Brain-derived neurotrophic factor (BDNF), a neutrophin highly expressed in the hippocampus, has been associated with hippocampal-dependent learning and memory processes.1 A polymorphism in the BDNF (BDNFval66met; rs6265) gene causing a valine (val)-to-methionine (met) substitution at codon 66 results in altered intracellular trafficking and packaging of BDNF, and in a reduction of its regulated secretion.1 BDNFval66met genotype predicts variation in human episodic memory, as well as in hippocampal anatomy and function.1,2 In rats, BDNF-mediated beneficial effects on neuroprotection, memory ability and learning decrease with advancing age.3 Interestingly, hippocampal activity in humans,4 as well as the expression of BDNF and its receptor5 in this region also decrease with age. In line with these findings, we expected that age-related decline of hippocampal function would be modulated by genetically determined variation in BDNF function, such that BDNF-met allele carriers (expressing diminished regulated secretion of BDNF) would show a more pronounced decrease in memory-dependent hippocampal activity with advancing age relative to BDNF-val/val individuals.
We studied 125 healthy, Caucasian participants (age: 19–85 years, 65 females, see Supplementary Materials, Table 1) during the performance of a simple declarative memory task that included incidental encoding and retrieval phases using whole-brain BOLD fMRI (blood–oxygen-level-dependent functional magnetic resonance imaging), as described previously.6 Participants were genotyped for BDNFval66-met, apolipoprotein E and Catechol-O-methyltransferase polymorphisms.6 On the basis of BDNFval66met genotype, they were subdivided into met-carriers (met/met = 9; val/met = 36) and val/val homozygote individuals (n = 80). Demographic features (IQ, education, age, gender and handedness), apolipoprotein-E and catechol-O-methyltransferase polymorphism genotypes, and retrieval performance (accuracy and reaction time) were not significantly different across BDNFval66met genotype groups. Random-effects general linear model simple regressions with age as a predictor on encoding > fixation and retrieval > fixation individual activation maps were computed in SPM5 (http://www.fil.ion.ucl.ac.uk/spm). On the basis of our a priori hypothesis for an age-by-BDNFval66-met interaction in hippocampal function, we extracted and analyzed BOLD signal change from the most significant clusters in left and right hippocampal regions, as identified by WFU-PickAtlas (http://www.fmri.wfubmc.edu), for each condition with a statistical threshold of P < 0.05 corrected for multiple comparisons using a false discovery rate with α = 0.05. The coefficients of the correlations between signal-change and age were then compared between BDNFval66met groups using one-tailed Fisher’s r-to-z′ transform. To ensure that these changes were not driven by BDNFval66met effects on structural changes in the hippocampus, we analyzed a subsample (17 met-carriers, 35 val/val; matched for demographics, and behavioral performance, apolipoprotein-E and catechol-O-methyl transferase polymorphism genotypes, see Supplementary Materials, Table 2) that also underwent three-dimensional structural magnetic resonance imaging using a T1-weighted SPGR sequence.2 Structural images were segmented, spatially normalized and modulated using the unified segmentation approach in SPM5. Hippocampal volumes were calculated from the resulting grey-matter images using WFU-PickAtlas. Multiple regressions on hippocampal activations for each condition, with age and hippocampal volume as covariates, were than computed and partial correlation coefficients were compared as above.
There was a significant age-related decline in activation of posterior hippocampal region bilaterally during encoding and retrieval phases (see Supplementary Materials, Table 3). When analyzed by BDNFval66-met genotype, met-carriers showed a significantly steeper slope in age-related decline in hippocampal activation bilaterally during encoding and retrieval phases relative to val/val individuals (Figure 1; Supplementary Materials, Table 4). These results remained significant in the hippocampal region bilaterally during encoding and in the left hippocampal region during retrieval, even after using the hippocampal volume as a covariate (see Supplementary Materials). We also found increased bilateral inferior frontal activity with increasing age during retrieval (see Supplementary Materials, Table 5, Figure S1A), possibly reflecting a compensatory response to maintain performance.4 Met-carriers showed greater activation in this region relative to val/val individuals, but there was no age by BDNFval66met interaction (see Supplementary Materials, Figure S1).
Consistent with our hypothesis, we found that BNDFval66met genotype modulates age-related changes in hippocampal function. Met-carriers showed a greater age-related decline in hippocampal activation during both encoding and retrieval tasks relative to val/val individuals, and this effect was independent of any potential structural2 or performance differences.6
These data are in line with earlier evidence for an effect of BNDFval66met polymorphismon hippocampal anatomy2 and function.1,6 Most importantly, they illustrate the modulatory effect of this polymorphism on the trajectory of age-related changes in the neurophysiology underlying episodic memory. The decline of hippocampal function is accelerated in met-carriers relative to val/val individuals, suggesting that the latter may be more resilient to age-related changes in hippocampal-dependent declarative memory. These findings add to evidence for a critical role of genes in the heterogeneity of age-related decline in cognition across individuals.