The BDNF gene Val66Met single-nucleotide polymorphism is carried by ~0.55% of Sub-Saharan Africans, 19.9% of Europeans and 43.6% of Asians, but may be carried by up to 72% of some Sub-Asian population groups.1 The switch from a guanine to adenine nucleotide at position 196 within the pro-region of the BDNF gene causes a Valine (Val) to Methionine (Met) amino-acid residue substitution at codon 66 (Val66Met), resulting in diminished activity-dependent secretion of BDNF at the synapse.2 The BDNF gene Val66Met polymorphism has been implicated as a modifier of hippocampal function and is a putative locus of risk for anxiety and affective disorders such as post-traumatic stress disorder (PTSD) and major depression.3 However, the literature suggesting that the loss of function 66Met variant is risk conferring for these disorders is inconsistent; and in some cases is even contradictory by suggesting that the wild-type 66Val allele provides risk as well.3 Likewise, a growing number of human studies have also failed to replicate the hippocampal deficits associated with the 66Met allele as reported by early studies, whereas meta-analyses have suggested that effect sizes of the Val66Met polymorphism on both cognition4 and hippocampal structure5 may only be small – if they exist at all. Many explanations for these discrepant results have been suggested, such as a lack of power, regression towards the mean or biased sampling.3 Another account is that discrepant data may be the result of failing to control for complex gene–environment interactions, which may determine or unmask BDNF phenotypes.

We hypothesized that stress, specifically the action of glucocorticoid stress hormones acting on the glucocorticoid receptor (GR), may be one such factor that interacts with the BDNF Val66Met polymorphism to determine hippocampus-dependent behavior. To test this hypothesis, we modeled the long-term effects of chronic stress exposure in a novel mouse model that has been genetically modified to express a humanized BDNF (hBDNF) coding transcript via endogenous mouse promoters that has yet to be behaviorally phenotyped. Specifically, this mouse model has the Val66Met polymorphism knocked-in including an extended sequence of 11 nucleotides across a 274-bp region that humanizes the coding exon of the rodent Bdnf gene.6 To model stress, we used a chronic corticosterone (CORT) exposure paradigm to specifically induce GR signaling without the confounding effects of other physiological components of the stress response axis. Exposure to CORT was time locked to a developmental period coinciding with late adolescence (weeks 6–8), with the behavioral phenotyping of hBDNFVal66Met mice occurring in adulthood (weeks 11–12) so to probe the long-term behavioral adaptation to mid-developmental GR recruitment.7 Tests of hippocampal function were the primary measures of the current study given that this is the primary phenotype of the Val66Met polymorphism and is a key component of the pathophysiology of both anxiety and affective disorders. Further details of our experimental design, the genetic construct of our hBDNFVal66Met mouse model and methodology can be found in the Supplementary Material.

To assess emotionally salient behavior, fear conditioning was used as a non-spatial, amygdala- and hippocampus-gated, memory paradigm. On day one, mice were conditioned using three tone-shock pairings, before being returned to their conditioning context 24 h later and to a novel context 48 h later to assess contextual and tone fear memory, respectively. A significant genotype × treatment interaction was observed for hippocampus-dependent contextual fear memory when analyzing males (F(2,76)=7.0, P=0.0016) and females (F(2,79)=14.14, P<0.0001) separately as well as when combined (F(2,161)=16.19, P<0.0001). As this interaction occurred independent of sex, the combined dataset was used for between-group comparisons so to increase power of detecting diminutive effect sizes as predicted by meta-analyses. Post hoc analyses of this dataset revealed that hBDNFMet/Met mice had significantly worse contextual fear memory relative to wild-type hBDNFVal/Val mice at baseline (P<0.01), however following the chronic CORT treatment this pattern was reversed with hBDNFMet/Met mice having significantly better contextual fear memory than mice carrying the hBDNFVal/Val genotype (P<0.01). Post hoc comparisons also revealed that the CORT-treated hBDNFMet/Met mice had significantly improved contextual fear memory than hBDNFMet/Met mice allocated to vehicle treatment (P<0.0001). For tone-elicited fear memory, none of the main effects reached significance, however, a significant genotype × treatment interaction, following the same direction as that reported for contextual fear memory, was also observed amongst the sex-collapsed data set (F(2, 161)=4.779, P=0.0096). The only comparison to reach significance was the enhanced tone-elicited fear memory of CORT-treated hBDNFMet/Met mice relative to hBDNFMet/Met controls (P<0.05). These results highlight that the chronic activation of glucocorticoid receptors during late adolescence potentiates the fear circuitry in adulthood according to hBDNFVal66Met genotype (see Figure 1).

