Abstract
Brain-derived neurotrophic factor (BDNF) plays an important role in hippocampal synaptic plasticity and long-term potentiation. A valine (Val) to methionine (Met) substitution in the BDNF gene (Val66Met) has been associated with episodic memory performance. This study aimed at fine-mapping the genomic region harbouring BDNF and the adjacent gene, BDNFOS, in order to identify other possible memory-related gene variants. Healthy young Swiss adults (n=333) underwent a verbal memory free-recall task which used words with both neutral and emotional content. Genetic variability of the BDNF and BDNFOS region was covered by analysing 55 single nucleotide polymorphisms (SNPs). Among all examined SNPs, the non-synonymous Val66Met SNP rs6265 showed the highest significant level of association with memory performance for words with emotional content. Recall performance for neutral words was unrelated to the analysed SNPs. Our results support a role for the Val66Met BDNF polymorphism in episodic memory and suggest a modulatory influence of emotional valence.
Introduction
Human episodic memory is a polygenic trait. Twin studies suggest that 50% of the phenotypic variance can be explained by inherited factors (McClearn et al.1997). Because episodic memory is heritable and polygenic, genetic association studies are a powerful tool to discover genes associated with this trait (de Quervain et al.2003; Egan et al.2003; Papassotiropoulos et al.2006).
Brain-derived neurotrophic factor (BDNF) is a member of the nerve growth factor family (Huang & Reichardt, 2001) and plays an important role in long-term potentiation (LTP), synaptic plasticity and influences the growth and maintenance of neurons (Martinowich et al.2007). Intracellular processing of BDNF is complex. At the transcriptional level numerous promoters and splicing variants are involved (Pezawas et al.2004). At the protein level, BDNF is first synthesized as a precursor protein, proBDNF, which is then proteolytically cleaved to become the final peptide termed mature BDNF (Pang et al.2004; Woo et al.2005). It has been shown that proBDNF and mature BDNF exert distinct effects on cellular function with proBDNF facilitating long-term depression via the p75NTR receptor, and mature BDNF facilitating LTP via the tyrosine kinase B receptor (TrkB), a phenomenon called the ‘yin and yang’ hypothesis of BDNF (Lu et al.2005).
The gene encoding BDNF is located on chromosome 11p14 and contains at least one non-synonymous single nucleotide polymorphism (SNP), which results in a valine (Val) to methionine (Met) substitution in codon 66. This SNP is located in the 5′ domain of proBDNF (dbSNP number rs6265). At the molecular level, this polymorphism influences intracellular trafficking and packaging of proBDNF and the secretion of the mature form of the protein. At the human cognitive and brain-function level carriers of the Met allele, which is associated with lower depolarization-induced secretion of the protein, showed poorer episodic memory performance, abnormal hippocampal activation and lower levels of hippocampal N-acetyl aspartate, a marker of neuronal integrity (Egan et al.2003; Stern et al.2008). Furthermore, it was shown that Met carriers have smaller hippocampi (Bueller et al.2006) and decreased grey-matter density of the prefrontal cortex (Szeszko et al.2005).
A number of studies have examined the association between the Val66Met polymorphism and episodic memory. For example Hariri et al. (2003), Tan et al. (2005), Dempster et al. (2005) and Goldberg et al. (2008) found impaired episodic memory performance in Met carriers compared to Val/Val homozygotes whereas Harris et al. (2006) and Strauss et al. (2004) did not find a significant association (see also Sabb et al.2009). These ambiguous results and the use of different psychological tests for the similar phenotype show the importance of further studies to examine whether memory performance is influenced by the Val66Met polymorphism and if so which subtype of memory may be influenced (Goldberg et al.2008).
Recent studies have shown an association of Val66Met with anxiety-related behaviour in mice, where BDNFMet mice exhibited significantly increased anxiety-related behaviour, which was refractory to antidepressants (Chen et al.2006). Therefore, BDNF is expected to play an important role in psychiatric disorders (Martinowich et al.2007; Pezawas et al.2008). From these genetic association studies and animal experiments it appears that BDNF may be associated with different and potentially overlapping phenotypic traits related to cognition and emotion. Importantly, human genetic association studies have focused on the non-synonymous SNP rs6265 and have not yet studied the impact of adjacent variations within and in the vicinity of BDNF on memory-related traits. Interestingly, a recent study has demonstrated the existence of several alternative BDNF transcripts, different transcription initiation sites and several functional promoters. Furthermore, BDNF transcripts form dsRNA molecules with anti-BDNF transcripts (BDNFOS), located on the antisense strand (Pruunsild et al.2007). In the present study we therefore performed fine-mapping of the entire BDNF and BDNFOS region and focused on the association of BDNF and BDNFOS SNPs on verbal episodic memory for both emotional and non-emotional material.
