Genome‐wide analysis of longitudinal Alzheimer’s disease biomarker endophenotypes

Abstract

AbstractBackgroundAlzheimer’s disease (AD) is characterized by a pathophysiological cascade, in part detectable using imaging and fluid biomarkers, that evolves over many years. Our goal was to study the interaction between genetic risk and longitudinal trajectories of selected AD endophenotypes using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset.MethodLongitudinal data included amyloid beta (PET and CSF), total tau and phosphorylated tau (CSF), FDG PET (3 ROIs), MRI measures of atrophy (8 ROIs) and white matter hyperintensities (WMHI), and composite scores from cognitive tests for memory and executive function (full list of endophenotypes, sample size, and covariates in Table 1). Genetic association with longitudinal phenotypes was performed using Linear Mixed Modelling (LMM; LME4 R package), modeling the SNP main effect and SNP x time effects, with time x subject as a random effect and participant age as the time variable. SNP P‐values from the LMM were analyzed for gene enrichment using MAGMA, with a 10kb boundary around genes for SNP to gene assignment. A stringent significance threshold was set to P ≤ 6.25 × 10−9 based on Bonferroni correction for the top 8 principal components explaining 85% of variance across all tested phenotypes. LD trimming was performed to identify top SNPs for genetic regions. MAGMA gene based association was FDR corrected to P ≤ 0.05.ResultMultiple genetic regions (26) were associated with the main SNP effect, including the APOE ε4 locus, which was significant on all measures except WMHI. Time interaction identified the APOE ε4 allele as study wide significant with cognitive measures and 5 MRI ROIs. Apart from APOE, 7 SNPs were identified that met genome‐wide significance, including SNPs intronic to BORCS5, HLA‐DPA1, and GRIN3A (Figure 1). Gene based association identified 21 genes, 14 primarily associated with MRI atrophy measures, 3 WMHI, 2 with CSF tau, 1 with amyloid PET, and 1 with cognitive measures (Figure 2).ConclusionGenetic association analysis of longitudinal trajectories of AD biomarker phenotypes identified both genes and pathways previously implicated with AD, as well as several novel loci. These findings may point to new targets for diagnostic and therapeutic development.