- Top of page
There is a strong etiological link between brain-derived neurotrophic factor and depression, but the neurocellular mechanisms and gene–environment interactions remain obscure. This study investigated whether one functional polymorphism in the brain-derived neurotrophic factor gene (BDNF Val66Met) modulates the influence of stressful life events on adolescent depressive symptoms. A total of 780 pairs of ethnic Han Chinese adolescent twins, 11–17 years of age, were randomly assigned to one of two subgroups (twin1 and twin2). All subjects were genotyped as Val/Val, Val/Met or Met/Met, and assessed for depressive symptoms using the Children's Depression Inventory. The level of environmental stress was estimated by the frequency of stressful life events using the Life Events Checklist. The frequency of stressful life events was significantly correlated with depressive symptoms (twin1: β = 0.21, P = 0.01; twin2: β = 0.27, P < 0.01), but there was no significant main effect of the BDNF Val66Met genotype on depressive symptoms. In both subgroups, however, the interaction between the BDNF Val66Met genotype and stressful life event frequency was significant (twin1: β = 0.19, P = 0.01; twin2: β = 0.15, P = 0.04); individuals with one or two Val alleles demonstrated a greater susceptibility to both the detrimental effects of higher stress and the beneficial effects of lower stress compared to the Met/Met genotype. These findings support the ‘differential-susceptibility’ hypothesis, whereby the BDNF Val allele modulates the influence of environmental stress on depression by enhancing the neuroplastic response to all life events.
Brain-derived neurotrophic factor (BDNF), a nerve growth factor supporting neuronal survival and synaptic plasticity (Cohen-Cory et al. 2010), has been strongly implicated in the pathophysiology of depression (Castrén 2010; Groves 2007). A single nucleotide polymorphism within the BDNF gene, which causes a valine-to-methionine substitution at codon 66 (Val66Met, rs6265) in the pro-domain of BDNF, has been shown to affect intracellular trafficking and activity-dependent secretion of BDNF (Egan et al. 2003). A meta-analysis of 14 associated studies, encompassing 2812 patients and 10 843 controls, revealed a modest association between the BDNF Val66Met polymorphism and major depressive disorder for males but not for females (Verhagen et al. 2010), while three other meta-analyses found no associations (Chen et al. 2008; Gratacòs et al. 2007; Lopez-Leon et al. 2008).
None of these studies, however, considered the interaction between stressful environmental factors or events and the BDNF Val66Met polymorphism, a confound that may account, at least in part, for these inconsistent findings. In fact, several empirical studies support such an interaction. Krishnan et al. (2007) found that Val/Val mice demonstrated significantly more depression-like behaviors than Met/Met mice when exposed to social defeat stress. Healthy adults with one or two Met alleles showed a significantly attenuated hypothalamic–pituitary–adrenal (HPA)-axis response to acute psychological stressors compared to adults with the Val/Val genotype (Alexander et al. 2010; Shalev et al. 2009). Several population-based studies also found a significant effect of the interaction between the BDNF Val66Met polymorphism and early adverse life events on adult depression (Aguilera et al. 2009; Carver et al. 2011; Gatt et al. 2009; Wichers et al. 2008).
The interaction of the BDNF Val66Met polymorphism with stressful environment factors can be explained by two theoretical frameworks, the diathesis-stress model (Caspi et al. 2003; Monroe & Simons 1991) and the differential-susceptibility hypothesis (Belsky et al. 2009). The former postulates that individuals with ‘vulnerability genes’ are more prone to psychopathology when they experience environmental adversity. The latter postulates that the interaction is due to ‘plasticity genes’ instead of ‘vulnerability genes’. That is, individuals with certain gene loci not only vary in the degree to which they are vulnerable to adverse environmental factors, but, more generally, in their developmental plasticity (Belsky & Pluess 2009). More ‘plastic’ individuals are more susceptible to both negative and positive environmental influences, while less susceptible individuals are less affected by changes in the environment.
The main aim of this study is to test whether the BDNF Val66Met polymorphism is associated with individual difference in vulnerability to stressful environmental factors, or associated with plasticity in response to changes in the environment among adolescents. The number of recent stressful life events (SLEs) was used to represent the level of environmental stress, from relatively positive (no SLEs) to relatively negative (several SLEs). If individuals with a particular genetic makeup demonstrated more sensitivity to changes of stress level, then the differential-susceptibility hypothesis was supported. Alternatively, if they were more affected only under higher stress levels, then the diathesis-stress theory was supported.
