Interactions between BDNF Val66Met polymorphism and early life stress predict brain and arousal pathways to syndromal depression and anxiety


Individual risk markers for depression and anxiety disorders have been identified but the explicit pathways that link genes and environment to these markers remain unknown. Here we examined the explicit interactions between the brain-derived neurotrophic factor (BDNF) Val66Met gene and early life stress (ELS) exposure in brain (amygdala–hippocampal–prefrontal gray matter volume), body (heart rate), temperament and cognition in 374 healthy European volunteers assessed for depression and anxiety symptoms. Brain imaging data were based on a subset of 89 participants. Multiple regression analysis revealed main effects of ELS for body arousal (resting heart rate, P=0.005) and symptoms (depression and anxiety, P<0.001) in the absence of main effects for BDNF. In addition, significant BDNF–ELS interactions indicated that BDNF Met carriers exposed to greater ELS have smaller hippocampal and amygdala volumes (P=0.013), heart rate elevations (P=0.0002) and a decline in working memory (P=0.022). Structural equation path modeling was used to determine if this interaction predicts anxiety and depression by mediating effects on the brain, body and cognitive measures. The combination of Met carrier status and exposure to ELS predicted reduced gray matter in hippocampus (P<0.001), and associated lateral prefrontal cortex (P<0.001) and, in turn, higher depression (P=0.005). Higher depression was associated with poorer working memory (P=0.005), and slowed response speed. The BDNF Met–ELS interaction also predicted elevated neuroticism and higher depression and anxiety by elevations in body arousal (P<0.001). In contrast, the combination of BDNF V/V genotype and ELS predicted increases in gray matter of the amygdala (P=0.003) and associated medial prefrontal cortex (P<0.001), which in turn predicted startle-elicited heart rate variability (P=0.026) and higher anxiety (P=0.026). Higher anxiety was linked to verbal memory, and to impulsivity. These effects were specific to the BDNF gene and were not evident for the related 5HTT-LPR polymorphism. Overall, these findings are consistent with the correlation of depression and anxiety, yet suggest that partially differentiated gene–brain cognition pathways to these syndromes can be identified, even in a nonclinical sample. Such findings may aid establishing an evidence base for more tailored intervention strategies.


Depression will be the second leading contributor to the burden of disease by 2020 ( There is therefore an urgent need to identify objective indicators of risk for depression, and to understand their link to underlying biological mechanisms and treatment.1, 2 To date, candidate risk markers such as genetic polymorphisms, vulnerability to stress, personality traits, and alterations in brain function, structure and cognition have tended to be assessed in separate studies.3 The integrative relationships between these factors, and how they may interact to predict depression and anxiety, remain unknown. Here, we used path modeling to test an integrative model of risk for depression, based on integrative neuroscience.4, 5, 6, 7, 8 Given the overlap in depression and anxiety, we drew on behavioral models that identify distinct dimensions of syndromal depression and anxiety, along with a common dimension of general distress.9, 10, 11

There is overwhelming evidence that depression can be precipitated by stressful life events12, 13, 14 particularly when exposure to stress is early in life. The influence of adverse events on depression is clearly moderated by genetic vulnerability. Much attention has been given to the role polymorphisms of the serotonin transporter (5HTT-LPR) in increasing (S/S genotype) or decreasing (L/L genotype) the depressogenic effects of life stress.15 These several observations by Caspi et al.15 have been largely confirmed in several subsequent studies16, 17 and are also congruent with functional brain imaging studies that have demonstrated reduced 5HT binding in depressed patients.18 However, other genes that exhibit polymorphisms such as the CRF-1 receptor19 and brain-derived neurotrophic factor (BDNF)20 have also been implicated in clinical depression, and risk for depression in healthy individuals.4, 5 Indeed, meta-analytical studies have questioned the strength of this 5HTT-LPR polymorphism finding, suggesting that alterations in serotonergic activity may actually reflect the beginning of more gradual plasticity changes, with the underlying mechanism being BDNF.20, 21 In line with this view, one study has suggested that the BDNF Met allele may have protective effects against the adverse neural effects of the 5HTT-LPR Short allele in healthy individuals.22 In contrast, when adverse environment is considered, the combination of BDNF Mets with 5HTT-LPR Short alleles actually increases risk for depression in maltreated children,23 highlighting the importance of early life trauma in depression/anxiety pathways. Indeed, association studies have observed a significant excess of the Methionine (Met) allele of BDNF Val66Met polymorphism in depression, including geriatric depression.24, 25 Although the Met allele of this polymorphism has also been associated with anxiety in humans24 and mice,26 at least one study has suggested the Val allele instead has a function in anxiety.27

Depression has been associated with heightened sensitivity on physiological measures of arousal and temperament, consistent with a biologically mediated vulnerability to stress. For instance, alterations in heart rate in both resting and activation (including startle) conditions characterize both depression and anxiety disorders,28, 29, 30 and poor emotion-related heart rate regulation is a promising predictor of antidepressant response.31 Similar alterations are observed in syndromal depression and anxiety in otherwise healthy individuals.32, 33 Notably, elevated heart rate is a key predictor of subsequent depression and anxiety disorder following exposure to a traumatic stressor.34 Likewise, the temperamental trait of neuroticism, which reflects emotional instability and stress sensitivity, is a highly heritable and robust indicator of genetic vulnerability for both depression and anxiety.13, 35

Excessive and/or repeated exposure to stress may contribute to neuronal loss in key regions of the limbic system; the hippocampus and amygdala.36 The dynamic interplay between the amygdala and hippocampus, and their projections to both the prefrontal cortex and autonomic arousal systems, is crucial for effective emotional function and memory of significant information.6, 7 Disruptions to emotion regulation and memory are defining features of both depression and anxiety disorders,37, 38 and these disorders have in turn been associated with alterations in amygdala and hippocampal volume, and of the related prefrontal regions.3, 36, 38, 39, 40, 41, 42, 43 Moreover, twin studies show that smaller hippocampal volume is implicated in genetic risk for these disorders in healthy individuals, particularly when coupled with exposure to stress.44, 45

BDNF has a direct impact on neuronal growth and plasticity in frontohippocampal46, 47 and amygdala48 networks, and reduced serum concentrations of BDNF have been reported in major depression.49, 50, 51 The BDNF (Val66Met) Met variant is associated with a functional alteration; that is, a decrease in activity-dependent secretion of BDNF.52 Met allele carriers show comparatively lower hippocampal gray matter53, 54 and poorer cognitive performance.2, 3, 52, 54 It has also been shown that, in otherwise healthy individuals with the BDNF Met variant, low emotional stability (higher neuroticism) with higher syndromal depression is associated with lower hippocampal volume.55 Animal evidence indicates that stress exacerbates the effects of reduced BDNF on both hippocampal networks and autonomic arousal. For instance, stress in laboratory animals reduces BDNF synthesis, resulting in atrophy of CA3 pyramidal hippocampal neurons.56 Further, BDNF knockout mice demonstrate impaired arousal responses57 and increased anxiety-related behavior.26

If our understanding of gene–environment interactions and their contribution to risk for depression and anxiety is to be improved beyond the phenomenological level, it is imperative that we understand these interactions in relation to their impact on brain structure and function.36, 58 The interaction of BDNF Val66Met and stress and its impact on brain, arousal and cognition markers in risk for depression has not been examined. In this study, we used structural equation modeling (SEM) to evaluate the explicit pathways by which this BDNF gene–stress interaction impacts the brain and arousal and, in turn, predicts the experience of depression and anxiety symptoms (Figure 1). The specificity of these effects to the BDNF gene was then confirmed by replicating the final model but with the 5HTT-LPR variants. An advantage of SEM over more commonly used multivariate methods (such as factor analysis) is that it provides a powerful means to examine both direct and mediated relationships between measures within one model.59

Figure 1

Graphic depiction of the hypothesized conceptual model by which the interaction of brain-derived neurotrophic factor (BDNF) Val66Met variants and exposure to early life stress (gene–environment interaction) predicts depression and anxiety and associate cognitive features (phenotype) through its impact on gray matter measures of brain structure and heart rate measures of body arousal (endophenotypes). For details of the specific measures included in the model, see Supplementary Figure 1.

In this investigation we drew upon both behavioral9 and neurobiological6, 7, 19 frameworks for understanding anxiety and depression to test an integrative model in which gene–stress interactions impact the brain, arousal and neurocognitive mechanisms of risk for depression or anxiety. The complete hypothesized model (shown in Supplementary Figure 1A) was tested using the markers for depression and anxiety summarized above. In this model, it was hypothesized that the interaction of the BDNF Met allele and early life stress (ELS) predicts symptom and cognitive features of depression and anxiety, by reductions in limbic–prefrontal gray matter volume and subsequent alterations in body arousal (assessed by autonomic measures of heart rate), which in turn reduce emotional stability (reflected in higher neuroticism). Heritability estimates were also used to specify order of effects given the cross-sectional nature of the data, with more heritable (or stable) measures ordered causally before less heritable measures (therefore more variable and susceptible to influence). Thus, the predicted sequence in which genetic variation has the strongest effect on gray matter followed by body arousal and then neuroticism is consistent with the greater heritability estimates for brain structure (for example, 75% for right frontal volume)60 compared to stress-related body arousal (51%)61 and neuroticism (40–50%).62

Three levels of analyses were used to test the model. First, we evaluated BDNF Val66Met variants and ELS in regard to their relationship with each of the individual brain, arousal and behavioral measures, with reference to published findings. Second, the interaction of BDNF variants and ELS was examined in relation to each of these individual measures. Together, these findings provided a framework for the third level of analysis. At the third level, SEM was used to evaluate the explicit pathways by which BDNF variants and ELS interact to produce effects on brain structure (gray matter), body arousal (heart rate), emotional stability (neuroticism), and in turn predict alterations in the symptoms and cognitive features of depression and anxiety. The specificity of these effects to BDNF was verified by retesting the model with the 5HTT-LPR polymorphism.

