Exome-sequencing based gene discovery and systems biology of autism spectrum disorders

Background: De novo loss of function (nonsense, canonical splice site, and frameshift) mutations are associated with ASD based on the observation of a higher rate in cases versus unaffected controls via exome-wide analyses. Furthermore, identification of multiple de novo LoF mutations in the same gene in unrelated individuals has emerged as a reliable approach to identifying ASD-associated genes at genome-wide significance. Previously, based on an initial set of nine ASD-associated genes emerging from exome analyses of 1,043 families, we conducted a co-expression analysis, leveraging spatio-temporally rich data from developing human brain, that has identified layer 5/6 glutamatergic neurons in the prefrontal cortex during human mid-fetal development as relevant for ASD pathophysiology.

Objectives: To identify additional ASD-associated genes in a hypothesis naïve manner by conducting further whole-exome sequencing, and to use these genes as inputs for an expanded gene co-expression analysis aimed at refining spatio-temporal understanding of ASD pathophysiology.

Methods: We present an analysis of whole-exome sequencing of 457 previously un-reported simplex families from the Simons Simplex Collection (SSC), contributing to a combined analysis of 1,500 simplex ASD families. Two detection algorithms are used to identify single nucleotide variants (SNVs) and insertion/deletions (indels) separately. All putative de novo LoF mutations are confirmed through PCR and Sanger sequencing. Genes with multiple de novo LoF mutations are used as seed genes for co-expression analyses aimed at refining our recently reported spatial and temporal data regarding convergence of ASD related genes in mid-fetal human cortical development.

Results: A total of 196 genes were observed to have at least one de novo LoF variant in an affected individual. Of these genes, 18 show at least two de novo LoF variants in affected individuals. Comparison with the distribution of de novo LoF variants in the unaffected siblings demonstrates that the 196 have a >50% probability of being associated with ASD while the 18 genes with multiple LOF mutations have a >95% of being associated with ASD risk. Eight of the 18 genes with multiple de novo mutations are novel ASD-associated genes. Co-expression network analysis based on the expanded set of genes validates our previously observed findings pointing to a convergence of ASD-related mutations in layer 5/6 glutamatergic neurons in the prefrontal cortex during mid-fetal development.

Conclusions: Ongoing exome sequencing of ASD trios is rapidly expanding the list of ASD-associated genes. Identification of these genes facilitates analyses of the biological systems underlying ASD pathology. The discovery of additional ASD genes and an expanded co-expression network analysis reported here highlights the convergence of ASD genes in deep layer glutamatergic neurons in the prefrontal cortex during mid-fetal development, strengthening this association. As expected, the data also point to additional possible points of spatio-temporal convergence of ASD mutations, underscoring the importance of integrating other gene discovery methods and genomic datasets such as chromatin state and gene regulation.