|Home | About | Journals | Submit | Contact Us | Français|
In this issue of Neuron, two papers (Lim et al. 2013, Yu et al. 2013) use whole exome sequencing (WES) to elucidate the contribution of inherited variation to the risk for autism by leveraging the increased penetrance of homozygous and compound heterozygous rare variants in autosomes and hemizygous rare variants in the X chromosome of males. Together, they expand our knowledge about the genetic architecture of ASD, verify previously identified genes, and identify novel mutations that will guide the discovery of the critical biological processes disrupted in autism.
Autism is a spectrum of neurodevelopmental disorders (referred to as autism spectrum disorder, ASD) affecting around 1 in 88 individuals (2012 CDC estimate). Genetic risk factors for ASD play a substantial role and come in many forms: those transmitted from parents to affected child or those appearing de novo in the germline, as well as those found commonly in the population, or those only rarely observed (Figures 1 and and2).2). Genetic investigations have revealed that hundreds of genomic loci are likely involved and it is now important to advance our understanding of ASD’s genetic architecture while simultaneously identifying specific deleterious variants in genomic loci to understand the biological pathways affected(Berg & Geschwind 2012). Inherited variation from both common and rare alleles provides the largest genetic contribution to population-wide ASD risk, explaining ~40% (95% Confidence Interval: 8%–84%) of the risk for developing ASD (Hallmayer et al. 2011). Though common variants are estimated to be a large driving factor for the disorder (40–60% variance explained; Klei et al. 2012), the effect size of individual common variants is small (estimated to be odds-ratio < 1.2; Anney et al. 2012). This observation has led to a search for rare variants, which may exhibit larger individual effect sizes.
To this end, cost-effective high-throughput genome sequencing of ASD patients analyzed in paradigms where the effect of a mutation can be seen over random chance plays a critical role(Iossifov et al. 2012; Neale et al. 2012; O’Roak et al. 2012; Sanders et al. 2012). However, this process is complicated by the fact that each of us inherits over 100 nonsense, loss-of-function mutations in our genomes, leading to about 20 completely inactivated genes and several dozen de novo variants, some of which may be functional, but none of which are clearly associated with disease (MacArthur et al. 2012). In the latter category, predicted protein structure altering rare variants have been observed to be more frequent in ASD cases, and recent work suggests non-transmitted, de novo mutations that delete genic regions, perturb splicing, or truncate protein products may contribute to the development of ASD in 15–20% of cases(Devlin & Scherer 2012). Although inherited mutations have been identified in rare families, no population-based exome sequencing study has demonstrated a significant role for inherited deleterious variants in ASD risk, leaving the contribution of this class of genetic variation unknown. Now, as shown in two papers in this issue (Lim et al. 2013, Yu et al. 2013), it is clear that ASD risk is increased when two rare variants found within the same gene deleteriously affect both copies of a protein, consistent with a role for recessive inheritance of non-synonymous mutations in ASD.
Starting with a population-based approach, Lim and colleagues performed WES in 933 ASD cases and 869 controls, allowing them to see the complete genome sequence in protein-coding and flanking regions. They focused on the most damaging type of mutations (nonsense and essential splice site) transmitted from the mother and father to the child. In autosomes they studied loci where the mutations truncate both copies of a protein, and in the X chromosome in males, they studied mutations that truncate the only copy. Only rare alleles of this type (minor allele frequency ≤ 5%) were studied, based on the hypothesis that double mutations commonly found in the population are not pathological. These rare, putative recessively acting mutations were observed twice as many times in ASD patients as controls in autosomes (6% of cases and 3.3% of controls carried such a mutation). When overlapped with a dataset of brain gene expression, the overall odds-ratio (OR)for ASD increased to 2.7. Importantly, this enrichment of double mutation variants was replicated in an independent dataset, this time showing a smaller, but still significant effect (7.6% of cases and 5.5% of controls). The authors estimate an overall 3% contribution to risk for ASD from this class of mutations. Rare hemizygous mutations on the X chromosome, which would also be depleted of the normal protein in males, were also enriched in male ASD cases compared to controls (4.8% vs 3.1%), revealing involvement in 1.7% of male ASD cases. This study fills an important gap in our knowledge of the genetic architecture of ASD by estimating that about 5% of ASD cases may be affected by rare inherited loss-of-function homozygous, compound heterozygous, or X chromosome mutations in males. In this way, Lim et al., significantly advance our population-level understanding of risk for ASD.
One key issue is that most current studies on the contribution of rare variants to genetic risk are not well powered to identify specific genes. Typically, functional studies or repeated instances of rare mutations in the same gene are needed to ensure that a variant is indeed functional and associated with ASD. Alternatively, other data, such as gene expression data can be used to functionally prioritize variants, as many mutations, including but not limited to de novo deletions (e.g. CNV), or recessive mutations causing intellectual disability (ID) or ASD are expected to significantly reduce RNA levels even in peripheral blood (Luo et al. 2012). We predict that future studies combining genome sequencing with gene expression will have increased power to detect pathogenic inherited or de novo genetic variation. Short of this, very large sample sizes (larger than current GWAS samples ranging ~10,000 subjects), or unique families with multiple affected children (multiplex families)will be necessary.