Materials and methods


An initial pool of 374 healthy Caucasian subjects of self-reported (as ancestry-informative markers were not typed (due to insufficient DNA), geographical origins of the ancestors of participants could not be confirmed) European ancestry (190 female, mean age of 36.2±12.7 years, mean education of 14.4±2.7 years) participated in collaboration with the Brain Resource International Database (BRID; Gordon et al.8; see Supplementary Table 2 for sample characteristics and Supplementary Material for power calculations). A standardized battery of assessments including items for Axis 1 symptom screen derived from the patient health questionnaire, AUDIT (WHO Alcohol Use Disorders Identification Test, WHO63) and the Fagerstrom Tobacco Dependency Questionnaire64 was used to assess exclusion criteria, including Axis I symptoms, physical brain injury (causing loss of consciousness for 10 min or more), neurological disorder and other serious medical or genetic conditions, and history of harmful drug or alcohol use. For all subjects who reported infrequent alcohol/drug use, the most recent use occurred at least a week before testing. On average, there was less than 10% missing data. A subset of 89 participants completed all assessments, including magnetic resonance imaging (MRI). This subset did not differ from the original participant pool in age, sex or genotype distribution, or descriptive characteristics on each of the candidate markers (see Supplementary Table 2 for details).

Informed written consent was provided in accordance with human research ethical requirements.

BDNF Val66Met genotypes

DNA was extracted from cheek swab samples and BDNF Val66Met genotypes were determined as described previously.3 The genotype frequencies were 64.7% V/V (n=242), 31.3% V/M (n=117) and 4% M/M (n=15), which were in Hardy–Weinberg equilibrium (χ2=0.03, P=0.862). These genotype frequencies conform to expected population rates for the BDNF Val66Met genotype and are similar to previously reported distributions.24, 27 Owing to limited M/M subjects, M/M and V/M subjects were combined to form the ‘Met allele carriers’ group (n=132). V/V and Met allele carriers did not differ in mean age (t=0.914, P=0.361; V/V=36.61±12.37, Met=35.35±13.22), gender (χ2=0.438, P=0.508; V/V: M/F=116/126, Met: M/F=68/64), years of education (t=−0.151, P=0.880; V/V=14.34±2.69, Met=14.39±2.60), body mass index (t=0.893, P=0.373) or physical activity levels (χ2=0.008, P=0.927). The BDNF groups did not differ in serotonin transporter promoter polymorphism (5HTT-LPR) distribution (χ2=1.521, d.f.=1, P=0.218). The subset of 89 participants who had complete data on all measures had an equivalent distribution of BDNF variants (V/V, n=63; Met, n=26).

Early life stress

Early life stress was measured using a 19-item Early Life Stress Questionnaire, which assesses the occurrence of specific early life stressors up to 18 years shown to have a psychological impact in childhood, including abuse, neglect, family conflict, illness/death and natural disasters.65 This scale is based on the Child Abuse and Trauma Scale,65 which has been shown to have strong reliability and validity, and correlates with adult outcome and psychopathology.

Syndromal depression and anxiety symptoms

Depressed mood and anxiety were assessed using the DASS-21 (a short form of the Depression Anxiety Stress Scales11), which assesses the severity of core symptoms of depression and anxiety. This scale is a psychometrically sound measure of trait depression in nonclinical and clinical populations, and has established norms that include Australian, US and UK populations.11, 65 Total scores were doubled for comparison with DASS-42 profiling. For depressed mood and anxiety, scores of 0–9 and 0–7 are considered ‘normal’, 10–13 and 8–9 ‘mild’, 14–20 and 10–14 ‘moderate’, 21–27 and 15–19 ‘severe’, 28+ and 20+ ‘extremely severe’.


Emotional stability was assessed using the Neuroticism scale from the 60-item NEO-Five Factor Inventory, a short form of the NEO-PI.66


The computerized and standardized ‘IntegNeuro’ battery was used to assess domains of cognitive function (see Supplementary Table 1 for the tests included in this battery). The IntegNeuro battery has sound psychometric properties in terms of reliability, validity, cross-cultural consistency, and age, sex and education norms.67, 68, 69 Core cognitive domains were established previously using factor analytic techniques: information processing speed, verbal memory, response speed (sensori-motor-function), sustained attention, working memory, verbal processing and executive function70 (see Supplementary Table 3 for details of factor loadings and cognitive measures defining each domain). Factor scores were inverted (where necessary) such that a high score reflected superior performance (that is, fewer errors and faster reaction time).

Autonomic arousal: heart rate

Average heart rate (beats per minute, b.p.m.) and heart rate variability (HRV) were recorded during both resting conditions and activation tasks using an electrocardiogram recording electrode positioned on the inner left wrist at the radial pulse. These conditions and tasks were designed to elicit increasing levels of arousal; from resting through to presentation of facial emotion and startle stimuli (see Supplementary Table 1).

Structural magnetic resonance imaging

Structural MRI was undertaken using a 1.5 T Siemens Vision Plus and Siemens Sonata systems (Siemens, Erlangen, Germany). High-resolution T1-weighted images were acquired using a three-dimensional magnetization-prepared rapid acquisition gradient echo sequence in the sagittal plane, with 180 slices, 1 mm cubic voxels, 256 × 256 matrix, repetition time (TR)=9.7 ms, echo time (TE)=4 ms, TI=200 ms and flip angle=12°.

Segmentation and spatial normalization of MRI data was performed using voxel-based morphometry in SPM2 (, using a protocol described in detail previously.71, 72 Images were spatially normalized by transforming each brain to a standardized stereotactic space based on the ICBM 152 template (Montreal Neurological Institute). Images were segmented into gray, white, CSF and non-brain portions based on a cluster analysis method to separate pixels based on intensity differences, together with a priori knowledge of spatial tissue distribution patterns in normal subjects.73 Customized templates created from the BRID subjects were used for normalization and segmentation processes.72 A correction was made to preserve quantitative tissue volumes following the normalization procedure.71

Regions of interest for volumetric analysis were those implicated in depression and anxiety, and were defined using a standardized neuroanatomical atlas:74 prefrontal cortex (lateral and medial portions), hippocampus and amygdala. We defined the medial prefrontal cortex (MPFC) as including the anterior cingulate (Brodmann areas, BA24/32), and medial orbital to superior frontal structures (extending to BA9/10). The lateral prefrontal cortex (LPFC) included the superior frontal gyrus, limited by the superior frontal sulcus externally, the middle frontal gyrus, limited caudally by the precentral sulcus, ventrally by the middle frontal sulcus and dorsally by the superior frontal sulcus, and the inferior frontal gyrus, limited dorsally by the inferior frontal sulcus and rostrally by the precentral sulcus.74

Statistical analysis

Multiple regression analysis was first used to evaluate the main effects of BDNF and ELS on candidate markers and syndromal depression and anxiety. Second, regression analyses were undertaken to evaluate the interaction effects of BDNF and ELS on these measures. In addition, regression analysis was used to evaluate the relationships between candidate markers. Together, these analyses provided a framework for the subsequent structural equation model.

Analyses of main effects for BDNF and ELS included age and sex as covariates (n=374 for body and arousal measures, n=89 for gray matter volume measures), and used a regression equation of the form: dependent measure=b0+b1(BDNF or ELS)+b2(Sex)+b3(Age); where b0 is the intercept, b1 the (unstandardized) BDNF or ELS coefficient (BDNF: 1=V/V, 2=V/M or M/M. Because the BDNF polymorphism was coded as a binary variable, it is more readily interpretable as a categorical variable, equivalent to the interpretation of sex; that is, the categories of female and male, rather than a continuum of ‘maleness’ or ‘femaleness’.); ELS: 0=no ELS, 1=1 ELS, 2=2 ELS, 3=3 ELS, 4=4 ELS, 5=5 or more ELS. Initial curvilinear regression analyses of ELS on all variables confirmed strongest effects for linear trends over and above curvilinear trends.), b2 the coefficient for sex (1=female, 2=male) and b3 the age coefficient (years). Multiple regression analysis was then used to examine the effects of the interaction of BDNF Val66Met and ELS on brain and arousal markers, and syndromal depression and anxiety. The regression equation took the form: dependent measure=b0+b1(BDNF)+b2(Sex)+b3(Age)+b4(ELS)+b5(BDNF × ELS); where b0 is the intercept, b1 the (unstandardized) BDNF coefficient (1=V/V, 2=V/M or M/M), b2 the coefficient for sex (1=female, 2=male), b3 the age coefficient (years), b4 the ELS coefficient (0=no ELS, 1=1 ELS, 2=2 ELS, 3=3 ELS, 4=4 ELS, 5=5 or more ELS) and b5 the product of BDNF and ELS. To control for multiple comparisons, an α level of 0.01 was used for regression analyses examining effects of gene or environment. For regression analyses examining relationships between brain and arousal markers, an α level of 0.05 was adopted given minimal number of tests per predictor were performed.