Methods
Participants
Memory testing and genotyping were performed in 333 healthy young Swiss subjects [232 females, 101 males, age range 18–45 yr, 22.8±3.8 yr (mean±s.d.)]. To control for effects of educational level on memory, we recruited both university students (academic group) and age-matched employees/trainees who were not studying at the university and did not have a university degree (non-academic group). All subjects were of Caucasian genetic background. After complete description of the study to the subjects, written informed consent was obtained. The study protocol was approved by the ethics committee of the Canton of Zurich, Switzerland. Blood was drawn using 2×9 ml EDTA tubes (Sarstedt, Germany) and column-based DNA isolation was performed according to Qiagen's (Hilden, Germany) standard protocols.
Assessment procedure
Subjects viewed six series of five semantically unrelated nouns presented at a rate of one word per second with the instruction to learn the words for immediate free recall after each series. Five minutes and 24 h after presentation of these words, subjects underwent unexpected delayed free-recall tests for the same words. The design of an unexpected free-recall test was chosen to prevent confounding factors such as different cognitive strategies and unbalanced reiteration of the learned words. The German words were taken from the collections of Hager & Hasselhorn (1994) and consisted of 10 neutral words [valence 0.2±0.3 (mean±s.e.m.), rating on a scale from −10 (most negative valence) to +10 (most positive valence)], 10 positive words (valence 7.9±0.4) and 10 negative words (valence −8.5±0.3). Subjects were instructed to write down the previously learned words, each word on a separate sheet. The number of correctly recalled words (hits) was the relevant output.
SNP selection and genotyping
The selection of SNPs spanning the BDNF and BDNFOS region was performed with SNP browser 3.5 (De La Vega, 2007). Initially, 78 SNPs were chosen spanning the genomic region of BDNF and BDNFOS (247 kb) on chromosome 11. Twenty-three of the selected SNPs did not meet the criterion of a minor allele frequency (MAF) of ⩾0.10 and were therefore excluded. Selected SNPs covered four independent haplotype blocks with an average between-SNP r2 of 0.8. SNPs were genotyped with pyrosequencing on a PyroMark™ ID system (Biotage, Sweden) and with the Genome-Wide Human SNP Array 6.0 (Affymetrix, USA). Genotype distributions were in Hardy–Weinberg equilibrium. dbSNP numbers, minor allele frequencies and the physical position of the SNPs on chromosome 11 are listed in Appendix A (available online).
Data analysis
Statistical analysis was performed using Golden Helix SNPs and variation software (Helix Tree 6.4.0, www.goldenhelix.com) and SPSS version 16.0 (SPSS Inc., USA). Associations were calculated using both univariate and repeated-measures analysis of variance (ANOVA) controlling for gender and education. Statistical significance was assumed when p 0.05. Given an average effect size of d=0.3 for the rs6265 SNP (Sabb et al.2009) the power of our study to detect a significant effect at an alpha error probability of 0.05 (one-tailed) and an allocation ratio of 0.4 was 85%. Power calculation was done with G*Power 3 (Faul et al.2007).
Results
Of the 55 SNPs analysed in the 333 individuals, five were significantly associated with free recall performance for words with positive emotional content after 24 h. The most significant SNP was rs6265, which predicts the valine to methionine substitution (Val66Met) in the pro region of BDNF. From the 333 subjects included in the analysis 203 were Val/Val homozygous, 115 Val/Met heterozygous and 15 Met/Met homozygous. Val/Val homozygous subjects showed better free recall performance for words with positive emotional content than Met carriers (p=0.028) (Table 1). Repeated-measures ANOVA with time (i.e. 5 min and 24 h) and emotional valence (i.e. positive, negative, neutral) as within-subjects effects confirmed this result (p=0.008 for time×valence×SNP effect). Time and valence effects separately were also highly significant (p<0.001). A further significant association with the same phenotype was observed in the BDNFOS region that included a cluster of three SNPs (rs7127239, rs7125904, rs10835190) reaching a nominally significant level (p=0.048, 0.048, 0.049, respectively). This cluster is in the same haplotype block as the Val66Met polymorphism (Fig. 1). Finally, rs10835218, located downstream of BDNF showed a similar significant association (p=0.045). Free recall performance for words independent of their emotional content, neutral words and words with negative emotional content 24 h after presentation was not related to the SNPs analysed herein. None of the 55 SNPs spanning the BDNF and BDNFOS region was associated with free recall after 5 min (total word hits, words with positive, negative or neutral content). Neither gender nor educational status interfered with the effect of SNP rs6265 on free recall performance for words with positive emotional content (all interactions p>0.5).