- Top of page
The descriptive statistics of the studied variables are shown in Table 1. The three genotype groups (Val/Val, Val/Met and Met/Met) were tested in Hardy–Weinberg equilibrium (twin1: χ2 = 0.74, df = 2, P > 0.10; twin2: χ2 = 1.10, df = 2, P > 0.10), and there was no significant sex difference in genotype frequency distributions (twin1: χ2 = 2.67, df = 2, P = 0.26; twin2: χ2 = 2.42, df = 2, P = 0.30). The genotype frequencies found in our study are consistent with those reported in other East Asian populations (Jia et al. 2011; Kim et al. 2010), but are distinct from Caucasians (Egan et al. 2003; Kaufman et al. 2006). Higher Met allele frequencies were observed in Asian populations than in Caucasian populations (Petryshen et al. 2009).
Table 1. Study variable means within genotypes
|Val/Val (n = 226)||Val/Met (n = 398)||Met/Met (n = 156)||Val/Val (n = 220)||Val/Met (n = 403)||Met/Met (n = 157)|
|Age in years (Mean ± SD)||13.55 ± 1.86||13.71 ± 1.81||13.55 ± 1.81||13.49 ± 1.82||13.76 ± 1.86||13.49 ± 1.75|
|CDI score (Mean ± SD)||36.98 ± 6.46||36.74 ± 6.79||36.82 ± 6.40||37.07 ± 6.88||37.38 ± 6.84||38.02 ± 6.52|
|Stressful life events (Mean ± SD)||2.66 ± 2.09||2.60 ± 2.13||2.58 ± 2.03||2.57 ± 2.07||2.63 ± 2.15||2.74 ± 2.12|
One-way anova revealed no significant main effect of BDNF genotype on depressive symptoms as measured by CDI scores (twin1: F = 0.09, df = 2, P = 0.92; twin 2: F = 0.88, df = 2, P = 0.41), suggesting that BDNF Val66Met genotype does not directly influence depressive symptoms. The main effect of BDNF genotype on SLE frequency was also not significant (twin1: F = 0.08, df = 2, P = 0.93; twin2: F = 0.31, df = 2, P = 0.73), suggesting no gene–environment correlation. Furthermore, results of t-tests revealed no significant differences in the mean depressive symptom scores between males and females (twin1: t = −1.5, P = 0.15; twin2: t = .30, P = 0.76).
We next examined whether the linear relationship between SLE number and depressive symptom differed significantly between BDNF genotypes. As shown in Table 2, the correlation coefficients in Val/Val and Val/Met groups were similar and both were larger than the correlation coefficient in the Met/Met group, so the Val/Val and Val/Met groups were combined into a Val+ group. After Fisher Z transformation, we found a significant differences between the Met/Met and Val+ group correlation coefficients for both subgroups (twin1: Zr = 2.44, P < 0.05, twin2: Zr = 2.31, P < 0.05).
Table 2. Person correlations between stressful life event (SLE) number and self-reported Children's Depression Inventory (CDI) scores in the different genotype groups
Finally, multiple linear regression analysis was conducted to examine the main effects and interactive effects of study variables. Sex, age and age squared were entered first, BDNF genotype and SLE numbers were entered as the second step, two-way interactions were entered as the third step, and the three-way interaction (sex × SLEs × BDNF genotype) was entered as the final step. Results showed that only the interaction between BDNF genotype and SLE frequency was significant. The other two-way interaction terms and the sex × SLE × BDNF genotype term were not significant and thus removed from the final models. The results of the final models are shown in Table 3. The main effects of age and age squared were significant in both subgroups. However, the main effect of sex on depressive symptoms was only observed in the twin1 subgroup. The main effect of SLE number was significant. Although the main effect of BDNF genotype on depressive symptom score was not significant, the interaction between BDNF genotype and SLE frequency was significant in both subgroups (twin1: β = 0.19, P = 0.01; twin2: β = 0.15, P = 0.04).