Structural equation modeling with maximum likelihood estimation (in AMOS 575) was then used to test the hypothesized path model in Supplementary Figure 1 (n=89; see Supplementary Materials for details on SEM method). SEM tested the prediction that the interaction of BDNF polymorphism and ELS produces effects on limbic (hippocampus, amygdala) circuitry and the prefrontal regions that regulate this circuitry, with subsequent alterations in body arousal and elevations in neuroticism, which in turn produce increases in syndromal depression and anxiety and related cognitive features.20 In the path model, candidate markers were ordered according to the hypothesized integrative model, drawing on previous behavioral9 and neurobiological6, 7 frameworks, heritability estimates and the published literature it draws on. The order of measures may be summarized according to three levels:

First level ‘gene–environment’ variables included BDNF (V/V vs Met), ELS and their interaction (BDNF × ELS). Age and sex were included in this level to control for any covariation due to these constitutional variables. First-level measures were treated as ‘exogenous’ variables (not caused by other variables in the model), whereas those at subsequent levels were treated as ‘endogenous’ variables (caused by one or more variables in the model).

Second level ‘endophenotype’ variables comprised (1) brain structure (right hippocampal, amygdala, LPFC and MPFC volume) and (2) autonomic arousal (startle-elicited heart rate and HRV), verified in initial multiple regression analyses (see Results). The order of these measures drew on both theory6, 7 and heritability estimates.60, 61

Third level ‘phenotypic’ variables first included the measure of neurotic temperament, associated with emotional stability and stress sensitivity. The outcome measures were the severity of symptoms of depression and anxiety, and associated cognitive features. Six of the cognitive domains were represented by two latent factors: ‘working memory’ (encompassing domains of working memory accuracy, memory span and sustained attention) and ‘executive cognitions’ (defined by information processing speed, executive function and verbal processing), whereas the remaining factor formed its own ‘verbal memory’ domain (defined by items measuring verbal recall, memory recognition, delayed recall; Supplementary Tables 3, 4; Supplementary Figure 1). This model was determined in a subanalysis before calculation of the full model.

Directed arcs (that is, single-headed arrows) were fitted to the model to examine the direct effects of the first-level variables onto the second- and third-level variables, and the second-level variables onto the third-level variables as shown in Supplementary Figure 1A. Correlated disturbances (that is, double-headed arrows) were also fitted to account for random variation: (1) between BDNF and BDNF × ELS, and between ELS and BDNF × ELS; (2) between the four brain structure measures (hippocampal, amygdala, lateral and medial prefrontal cortical volume) and (3) between depressed mood and anxiety. A correlated disturbance was not fitted to account for random variation between the two heart rate measures as initial bivariate correlations suggested these two variables were not significantly correlated. Residual variances were estimated for all endogenous variables (a variable caused by one or more variables in the model).

Overall model fit was evaluated by observing the following goodness-of-fit (GOF) indices: the χ2-statistic (χ2 with good fit indicated by a P-value >0.05), the root mean square error of approximation (RMSEA with good fit indicated by an index <0.05 and a P-value >0.50), the comparative fit index (CFI with good fit indicated by an index >0.95) and the parsimony goodness-of-fit index (PGFI with good fit indicated by an index >0.5076) to ensure the model was not too complex and therefore replicable in another sample. Unstandardized (b) path coefficients were observed. When overall model fit was poor, model respecifications were made by removing nonsignificant directed arcs or correlated disturbances (P>0.05), and adding directed arcs or correlated paths as indicated by modification indices that were consistent with hypotheses. As the models were nested (that is, the models’ parameters were subsets of one another76), the statistically significant improvements between models could be assessed using the likelihood ratio (calculated as the difference in χ2 and d.f. between the models of interest77). Model respecifications were undertaken to determine satisfactory model fit with the BDNF gene. To establish mediation, indirect paths were tested for significance using the Sobel test of estimated standard error.78

Three models (referred to as confirmation models) were examined to confirm the validity of the final model. First, to examine the specificity to BDNF, the final model was retested with 5HTT-LPR (abbreviated as ‘5HTT’) variants (LL vs SL/SS) replacing BDNF variants, and the computed interaction of ‘5HTT × ELS’ replacing the interaction of BDNF × ELS. Second, given the arguably similar heritability estimates of body arousal (51%) and neuroticism (40–50%), we tested an alternative path model in which the direction of effect was brain–body arousal–neuroticism, rather than brain–neuroticism–body arousal. Third, we further examined the validity of the predicted phenotypes of depression and anxiety, by determining if the pathways predicting these phenotypes were also distinguished by other behavioral indices of depression vs anxiety, established in the literature.


Exposure to early life stress

The top five ELS events reported in the current sample were family conflict (25.2%), divorce of parents (22.1%), being bullied at school (19.1%), a family member's life being threatened (16.9%) and separation from kin (15.6%), with the mean number of reported ELS events being 2.16±2.23 events. BDNF genotypes did not differ in number of reported ELS events (χ2=2.771, d.f.=5, P=0.735; V/V: 2.18±2.22, Met: 2.12±2.24).

Multiple regression: main effects of BDNF Val66Met and early life stress

Consistent with previous findings, greater exposure to ELS predicted an elevation in symptoms of depression (unstandardized b=0.745, t=4.50, P<0.001) and anxiety (b=0.335, t=3.52, P<0.001), along with a reduction in associated cognitive performance for information processing, within the executive cognitions domain (b=−0.080, t=−2.16, P=0.031), and memory span (b=−0.110, t=−3.25, P=0.001) and sustained attention (b=−0.100, t=−2.16, P=0.032), within the working memory domain.

Greater ELS also predicted an elevation in arousal-related candidate markers, including the temperamental trait of neuroticism (b=0.762, t=3.09, P=0.002) and mean heart rate during resting (eyes open: b=0.906, t=2.74, P=0.006; eyes closed: b=0.927, t=2.82, P=0.005), and activation tasks designed to increase arousal (facial emotion perception: b=0.948, t=2.72, P=0.007; working memory: b=1.281, t=3.63, P<0.001; auditory oddball: b=0.986, t=2.93, P=0.004; startle: b=0.906, t=2.09, P=0.038). With greater ELS, HRV was reduced during the executive maze task (b=−1.060, t=−3.07, P=0.002).

BDNF Val66Met status, by contrast, predicted only one effect, for HRV during the startle task for Met allele status (b=1.634, t=1.98, P=0.049). However, this effect was not significant (P=0.075) when BDNF genotypes were compared using t-tests without the confounds age and sex included (Supplementary Table 5). Consistent with the focus of the study, a greater number of effects were apparent for the interaction of BDNF status and ELS, as outlined in the following section.

Multiple regression: interaction effects of BDNF Val66Met variants and early life stress

When the interaction of BDNF Val66Met status with ELS was considered, there was an even more robust prediction of cognitive features associated with depression and anxiety. Specifically, the interaction of Met carrier status and exposure to at least two ELS events predicted poorer working memory performance in terms of accuracy (unstandardized b=−0.156, t=−2.23, P=0.026) and reaction time (b=−0.164, t=−2.30, P=0.022) (see Figure 2a for fitted regression plot and Supplementary Figure 3A for scatter plots).

Figure 2

Significant effects for the interaction of brain-derived neurotrophic factor (BDNF) Val66Met polymorphism (BDNF) with early life stress (ELS). Fitted regression plots based on estimated marginal means from multiple regression analyses showing significant interactions between BDNF Val66Met polymorphism and number of ELS events (correcting for main effects of BDNF and ELS, age and gender) with regard to (a) working memory accuracy and reaction time (n=374); (b) right hippocampal and amygdala gray matter volume (ml) (n=89) and (c) average heart rate (beats per minute) during resting (eyes closed, with similar effects evident during the eyes open condition), startle, facial emotion perception, and tests of working memory, executive function (executive maze task) and selective attention (auditory oddball task) (n=374). The R2 and P-values for the full model (main and interaction effects of BDNF and ELS, and covarying effects of age and gender) are indicated in each figure. For corresponding scatter plots, see Supplementary Figure 3.

Similarly, the interaction of ELS with BDNF Val66Met status predicted more pronounced effects on heart rate, especially for the most arousing (startle) task (b=3.494, t=3.776, P=0.0002), but also across resting conditions (eyes open: b=1.582, t=2.287, P=0.023; eyes closed: b=1.912, t=2.768, P=0.006) and other activation tasks (facial emotion perception: b=1.563, t=2.095, P=0.037; working memory: b=1.828, t=2.480, P=0.014; executive maze: b=2.088, t=2.304, P=0.022; oddball: b=1.873, t=2.644, P=0.009) (Figure 2c; Supplementary Figure 3C). Averaging across the conditions, the interaction of Met carrier status and exposure to a high level of ELS (five or more events) predicted a mean increase that was 12.0 b.p.m. higher than in those without exposure to ELS (Figure 2). In contrast, a mean increase of only 1.3 b.p.m. was predicted for V/V subjects with the same level of stress exposure.