In the second study, Yu and colleagues took an elegant approach to identify specific inherited rare variants by studying multiplex families, combining multiple gene-mapping modalities, and validating variant pathogenicity experimentally. Specifically, the authors searched consanguineous and/or multiplex families using a combination of heterozyogsity mapping and linkage directed WES analysis. This identified regions in three families with high linkage scores (likelihood > 600:1 of the genetic region containing a variant tracking ASD in that family) where specific nonsynonymous or frame shift rare variants were also found.
This analysis identified three genes, AMT, PEX7, and SYNE1 as affected by homozygous or compound heterozygous rare variation linked to ASD. To demonstrate the pathogenicity of each of these rare variants, the authors perform detailed in vitro functional analysis, which has not been previously performed in population-based WES studies. For example, one family contained a homozygous I308F change in AMT, a gene coding for a glycine-degradation enzyme in which severe mutations have been shown to lead to neonatal nonketotic hypoglycemia (NKT), a severe life-threatening neonatal metabolic syndrome. This mutation was analyzed biochemically and its effect on protein solubility was evaluated in bacteria, suggesting that the mutation may induce a protein-folding defect and result in a functional hypomorph. Indeed, the family contained three affected individuals with an atypical, milder manifestation of NKT symptoms in addition to ASD. The authors propose that, in general, less severe mutations in genes involved in recessive neurodevelopmental ID syndromes may lead to ASD.
A similar inheritance pattern and relationship to syndromic disease was observed with PEX7, which is the causative gene for rhizomelic chondrodysplasia punctata and SYNE1, where null mutations have been linked to cerebellar ataxia and a recessive form of arthrogryposis multiplex congenita. The functional data presented by the authors and absence of the strong classical phenotypes suggested these changes were also hypomorphic, and their experimental investigations supported this. Yu and colleagues’ approach exemplifies how various lines of direct and indirect evidence - genome-wide linkage, exonic variation, biological validation, and clinical relevance to existing syndromes - may be used to convincingly implicate a rare variant, which itself may never reach population-wide genome-wide significance.
Yu and colleagues further generalized this mutational pattern and its relationship to syndromic disease by compiling a set of risk genes comprising these 3 and 70 other previously implicated ID-related genes. They then compared rare inherited variants from WES in 163 consanguineous and/or multiplex families to 831 population-matched individual exomes to find additional homozygous, compound heterozygous, or hemizygous rare mutations. They find 5 families with previously unidentified nonsense or frameshift variants and 11 with previously unidentified missense variants in their set of risk genes. Intriguingly, two families had nonsense or frameshift mutations in PAH and one had a likely functional missense mutation in AMT suggesting a potential metabolic contribution to their phenotype and perhaps autistic symptoms. This highlights the potential for better genetic diagnoses and treatment by immediate intervention in a subset of ASD, as was recently highlighted in a recently discovered metabolic form of ASD (Novarino et al. 2012). In total, this study’s cohort found novel changes linked to ASD in the novel genes AMT, PEX7, SYNE1, VPS13B/COH1, PAH, and POMGNT1, as well as previously implicated genes NLGN4X and MECP2.
These studies provide two unique vistas on ASD genetics. Given an etiologically and phenotypically heterogeneous neurodevelopmental disorder that may involve hundreds of genes, the field has attempted to gain a foothold on biological pathways by identifying genes using highly penetrant mutations that are linked to ASD. These studies also provide convincing statistical evidence for the role of homozygous or compound heterozygous loss of function mutations in ASD risk. Moreover, the results emphasize how difficult it can be to assign blame to a single gene, given the rarity of events, their occurrence in controls, and the numbers necessary to attain genome-wide significance. Observing the event more times in affected individuals is still necessary to provide definitive proof of genetic association with disease. This is further complicated by the several order variation in mutation rates across the genome, which suggests the need for locus specific calculations (Michaelson et al. 2012). Furthermore in most cases, it is becoming clear that ASD is a multigenic disorder.
Finally, the clearly defined role for CNVs and SNVs that delete or alter either one or both copies of a gene or its isoforms indicates that transcript levels play a significant role in ASD susceptibility. From this perspective, we predict that future work will identify variants in regulatory regions that affect transcript levels, such as promoters and enhancers, and that noncoding regions will be just as functionally important to ASD as currently implicated protein-coding regions. This is also supported by the high mutability of DNAse hypersensitivity sites and regions containing CpGs (Michaelson et al. 2012). So far the contribution of regulatory elements is unexplored at a population-level, as whole-exome sequencing does not effectively measure these regions. Nevertheless, these current approaches provide unprecedented insight that will aid in explaining ASD’s pathobiology.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.