Longitudinal GWAS Identifies Novel Genetic Variants and Complex Traits Associated with Resilience to Alzheimer’s Disease

Abstract Background We completed a large genetic analysis of resilience to cognitive decline in Alzheimer’s Disease (AD) and discovered novel variants, genes, and complex traits associated with better‐than or worse‐than‐expected cognitive performance given an individual’s age, sex, and APOE genotype. Method Leveraging 15,933 non‐Hispanic white participants across four longitudinal cohort studies of aging and AD (Figure 1), our group determined the effects of genetic variants on resilience metrics using mixed‐effects regressions. Models adjusted for age, sex, APOE ε4 allele count, presence of the APOE ε2 allele and all covariate interactions with interval (years from baseline). The outcomes of interest were residual cognitive resilience, quantified from residuals in three cognitive domains (memory, executive function, and language), and combined resilience, summarized as the covariance of educational attainment with residual cognitive resilience. Post‐GWAS analyses included gene tests using MAGMA and estimates of genetic correlation with 65 complex traits using GNOVA. Result We observed genome‐wide significant associations at multiple established AD loci, including BIN1 and CR1 (Figure 2). We observed a novel association with combined resilience on chromosome 13 (top SNP: rs11838654, MAF = 0.06, P = 4.7×10 −8 ; Figure 3) and a novel signal on chromosome 1 approaching significance (top SNP: rs2817183, MAF = 0.41, P = 5.1×10 −8 ). Interestingly, rs11838654 is an eQTL for four genes in hippocampus ( WBP4 , COG6 , MRPS31 , and NHLRC3I ; Braineac database). We also observed an association with residual cognitive resilience on chromosome 5 that approached genome‐wide significance (top SNP: rs4482935, MAF = 0.25, P = 5.5×10 −8 ; Figure 2). Gene‐level tests identified associations of CD2AP (P.fdr = 0.027) and ZNF146 (P.fdr = 0.049) with residual cognitive resilience and combined resilience, respectively. Additionally, we identified negative genetic correlations of combined resilience with ischemic stroke and coronary artery disease (all P.fdr<2.5×10 −2 ; Figure 4). Conclusion Compared to models of resilience that regress out the effects of AD neuropathology on cognition, the present models benefit from larger sample size at the cost of molecular precision. Although the genetic architecture of resilience from these less precise models more closely resembles that of clinical AD, we uncovered novel genetic drivers of resilience through this approach. Such findings will require future replication but suggest a trajectory‐based definition of resilience holds substantial promise for discovery.