Table 3. The multiple linear regressions on self-reported depressive symptoms
|Constant||2.65||1.88|| ||0.00||3.99||1.95|| ||0.00|
|BDNF × Life stress||0.67||0.26||0.19||0.01||0.53||0.26||0.15||0.04|
For both BDNF genotype groups (Met/Met and Val+), depressive symptom scores increased with SLE frequency (stress level). However, depressive symptom scores increased more quickly with stress level in the Val+ group (twin1: β = 0.41, P < 0.01; twin2: β = 0.43, P < 0.01) than in the Met/Met group (twin1: β = 0.21, P < 0.01; twin2: β = 0.28, P < 0.01). We further compared the mean CDI scores of the two genotype groups in the low stress condition (0 or only 1 SLE) and the high stress condition (5 or more SLEs), respectively. In the low stress condition, Met/Met homozygotes demonstrated significantly higher depressive symptom scores than individuals with one or two Val alleles (twin1: t = 2.67, df = 264, P < 0.01; twin2: t = 2.00, df = 272, P < 0.05), while in the high stress condition, Val allele carriers exhibited significantly higher depressive symptom scores than the Met/Met group (twin1: t = −1.98, df = 140, P < 0.05; twin2: t = −2.05, df = 149, P < 0.05). Thus, the interaction of BDNF genotype with SLE frequency (stress level) appears to conform to the ‘differential-susceptibility’ hypothesis, whereby the BDNF Val allele confers higher sensitivity to changes in environmental stress (both increasing stress and reduced stress) rather than the diathesis-stress model, which predicts that the Val allele renders individuals more sensitive to an adverse environment (high stress) only (Figure 1).
Figure 1. Interaction effect of BDNF Val66Met polymorphism and stressful life events on adolescent depressive symptoms. Results from the two subgroups (twin1 and twin2) are presented separately.
Download figure to PowerPoint
To examine whether the results are confounded by other factors, such as family socioeconomic status and substance use, we entered perceived family economic status (1 = rich, 2 = average, 3 = below average, 4 = very poor), smoking (0 = smoke regularly, 1 = never smoked) and frequency of drinking (1 = every day, 2 = 2 to 6 times per week, 3 = once per week, 4 = 1 to 3 times per month, 5 = less than once per week, 6 = never drank) into the regression model as independent variables. Although significant main effects of family economic status (twin1: β = 0.07, P = 0.04; twin2: β = 0.12, P < 0.01) and smoking (twin1: β = −0.08, P = 0.03; twin2: β = −0.12, P < 0.01) were observed, the main effects of SLE frequency (stress level) and the interaction of BDNF × SLEs were still significant and the original model was retained for simplicity.
Genetics can influence stress exposure (Kendler & Baker 2007), for example, through effects on behavior. Thus, SLE frequency may reflect both uncontrollable environmental stressors and stressors that depend, at least in part, on individuals' behavior. To minimize the potential confounds of gene–environment correlation, the SLEs were further classified as independent (uncontrollable) stressors not likely related to behavior, such as ‘death of a family member’ or ‘severe family economic losses’, and dependent (controllable) stressors likely dependent on behavior, such as ‘trouble with teachers’ or ‘lack of friends’ (Brown & Harris 1978; Silberg et al. 1999). Taking advantage of the twin design, we estimated the genetic influence of these two types of SLEs. Data were fit into classic univariate ACE model for biometric analysis. This analysis revealed a low level genetic influence on dependent SLEs (heritability = 11%) but no genetic influence on independent SLEs (heritability = 0%), suggesting that the subset of independent SLEs can be regarded as purely environmental factors. We then examined the effect of the BDNF × independent SLE interaction on depressive symptoms by multiple linear regression analysis. After age, age squared, sex and dependent SLEs were controlled, the effect of BDNF genotype × independent SLEs on depressive symptoms remained significant (twin1: β = 0.15, P = 0.04; twin2: β = 0.19, P = 0.02), indicating that an interaction between BDNF genotype and purely environmental stressors influences depressive symptoms in Chinese adolescents.
- Top of page
The main goal of this study was to examine how SLE frequency, an estimate of life stress, and the BDNF Val66Met polymorphism interact to impact adolescent depression. Furthermore, we tested predictions of the diathesis-stress and differential-susceptibility theories to glean additional mechanistic insight into the effect of BDNF on depression and how BDNF may regulate the relationship between stress and depression. Statistical analysis revealed no significant relationship between the BDNF Val66Met polymorphism and adolescent depressive symptom score. Instead, the Val allele enhanced the correlation between SLE frequency and adolescent depressive symptoms, enhancing CDI score at higher SLE frequencies and reducing CDI score below that of Met/Met carriers at lower SLE frequencies. In other words, individuals with one or two Val alleles demonstrated higher susceptibility to changes in environmental stress than those having no Val alleles.
