Characterizing missing heritability of late-onset Alzheimer’s disease: An exome array study

Abstract

Alzheimer's disease (AD) is highly heritable (h2=58-79%), but despite the identification of more than 25 risk loci with common variants, the known the heritability accounted for by the genome-wide association studies and the APOE locus is estimated at less than 40%. A substantial proportion of this missing heritability is likely to be accounted for by rare variants. Some next generation sequencing approaches have proved successful in identifying rare risk variants for disease, however they lack sufficient statistical power to detect variants of modest effect. Here we utilize exome arrays to identify rare genetic variants that influence disease risk. Rare variant exome chip analyses used genotype data passing stringent quality control on 16,173 cases and 18,116 controls from the Illumina HumanExome BeadChip. Single-variant association analyses employing score tests and SKAT-O as well as burden analyses based on gene regions were performed using SeqMeta/R, with unadjusted and adjusted models (including population substructure, age, and sex). Only SNPs polymorphic in at least two studies with a MAC ≥5 and MAF ≤0.05 were considered for further follow-up analyses. Our exome chip analyses confirmed previously identified associations with rare variants at the TREM2 (P=3.5×10−9) and APOE [ε2] (P=2.3×10−80) loci. In the single-variant meta-analysis, one rare missense variant in the gene coding for ATP synthase, H+ transporting, mitochondrial F1 complex, gamma polypeptide 1 (ATP5C1) showed genome-wide significant evidence for association (P=1.6×10−9) and the gene coding for cerebellin 3 precursor (CBLN3) showed gene-wide significant evidence of association (P=7.6×10−6). We identified a further seven loci showing suggestive evidence for association at P<10−6 in the single-variant analyses. Replication studies are under way in 11,272 cases and 14,708 controls using the Sequenom MassArray platform to confirm these findings. This study has identified two novel loci (a rare variant in the ATP5C1 gene and the CBLN3 gene), which show significant association with increased AD risk. A further seven rare variants show suggestive evidence for influencing AD risk. Work is ongoing to confirm these associations in independent samples.