The association between polygenic hazard scores and clinical markers of Alzheimer’s disease following stratification for APOE genotype

Abstract

AbstractBackgroundGenome‐wide association studies have confirmed the APOE ε4 allele as the single‐nucleotide polymorphism (SNP) best associated with Alzheimer’s disease. These same studies, however, have found that a considerable number of other SNPs are associated with Alzheimer’s disease. The statistical association yielded by these other SNPs, however, is considerable weaker than that of the APOE ε4. It is unclear, therefore, whether these non‐APOE SNPs have similar effects in APOE ε4 carriers and non‐carriers. In this study, we investigated the association between a polygenic hazard score (PHS) for sporadic Alzheimer’s disease and clinical profiles of a large cohort non‐demented participants stratified by APOE ε4 status.MethodA cohort of 776 participants was shortlisted from the entire ADNI database: 262 cognitively unimpaired participants (185 with a ε3ε3 genotype and 77 with a ε4ε3 genotype) and 514 participants with a diagnosis of MCI (241 with a ε3ε3 genotype, 212 with a ε4ε3 genotype and 61 with a ε4ε4 genotype). This same cohort was also split according to cerebrospinal amyloid (Aβ) positivity status, resulting into 275 Aβ‐ and 501 Aβ+ participants. Genome‐wide data, a structural brain MRI and cognitive profiles were available for each dataset. Genetic information was used for the calculation of a PHS, following the methodology by Desikan and colleagues [1]. Memory and executive composite scores were calculated for each dataset based on previously described procedures [2,3]. A voxel‐based morphometry was carried out to process T1‐weighted MRI images and obtain grey‐matter maps analysable at a group level. Linear models were devised in each sub‐cohort and for each APOE ε4 status to test the association between PHS and cognitive composites and between PHS and volumetric grey‐matter maps.ResultPHS was a predictor of mediotemporal volume in MCI patients and a predictor of memory and executive functioning both in MCI and Aβ+ participants. Following APOE stratification, PHS predicted memory and executive functioning in ε4ε3 MCI patients, memory in ε3ε3 MCI patients, and memory in ε4ε3 Aβ+ participants. PHS also predicted volume in sensorimotor regions in ε3ε3 MCI and ε3ε3 Aβ+ participants.ConclusionThese results suggest that clinical phenotypes might be influenced by complex genetic interactions.