Association between brain-derived neurotrophic factor genetic polymorphism Val66Met and susceptibility to bipolar disorder: a meta-analysis

Bipolar disorders (BPD) are chronic, recurrent, debilitating disorders with high lifetime prevalence and significant disease burden across different populations [1]-[3]. However, recent advances in pharmacological treatment for BPD remained quite modest. The treatment of bipolar depression is still a major challenge [4],[5]. Moreover, BPD is frequently unrecognized and misdiagnosed, particularly in patients presenting with their first-episode of depression. These patients are often treated with inappropriate and costly regimens [6]-[9]. Thus, there is an urgent need to understand the pathophysiology of BPD in order to develop earlier diagnoses and more effective treatments [10]. Family, twins and epidemiological studies unequivocally demonstrate that BPD is a highly heritable disease with a heritability of more than 85%, and involves the interaction of multiple genes or more complex genetic mechanisms [11]-[13]. To date, association studies support a possible role for several candidate genes in BPD, including brain-derived neurotrophic factor (BDNF), but consistent direction of effects and alleles have not been established [14]. These inconsistent findings from previous genetic association studies may be related to variation in ascertainment, phenotype definition and control selection, limited power and possibly confounded by ethnic heterogeneity and population substructure [10],[15].

The hypothesis of neuronal plasticity involved in mood disorders has been supported by the use of antidepressants and mood stabilizers, e.g. lithium and valproate, inducing the expression of neurotrophins (e.g. BDNF) and synaptic changes [16],[17]. Moreover, BDNF gene has been implicated in the etiology of BPD by linkage studies [18]. Position 196 in exon 5 of the BDNF gene contains a G to A transition (dbSNP: rs6265) that results in an amino acid substitution (valine to methionine) at codon 66 in the precursor BDNF peptide sequence [19]. This change results in BDNF functional polymorphisms in this region as Val66Val, Val66Met, and Met66Met. Two previous family-based association studies found that this functional polymorphism Val66Met was significantly associated with the susceptibility to BPD [20],[21]. Following these two reports, a large number of association studies between BDNF gene polymorphisms and BPD have been published. Most of them specifically focused on the Val66Met polymorphism, but yielded conflicting results [10],[15],[22]. In view of the conflicting results, a meta-analysis on all original case–control association studies was performed to comprehensively determine the overall strength of associations between BPD and Val66Met polymorphism.

The process of identifying studied included in this meta-analysis is shown in Figure 1. Twenty-one case–control association studies met our criteria for the meta-analysis (Table 1). Data from four studies [26]-[29] were excluded due to the partial overlap with a larger sample size case–control study [30], and data from six studies were excluded due to less than 100 subjects in either patient group or control group [31]-[36]. Additionally, the case–control sample from a genome-wide association study was also excluded from current meta-analysis [37].

Overall, the data from 7219 BPD cases and 9832 healthy controls were analyzed. The mean genotype distribution in Caucasian and Oriental population was presented in Table 2. There was no significant difference in allelic distribution of Val66Met polymorphism between patients and controls. The pooled OR was 1.03 (95% CI: 0.98-1.08, Z = 1.00, P = 0.32) (Figure 2). Similarly, there was also no significant difference in allelic distribution of Val66Met polymorphism between patients and controls in Oriental population, with a pooled OR of 0.96 (95% CI: 0.89-1.05, Z = 0.82, P = 0.41) for Han Chinese population, 0.99 (95% CI: 0.85-1.15, Z = 0.12, P = 0.90) for Japanese population, and 1.06 (95% CI: 0.89-1.25, Z = 0.67, P = 0.50) for Korean population, respectively (Figure 2). However, there was a trend towards significant difference in Caucasian population with a pooled OR of 1.08 (95% CI: 1.00-1.16, Z = 1.97, P = 0.05) (Figure 2). The sensitivity analysis showed that the results were not being driven by any one specific study, and the funnel plot did not detect there was evidence of publication bias (Figure 3).

Among 21 included studies, six case–control association studies made a distinction for clinical phenotypes between bipolar I disorder (BP I) and bipolar II disorder (BP II), and two other studies only recruited BP I patients (Table 3). A further meta-analysis of the data from the aforementioned eight studies did not find a significant difference in allelic distribution of Val66Met polymorphism between BP I patients and healthy controls with a pooled OR of 1.00 (95% CI: 0.93-1.08, Z = 0.04, P = 0.97) (Figure 4). However, there was a significant difference in allelic distribution of Val66Met polymorphism between BP II patients and healthy controls with a pooled OR of 0.88 (95% CI: 0.78-0.99, Z = 2.20, P = 0.03) (Figure 4). A post-hoc analysis did not find a significant difference in allelic distribution of Val66Met polymorphism between BP I patients and BP II patients with a pooled OR of 1.10 (95% CI: 0.98-1.25, Z = 1.58, P = 0.12) (Figure 5).

