Exome Array Analysis of 9,721 ischemic stroke cases from the SiGN Consortium

Abstract

Recent studies have identified > 40 genetic variants robustly associated with ischemic stroke, most identified through genome wide association studies and primarily marking common variants in non-coding regions presumed to have regulatory roles on gene and protein expression. To evaluate the contribution of coding variants, which are mostly rare, to the etiology of ischemic stroke, we performed an exome array analysis of 9,721 ischemic stroke cases with mean age of onset at 67.1 years from the SiGN Consortium, and 12,345 subjects with no history of stroke (mean age 67.0 years) from the Health Retirement Study and SiGN consortium. Both cohorts included people with diverse ancestries. Genotyping for both SiGN and HRS was performed using similar array content at the Center for Inherited Disease Research (CIDR), albeit as two separate studies. Following extensive SNP- and sample-level quality control, a total of 106,101 SNPs from the exome content was used for exome association analysis. We identified 15 coding variants significantly associated with all ischemic stroke at array-wide threshold for statistical significance (i.e., p < 3.6 x 10-7) that also showed good genotyping quality, including two common SNPs in ABO that have previously been associated with stroke. Twelve of the remaining 13 variants were extremely rare in European Caucasians (MAF<0.1%) and the associations were driven by substantially higher allele frequencies in African American cases than in African American controls. A variant in PRIM2, rs199585353, was present exclusively in the stroke cases of European Caucasians while absent in all other samples from our data. There was no evidence for replication of these associations in either TOPMed Stroke samples (n = 5613 cases) or UK Biobank (n = 5,874 stroke cases), although power to replicate was very low given the low allele frequencies of the associated variants. In conclusion our analyses revealed 13 novel associations, but the low allele counts of associated variants and difficulty in acquiring large, well-powered replication highlight the challenges of rare variant association analysis, especially using array-based genotyping technologies.