Whole Genome Sequencing of Extended Families Reveals Novel ASD Risk Variants

Background: Massively parallel sequencing in autism spectrum disorder (ASD) has focused on whole exome sequencing (WES) and whole genome sequencing (WGS) in trio cohorts for identifying de novoprotein coding variants. These studies have largely utilized simplex families or siblings with ASD, and a majority of WGS analyses have reported only on protein coding variants.

Objectives: Our study applies WGS to extended, multiplex families with at least two cousins with ASD likely to carry rare, inherited and partially penetrant alterations. We hypothesize that identical by descent (IBD) filtering in these pedigrees would define genomic regions of shared ASD risk and allow for the identification of variants in noncoding regions, coding variants missed by exome sequencing, and structural variants, which could potentially isolate new ASD loci.

Methods: We performed WGS on at least two cousins with ASD across six extended families (15 individuals). Sequencing was performed on the Illumina HiSeq2500 and analyzed through pipelines including BWA-MEM alignment, quality recalibration by GATK, and variant calling with the GATK HaplotypeCaller and FreeBayes. Structural variants (SVs) were called with the SWAN and GenomeSTRiP algorithms. Annotations were applied with ANNOVAR including functional predictions for noncoding variants (CADD, FATHMM-MKL, and Eigen). We determined IBD regions using whole genome genotyping data and the MERLIN package. Variants were prioritized which were shared in all individuals with ASD within each family, rare in the population (<1%), and predicted to have functional significance through in silicoprograms (CADD >10, FATHMM-noncoding >0.5, and Eigen >1).

Results: We sequenced each genome to ~40x coverage and identified >4 million single nucleotide variants (SNVs) and small indels and >100 SVs per individual. Variant calls between HaplotypeCaller and FreeBayes were >95% concordant. IBD filtering in each family limited the total number of SNVs and short indels for analysis to between 732 and 1,020,713, depending on the family’s structure. Among coding SNVs, ~94% concordance was found with existing whole exome data (Cukier, et al, 2014); however, WGS identified ~10% more coding variant calls. These include a family with a rare missense mutation in the neurogenesis growth factor GDF11 and another with a frameshift in the axonal development gene SLAIN1. In silico prioritization of noncoding regions revealed several variants of interest. For example, two variants were identified in the putative promoter of the chromatin remodeling gene ARID1B in one family, and two other variants upstream of ankyrin repeat gene KANK1 were present in another family. Finally, rare copy number variants were found; one CNV disrupted the promoter of the neurodevelopmental WWOX gene and another deleted an exon of the lincRNA FIRRE, which is involved in chromosomal organization.

Conclusions: By studying these unique pedigrees, applying cutting edge sequencing and analysis methods, and employing IBD filtering, we establish that WGS of extended families can identify inherited ASD risk alterations. These methods extend the scope of WGS beyond de novo protein coding variants to functional noncoding SNVs, SNVs not captured by exome sequencing, and SVs that may contribute to ASD. Taken together, WGS identifies new ASD candidate genes and pathways.