distinct mass of the extended primer identifies the allele [27].

PCR-RFLP (restriction fragment length polymorphisms) techniques were utilized for the determination of the Val66Met and C-270T SNPs as described elsewhere [9]. The analysis of samples with unclear restriction patterns was confirmed by direct sequencing (Applied Biosystems, Foster City, CA, USA).

Psychometric evaluation

The examination of personality characteristics and psychopathological symptoms in ED patients was performed on the day of their first visit to the Eating Disorders Unit by two self-reported questionnaires, namely the Eating Disorders Inventory Test-2 (EDI-2) and the Symptom Checklist 90 Revised (SCL-90R).

The EDI-2 is a 64-item questionnaire designed to assess the cognitive and behavioral features characteristic of ED patients [28]. It is composed of eight main subscales (Drive for Thinness, Bulimia, Body Dissatisfaction, Ineffectiveness, Perfectionism, Interpersonal Distrust, Interoceptive Awareness and Maturity Fears) and three additional ones (Asceticism, Impulse Regulation and Social Insecurity). This inventory has been adapted for the Spanish population showing high internal consistency between the different subscales [29]. The SCL-90R is a 90-item psychiatric self-reported inventory that evaluates a broad range of psychological problems and symptoms of psychopathology through three Global Indexes (Global Severity Index, GSI; Positive Symptom Distress Index, PSDI; and Positive Symptom Total, PST) and nine primary symptom dimensions (somatization, depression, anxiety, hostility, phobic anxiety, paranoid ideation and psychoticism) [30]. This inventory has also been previously validated in the Spanish population [31].

Statistical analyses

The comparison of allelic and genotypic frequencies in cases and controls was performed with the χ2 or Fisher’s exact test as appropriate. The association between the analyzed polymorphisms and the risk to develop AN or BN was estimated by odds ratio (OR) with 95 % confidence interval (CI). The distribution of the quantitative variables across the different genotypes was compared by T test or ANOVA, as appropriate, and the results were corrected for multiple testing by the Bonferroni method, adjusting for the number of studied SNPs (significant p value <0.0083). To assess the association between BDNF variants and anthropometric parameters or personality characteristics and psychopathological symptoms, genotypes were considered as a dichotomous variable: value 0 was attributed to genotypes Val66Val, -270 CC; rs11030102 CC, rs126917237 GG, rs11030119 GG and rs10835210 CC as reference; and value 1 to heterozygous and mutant homozygous genotypes, assuming an effect for the presence of at least one allele at risk.

The statistical power of the sample size was evaluated with a log-additive genetic model, analyzing the frequency for carriers of the variant alleles with arbitrarily established effect size at 2.0 (type I error = 0.05). With the available sample size and minor allele frequencies ranging from 0.1 to 0.44, the statistical power for detecting associations with categorical variables in the AN and BN groups was 0.85–0.99 and 0.70–0.93, respectively (Quanto software v. 1.2.4, University of Southern California).

The SNPstats platform [32] was utilized to estimate general haplotype frequencies and to obtain linkage disequilibrium data. To refine the haplotype construction and to identify core regions with the potential to affect either the susceptibility for ED or the anthropometric/psychological parameters in the patients, we used a sliding window approach to construct successive and adjacent 3-SNP haplotypes. Subsequent case/control and quantitative associations were assessed using PLINK software v1.07 with a haplotype frequency cut-off of 0.1 [33].

In all instances, differences were considered to be significant when p values were under 0.05. Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) version 15.0 for Windows (SPSS Inc., Chicago, IL, USA).

Results

Table 2 shows descriptive and clinical characteristics of ED patients and healthy subjects, who displayed higher weight and BMI than AN patients (p < 0.05).

Table 2 Descriptive and clinical variables of patients with anorexia nervosa (AN) or bulimia nervosa (BN) and healthy controls
Full size table

Single-SNP study

Genotyping analyses were successful in all the patients, whilst ten samples for SNP rs11030119, four for SNP rs16917237 and one for SNP rs11030102 failed to be genotyped in the control group. The frequencies of the six studied polymorphisms in the control group showed no departures from the Hardy–Weinberg equilibrium (p > 0.05 in all cases). The distribution of the different genotypes both in control subjects and ED patients is depicted in Table 3. None of the BDNF SNPs studied were associated with a greater risk of AN or BN (Table 3).

Table 3 Genotype frequencies of the analyzed polymorphisms in patients with anorexia nervosa (AN) or bulimia nervosa (BN) and healthy controls
Full size table

With regard to anthropometric parameters in the ED patients, analyses conducted using a recessive model of inheritance revealed that the TT mutant homozygous genotype of the rs16917237 SNP was associated with higher BMI (p < 0.01) and higher minimum weight (p < 0.01) in the BN patients (Table 4). The statistical significance of the association remained after correction for multiple testing. No other SNP showed a relevant effect in any of the two study groups.

