Association of Common and Rare Variants with Alzheimer’s Disease in 16,905 individuals with Whole‐Genome Sequence Data from the Alzheimer’s Disease Sequencing Project

Abstract

AbstractBackgroundWhole genome sequencing (WGS) data enables assessment of common and rare variants for association with human diseases. Here we report our analysis using ADSP R3 WGS data to discover genetic variants associated with AD.MethodWe performed association testing of common variants (minor allele frequency [MAF] > 0.5%) as well as aggregates of rare (MAF < 1%) coding and non‐coding variants. Given that ADSP is composed of multiple ancestry groups, we performed association testing across all participants (pooled population [PP], Ncases = 6,519 and Ncontrol = 6,852) and within the following populations: African Americans (AA, Ncases = 1,137 and Ncontrol = 1,707), Hispanics (HIS, Ncases = 1,021 and Ncontrol = 1,988), and Non‐Hispanic White (NHW, Ncases = 4,230 and Ncontrol = 3,109) defined by self‐reported race and ethnicity (Table 1).ResultAs expected, our most significant associations are at the APOE locus (chr19:44,905,796‐44,909,393) (Figure 1). In the pooled‐population analysis, we observed known AD associations for BIN1 (p = 3.2×10−09) variants with AD status. In the NHW analysis, we observed AD associations for variants in CR1 (p = 7.97×10−08), BIN1 (p = 2.16×10−08), ARHGEF33 (p = 4.85×10−8), ANK3 (p = 1.43×10−12) and MYCBP2‐AS1 (p = 4.59×10−08), and in the AA analysis, we observed a significant variant near LINC00320 (p = 1.5×10−8). In the HIS subset, a region (chr14:72,698,038‐75,846,454) has 31 significant variants (p = 3.9×10−11) crossing multiple genes. This region covers PSEN1 G206A (rs63750082), a well‐known early‐onset AD mutation. Local‐ancestry analysis shows that the mutations, including G206A, are of African ancestry. Using kinship coefficients calculated within this dataset, it was determined that the signals are not driven by a familial linkage disequilibrium.We observed gene‐based aggregation of coding variants in PSEN1 in the pooled‐population and HIS analyses (p<5×10−8) (Figure 2). In our gene‐centric noncoding aggregates analysis, we identified noncoding variants in the promotor of TOMM40 in the pooled‐population analysis (p = 7.2×10−08). After adjusting for APOE e4 (rs429358) and e2 (rs7412) alleles, there was an attenuation, but the association remained suggestive (p = 7.1×10−06).ConclusionWe performed common and rare association analysis of AD on ADSP WGS data. We observed associations with variants in known AD loci. We also observed population‐specific associations and an association of AD with rare noncoding variants in TOMM40, indicating WGS in diverse populations can lead to novel genetic associations with AD.