This study has several methodological strengths. First, it is based on a relatively large community sample of Chinese Han adolescents, which enabled us to investigate the gene (BDNF) × environment (SLE) interaction with good statistical power and to exclude the confound of population stratification. Second, by enrolling only twins and then randomly assigning them to the different subgroups (twin1 and twin2) we could replicate our results in one study. Third, instead of childhood adversity, we measured recent SLEs, which have been found to rise in frequency during adolescence and exert great impact on adolescent depressive symptoms (Ge et al. 1994; Grant et al. 2004), while being less affected by recall bias. Furthermore, measuring the effects of the BDNF Val66Met polymorphism may be particularly appropriate for an experimental design employing recent SLEs. Indeed, the Met allele significantly attenuated the activity-dependent form of BDNF secretion without affecting constitutive secretion (Egan et al. 2003).
To our knowledge, this is the first report of a modulatory role of the BDNF Val66Met polymorphism on the relationship between recent SLEs and depressive symptoms in an adolescent population. Our finding that the BDNF Val allele enhanced the depressive response to stress is consistent with several animal and human studies. For instance, Val/Val mice displayed depression-like behaviors, while Met/Met mice exhibited an unsusceptible phenotype, when exposed to chronic social stress (Krishnan et al. 2007). In two studies of healthy adult males, Val/Val carriers showed a significantly higher HPA-axis reactivity to psychological stressors compared to Met allele carriers (Alexander et al. 2010, Shalev et al. 2009). In addition, the Val/Val genotype has been linked to other depression-related traits, including neuroticism (Frustaci et al. 2008; Lang et al. 2005; Sen et al. 2003) and rumination (Hilt et al. 2007; Juhasz et al. 2011), and traits that increase vulnerability to depression especially under stressful conditions (Brown & Rosellini 2011; Kercher et al. 2009; Nolen-Hoeksema et al. 2008).
Our finding that the effect of the BDNF Val66Met × SLE interaction on CDI score conformed to the ‘differential-susceptibility’ model has important mechanistic and clinical implications. Individuals carrying the Val allele were more susceptible to both the deleterious effects of higher stress and the beneficial effects of lower stress, suggesting that BDNF may be a ‘plasticity gene’ instead of a ‘vulnerability gene’ for adolescent depression. Thus, our findings provide evidence supporting the differential-susceptibility model of gene–environment interactions (Belsky et al. 2009).
The exact neural mechanisms underlying the association between the specific BDNF Val66Met genotype and differential susceptibility to recent stress remain to be elucidated. One possibility is a change in BDNF signaling within the mesolimbic dopamine circuit. The pathway from the dopamine-producing ventral tegmental area (VTA) to the nucleus accumbens (NAc) (the VTA–NAc pathway) has been shown to be involved in rewarding behavior (Spanagel & Weiss 1999) and stress responses (Trainor 2011), as well as the pathophysiology of depression (Blood et al. 2010; Nestler & Carlezon 2006). In a social defeat paradigm, Berton et al. (2006) showed that intact BDNF signaling in the VTA–NAc pathway was necessary for establishing neural and behavioral plasticity in response to aversive social experiences. Furthermore, Val/Val mice had ~50% higher levels of BDNF protein in the NAc and displayed more depression-like symptoms after chronic social stress than Met/Met mice (Krishnan et al. 2007). In a human imaging study, Val/Val individuals exhibited stronger activation of reward circuitry in response to aversive stimuli (e.g. angry, fearful and sad faces) compared to Met allele carriers (Gasic et al. 2009). Thus, increased activity of the VTA–NAc circuit, which subsequently gives rise to increased activity-dependent release of BDNF, may mediate the association between the Val allele and stronger depressive reactions to negative experiences (Feder et al. 2009).
In addition to stronger depressive responses, the Val/Val genotype was also associated with reward-related behaviors, such as substance-related disorders (Gratacòs et al. 2007; Wojnar et al. 2009), drug dependence (Jia et al. 2011; Lotfipour et al. 2009) and novelty-seeking traits (Montag et al. 2010). Thus, the BDNF Val allele may enhance sensitivity to both aversive and rewarding stimuli by increasing neural plasticity within mesolimbic dopamine circuits. Individuals with the Val allele may also show greater sensitivity to reward, possibly compensating for the increased vulnerability to stress-related disorders like depression. This notion corresponds well with the ‘differential-susceptibility’ pattern of stress responses we observed in the different BDNF Val66Met genotypes.
