Whole-genome sequencing in an autism multiplex family

Abstract

Background: Autism spectrum disorders (ASDs) represent a group of childhood neurodevelopmental disorders that affect 1 in 88 children in the US. Previous exome sequencing studies on family trios have implicated a role for rare, de-novo mutations in the pathogenesis of autism.

Methods: To examine the utility of whole-genome sequencing to identify inherited disease candidate variants and genes, we sequenced two probands from a large pedigree, including two parents and eight children. We evaluated multiple analytical strategies to identify a prioritized list of candidate genes.

Results: By assuming a recessive model of inheritance, we identified seven candidate genes shared by the two probands. We also evaluated a different analytical strategy that does not require the assumption of disease model, and identified a list of 59 candidate variants that may increase susceptibility to autism. Manual examination of this list identified ANK3 as the most likely candidate gene. Finally, we identified 33 prioritized non-coding variants such as those near SMG6 and COQ5, based on evolutionary constraint and experimental evidence from ENCODE. Although we were unable to confirm rigorously whether any of these genes indeed contribute to the disease, our analysis provides a prioritized shortlist for further validation studies.

Conclusions: Our study represents one of the first whole-genome sequencing studies in autism leveraging a large family-based pedigree. These results provide for a discussion on the relative merits of finding de-novo mutations in sporadic cases versus finding inherited mutations in large pedigrees, in the context of neuropsychiatric and neurodevelopmental diseases.

Figure 1

Analysis of copy number variations (CNVs) in the family with autism. (A) The five inherited CNVs inferred from SNP arrays are depicted with family structure, but none of the CNVs segregate with disease status. (B) Signal intensity (Log R Ratio and B Allele Frequency) plot from SNP arrays validates the 1q31 deletion in sample 10. In the deletion (dots between the two vertical lines), Log R Ratio values for SNP markers drop, and B Allele Frequency values cluster around 0 or 1. (C) PennCNV-Seq signal (sequence count and B Allele Frequency) plot on WGS data validates the 1q31 deletion in sample 10. In the deletion, the sequence counts tend to be lower than neighboring regions, and very few B Allele Frequency values cluster around 0.5. (D) Signal intensity plot from SNP arrays did not indicate the presence of the 1q31 deletion in sample 8. (E) PennCNV-Seq signal on WGS data did not indicate the presence of a 1q31 deletion in sample 8.

Figure 2

Illustration of the variants reduction procedure on two probands with autism in the pedigree. Applying a recessive model of disease inheritance, we identified 22 and 23 candidate genes in the two probands, including seven shared genes.

Figure 3

Illustration of the non-synonymous mutation in ANK3. (A) A UCSC genome browser shot of the ANK3 gene and the location of the mutation, together with sequence conservation patterns across 28 vertebrate species. (B) Validation of the mutation by Sanger sequencing in the family. The primers used are CTTCATGGTCATGGTGGATG (forward) and AGGGGGAAGGGGATAAAAGT (reverse).

Figure 4

UCSC genome browser shots of prioritized non-coding variants, demonstrating the sequence conservation levels and the predicted functionality in ENCODE lymphoblastoid cell lines. The ‘prioritized variants’ track shows the location of the prioritized non-coding variants shared by both probands. (A) A prioritized variant is located in a predicted ‘active promoter’ for COQ5. (B) A prioritized variant is located in the intronic region of SMG6, and is predicted to be a ‘strong enhancer’.