APOE*4-stratified genome-wide association study of Alzheimer's disease in over 350,000 individuals.

APOE*4 is the strongest genetic risk factor for late-onset Alzheimer's disease (AD) and is highly pleiotropic, such that it may be considered as a biological factor that can affect overall genetic risk for AD. To advance our understanding of the genetic architecture of AD, we sought to perform the largest APOE*4-stratified genome-wide association study (GWAS) of AD.Twenty-five publicly available AD GWAS datasets provided case-control diagnoses for phase-1 samples (imputed to the HRC r1.1 reference panel). The UK Biobank provided subjects with family history of AD status, transformed into an AD phenotype as described previously (Jansen et al., 2019) for phase-2 samples. Linear mixed model regressions were performed on case-control status (LMM-BOLT v.2.3.4), adjusting for age (age-at-onset in cases; age-at-last-exam in controls), sex, APOE*4 and APOE*2 dosage, the first 12 genetic principal components, array/batch, cohort in phase-1, and assessment center in phase-2. In phase-3, phase-1 and phase-2 findings were combined using multivariate genome-wide meta-analysis (Jansen et al., 2019). APOE*4+ heterogeneity tests were evaluated per phase and meta-analyzed in phase-3.Participant demographics are in Table 1. Combining results from both APOE*4-stratified analyses, 106 lead variants across 98 loci passed suggestive significance (p<10-5 ; Figure 1). Although most variants reached only suggestive significance, 28 loci were previously reported at genome-wide significance in large-scale GWAS of AD (Kunkle et al., 2019, Jansen et al., 2019, Bellenguez et al., 2020), supporting that we identified potentially relevant AD loci. APOE*4-stratified effects were observed for 28 variants/loci covered across both phase-1 and phase-2 (NAPOE4+ =19; NAPOE4- =9; Table 2), and 25 variants/loci seen only in phase-2 (NAPOE4+ =17; NAPOE4- =8; Table 3). Notably, a genome-wide significant APOE*4+ heterogeneity effect was observed for the USP17L13 locus (a regulator of deubiquitination), while PPP1R12A, BRINP1, PCBD1, and SESN2 loci passed suggestive significance.Our findings revealed novel AD risk loci/genes and characterized which of these associated with AD risk differentially across APOE*4 status. This contributes highly to personalized genetic medicine and paves the way towards new potential AD drug targets. Ongoing work is adding samples for phase-1 analyses (imputing data to TOPMed) and pursuing both multi-omics and AD endophenotype validation efforts for variant prioritization.