The interaction of BDNF Val66Met status and ELS also predicted an impact on brain structure; a reduction in gray matter for the right hippocampus (b=−0.115, t=−2.528, P=0.013) (A similar interaction effect was found for the left hippocampus at weak trend level (P=0.089).) and right amygdala (b=−0.034, t=−2.533, P=0.013) was comparatively greater for Met carriers with higher exposure to ELS (Figure 2b; Supplementary Figure 3B).

Multiple regression: relationships between candidate brain-arousal markers

Consistent with the literature, both depression (b=0.365, t=9.36, P<0.001) and anxiety (b=0.152, t=6.64, P<0.001) were predicted by higher neuroticism (lower emotional stability). Moreover, anxiety predicted poorer cognitive performance characteristic of depressive and anxiety disorders; including information processing (b=−0.049, t=−2.02, P=0.044) and executive function (b=−0.076, t=−2.54, P=0.012), within the executive cognitions domain, and verbal memory (b=−0.072, t=−3.17, P=0.002) and memory span (b=−0.054, t=−2.81, P=0.005), within the working memory domain.

In accord with the predictions of the model, in which gray matter alterations in the limbic-prefrontal circuits that regulate emotional arousal will in turn produce alterations in autonomic arousal, lower right hippocampal gray matter volume was found to predict higher mean heart rate for the startle task (b=−11.342, t=−2.38, P=0.021). To a lesser extent, lower left medial prefrontal volume also predicted higher startle-elicited heart rate (b=−2.063, t=−2.04, P=0.046).

Structural equation modeling: BDNF Val66Met–early life stress interaction and brain–arousal pathways predict depression and anxiety

Drawing on both published studies and the above findings from multiple regression analyses, SEM focused on the interaction of BDNF Val66Met (referred to hereafter as ‘BDNF’) and ELS. SEM was undertaken with those candidate brain and arousal markers shown to be predicted most robustly by BDNF variants and ELS (and outlined in Supplementary Figure 1). These were gray matter for the right-sided hippocampus, amygdala and prefrontal regions, and startle-elicited heart rate measures (although we note that there were correlations of greater than 0.9 between right and left hemispheres for gray matter, and between startle and other tasks for heart rate measures, indicating that the model is generalizable to these measures).

Table 1 describes the GOF indices for the initial base model and subsequent respecifications. The initial base model (Table 1, 1i) tested the hypothesized pathways, with directed paths (or arcs) fitted in one direction—top to bottom (see Supplementary Figure 1A for details). Directed paths were fitted specifically to represent the prediction that hippocampal–LPFC networks would predict depression (and not anxiety), whereas amygdala–MPFC networks would predict anxiety (and not depression). At the phenotypic level, relationships between symptoms and cognition were left free to vary as it was difficult to decipher the direction of effects at this level in a nonclinical sample. The second model (Table 1, 1ii) used modification indices to test the addition of six paths to the initial model to account for these phenotypic relationships. Two of these paths included a correlation between depression and working memory accuracy, and a directed arc between anxiety and verbal memory. The addition of these six paths resulted in a significant improvement in model fit (significant likelihood ratio, P<0.001). Model three (Table 1, 1iii) tested the improvement in model fit once nonsignificant paths were removed. Removal of these paths reduced the complexity of the model (PGFI increased from 0.393 to 0.615). The fourth model (Table 1, 1iv) used modification indices to test the addition of a correlated path between LPFC and depression, which further improved the fit of the model (significant likelihood ratio, P<0.01). In the final model (Table 1, 1v), remaining nonsignificant paths were removed, and model fit was further improved (PGFI further increased from 0.616 to 0.626). The overall model fit of this final BDNF model was satisfactory (χ2=193.23, P>0.05).

Table 1 GOF indices for structural equation models: hypothesized BDNF Val66Met model and confirmatory models

In terms of variance accounted for (R2), the final respecified model (Table 1; 1v) accounted for 17.9% of depression symptoms and 13.3% of anxiety symptoms, and these symptoms were correlated with poorer cognitive performance on working and verbal memory domains, respectively. These behavioral outcomes were predicted by the impact of BDNF Val66Met and its interaction with ELS on gray matter, heart rate and neuroticism. For brain structure, a large amount of variance for hippocampal (38.5%), amygdala (52.2%), and related LPFC (45.9%) and MPFC (34.4%) cortical regions was accounted for. Effects on heart rate (11.5% for mean b.p.m., 5.3% for variability for startle task) and neuroticism (5.7%) were smaller yet important in the mediation of depression and anxiety.

The significant pathways of interest in the final BDNF model are shown in Figure 3 (see Supplementary Figure 2 and Supplementary Table 4 for further details). The interaction of BDNF Met status with exposure to ELS contributed to distinctive brain–arousal pathways to syndromal depression and anxiety. Elevated depression was predicted by two parallel systems. The interaction of BDNF Met status and ELS directly predicted reduced hippocampal volume (b=−0.11, P<0.001; Figure 3A). This reduction was correlated with reductions in LPFC volume (r=0.68, P<0.001; Figure 3B). Loss of hippocampal–LPFC volume was associated with greater depressed mood (r=−0.09, P=0.005; Figure 3C), which was associated in turn with reductions in working memory accuracy (r=−0.75, P=0.005; Figure 3G). Second, the interaction of BDNF–ELS predicted increases in startle-elicited heart rate (b=4.59, P<0.001; Figure 3D), which in turn predicted increased neuroticism (b=0.19, P=0.019; Figure 3E). Consistent with a multidimensional model of depression and anxiety,11, 12, 13, 14 neuroticism predicted increased depression (b=0.34, P<0.001; Figure 3F) along with increases in anxiety (b=0.11, P<0.001; Figure 3H). Depression and anxiety were positively correlated (r=0.42, P<0.001; Figure 3J). The Sobel test of mediation confirmed a significant indirect pathway from the BDNF–ELS interaction down to the brain level and symptom-cognition outcomes (see pathways A and B in Table 2).

Figure 3

Structural equation model summary for the effects of brain-derived neurotrophic factor (BDNF) Val66Met polymorphism and its interaction with early life stress (ELS) on depression and anxiety (and associated cognitive function) by its effects on brain structure and arousal. These effects are presented as model estimates (unstandardized path coefficients) (n=89). Solid lines represent positive path coefficients, dashed lines represent negative path coefficients, single-headed arrows represent direct effects and double-headed arrows reflect correlation coefficients. The thickness of the line represents the significance of the effect (thin, medium and thick lines corresponding to P<0.05, 0.01 and 0.001, respectively). The interaction of BDNF Met status with exposure to ELS contributed to distinctive brain–arousal pathways to syndromal depression and anxiety. The interaction of BDNF Met status and ELS directly predicted reduced hippocampal volume (A, b=−0.11, P<0.001). This reduction was correlated with reductions in lateral prefrontal cortical volume (B, r=0.68, P<0.001), which in turn was associated with greater depressed mood (C, r=−0.09, P=0.005). Second, the interaction of BDNF–ELS predicted increases in startle-elicited heart rate (D, b=4.59, P<0.001), which predicted increased neuroticism (E, b=0.19, P=0.019), and subsequently increased depression (F, b=0.34, P<0.001) and anxiety (H, b=0.11, P<0.001). Syndromal depression was associated with poorer working memory accuracy (G, r=−0.75, P=0.005), whereas higher anxiety predicted poorer performance on a distinct aspect of memory; verbal memory (I, b=−0.08, P=0.003). Working memory was also correlated with slowed response speed (r=0.229, P=0.030), whereas anxiety independently predicted increased impulsivity (b=0.021, P=0.018). Depression and anxiety were positively correlated (J, r=0.42, P<0.001). In contrast, the interaction of BDNF V/V genotype with exposure to ELS predicted elevated syndromal anxiety through an alternative neural system. Compared to Met carriers, the interaction of BDNF V/V status and ELS predicted enlarged amygdala volume (K, b=−0.03, P=0.003), which correlated positively with medial prefrontal volume (L, r=0.65, P<0.001). Enlarged medial prefrontal volume predicted elevations in startle-elicited heart rate variability (M, b=3.53, P=0.026), which was associated with elevated anxiety (N, b=0.08, P<0.026). For details of the full final model estimated see Supplementary Figure 2, and for details of the path coefficients of other model predictors see Supplementary Table 4.

