Genome‐Wide Mediation Analysis Reveals the Genetic Underpinnings of the Neuromorphometric Endophenotypes of Alzheimer’s Disease

Abstract

AbstractBackgroundAlzheimer’s disease (AD) is a common neurodegenerative disorder. AD has high heritability, estimated to be 58%‐79% from twin studies. The strong genetic basis motivates the investigation of AD etiology. AD progression can be captured, in part, by brain volumetric variations, quantified by magnetic resonance imaging (MRI). Brain volumetric variations have been used to capture in vivo imaging biomarkers of AD. Advances in recent brain imaging genetics studies provide an opportunity to elucidate the neuropathogenesis of AD.MethodWe performed a genome‐wide mediation study (GWMS) investigating genetic effects on AD classification that are mediated by regional brain volumes. Specifically, we downloaded the genotyping and raw T1‐weighted brain MRI data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. We extracted 145 regions of interest (ROIs) using a multi‐atlas parcellation method (MUSE) from voxel‐wise regional volumetric maps (RAVENS) generated from the MRI scans. We downloaded, integrated, aligned, imputed, and annotated the genotyping data from ADNI 1, GO, 2, and 3 studies. After quality control that excluded low‐quality SNP variants and genetically similar subjects, we performed a genome‐wide association study (GWAS) on 18,366,758 variants and 145 ROI‐level volumetric QTs with 1,382 individuals adjusted for age, sex, and population structure. We prioritized 12,733 significant SNP‐ROI GWAS signals and performed GWMS of AD on those prioritized SNPs with ROIs as mediators.ResultWe found that 35 unique SNPs significantly influenced AD classification mediated by 13 ROI‐level volumetric QTs after false discovery rate correction – in total 218 significant mediation signals. Among all the significant mediation signals, there were 214 partial mediations and 4 full mediations. Chromosome 19 showed the most significant mediation signals. The hippocampus and amygdala in both hemispheres had the most (n = 30) mediation signals, followed by the left entorhinal area with 28 signals (Figure 1).ConclusionOur GWMS of AD identified various SNP variants influencing AD classification via the mediation of ROI‐level volumetric QTs. Complementary to the traditional GWAS analyses, the GWMS analyses can help to explain the genetic underpinnings of the neuromorphometric endophenotypes of AD and provide valuable guidance for AD precision medicine.