Figure 1
figure 1

Fear conditioning was used as a measure of emotionally salient memory function. A significant genotype x treatment interaction was observed for contextual and tone fear memory. At baseline, hBDNFMet/Met mice had the worst task performance of all genotype groups, having significantly worse contextual fear memory than the wild-type hBDNFVal/Val controls. Following CORT, this pattern of results was reversed. For both contextual and tone fear memory, the chronic CORT treatment selectively enhanced the memory function of hBDNFMet/Met mice. The Y-maze, utilized here as a further test of hippocampus-dependent memory, revealed that a significant preference for the novel arm was absent amongst the hBDNFMet/Met genotype group at baseline. Following the chronic CORT treatment, this deficit was reinstated to levels consistent with hBDNFVal/Val controls. All data presented as mean±s.e.m.; *P<0.05, **P<0.01, and ****P<0.0001. Per group, n=22–41.

PowerPoint slide

Full size image

We next examined short-term spatial memory using the Y-maze as an alternative test of hippocampus-dependent memory function that is independent of the fear circuitry. Briefly, mice were allowed to explore two open arms of a three-arm maze for 10 min. One hour later, mice were returned to the maze but allowed to explore all three arms, with intact spatial memory being quantified by exploration time of the previously blocked ‘novel’ arm. There was no significant main effect or interaction comprising sex on time spent in the novel or other arms, so males and females were once more analyzed together to increase power. A significant interaction between the factors of ‘Group’ and ‘Arm’ (F(2,160)=5.2, P=0.0065) was detected. Within-group analysis revealed that hBDNFVal/Val (P<0.0001) and hBDNFVal/Met (P<0.0001) mice showed a highly significant preference for exploring the novel arm, relative to the other arms, indicating intact short-term spatial memory performance. On the other hand, hBDNFMet/Met mice showed a lack of preference for exploring the novel arm relative to the other arms, suggesting that the short-term spatial memory of these mice is subtly disrupted at baseline. Although, the chronic CORT treatment had no effect on Y-maze performance of hBDNFVal/Val and hBDNFVal/Met mice, the disrupted short-term spatial memory of hBDNFMet/Met mice was rescued by CORT treatment (P<0.0001) to levels consistent with hBDNFVal/Val controls. Further experimentation determined that this effect was not the result of altered anxiety-related exploratory drive (see Supplementary Material). This subtle Y-Maze result confirms the specificity of late adolescent glucocorticoid signaling as a long-term modifier of hippocampus-dependent behavior in hBDNFVal66Met mice.

The implications for the novel data reported here is that it is the first to provide experimental evidence that a history of glucocorticoid signaling during adolescence, a bottom-up model of chronic stress, determines adult hippocampus-dependent memory function according to BDNF Val66Met genotype. Outside of already identified sampling factors (e.g., underpowered clinical studies),3 these results suggest that discrepant clinical data on the topic of hippocampus-dependent memory function may be explained by a failure to stratify samples for stressful life events, in that the 66Met allele may be associated with poor memory function at baseline but may recover to the levels similar to, if not better than, controls following a history of stress. In particular, it appears although this effect occurs via the innate susceptibility of Met/Met homozygotes to CORT due to an increase in the expression of glucocorticoid receptors in the dorsal hippocampus during adolescence (see Supplementary Figure 2), which has long-lasting effects on hippocampus-dependent behavior into adulthood.

The data reported here could be interpreted as a protective mechanism of stress in 66Met carriers; whereby, in the absence of glucocorticoid exposure during late adolescence the brain may be more vulnerable to stress in adulthood. However, the clinical literature would suggest that the data reported here is more likely to represent an ‘undesirable gain of function’ than a ‘positive adaptation’. Specifically, risk of depression has been selectively linked to the 66Met allele in females with a history of childhood stress,8 while a history of adverse events interacts with 66Met genotype to increase rumination.9 In non-humanized 66Met mice exposed to acute stress, there is also an increase in depression-related behavioral markers such as learned helplessness on the forced-swim test.10 Ultimately, the enhanced memory of fear – while possibly related to depressive disorders – is likely to hold more relevance to anxiety disorders such as PTSD where stress is a requisite factor in the pathogenesis of the disorder. Although the Val66Met polymorphism has been only briefly investigated in PTSD patients, there is evidence that the 66Met allele is carried with a two- to threefold higher frequency in ‘probable’ PTSD probands11 and confers resistance to exposure therapy.12 Further, the 66Met allele has been associated with the persistence of fear in an extinction-learning paradigm in both man and mouse.13 Adding to this literature, our data suggest that there is a long-term effect of glucocorticoids in 66Met carriers that potentiates the fear circuitry into adulthood, which may increase susceptibility to trauma, events with negative emotional valence and related psychopathology.

Although clinical studies are required to confirm that the phenotype described here replicates in humans, the current data provides the first evidence for a long-term glucocorticoid-mediated ‘switch’ of hippocampus-dependent behavior in the hBDNFVal66Met mouse, as well as a theoretical framework from which to resolve a putative role of BDNF in anxiety and affective disorders.