Table 2 Mediation tests for indirect pathways in the final BDNF model

In contrast, the interaction of BDNF V/V genotype with exposure to ELS predicted elevated syndromal anxiety through an alternative neural system. The interaction of BDNF V/V status and ELS predicted greater amygdala volume (b=−0.03, P=0.003; Figure 3K), which correlated positively with medial prefrontal volume (r=0.65, P<0.001; Figure 3L). In turn, greater amygdala-medial prefrontal volume predicted elevations in startle-elicited HRV (b=3.53, P=0.026; Figure 3M), which were associated with higher anxiety (b=0.08, P=0.026; Figure 3N). Syndromal anxiety predicted poorer performance on a distinctive aspect of memory; verbal memory (b=−0.08, P=0.003; Figure 3I). Mediation was generally confirmed for these BDNF V/V-ELS pathways to the brain, arousal and symptom-cognition levels, with a trend effect (P=0.056) evident for the contribution of HRV in the link between MPFC volume and anxiety suggesting other potential variables may have a function (see pathway C in Table 2).

Structural equation modeling: confirmatory models

To examine the specificity of the path model for the BDNF polymorphism, the final model was retested with the 5HTT-LPR polymorphism (Table 1; model 2). Relative to the BDNF model, the overall fit of the 5HTT-LPR model was reduced in terms of both overall χ2 and GOF indices. Further, neither 5HTT-LPR nor 5HTT-LPR × ELS predicted the pathways predictive by the equivalent BDNF and BDNF-ELS effects, suggesting a preferential role for the BDNF polymorphism in these pathways.

In a third model (Table 1; model 3), the direction of the effect between heart rate and neuroticism was verified by examining model fit when the path between these two variables was reversed. No significant improvement in model fit was evident, and the strength of the path coefficient from neuroticism to heart rate was weaker (standardized β=0.201) in comparison to the original BDNF model (Table 1; model 1, v) in which heart rate predicted neuroticism (standardized β=0.238). Nonetheless, given both effects were significant (albeit stronger in the original model), the direction of these effects requires confirmation in future research.

Because the final BDNF model indicated that both BDNF Met and BDNF V/V may contribute to risk (depression and anxiety, respectively), a third confirmatory analyses examined if these pathways were distinguished by correlations with other behavioral indicators of depression vs anxiety, established in the literature. Slowed response (or psychomotor) speed is part of the diagnostic criteria for major depression, particularly with melancholia, and has been distinguished from poor inhibition (referred to here as ‘impulsivity’) that defines an anxiety dimension.79 From our cognitive battery, we extracted measures of response speed (psychomotor task) and impulsivity (‘go-nogo’ task) and included them in the final BDNF model, to test correlation coefficients specified between the working memory domain and response speed (linked to depression), and between verbal memory and impulsivity (linked to anxiety) (for further details see Supplementary Materials). The resulting model demonstrated good overall fit (χ2=240.59, P=0.241; RMSEA=0.026, P=0.932; CFI=0.987, PGFI=0.633). Response speed correlated with working memory (r=0.229, P=0.030), but not verbal memory, consistent with a link to depression, and impulsivity was predicted by anxiety (b=0.021, P=0.018), but not depression. These findings provide further support for the separation of pathways to depression and anxiety, from the interaction of BDNF and ELS.


These findings demonstrate for the first time the neurobiological pathways that predict syndromal depression and anxiety, consistent with a multidimensional model of depression and anxiety.9 Using SEM, we identified the effects of the interaction of the BDNF Val66Met polymorphism and exposure to ELS on neural circuitry and autonomic arousal pathways that in turn predict syndromal depression and anxiety, and associated alterations in cognition. Further, these effects appear to involve preferentially the BDNF polymorphism, compared to the 5HTT-LPR polymorphism, which has been implicated in risk for depression in some studies15, 16, 17, 18 but not all.80, 81

In BDNF Met carriers, exposure to ELS predicted higher syndromal depression through the loss of hippocampal-lateral prefrontal gray matter. In addition, the BDNF Met–stress interaction predicted greater general distress (higher depression with anxiety) by elevations in body arousal and the temperamental trait of neuroticism.

The specific pathway to syndromal depression in this study accords with previous integrative findings,82 and with behavioral formulations that distinguish a depression factor defined by anhedonia and lack of positive emotion.9 This pathway also accords with gene–environment models of affective disorder58 that implicate both exposure to stress early in life and its effects on hippocampal circuits in susceptibility to depression. SEM showed that the BDNF Met–ELS interaction predicted a reduction in hippocampal gray matter volume, and its lateral prefrontal projections, which was associated with elevated trait depression. This association is consistent with the reductions in hippocampal volume commonly found in depressed patients,39, 40, 41 particularly those with a history of early life trauma.83 Because BDNF has an important role in long-term potentiation46 associated with hippocampal networks, altered release of BDNF during development could affect the formation of neural networks84 through reduced neural plasticity.

By contrast, the BDNF V/V genotype contributed specifically to elevated syndromal anxiety in the absence of elevated depression in this study, by a parallel anatomical pathway. This pathway accords with behavioral formulations that differentiate a specific anxiety dimension defined by somatic features. It also suggests that BDNF V/V status might have a role in the disposition toward specific anxiety disorders. Our observation that the BDNF V/V–ELS interaction predicted greater anxiety by an increase in amygdala and associated medial prefrontal volume is concordant with evidence from laboratory animal studies. Chronic stress has been found to enhance dendritic arborization of neurons in the rat amygdala,85 serving to augment amygdala-dependent fear conditioning toward specific stimuli. Moreover, the linkage between enlarged amygdala volume and elevated anxiety is consistent with evidence for enhanced right amygdala volume in anxiety disorder patients86 or anxiety scores in autistic children87 as well as the converse reduction in fear responses following amygdala lesions.88 The enlarged amygdala-medial prefrontal of ‘stressed’ V/V genotypes was in turn associated with elevated HRV, which was then associated with increased anxiety. Although a relationship between reduced HRV and anxiety and poor health has been observed in many studies,89 the precise nature of this relationship remains unclear. Extension of the present study in a sample of patients with syndromal anxiety disorder is required to further delineate the direction of this association.

The additional role of the BDNF Met allele and stress in predicting syndromal depression and associated anxiety accords with the behavioral formulation of a mixed depression–anxiety or ‘general distress’ dimension.13 The findings indicate that the interaction of BDNF Met status and exposure to stress contributes, in addition to loss of hippocampal-LPFC, to an elevation of body arousal and neuroticism, which have been implicated in risk for depressive and comorbid anxiety conditions. ‘Stressed’ BDNF Met carriers showed striking elevations in heart rate (16.5% in Met carriers compared to 1.7% in V/V homozygotes, each exposed to five or more stressful events). Heart rate elevations contributed to higher neuroticism, which in turn predicted higher levels of depression and anxiety. This pathway is consistent with previous evidence linking the BDNF Met allele with both depression and anxiety,24, 25 and with evidence that heightened arousal through hypothalamic–pituitary–adrenal axis dysregulation is a major clinical factor in determining severity of depression and associated anxiety.36, 58 Stress coupled with BDNF Met status may have a compound effect on these circuits. Moreover, the link between BDNF status and heart rate may have further implications for heightened vulnerability to cardiovascular disease commonly documented in depressed patients. Disrupted cardiovascular function has been similarly observed in BDNF-deficient mice.90

Contrary to the view that the BDNF Met allele may confer ‘risk’ and the Val allele ‘protection’, our findings suggest that both alleles may contribute to distinct mechanisms of risk when an individual is also exposed to early life stressors. The interaction of BDNF V/V homozygote status and ELS predicted greater amygdala-medial prefrontal gray matter, which in turn was associated with higher HRV and syndromal anxiety. This finding is not unusual but suggests a possible presence of allelic heterogeneity; namely, that distinctive patterns of traits that may be produced by different mutations within a gene. One example is the MAPT gene, encoding for the microtubule-associated protein tau. Different mutations in this gene can cause progressive supranuclear palsy or frontotemporal dementia with parkinsonism, as well as other disorders.91 Exposure to stress may differentially impact the function of brain–body pathways involved in regulating syndromal depression and anxiety, depending on an individual's level of BDNF (according to the presence of Met or Val alleles). Of course, this possibility needs to be examined in clinical samples, to determine if these pathways predict depression and anxiety at the level of disorder.

The findings of this study highlight the importance of investigations that focus on integrating candidate gene, brain and behavior risk markers for affective disorder.1, 4, 5, 8, 55 These findings provide new evidence to suggest that a chain of events commencing with gene–ELS interactions and their impact on the brain and autonomic arousal systems define two parallel pathways to depression and anxiety, and account for the overlapping presentation of these conditions. It is important to recognize however that this model is a statistical one, and cannot consider all candidate markers for depression and anxiety. Yet, an advantage of SEM over conventional statistical methods is that residual error terms can partly take account of absent measures. In this first study using path modeling, we focused on candidate markers observed in previous individual studies, and associated with stress vulnerability mechanisms of risk for these disorders. With larger sample sizes, the effects of other variables (such as other genes or health variables) could be examined. Investigations with larger samples, including with clinical patients, are underway in which gene–gene and gene–gene–environment variables are being included within the integrative model. These include the international study to predict optimized treatment in depression (iSPOT-D).2 Further, to explore temporal dynamics and reciprocal relations, longitudinal or experimental data need to be applied; for instance to confirm the direction of effects between body arousal and the neuroticism measure of emotional stability. It would also be interesting to examine if there are additional interactions between BDNF alleles and sex differences, given previous indications of such differences in relation to temperament.92 These insights are crucial to targeting the earliest risk factors for depression and anxiety, and thus early intervention. Given the involvement of BDNF, arousal and hippocampal–amygdala circuits in modulating the action of antidepressants, the present findings also have implications for identifying gene–brain markers of individual response to pharmacological targets for the treatment of depression and anxiety.2, 93


  1. 1

    Gordon E . Integrating genomics and neuromarkers for the era of brain-related personalized medicine. Personalized Medicine 2007; 4: 201–215.