Table 4 Effect of the rs16917237 SNP (G/T) on physiological parameters measured in patients with anorexia or bulimia nervosa
Full size table

The study of the associations between BDNF genetic variants and personality characteristics and psychopathological symptoms in the AN group showed that patients carrying the -270CC wild type genotype scored significantly higher than CT carriers in the Interpersonal Distrust scale of the EDI-2 test (6.91 ± 4.89 vs. 2.67 ± 3.12, Table 5). In turn, BN patients who were carriers of the same genotype showed higher values in two of the three scales of the SCL-90R inventory, namely GSI (2.08 ± 0.81 vs. 1.44 ± 0.68, Table 6) and PST (73.31 ± 15.69 vs. 59.91 ± 20.0; Table 6). Finally, BN patients’ carriers of the rs11030102 mutant homozygous genotype showed significantly higher scores in social insecurity than patients who harbored at least one wild type allele (7.83 ± 4.75 vs. 15.0 ± 1.41; Table 6). The p value for all these associations was significant after adjusting the results for multiple testing.

Table 5 Association between the studied BDNF genotypes (recessive model) and psychopathological traits in anorexia nervosa patients measured by the EDI-2 and SCL-90R inventories
Full size table
Table 6 Association between the studied BDNF genotypes (recessive model) and psychopathological traits in bulimina nervosa patients measured by the EDI-2 and SCL-90R inventories
Full size table

Haplotype study

Linkage disequilibrium (LD) plot and data (D, D′ r and p) for the SNPs considered are depicted in supplementary Figure S1.

The haplotype analysis carried out by the sliding window approach using three contiguous SNPs produced four loci of interest, namely locus 1, with SNPs rs6265, rs11030102 and rs10835210; locus 2, which included rs11030102, rs10835210 and rs16917237; locus 3, with rs10835210, rs16917237 and C-270T and locus 4, containing the rs16917237/C-270T/rs11030119 combination.

As it was observed in the single-SNP study, there were no relevant differences in the distribution of these haplotypes between AN or BN patients and control subjects (data not shown). Only the haplotype GCG in locus 4 was found to be slightly more frequent in the control group (frequency = 0.527) than in BN patients (frequency = 0.428; Chi square p = 0.044). The complete analyses of the distribution of BDNF haplotypes in the study population are shown in supplementary tables S1 and S2.

The regression analyses of BDNF haplotypes in relation to psychopathological symptoms in ED patients revealed that in the AN group the allele combination CGC within locus 3 was associated with all three scales of the SCL-90R inventory, namely GSI (BETA = 0.34, r 2 = 0.07, p < 0.01), PST (BETA = 6.80, r 2 = 0.04, p < 0.05) and PSDI (BETA = 0.20, r 2 = 0.04, p < 0.05). Furthermore, this locus 3 (haplotype CGT) was also observed to be associated with the same three scales in the BN patients GSI (BETA = −0.65, r 2 = 0.09, p = 0.01), PST (BETA = −13.52, r 2 = 0.09, p = 0.01) and PSDI (BETA = −0.39, r 2 = 0.06, p = 0.06). After adjusting for multiple testing considering the four haplotypes involved (level of significance: p < 0.0125), only the association with the GSI scale in AN patients remained relevant. Notwithstanding, it should be noted that p values for both the GSI and PST scales in BN patients were on the border of statistical significance.

The scores of the EDI-2 inventory were largely unaffected by the genetic variability in the four loci considered (data not shown). Only the association previously observed for the interpersonal distrust subscale and the C-270T SNP in AN patients was maintained for two haplotypes that harbored this SNP, namely CGT in locus 3 (BETA = −4.559, r 2 = 0.0516, p = 0.023) and GTA in locus 4 (BETA = −4.292, r 2 = 0.0503, p = 0.025), although significance was lost after Bonferroni correction.

Discussion

Our results did not show any significant associations between any SNP in the BDNF gene and the risk of AN or BN. This finding is consistent with the currently only genome-wide association (GWAS) study performed in AN, which could not find any individual SNP associated with any definition of illness [34]. With regard to genetic association studies, most of the available information refers to either the Val66Met or the C-270T polymorphisms. The former has been associated with a higher risk for these disorders, particularly restrictive AN (ANR) [9, 35–37]. However, we and others [10, 38–41] could not reproduce these results. Moreover, the analysis of a subset of ANR patients in our population (data not shown) did not reveal further significant associations. In the same manner, and consistent with previous studies [35, 40], we did not detect any association between the risk of ED and the C-270T SNP. The present controversy on the role of BDNF genetic variants in the susceptibility to ED has recently been stressed by Hong et al., who suggest that phenotype assessment may play an important role [42]. Haplotype analyses in the BDNF gene could help explain these inconsistencies, but these studies are still scarce in the ED setting. Only a previous family-based association study has identified a 2-SNP haplotype associated with AN [43]; however, we could not confirm this finding in our sample.

