Whole genome sequencing resource identifies 18 new candidate genes for autism spectrum disorder

Abstract

We are performing whole-genome sequencing of families with autism spectrum disorder (ASD) to build a resource (MSSNG) for subcategorizing the phenotypes and underlying genetic factors involved. Here we report sequencing of 5,205 samples from families with ASD, accompanied by clinical information, creating a database accessible on a cloud platform and through a controlled-access internet portal. We found an average of 73.8 de novo single nucleotide variants and 12.6 de novo insertions and deletions or copy number variations per ASD subject. We identified 18 new candidate ASD-risk genes and found that participants bearing mutations in susceptibility genes had significantly lower adaptive ability (P = 6 × 10^-4). In 294 of 2,620 (11.2%) of ASD cases, a molecular basis could be determined and 7.2% of these carried copy number variations and/or chromosomal abnormalities, emphasizing the importance of detecting all forms of genetic variation as diagnostic and therapeutic targets in ASD.

Figure 1. Schematic of sample and data processing in MSSNG

An executive committee oversees the project. The parameters for DNA sample selection and (genetic and phenotypic) data are managed by the committee, including consenting and ethics protocols. Coded identifiers’ for samples selected for WGS are posted as they are identified at MSSNG portal (http://mss.ng/research), so the ASD research community can monitor progress. Phenome data include subject information (identity number, year of birth, sex), family code (proband, parent, sibling), results of diagnostic tests (e.g. ADOS, ADI-R, age at diagnosis, functional assessments, intelligence tests, body measurements and dysmorphic features). The database accommodates as much of this information as is available for each sample, but that varies widely. Future plans include incorporation of fields for co-morbidities, related conditions, exposures, extended family history, interventions, and other parameters that become apparent. WGS technologies were Complete Genomics and Illumina HiSeq (2000 and X). WGS data are transferred to Google Genomics for data processing through the Google Cloud. Ref-blocked gVCFs were generated and stored in Google Cloud Storage, which were also processed for de novo mutation detection in the local cluster (for Complete Genomics data using filtering method) and Google Compute Engine (for Illumina data using DenovoGear). The Ref-blocked gVCFs and the de novo mutations were annotated through the Google Compute Engine (using Annovar), which can be accessed through the BigQuery tables. Quality controls of the genomic data were performed in the local cluster and the Google Cloud. The processed genetic data and the phenotypic data are accessible through the MSSNG Portal interface. The MSSNG database is designed to support incremental addition of data without changes in architecture, scaling to at least tens of thousands. New WGS and phenotype data are continually added to MSSNG as new batches of 1,000 samples are processed. DACO: Data access committee; UPD: Uniparental disomy; Ti/Tv: Transition to transversion ratio; IBS: Identity by state.

Figure 2. Characteristics and quality of WGS from different sequencing platforms

(a) Number of SNVs detected per genome. (b) Number of indels detected per genome. (c) Number of rare coding SNVs detected per genome after quality filtering. (d) Number of rare coding indels detected per genome after quality filtering. Genomes sequenced by Complete Genomics with 2.0 pipeline version are colored in orange, by Complete Genomics with 2.4 pipeline version are colored in brown, by Complete Genomics with 2.5 pipeline version are colored in green, by Illumina HiSeq 2000 are colored in blue, and by Illumina HiSeq X are colored in purple. Details of quality for individual samples can be found in Supplementary Table 1.

Figure 3. ASD-susceptibility genes/loci

(a) ASD-risk genes with higher than expected mutation rate from MSSNG integrated with other large-scale high-throughput sequencing projects. ASD-risk genes are ranked in descending order of the number of mutations found for each gene. Other LOF mutations, including inherited LOF mutations and LOF mutations with unknown inheritance (where parents are unavailable for testing), and CNVs found in the MSSNG cohort are indicated (except for genes found by higher than expected de novo missense mutation rate). MSSNG data are in green and published data are in yellow. Novel putative ASD-risk genes identified in this study carry an asterisk. Δ indicate genes with druggable protein domains identified (Supplementary Table 6). (b) Pathogenic chromosomal abnormalities and CNVs identified falling into one of four categories: Chromosomal abnormalities; DECIPHER loci and other genomic disorders associated with ASD; large rare CNVs between 3–10Mb and CNVs disrupting ASD candidate genes not described above in Figure 3a. Deletions are in red, duplications are in blue and complex variants are in purple. # indicate CNVs shared between affected siblings; ‡ indicates a CNV carried by an individual with a second pathogenic CNV; † indicates a CNV shared between individuals within an extended pedigree. Details can be found in Supplementary Table 8. Examples of CNVs affecting the NRXN1 and CHD8 genes, and the PTCHD1-AS non-coding gene identified from the WGS are shown in Figure 4.

Figure 4. CNV characterization via WGS reads in the MSSNG Portal

(a–c) Visualization of CNVs in WGS data. (a) A heterozygous 246kb deletion of three exons of NRXN1 at chromosome 2p16.3 in subject 2-1428-003 (average 50% decrease in sequence read-depth); (b) a 31.1kb duplication within CHD8 at chromosome 14q11.2 in subject 2-1375-003 (average of 50% increase in sequence read-depth) and (c) 125kb deletion of exon 3 of the non-coding gene PTCHD1-AS at Xp22.11 in male subject 1-0277-003 (no reads apparent, other than a small stretch of likely mis-aligned repetitive sequences). Left and right panels show the proximal and distal breakpoints of the CNVs respectively. Aligned reads viewed from the BAM files in the MSSNG browser are shown indicating the read depth. Genome co-ordinates are shown above and impacted genes below. The predicted CNVs visible from the WGS data and high-resolution microarray are shown by the red (deletion) and blue (duplication) bars. For 32 CNVs described in Figure 3b plus 17 additional CNVs, we derived a more accurate estimate of the breakpoints by visual inspection of read depth from the BAM file in the MSSNG browser. On average, the size difference between the CNV predicted by microarray data and the estimated size from WGS data was 6.9kb and for 31/49 (63%) CNVs, the size of the CNV was smaller in the microarray data than WGS. For four CNVs, the WGS-resolved breakpoints altered the exons of genes being annotated as deleted or duplicated. In another case, this resulted in a CNV from microarray no longer being classified as pathogenic as the revised breakpoints no longer included coding sequence.

Figure 5. Phenotype comparison for the samples with and without identified mutations

Standard score of (a) IQ Full scale and (b) Vineland Adaptive Behavior were compared between samples with pathogenic CNVs, de novo LOF mutations, mutations in ASD-risk genes and other samples without any of these mutations.

Figure 6. Interaction similarity network of ASD-risk genes

Connections represent gene similarity based on physical protein interactions and pathway interactions. Connection thickness is proportional to the fraction of interaction partners shared by the connected genes. The size of the node for each gene is proportional to the total mutation count (Figure 3). Genes associated with LOF mutations are in circle shape, while genes associated with missense mutations are in diamond shape. The node color corresponds to the BrainSpan brain expression Principal Component 1 (prenatal in yellow, postnatal in blue, balanced in light blue, undetermined in grey). The labels of novel ASD-risk genes are displayed in red. The network was visualized using Cytoscape.