GENOME-WIDE ASSOCIATION STUDY OF A MULTIMODAL IMAGING BIOMARKER IN THE ADNI COHORT

Abstract

Genome-wide association studies (GWAS) in Alzheimer's disease (AD) using neuroimaging-based phenotypes are typically derived from a single imaging modality. However, AD is a complex disorder involving different inter-linked pathogenic pathways. To explore genetic influences on disease progression, we generated a novel disease progression score (DPS) comprising cortical β-amyloid levels and hippocampal volume, then using it as a quantitative phenotype in GWAS. Study participants were part of the AD Neuroimaging Initiative (ADNI). Hippocampal volumes were computed with FreeSurfer from T1-weighted MRI scans. Cortical Aβ42 levels were computed as standardised uptake values ratio (SUVR) from florbetapir PET scans. The AD-DPS was computed by jointly modelling the long-term time evolution of hippocampal volume and cortical SUVR for 1088 individuals, using the growth models via alternating conditional expectation (GRACE) algorithm. We used this DPS as a continuous phenotype for a GWAS in 846 subjects of Caucasian ancestry, covarying for sex, age, APOE4 status, baseline SUVR and 2 principal components of population structure. Additionally, we repeated the analysis stratified by APOE4 status. Progression curves were in good agreement with proposed models of AD biomarker evolution (Figures 1–2). A genome-wide significant association with AD-DPS was found on chromosome 4 for rs3733541 (p=9.45e-09), located 300 kbp upstream of the KIT gene (Figure 3). This association was not found when testing hippocampal volume and Aβ42 burden separately. There were no genome-wide significant SNPs in the stratified analyses. However, the standardised effect size for rs3733541 was greater in APOE4 non-carriers (β=-0.11± 0.02, p=9.5e-5) than in APOE4 carriers (β=-0.08±0.02, p=3e-3). We describe the first use of a DPS derived from multiple imaging modalities in a quantitative trait-GWAS. The KIT gene and its ligand stem cell factor (SCF) are involved in a pathway leading to migration and differentiation of neural stem cells to sites of brain injury. Significantly lower levels of SCF have been found in the plasma and CSF of patients with early AD. If validated in future studies, this novel susceptibility locus may influence the joint occurrence of hippocampal atrophy and amyloid deposition in AD. Long-term progression curves for two AD biomarkers. Left, florbetapir PET SUVR; right, bilateral hippocampal volume. Disease progression score stratified by diagnostic group (left) and by MCI progression status (right). Manhattan plots for the GWAS of hippocampal volume (top), amyloid burden (middle), disease progression score (bottom). The blue line is the threshold for suggestive associations at p = 1e-5; the red line is the threshold for genome-wide significant associations at p = 5e-8.