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  2. 2

    Kemp AH, Gordon E, Rush AJ, Williams LM . Improving the prediction of treatment response in depression: Integration of clinical, cognitive, psychophysiological, neuroimaging and genetic measures. CNS Spectrums 2008; 13: 1066–1086.

    PubMed  Article  PubMed Central  Google Scholar 

  3. 3

    Hasler G, Drevets WC, Manji HK, Charney DS . Discovering endophenotypes for major depression. Neuropsychopharmacology 2004; 29: 1765–1781.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  4. 4

    Gatt JM, Clark CR, Kemp AH, Liddell BJ, Dobson-Stone C, Kuan SA et al. A genotype–endophenotype–phenotype path model of depressed mood: integrating cognitive and emotional markers. J Integr Neurosci 2007; 6: 75–104.

    PubMed  PubMed Central  Article  Google Scholar 

  5. 5

    Gatt JM, Kuan SA, Dobson-Stone C, Paul RH, Joffe RT, Kemp AH et al. Association between BDNF Val66Met polymorphism and trait depression is mediated via resting EEG alpha band activity. Biol Psychol 2008; 79: 275–284.

    PubMed  Article  PubMed Central  Google Scholar 

  6. 6

    Williams LM, Gordon E . The dynamic organization of the emotional brain: responsivity, stability and instability. Neuroscientist 2007; 13: 349–370.

    PubMed  PubMed Central  Article  Google Scholar 

  7. 7

    Williams LM, Gatt JM, Hatch A, Palmer DM, Nagy M, Rennie C et al. The Integrate model of emotion, thinking, feeling and self regulation. An application to the ‘paradox of aging’. J Integr Neurosci 2008; 7: 367–404.

    PubMed  PubMed Central  Article  Google Scholar 

  8. 8

    Gordon E . Integrative Neuroscience. Neuropsychopharmacology 2003; 28: S2–S8.

    PubMed  PubMed Central  Article  Google Scholar 

  9. 9

    Clark LA, Watson D . Tripartite model of anxiety and depression: psychometric evidence and taxonomic implications. J Abnorm Psychol 1991; 100: 316–336.

    CAS  Article  Google Scholar 

  10. 10

    Mineka S, Watson D, Clark LA . Comorbidity of anxiety and unipolar mood disorders. Annu Rev Psychol 1998; 49: 377–412.

    CAS  PubMed  Article  Google Scholar 

  11. 11

    Lovibond SH, Lovibond PF . Manual for the Depression Anxiety Stress Scales. Psychology Foundation: Sydney, 1995.

    Google Scholar 

  12. 12

    Kendler KS, Gardner CO, Prescott CA . Toward a comprehensive developmental model for major depression in women. Am J Psychiatry 2002; 159: 1133–1145.

    PubMed  PubMed Central  Article  Google Scholar 

  13. 13

    Kendler KS, Kuhn J, Prescott CA . The interrelationship of neuroticism, sex, and stressful life events in the prediction of episodes of major depression. Am J Psychiatry 2004; 161: 631–636.

    PubMed  PubMed Central  Article  Google Scholar 

  14. 14

    Kessler RC . The effects of stressful life events on depression. Annu Rev Psychol 1997; 48: 191–214.

    CAS  Article  Google Scholar 

  15. 15

    Caspi A, Sugden K, Moffitt TE, Taylor A, Craig IW, Harrington H et al. Influence of life stress on depression: moderation by a polymorphism in the 5-HTT gene. Science 2003; 301: 386–389.

    CAS  Google Scholar 

  16. 16

    Wilhelm K, Mitchell PB, Niven H, Finch A, Wedgewood L, Scimone A et al. Life events, first depression onset and the serotonin transporter gene. Br J Psychiatry 2006; 188: 210–215.

    PubMed  PubMed Central  Article  Google Scholar 

  17. 17

    Kaufman J, Yang B-Z, Douglas-Palumberi H, Houshyar S, Lipschitz D, Krystal JH et al. Social supports and serotonin transporter gene moderate depression in maltreated children. Proc Natl Acad Sci 2004; 101: 17316–17321.

    CAS  PubMed  Article  Google Scholar 

  18. 18

    Mann JJ, Huang Y-Y, Underwood MD, Kassir SA, Oppenheim S, Kelly TM et al. A serotonin transporter gene promoter polymorphism (5-HTTLPR) and prefrontal cortical binding in major depression and suicide. Arch Gen Psychiatry 2000; 57: 729–738.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  19. 19

    Nemeroff CB, Vale WW . The neurobiology of depression: inroads to treatment and new drug discovery. J Clin Psychiatry 2005; 66: 5–13.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  20. 20

    Urani A, Chourbaji S, Gass P . Mutant mouse models of depression: candidate genes and current mouse lines. Neurosci Biobehav Rev 2005; 29: 805–828.

    CAS  PubMed  Article  Google Scholar 

  21. 21

    Munafò MR, Clark T, Flint J . Does measurement instrument moderate the association between the serotonin transporter gene and anxiety-related personality traits? A meta-analysis. Mol Psychiatry 2005; 10: 415–419.

    PubMed  Article  CAS  Google Scholar 

  22. 22

    Pezawas L, Meyer-Lindenberg A, Goldman AL, Verchinski BA, Chen G, Kolachana BS et al. Evidence of biologic epistasis between BDNF and SLC6A4 and implications for depression. Mol Psychiatry 2008; 13: 709–716.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  23. 23

    Kaufman J, Yang B-Z, Douglas-Palumberi H, Grasso D, Lipschitz D, Houshyar S et al. Brain-derived neurotrophic factor-5-HTTLPR gene interactions and environmental modifiers of depression in children. Biol Psychiatry 2006; 59: 673–680.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  24. 24

    Jiang X, Xu K, Hoberman J, Tian F, Marko AJ, Waheed JF et al. BDNF variation and mood disorders: a novel functional promoter polymorphism and val66met are associated with anxiety but have opposing effects. Neuropsychopharmacology 2005; 30: 1353–1361.

    CAS  Article  PubMed  Google Scholar 

  25. 25

    Hwang J-P, Tsai S-J, Hong C-J, Yang C-H, Lirng J-F, Yang Y-M . The Val66Met polymorphism of the brain-derived neurotrophic factor gene is associated with geriatric depression. Neurobiol Aging 2006; 27: 1834–1837.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  26. 26

    Chen Z-Y, Jing D, Bath KG, Ieraci A, Khan T, Siao C-J et al. Genetic variant BDNF (Val66Met) polymorphism alters anxiety-related behavior. Science 2006; 314: 140–143.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  27. 27

    Lang UE, Hellweg R, Kalus P, Bajbouj M, Lenzen KP, Sander T et al. Association of a functional BDNF polymorphism and anxiety-related personality traits. Psychopharmacology 2005; 180: 95–99.

    CAS  PubMed  Article  Google Scholar 

  28. 28

    Carney RM, Saunders RD, Freedland KE, Stein P, Rich MW, Jaffe AS . Association of depression with reduced heart rate variability in coronary artery disease. Am J Cardiol 1995; 76: 562–564.

    CAS  PubMed  Article  Google Scholar 

  29. 29

    Dawson ME, Schell A, Braaten JR, Catania JJ . Diagnostic utility of autonomic measures for major depressive disorders. Psychiatry Res 1985; 15: 261–270.

    CAS  PubMed  Article  Google Scholar 

  30. 30

    Pole N . The psychophysiology of posttraumatic stress disorder: a meta-analysis. Psychol Bull 2007; 133: 725–746.

    PubMed  Article  PubMed Central  Google Scholar 

  31. 31

    Fraguas Jr R, Marci C, Fava M, Iosifescu DV, Bankier B, Loh R et al. Autonomic reactivity to induced emotion as potential predictor of response to antidepressant treatment. Psychiatry Res 2007; 151: 169–172.

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  32. 32

    Greaves-Lord K, Ferdinand RF, Sondeijker FE, Dietrich A, Oldehinkel AJ, Rosmalen JG et al. Testing the tripartite model in young adolescents: is hyperarousal specific for anxiety and not depression? J Affect Disord 2007; 102: 55–63.

    PubMed  Article  PubMed Central  Google Scholar 

  33. 33

    Brown TA, Chorpita BF, Barlow DH . Structural relationships among dimensions of the DSM-IV anxiety and mood disorders and dimensions of negative affect, positive affect, and autonomic arousal. J Abnorm Psychol 1998; 107: 179–192.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  34. 34

    Shalev AY, Freedman S, Peri T, Brandes D, Sahar T, Orr SP et al. Prospective study of posttraumatic stress disorder and depression following trauma. Am J Psychiatry 1998; 155: 630–637.

