Objectives: Given that structural variation can be causal in some cases of autism, we reasoned that CNVs would significantly up- or down-regulate the expression of specific genes, and that these changes could be detected by expression arrays. Furthermore we reasoned that recurrent CNVs would result in convergent expression changes, and that these changes would associate with measurable phenotypes.
Methods: We performed functional genomic analyses in lymphoblast cell lines from 439 discordant siblings over 244 Simons Simplex families on Illumina Human Ref8 version2 chip. The distribution of gene expression was analyzed in autistic probands and unaffected sibling populations. Genes that were 2 or 3 standard deviations (SDs) further from average expression levels were deemed potential outliers genes (dysregulated genes). We compared the dysregulated molecular pathways in affected versus unaffected siblings by this outlier analysis. We then integrated the transcriptome profiling with the copy number variations (CNVs) identified in the same population via different statistical analyses including odd ratios, multivariate linear regression. We further explored the functional study of CNVs via analyzing the genes’ haploinsufficiency. Also, we examined the relationship between gene expression and head circumference phenotype via multivariable linear regression.
Results: Our results show that outlier genes identified in probands, but not in unaffected siblings, fall into neural-related pathways as development/ neurogenesis/ synaptogenesis (p = 9.54E-03), and synaptic cell adhesion (p = 2.0E-02). Intersection of the expression data with the CNV data on the same population (Sanders et al. 2011, in press), we demonstrate that outlier genes show significant enrichment within the most pathogenic CNVs (rare de novo CNVs). For rare non-recurrent CNVs not known to be associated with autism, by permutation test, we prioritized deletions at 3q27, 3p13 and 3p26 and duplications at 2p15 and 3q14, all of which show significant enrichment of outlier genes compared to genome background. For recurrent CNVs known to be associated with autism, we demonstrate that dysregulated genes enrich in distinct pathways in 16p11.2 microdeletions, microduplications, and 7q11.23 duplications, which provides a potential molecular explanation of their different penetrance. Our analyses also show that specific genes, including TAOK2, CORO1A, KCTD13 and QPRT within the 16p CNV interval are correlated with differences in head circumference.
Conclusions: This study is the first to provide evidence that pathogenic structural variants lead to transcriptome alterations in ASD at a genome-wide level and demonstrate the utility of this approach for prioritization of genes for further downstream functional analysis subsequent to a whole genome screen.