O3–01–05: The role of functional genetic variation in Alzheimer's disease: The CHARGE consortium

The heritability of Alzheimer's disease (AD) is substantial and estimated to be around 60%. Relatively common SNPs with modest effects on AD risk have been identified but they explain only a fraction of the heritability. There is increasing evidence that the “missing” heritability may be explained in part by the AD risk being modulated by lower frequency variants, not routinely available on genome-wide association arrays. The Exome-Chip (EC) is a comprehensive genotyping array of functional, mostly rare, variation in the human exome. Within the CHARGE consortium, four studies with 1,273 AD cases and 10,531 controls of European ancestry were genotyped and jointly called using the Illumina V1.0 EC. The calling was based on over 60,000 samples, which significantly increased genotype accuracy for the very rare variants that are challenging to call in cohorts with small sample sizes. The CHARGE sample includes genotypes for 238,876 variants with 82.9%, 5.7%, 11.5% having minor allele frequency (MAF) ≤1%, 1 to 5%, and >5%, respectively. 90% of the variants are predicted to alter the protein structure and 45% are to be damaging. We investigated the variation available on the EC in 16 previously discovered AD genes: ABCA7, APOE, APP, BIN1, CD2AP, CD33, CLU, CR1, EPHA1, MS4A4A, MS4A6A, PICALM, PSEN1, PSEN2, SORL1, TREM2. Fixed effects meta-analysis was applied to logistic regression results from individual cohorts. The meta-analysis included 189 polymorphic SNPs and replicated association of several genes. As expected, the strongest association is found with a common (MAF 11%) intronic variant, rs769449, at APOE(OR 2.4, p-value 10 -30). Importantly, associations to very rare variants (MAF<0.5%) were also found. Of particular interest are two predicted damaging SNPs: rs138650483 in MS4A6A is splice variant which was observed with the same direction of effect in all four cohorts (OR=2.3, p-value 0.01) and rs117411388 in PICALM is a predicted damaging missense variant (OR=5.4, p-value 0.006). We have identified associations of novel rare functional SNPs with AD in genes previously associated with AD, strengthening the hypothesis of the involvement of those genes in AD. Further study and sequencing may identify additional rare variants in these genes and other novel regions.