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  35. 35

    Fullerton J, Cubin M, Tiwari H, Wang C, Bomhra A, Davidson S et al. Linkage analysis of extremely discordant and concordant sibling pairs identifies quantitative-trait loci that influence variation in the human personality trait neuroticism. Am J Hum Genet 2003; 72: 879–890.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  36. 36

    Heim C, Nemeroff CB . The impact of early adverse experiences on brain systems involved in the pathophysiology of anxiety and affective disorders. Biol Psychiatry 1999; 46: 1509–1522.

    CAS  Article  Google Scholar 

  37. 37

    Mayberg HS . Defining the neural circuitry of depression: toward a new nosology with therapeutic implications. Biol Psychiatry 2007; 61: 729–730.

    PubMed  PubMed Central  Article  Google Scholar 

  38. 38

    Drevets WC . Neuroimaging studies of mood disorders. Biol Psychiatry 2000; 48: 813–829.

    CAS  Article  PubMed  Google Scholar 

  39. 39

    Bremner JD, Narayan M, Anderson ER, Staib LH, Miller HL, Charney DS . Hippocampal volume reduction in major depression. Am J Psychiatry 2000; 157: 115–117.

    CAS  Article  PubMed  Google Scholar 

  40. 40

    Von Gunten A, Fox NC, Cipolotti L, Ron MA . A volumetric study of hippocampus and amygdala in depressed patients with subjective memory problems. J Neuropsychiatry Clin Neurosci 2000; 12: 493–498.

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  41. 41

    Sheline YI, Gado MH, Price JL . Amygdala core nuclei volumes are decreased in recurrent major depression. Neuroreport 1998; 9: 2023–2028.

    CAS  Article  Google Scholar 

  42. 42

    Landro NI, Stiles TC, Sletvold H . Neuropsychological function in nonpsychotic unipolar major depression. Neuropsychiatry Neuropsychol Behav Neurol 2001; 14: 233–240.

    CAS  PubMed  PubMed Central  Google Scholar 

  43. 43

    De Bellis MD, Keshavan MS, Shifflett H, Iyengar S, Beers SR, Hall J et al. Brain structures in pediatric maltreatment-related posttraumatic stress disorder: a sociodemographically matched study. Biol Psychiatry 2002; 52: 1066–1078.

    PubMed  PubMed Central  Article  Google Scholar 

  44. 44

    De Geus EJC, van’t Ent D, Wolfensberger SPA, Heutink P, Hoogendijk WJG, Boomsma DI et al. Intrapair differences in hippocampal volume in monozygotic twins discordant for the risk for anxiety and depression. Biol Psychiatry 2007; 61: 1062–1071.

    PubMed  Article  PubMed Central  Google Scholar 

  45. 45

    Gilbertson MW, Shenton ME, Ciszewski A, Kasai K, Lasko NB, Orr SP et al. Smaller hippocampal volume predicts pathologic vulnerability to psychological trauma. Nat Neurosci 2002; 5: 1242–1247.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  46. 46

    Black IR . Trophic regulation of synaptic plasticity. J Neurobiol 1999; 41: 108–118.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  47. 47

    Conner JM, Lauterborn JC, Yan Q, Gall CM, Varon S . Distribution of brain-derived neurotrophic factor (BDNF) protein and mRNA in the normal adult rat CNS: evidence for anterograde axonal transport. J Neurosci 1997; 17: 2295–2313.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  48. 48

    Rattiner LM, Davis M, French CT, Ressler KJ . Brain-derived neurotrophic factor and tyrosine kinase receptor B involvement in amygdala-dependent fear conditioning. J Neurosci 2004; 24: 4796–4806.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  49. 49

    Karege F, Bondolfi G, Gervasoni N, Schwald M, Aubry J-M, Bertschy G . Low brain-derived neurotrophic factor (BDNF) levels in serum of depressed patients probably results from lowered platelet BDNF release unrelated to platelet reactivity. Biol Psychiatry 2005; 57: 1068–1072.

    CAS  Article  PubMed  Google Scholar 

  50. 50

    Karege F, Perret G, Bondolfi G, Schwald M, Bertschy G, Aubry J-M . Decreased serum brain-derived neurotrophic factor levels in major depressed patients. Psychiatry Res 2002; 109: 143–148.

    CAS  Article  PubMed  Google Scholar 

  51. 51

    Shimizu E, Hashimoto K, Okamura N, Koike K, Komatsu N, Kumakiri C et al. Alterations of serum levels of brain-derived neurotrophic factor (BDNF) in depressed patients with or without antidepressants. Biol Psychiatry 2003; 54: 70–75.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  52. 52

    Egan MF, Kojima M, Callicott JH, Goldberg TE, Kolachana BS, Bertolino A et al. The BDNF val66met polymorphism affects activity-dependent secretion of BDNF and human memory and hippocampal function. Cell 2003; 112: 257–269.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  53. 53

    Pezawas L, Verchinski BA, Mattay VS, Callicott JH, Kolachana BS, Straub RE et al. The brain-derived neurotrophic factor val66met polymorphism and variation in human cortical morphology. J Neurosci 2004; 10: 10099–10102.

    Article  CAS  Google Scholar 

  54. 54

    Schofield PRS, Williams LM, Paul RH, Gatt JM, Brown K, Luty A et al. Disturbances in task-relevant information processing associated with the brain-derived neurotrophic factor Val66Met polymorphism: evidence from cognition, the P300 endophenotype and neuroimaging. Biol Psychol 2009; published online 11 June 2008 (in press).

  55. 55

    Joffe RT, Gatt JM, Kemp AH, Grieve S, Dobson-Stone C, Kuan S et al. Brain derived neurotrophic factor Val66Met polymorphism, the five factor model of personality and hippocampal volume: implications for depressive illness. Hum Brain Mapp 2009; published online 16 September 2008 (in press).

  56. 56

    Duman RS, Malberg J, Thome J . Neural plasticity to stress and antidepressant treatment. Biol Psychiatry 1999; 46: 1181–1191.

    CAS  Article  PubMed  Google Scholar 

  57. 57

    Carney RM, Freedland KE, Veith RC . Depression, the autonomic nervous system, and coronary heart disease. Psychosom Med 2005; 67: S29–S33.

    PubMed  PubMed Central  Article  Google Scholar 

  58. 58

    Heim C, Owens MJ, Plotsky PM, Nemeroff CB . The role of early adverse life events in the etiology of depression and posttraumatic stress disorder. Ann NY Acad Sci 1997; 821: 194–207.

    CAS  PubMed  Article  Google Scholar 

  59. 59

    Tomarken AJ, Waller NG . Structural equation modeling: strengths, limitations, and misconceptions. Annu Rev Clin Psychol 2005; 1: 31–65.

    Article  Google Scholar 

  60. 60

    Carmelli D, Swan GE, DeCarli C, Reed T . Quantitative genetic modelling of regional brain volumes and cognitive performance in older male twins. Biol Psychol 2002; 61: 139–155.

    PubMed  Article  Google Scholar 

  61. 61

    Boomsma DI, van Baal GC, Orlebeke JF . Genetic influences on respiratory sinus arrhythmia across different task conditions. Acta Genet Med Gemellol (Roma) 1990; 39: 181–191.

    CAS  Article  Google Scholar 

  62. 62

    Ormel J, Rosmalen J, Farmer A . Neuroticism: a non-informative marker of vulnerability to psychopathology. Soc Psychiatry Psychiatric Epidemiol 2004; 39: 906–912.

    Article  Google Scholar 

  63. 63

    Babor TF, Higgins-Biddle JC, Saunders JB, Monteiro MG . AUDIT—The Alcohol Use Disorders Identification Test: Guidelines for Use in Primary Care. World Health Organization, Geneva, 2001.

    Google Scholar 

  64. 64

    Heatherton TF, Kozlowski LT, Frecker RC, Fagerström KO . The Fagerström test for nicotine dependence: a revision of the Fagerström tolerance questionnaire. Br J Addict 1991; 86: 1119–1127.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  65. 65

    Cohen RA, Paul RH, Stroud L, McCaffrey J, Sweet L, Hitsman B, Gunstad J et al. Early life stress and adult emotional experience: An international perspective. The International J of Psychiatry in Medicine 2006; 36: 35–52.

    Article  Google Scholar 

  66. 66

    Costa P, McCrae R . NEO PI-R Professional Manual. Psychological Assessment Resources Inc.: Odessa, FL, 1992.

    Google Scholar 

  67. 67

    Williams LM, Simms E, Clark CR, Paul RH, Rowe D, Gordon E . The test–retest reliability of a standardized neurocognitive and neurophysiological test battery: ‘NeuroMarker’. Int J Neurosci 2005; 115: 1605–1630.

    CAS  PubMed  Article  Google Scholar 

  68. 68

    Paul RH, Lawrence J, Williams LM, Clark CR, Cooper N, Gordon E . The validity of ‘IntegNeuro’: a new computerized and standardized battery of neurocognitive tests. Int J Neurosci 2005; 115: 1549–1567.

    PubMed  PubMed Central  Article  Google Scholar 

  69. 69

    Clark CR, Paul RH, Williams LM, Arns M, Fallahpour K, Handmer C et al. Standardized assessment of cognitive functioning during development and aging using an automated touchscreen battery. Arch Clin Neuropsychology 2006; 21: 449–467.