With a total of 7,219 patients and 9,832 control cases, our meta-analysis included an additional 4,076 BPD cases and 3,485 healthy controls compared to a previous meta-analysis of case–control studies [22]. Similar to this previous meta-analysis, we did not find significant associations between the Val66Met polymorphism and BPD susceptibility either in a combined population or a subgroup of Oriental patients. However, there was a trend towards significant association in Caucasian population (fixed-effects pooled OR = 1.08, P = 0.05). Subgroup meta-analyses also showed that the Val allele may be a protective factor for BP II (fixed-effects pooled OR = 0.88, P = 0.03).

Our overall finding, that no convincing evidence for association between the Val66Met polymorphism and BPD as a whole, is consistent with a previous meta-analysis reported by Kanazawa et al. [22], which included 11 case–control designed studies up to February 2006. However, our overall finding is inconsistent with the result of a meta-analysis conducted by Fan et al. [15], which included all original case–control and family-based association studies published up to May 2007 and found a modest but statistically significant association between the Val allele and BPD susceptibility [15]. Both our analysis and the analysis of Kanazawa et al. [22] exclusively included case–control studies. In contrast, the analysis of Fan et al. [15] included five family-based association studies with a total of 858 families and a genome-wide association study with 1866 BPD patients and 2932 controls. The inclusion of different studies in these meta-analyses may explain the discrepancy among these three meta-analysis studies. It is well known that BPD is a highly inheritable disorder. Inclusion of family-based association studies in the meta-analysis of Fan et al. [15] might increase the probability of detecting difference between patients and controls.

The finding of insignificant association between the Val66Met polymorphism and BPD in Oriental population populations (Chinese, Japanese and Korean populations), but a trend towards significance in Caucasian population suggest that ethnic heterogeneity may affect the results of these genetic association studies. Fan et al. [15] reported that the allele frequencies of the Val66Met polymorphism in BDNF gene across individual studies and four HapMap populations (European, Chinese, Japanese and Yoruban populations) had significant global variations, which raised concerns of possible population stratification among case–control studies [15]. A more recent population genetic study found that there were substantial variations in BDNF coding regions and haplotype frequencies between 58 global populations with the Met allele of Val66Met ranged from 0-72% frequencies [58]. As previously pointed out, unless all study participants are from a homogenous ethnic group, the confounding effect from different ethnic groups is inevitable [10]. Moreover, control selection (e.g. healthy control or family control) might also result in the discrepancy of findings in these genetic association studies. A significant discrepancy between the pooled ORs from case–control studies and family-based studies raised a concern regarding a more generalized transmission distortion at this locus that is not disease related [15].

The finding of significant associations between Val66Met polymorphism and BP II suggests that the strength of association between Val66Met polymorphism and BPD could depend on the clinical phenotypes or subtypes of BPD such as with rapid cycling course, early onset, or substance comorbidity [10],[28],[30]-[32],[34],[50],[55],[59]. Interestingly, the Val allele may have opposite associations with disease susceptibility in different bipolar subtypes. The findings of two previous studies in Caucasian population (a case–control study and a family-based study) appeared that the Val allele was associated with an increased disease risk for rapid-cycling bipolar disorder (RCBD) [46],[60], but this meta-analysis showed a decreased risk for BP II (especially in Han Chinese population). However, another family-based association analysis in Caucasian population did not replicate a significant association between the Val66Met polymorphism and BPD or RCBD. Thus the discrepancy of findings in these studies may stem from the differences of clinical phenotypes, ethnic origin and control selection.

Several limitations of this meta-analysis should be considered. One important limitation is that we only investigated relatively well-studied polymorphic variations in BDNF gene. Another limitation is that family-based studies and genome-wide association studies were excluded due to the heterogeneity of research methods. In a previous meta-analysis, the pooled OR derived from nine case–control studies is nominally significant (random-effects pooled OR = 1.07, P = 0.04), while the pooled OR derived from five family-based studies increases notably to an OR of 1.54 (P = 0.000019). Finally, since we primarily designed the current analysis to demonstrate the potential association(s) between Val66Met polymorphism and bipolar diagnostic boundaries, the association of this polymorphism with other characteristics of BPD such as sex, onset age, comorbidity, impairment in brain morphology and function, or treatment response, were not explored.

Taken together, the Val66Met polymorphism in BDNF gene may be involved in the pathogenesis of BPD by influencing the susceptibility of specific subtypes such as BP II, but there is no compelling evidence of BDNF gene playing an important role in susceptibility to BPD across different ethnicities. The associations observed in current meta-analysis should be interpreted with caution. Further large-scale studies with same definitions of phenotypes and controls in homogenous ethnic groups are warranted to elucidate the relevance of BDNF gene variations as a risk factor for BPD (or diagnostic subtypes) susceptibility.