We next tested the effect of BDNF genetic variability on the anthropometric parameters of the patients. The results showed that the TT mutant genotype of the rs16917237 SNP was consistently associated with higher weight and BMI in subjects with BN. There are no previously reported evidences linking this SNP to body weight in ED patients but, interestingly enough, the GIANT (Genetic Investigation of ANthropometric Traits) consortium (http://www.broadinstitute.org/collaboration/giant/index.php/GIANT_consortium) has recently included this polymorphism in a list of potential markers of obesity. In this regard, it should be mentioned that BN patients tend to display higher BMI than controls and higher rates of premorbid obesity [44, 45].

The putative impact of BDNF genetic variants on body weight has recently been pointed out by a GWAS study in healthy subjects, which reported the association of the rs10767664 SNP with higher BMI [46]. Although we did not test the effect of this variant, rs10767664 is in strong LD (r 2 > 0.75) with the Val66Met SNP, whose effect was evaluated with negative results by both the present work and previous studies in ED patients [9, 47]. Larger studies in the ED setting are, therefore, needed to confirm the observation reported by the afore-mentioned GWAS study for the general population.

On the other hand, Ribases et al. have reported an increase in body weight for BN patients (with no previous AN) who were carriers of the -270T allele [9, 47], an association we did not observe in our study. This discrepancy could reflect phenotypic heterogeneity of our bulimic sample, which did include women with a history of AN, the course of which often includes the emergence of bulimic symptoms and a crossover to the full syndrome of BN [48]. In addition, interactions of BDNF SNPs with dopaminergic polymorphisms holding the potential to affect body weight could have also influenced our observations [25, 49]. Overall, the explanation of the modulation of body weight by BDNF variants is still unclear. A putative hypothesis has been given by Saito et al., who found a positive correlation between serum BDNF levels and BMI among females with BN [50]. Although the precise biochemical effect of the rs16917237 SNP, which we found to be associated with weight, is presently unknown, the fact that other SNPs in the intronic regions of BDNF can affect circulating levels of the protein in BN patients [22, 43] suggests that this could be an underlying mechanism for the observed up-regulation of weight and BMI.

The main objective of this work was to investigate whether BDNF genetic variability correlates with personality dimensions and/or psychopathological symptoms in ED patients. To our knowledge, this has only been addressed by one previous study in Polish subjects with AN, albeit only the Val66Met and C-270T SNPs were analyzed [41]. The authors concluded that the -270T allele was related to certain personality traits. In our group of AN patients, it was the -270C allele that correlated with interpersonal distrust, meaning reluctance to form close relationships. The different inventories utilized, ethnicity, or the different setting in which the studies were conducted (outpatients vs. inpatients) might be behind this controversy.

The sliding window approach applied in this work revealed a broad effect of one of the resulting four loci on the scores of the SCL-90R test. This inventory measures psychopathological symptoms that do not necessarily have to be intimately related to ED and are common to various psychiatric disorders. The observed association of certain BDNF haplotypes with these symptoms, the negligible effect on the traits measured by the EDI-2 questionnaire, specifically developed for ED patients, and the fact that these haplotypes were not relevant for the susceptibility to ED, suggest altogether that BDNF genetic variability may be important for psychopathological problems that could actually reflect comorbid disorders in ED patients.

Why genetic variations in the BDNF gene could lead to psychopathological symptoms in ED patients? Compelling evidence suggests that the pathophysiology of several mental illnesses could be the result of an alteration in the synaptic plasticity caused by an altered expression and release of BDNF [51–53]. Furthermore, Mercader et al. [43] have reported that certain BDNF haplotypes are able to modulate plasma protein concentrations in BN patients. Now, these concentrations have in turn been shown to correlate negatively with the scores of psychological inventories in ED patients [50, 54, 55]; therefore, it is tempting to speculate that altered BDNF levels resulting from the presence of certain haplotypes in our patients could have contributed to the development of the observed psychopathological traits. A fact that supports this hypothesis is that the same SCL-90R scales that correlated with BDNF plasma levels in the study by Mercader et al. [55] were those affected by the locus identified in our study.

There are a number of limitations in the present work that should be considered. First, the limited size of our sample, particularly of the BN group, which may limit the generalizability of the results. On the other hand, this limited sample also allowed for all the patients to be diagnosed and treated by the same clinicians in the same facilities over a short period of time, which increases the homogeneity of our analyses. Second, we did not consider ED clinical categories to correct for multiple testing, as this has been suggested to be too stringent to detect a moderate correlation with different endophenotypes in similar studies [55]. Third, our control sample was composed by University students, which might constitute a potential selection bias in terms of differences in socioeconomic status between patients and controls.

In conclusion, the findings of the present study, preliminary as they are given the described limitations, indicate that variability in the BDNF gene locus may contribute to anthropometric features, but also to psychopathological symptoms in ED patients that are likely indicative of comorbid disorders. Further studies with larger and homogeneous populations of patients, evaluated with the same inventories, are needed to confirm the reported associations.