    Article  Google Scholar 

  70. 70

    Rowe DL, Cooper NJ, Liddell BJ, Clark CR, Gordon E, Williams LM . Brain structure and brain function correlates of general and social cognition. J Integr Neurosci 2007; 6: 35–74.

    PubMed  Article  PubMed Central  Google Scholar 

  71. 71

    Ashburner J, Friston KJ . Voxel-based morphometry—the methods. Neuroimage 2000; 11: 805–821.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  72. 72

    Grieve S, Clark CR, Williams LM, Gordon E . Limbic and paralimbic preservation with aging. Hum Brain Mapp 2005; 25: 391–401.

    PubMed  Article  PubMed Central  Google Scholar 

  73. 73

    Friston KJ, Holmes AP, Poline JB, Price CJ, Frith CD . Detecting activations in PET and fMRI: levels of inference and power. Neuroimage 1996; 4: 223–235.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  74. 74

    Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O, Delcroix N et al. Automated anatomical labelling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single subject brain. Neuroimage 2002; 15: 273–289.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  75. 75

    Arbuckle JL, Wothke W . AMOS 4.0 User's Guide. SPSS Inc.: Chicago, 1999.

    Google Scholar 

  76. 76

    Byrne BM . Structural Equation Modeling with AMOS: Basic Concepts, Applications, and Programming. Lawrence Erlbaum Associates: New Jersey, 2001.

    Google Scholar 

  77. 77

    Bollen KA . Structural Equations with Latent Variables. Wiley: New York, 1989.

    Google Scholar 

  78. 78

    Sobel ME . Asymptotic confidence intervals for indirect effects in structural equation models. In: S Leinhardt (ed). Sociological Methodology. American Sociological Association: Washington, DC, 1982, pp 290–312.

    Google Scholar 

  79. 79

    Pierson A, Partiot A, Jouvent R, Bungener C, Martinerie J, Renault B et al. Loss of control of pre-motor activation in anxious agitated and impulsive depressives. A clinical and ERP study. Prog Neuropsychopharmacol Biol Psychiatry 1994; 18: 1037–1050.

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  80. 80

    Gillespie N, Whitefield J, Williams B, Heath A, Martin N . The relationship between stressful life events, the serotonin (5-HTTLPR) genotype and major depression. Psychol Med 2005; 35: 101–111.

    PubMed  PubMed Central  Article  Google Scholar 

  81. 81

    Surtees PG, Wainwright NW, Willis-Owen SA, Luben R, Day NE, Flint J . Social adversity, the serotonin transporter (5-HTTLPR) polymorphism and major depressive disorder. Biol Psychiatry 2005; 59: 224–229.

    PubMed  Article  CAS  Google Scholar 

  82. 82

    Kemp AH, Hopkinson PJ, Stephan BCM, Clark CR, Gordon E, Bryant RA et al. Predicting severity of non-clinical depression: Preliminary findings using an integrated approach. J Integrative Neuroscience 2006; 5: 89–110.

    Article  Google Scholar 

  83. 83

    Vythilingam M, Heim C, Newport J, Miller AH, Anderson E, Bronen R et al. Childhood trauma associated with smaller hippocampal volume in women with major depression. Am J Psychiatry 2002; 159: 2072–2080.

    PubMed  PubMed Central  Article  Google Scholar 

  84. 84

    Angelucci F, Brene S, Mathe AA . BDNF in schizophrenia, depression and corresponding animal models. Mol Psychiatry 2005; 10: 345–352.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  85. 85

    Vyas A, Mitra R, Rao BSS, Chattarji S . Chronic stress induces contrasting patterns of dendritic remodeling in hippocampal and amygdaloid neurons. J Neurosci 2002; 22: 6810–6818.

    CAS  Article  PubMed  Google Scholar 

  86. 86

    De Bellis MD, Casey BJ, Dahl RE, Birmaher B, Williamson DE, Thomas KM et al. A pilot study of amygdala volumes in pediatric generalized anxiety disorder. Biol Psychiatry 2000; 48: 51–57.

    CAS  PubMed  PubMed Central  Article  Google Scholar 

  87. 87

    Juranek J, Filipek PA, Berenji GR, Modahl C, Osann K, Spence MA . Association between amygdala volume and anxiety level: magnetic resonance imaging (MRI) study in autistic children. J Child Neurol 2006; 21: 1051–1058.

    PubMed  PubMed Central  Article  Google Scholar 

  88. 88

    Kemble ED, Blanchard DC, Blanchard RJ . Effects of regional amygdaloid lesions on flight and defensive behaviors of wild black rats (Rattus rattus). Physiol Behav 1990; 48: 1–5.

    CAS  PubMed  Article  PubMed Central  Google Scholar 

  89. 89

    Friedman BH, Thayer JF . Autonomic balance revisited: panic anxiety and heart rate variability. J Psychosomatic Res 1998; 44: 133–151.

    CAS  Article  Google Scholar 

  90. 90

    Donovan M, Lin MI, Wiegn P, Ringstedt T, Kraemer R, Hahn R et al. Brain derived neurotrophic factor is an endothelial cell survival factor required for intramyocardial vessel stabilization. Development 2000; 127: 4531–4540.

    CAS  PubMed  Google Scholar 

  91. 91

    Rademakers R, Cruts M, van Broeckhoven C . The role of tau (MAPT) in frontotemporal dementia and related tauopathies. Hum Mutat 2004; 24: 277–295.

    CAS  PubMed  Article  Google Scholar 

  92. 92

    Itoh K, Hashimoto K, Kumakiri C, Shimizu E, Iyo M . Association between brain-derived neurotrophic factor 196 G/A polymorphism and personality traits in healthy subjects. Am J Med Genet B 2004; 124B: 61–63.

    Article  Google Scholar 

  93. 93

    Duman R . Novel therapeutic approaches beyond the serotonin receptor. Biol Psychiatry 1998; 44: 324–335.

    CAS  PubMed  Article  Google Scholar 

Download references


This project was supported by an ARC-linkage grant (LP0455104), with Brain Resource as industry partner. LMW had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. LMW holds a peer-reviewed Pfizer foundation Senior Research Fellowship, and PRS held an NHMRC Senior Principal Research Fellowship (no. 157209). CBN was supported by NIMH MH-42088, MH-52899 and MH-39415. CDS was supported by a European Molecular Biology Organisation postdoctoral fellowship (ALTF 166-2004). AHK holds an NHMRC Peter Doherty Fellowship (no. 358770). We acknowledge the support of the Brain Resource International Database (under the auspices of Brain Resource, for use of normative data. We thank the individuals who gave their time to participate in the database. Access to the database for scientific purposes was administered independently through the scientific network (BRAINnet,, which is coordinated independently of the commercial operations of BR. We also thank Scott Norrie ( for graphical design.

Author information



Corresponding author

Correspondence to C B Nemeroff.

Additional information

Disclosure/conflict of interest

Brain Resource Ltd (BR) was the industry partner on the ARC-linkage grant that funded this study, but had no further role in design or implementation of the project. EG is the CEO of BR, and holds significant equity and stock options in the company (however, scientific decisions about access to the Brain Resource International Database are made through an independently administered network of scientists; BRAINnet, PRS and JMG hold stock options, and LMW is a small equity holder in BR. JMG is employed as a postdoctoral researcher on the ARC-linkage grant that funded this project. LMW, RHP, JMG and AHK have received fees from BR for consultancies unrelated to this study. CBN is a member of the scientific advisory board of American Foundation for Suicide Prevention (AFSP), Johnson and Johnson, Forest Laboratories, AstraZeneca, Quintiles, Janssen/Ortho-McNeil and PharmaNeuroboost. He holds equity and/or stock options in Corept, CeNeRx, Reevax and NovaDel Pharma and PharmaNeuroboost. CBN is on the board of directors for AFSP, George West Mental Health Foundation, NovaDel Pharma and Mt Cook Pharma. His patents include method and devices for transdermal delivery of lithium (US 6,375,990 B1) and method to estimate serotonin and norepinephrine transporter occupancy after drug treatment using patient or animal serum (provisional filing April 2001).

Author Contributions

JMG undertook the statistical analyses, with a leading role in developing the structural equation model and theoretical approach, and was responsible for all sections of each draft of this paper. LMW developed the project, its theoretical basis and experimental design, with significant input into interpretation of analyses and each draft of the paper. PRS and RHP were principal investigators and EG, partner investigator, in developing the project and design. EG was responsible for establishing the testing platforms used in this study. CBN and RB contributed significantly to the theoretical basis of this study, and to interpretation of analyses. CD-S and PRS performed the DNA genotyping. All authors have contributed to the final paper.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Gatt, J., Nemeroff, C., Dobson-Stone, C. et al. Interactions between BDNF Val66Met polymorphism and early life stress predict brain and arousal pathways to syndromal depression and anxiety. Mol Psychiatry 14, 681–695 (2009).

Download citation


  • depression-anxiety
  • BDNF Val66Met
  • early life stress
  • hippocampal gray matter
  • heart rate
  • cognition

Further reading