Upper panel: the blue dots indicate the 55 genotyped SNPs and their significance of association with free-recall performance for words with positive emotional content 24 h after learning. Lower panel: vertical blue and grey bars indicate the genomic position of known SNPs. Genes are represented by green horizontal bars. Haplotype blocks are adapted from HapMap (CEU population). Red squares indicate high linkage disequilibrium (LD), light blue squares low LD.
Association of the Val66Met SNP with verbal episodic memory
SD, Short delay (word recall after 5 min); 24 h, word recall after 24 h; Val, valine; Met, methionine.
Values are means of word hits (±s.d.); p=level of significance.
Association of the Val66Met SNP with verbal episodic memory
SD, Short delay (word recall after 5 min); 24 h, word recall after 24 h; Val, valine; Met, methionine.
Values are means of word hits (±s.d.); p=level of significance.
Discussion
In the present study fine-mapping the genomic region harbouring BDNF and BDNFOS showed a significant association of the SNP rs6265 (Val66Met) with the recall of words with positive emotional content 24 h after learning. Specifically, Val/Val homozygous subjects had better memory performance than Met carriers. In contrast, the recall performance for words with neutral and negative valence was unrelated to the genotype.
The non-synonymous Val66Met SNP has been associated with a number of different phenotypes related to human memory performance (Bueller et al.2006; Chen et al.2006; Egan et al.2003; Goldberg et al.2008; Montag et al.2008). Linkage with other potentially functional polymorphisms, which may be slightly different from population to population, might partially explain this phenotypic heterogeneity. However, the present study shows that out of 55 SNPs the most significant memory-related SNP was the Val66Met polymorphism. This is indeed of note, and indicates that in the era of genome-wide studies, where SNPs are selected by position, hypothesis-driven studies based on a-priori biological information and evidence for SNP functionality are still a powerful and valid approach.
The effect of the Val66Met polymorphism was confined to free recall performance for words with positive emotional content. There was no significant association of any of the 55 SNPs spanning the BDNF and BDNFOS region with the total number of recalled neutral words after either 5 min or after 24 h. Thus, these results are consistent with the findings of Egan et al. (2003) who reported lack of association between the recall performance for single, unrelated words and SNP rs6265. The results of the present study rather suggest a role for BDNF in memory performance when the presented words have emotional content. Indeed, a number of studies support a role for BDNF in emotion-related traits. Genetically engineered mice with the Met variant show increased anxiety-related behaviour compared to Val-carrying mice (Chen et al.2006). In humans, carriers of the Met variant show stronger amygdala activation than Val carriers in the right hemisphere in response to emotional stimuli compared to neutral stimuli (Montag et al.2008). In the present study, words with positive but not with negative or neutral valence showed a significant association with the Val66Met polymorphism. Although causal and mechanistic interpretations of genetic association findings must be treated with caution, one could speculate that this observation is related to different neural processing of words with different emotional valence. Interestingly, switching between distinct behavioural states can be triggered by selective activation of specific neuronal circuits (Herry et al.2008).
In contrast to short-term memory, long-term memory (i.e. the ability to retain information for several hours, days and even years) depends on de-novo protein synthesis and on long-term changes in the molecular components of the neuronal synapse (Kandel, 2001). The fact that the Val66Met polymorphism was significantly associated with recall performance after 24 h but not after 5 min argues for a role of BDNF in protein synthesis-dependent memory consolidation. However, studies in larger populations should follow to rule out the possibility that the lack of associations presented herein is not merely due to low statistical power. Another possible limitation of our study is also the fact that the reported significances are not corrected for multiple comparisons (i.e. 55 SNPs, six phenotypes). However, given the strong a-priori biological hypothesis for the studied gene locus, the substantial inter-correlation of the SNPs and phenotypes assessed herein, and the confirmation of the hypothesis-driven, functional SNP being the most significant one, we believe that the results are not biased towards false-positivity.
Taken together, we provide further support for a role of BDNF in human memory that might have implications for psychiatric diseases related to encoding, storage, and retrieval of emotional information.
Note
Supplementary material accompanies this paper on the Journal's website.
Acknowledgements
Supported by grants from the Swiss National Science Foundation to D.Q. (PP00B-106708; CRSIK0_122691) and A.P. (CRSIK0_122691), from the European Science Foundation to D.Q. and A.P. (EUROStress) and from the 7th framework programme of the European Union (ADAMS, project no. 242257, FP7-HEALTH-2009). We also thank Ms. Kim-Dung Huynh for expert technical assistance.
Statement of Interest
None.
References