In contrast to our findings, four previous studies found that adult BDNF Met allele carriers were more likely to develop depression or depressive symptoms when exposed to early adversity (Aguilera et al. 2009; Carver et al. 2011; Gatt et al. 2009; Wichers et al. 2008). The discrepancy may be due to differences in conceptualization of stress, age, study design or ethnicity of the sample population. However, these studies examined the long-term effects of severe early adversity, such as abuse, neglect or parental loss on adult depression, whereas we investigated the short-term influences of recent mild or moderate SLEs (e.g. dissolution of friendships, argument with parents, failure of exam) on adolescent depression. Different neurobiological mechanisms may underlie these two associations. We speculate that exposure to stress during adolescence might trigger symptoms of depression over a shorter incubation period by affecting the VTA–NAc pathway, whereas exposure to childhood adversity might predispose to depression by altering hippocampal development (Eamon et al. 2010; Rao et al. 2010). Accumulating evidence suggests that BDNF exerts antidepressant effects in hippocampus but has a potent pro-depressant effect in the VTA–NAc (Groves 2007; Martinowich et al. 2007). These paradoxical effects of BDNF in the VTA–NAc pathway and hippocampus may account for the discrepant effects of early severe stress and recent mild stress on depression; the BDNF Val allele, which mediates higher activity-dependent BDNF release, conferred an increased risk for depression in adolescents experiencing a greater number of SLEs by increasing VTA–NAc activity but conferred resilience to adult depression following early severe adversity by providing trophic support to the hippocampus. Our study and others (Casey et al. 2009, Lenroot & Giedd 2011) indicate that the interactive effects of genetic and environmental factors on brain and behavior may depend on the stage of development. Furthermore, given the relatively more common Met allele in the Asian population compared to various Caucasian populations (Petryshen et al. 2009; Verhagen et al. 2010), ethnicity may also play a role.
In Western populations, it is well established that girls showed more depressive symptoms than boys, especially from mid- and late-adolescence (Ge et al. 1994; Hankin et al. 2007; Wade et al. 2002). In our study, however, we found no significant sex differences in mean CDI scores. In multiple regression modes, after controlling for other variables, sex significantly predicted youth depression in only one subgroup (twin1). This discrepancy between relatively well-matched populations typifies the inconsistent results found when examining sex differences in depression symptoms in Chinese adolescents. Tepper et al. (2008) reported no sex differences while Greenberger et al. (2000) found more depressive symptoms in girls than in boys. In contrast, another study observed more depressive symptoms in boys than girls (Hong et al. 2009). Several researchers have speculated that poor coping strategies in boys may account for the absence of consistently higher depressive symptoms in Chinese girls, but the question of sex differences in depression among Chinese children and adolescents is still unresolved.
There are several possible limitations to this study. First, we used a cross-sectional design that measured an association between previous (but recent) SLEs and current depressive symptoms. Second, the measurements were solely based on self-reports, so the relations between study variables might be due partly to shared method variance. Third, we examined the effect of the BDNF × SLE interaction on depression within a nonclinical community sample of adolescents, and the CDI is not very sensitive to mild depressive symptoms. Thus, the significance of these findings to diagnosed clinical depression is unclear. Fourth, we regarded the absence of SLE or fewer SLEs as ‘a positive environment’ while studies directly measuring positive experiences, such as warm parenting and social support, may provide a better test of the differential-susceptibility hypothesis. Finally, the age range of the sample is wide, and symptoms of clinical depression are rare in children as young as 11. Cohort studies are warranted to replicate the main findings of this study.
In conclusion, this study suggested that the influence of recent SLEs on adolescent depressive symptoms was enhanced by the BDNF Val allele. Individuals with one or two Val alleles exhibited higher susceptibility to both the detrimental effects of ‘higher stress’ (a greater number of SLEs) and the beneficial effects of ‘lower stress’ (fewer or no SLEs). Further studies are needed to replicate our results in other ethnic groups and in clinical populations, as well as to elucidate the underlying neurocellular mechanisms.