Probing the exome in Alzheimer disease and other neurodegenerative disorders.

Abstract

Alzheimer disease (AD) is a heterogeneous disorder with a substantial genetic component. A small number of cases (ie, early-onset familial AD) are caused by exceedingly rare but pathogenic and highly penetrant mutations, while most cases (ie, late-onset AD) are caused by an intricate—and still only partially understood—interplay of genetic and nongenetic risk factors.1 The past decade has seen unprecedented progress in deciphering the genetic underpinnings of lateonset AD. This advancement was achieved mostly by the application of high-throughput microarray genotyping in the context of genome-wide association studies (GWASs) comparing the allele status at millions of different base pairs on increasingly large samples of affected and unaffected individuals.2 Most AD GWAS findings to date were made with common (ie, frequency of the minor allele typically >5%) single-nucleotide polymorphisms (SNPs) typically exerting small genetic effect sizes (ie, odds ratios <1.3). In most cases, the pathogenic mechanisms underlying these associations have been difficult to discern owing to the fact thatmost common SNPs are located in noncoding regions of the genome. The study by Chen et al3 in this issue is different from most other GWASs in the field because it specifically focused on SNPs located in the coding—and thus potentially functional—portions of the genome (ie, the exome). They achieved this by using a microarray (the HumanExome Array, Illumina, Inc) specifically designed to capture the mostly rare (ie, frequency <5%, applicable to approximately 90% of all markers on the array) exome SNPs in a sample of 224 patients with AD and 224 healthy control subjects. Additional genotyping was completed on 168 patients with frontotemporal dementia and 48 patients with progressive supranuclear palsy. While the analyses of frontotemporal dementia and progressive supranuclear palsy did not yield any noteworthy new results, at least 2 novel putative loci, ie, DYSF and PAXIP1, were highlighted by the authors for AD, in addition to very strong signals near the well-established APOE locus on chromosome 19. Overall, Chen et al3 estimated that about 44% of the variance in case-control status in their sample could be explained by the SNPs typed on the exome array, a number substantially higher than the approximately 25% reported from an earlier and much larger GWAS metaanalysis mostly studying common variants.2 Themain strengthof the studybyChenet al3 is that it represents oneof the first publishedapplicationsof theexomearray technology to AD (and frontotemporal dementia and progressive supranuclearpalsy).This technology ismeant to serve as a cost-effective interim solution to studying the genetics of complex traits in times when more comprehensive approaches, such as whole-exome sequencing or wholegenome sequencing, are still relatively expensive. Thedesign of the exome array used by Chen et al3 resulted from a comprehensive analysis of whole-exome sequencing and wholegenome sequencing data of approximately 12 000 individuals with the aim to specifically capture all nonprivate and putatively functional (ie, nonsynonymous, nonsense, splice) base-pair changes (http://genome.sph.umich.edu/wiki/Exome _Chip_Design). It is estimated that 94% to 98% of these types of variants identified in an average genome by whole-exome sequencing are also captured by the exome array. While this coverage should suffice for most exome variant association studies, this typeofmicroarray technology—bydesignandunlike whole-exome sequencing or whole-genome sequencing—is not able to discover any novel sequence variants. Instead, theexomearray is limited togenotypingpolymorphisms previously shown to exist at sufficient frequency in the generalpopulationand,conceptually, is thereforenodifferent from a conventional (ie, focusing on commonvariants not selected based on functionality) GWAS or smaller-scale association study.Despite this limitation, almost90%of thevariantsgenotyped on the exome array used by Chen et al3 were not captured or analyzed in themost comprehensive AD GWAS published to date2 (variant overlap determined for the purpose of this editorial is basedondata releases fromLambert et al2 and http://www.illumina.com). Differently put, approximately 213 000 SNPs,most ofwhichwere predicted to be functional, genotyped in the study by Chen et al3 were not previously tested in patients with AD and controls of European descent, although similar studies are currently underway (see below). The only other exome genotyping study for AD published to date was performed on 400 patients with late-onset AD and 605 controls from Korea using a similar exome microarray (manufactured by Affymextrix, Inc).4 While the study by Chen et al3 thus uses state-of-the-art technology, its scientific implications currently remain limited. This is mostly owing to the small sample size investigated. In their discovery phase, Chen et al3 analyzed only 416 patients with AD and controls, followed by replication analyses in an independent set of 480 individuals. Thus, the overall sample size analyzed in the AD portion of this study was well below 1000 individuals. It was even smaller for the frontotemporal dementia (n = 168) and progressive supranuclear palsy (n = 48) analyses, forwhichno replicationanalyseswere Related article Opinion