Apolipoprotein E moderates the association between Non‐APOE Polygenic Risk Score for Alzheimer’s Disease and Aging on Preclinical Cognitive Function

Abstract

BackgroundApolipoprotein E4 (APOE‐ε4) is the strongest known risk factor for late‐onset Alzheimer’s disease (LOAD). Variation in cognitive decline patterns among cognitively unimpaired individuals suggests that additional LOAD‐related genetic factors may interact with APOE‐ε4 to influence cognitive decline during asymptomatic aging. In the present study, we aim to investigate whether any genome‐wide association studies (GWAS) and p‐value threshold (pT)‐informed non‐APOE polygenic risk scores (PRS) moderate associations between APOE‐ε4 and longitudinal cognition.MethodsWe leveraged 12 years of longitudinal data from the Wisconsin Registry for Alzheimer’s Prevention, and we built eight PRSs under various pTs for single nucleotide polymorphism selection using summary statistics from three GWAS meta‐analyses: International Genomics of Alzheimer’s Project (IGAP) case‐control study, UK‐biobank (UKBB) AD‐by‐proxy study, and a combined sample that merges IGAP and UKBB study results. All association analyses were fitted using a linear mixed effects model and adjusted for within‐individual/family correlation among 1,190 cognitively unimpaired individuals at baseline (Table 1). A set of sensitivity analyses were performed to assess the robustness of the main findings.ResultsWe found that the adverse effect of non‐APOE PRS on overall and memory‐related cognitive domains is more evident among APOE‐ε4 carriers as people age (Figure 1). There is a gradual, statistically significant difference in the rate of overall and memory‐related cognitive decline per unit change of non‐APOE PRSs between APOE‐ε4 carriers and non‐carriers for people over the age of 70 (Figure 2). Our study findings are consistent with a relatively stringent pT (p <= 5e‐8) and more evident when the PRSs are constructed using the case‐control GWAS study results, despite the smaller sample size (“Kunkle significant variants only” in Figures 1 and 2). Our findings are also robust to various sensitivity checks when we replace the dichotomous APOE‐ε4 predictor with the APOE genotype, use different PRS clumping parameters, and exclude cognitive measurements from the last test occasion, which has a smaller sample size.ConclusionWe recommend considering the interaction between non‐APOE PRS, APOE status, and age in modeling pre‐clinical cognitive phenotypes for LOAD. We also recommend leveraging a conservative p‐value threshold and using case‐control GWAS results in constructing the LOAD PRS.