22393
Homozygous Deletions of Non-Coding DNA Sequences in Autism Spectrum Disorder

Saturday, May 14, 2016: 2:52 PM
Hall B (Baltimore Convention Center)
K. Schmitz-Abe1,2, G. Sanchez-Schmitz1,2, E. M. Morrow3, M. Greenberg1,2, T. W. Yu1, C. A. Walsh1,2 and K. Markianos1,2, (1)Harvard Medical School, Boston, MA, (2)Children's Hospital Boston, Boston, MA, (3)Department of Molecular Biology, Cell Biology and Biochemistry and Institute for Brain Science, Brown University, Providence, RI
Background:  

Noncoding DNA comprises 99% of the genome but methods for identifying its contribution to disease have greatly lagged our understanding of protein-coding mutations. Autism Spectrum Disorder (ASD), associated with defects in social and/or cognitive function, has previously been linked to de novo Copy Number Variants (CNVs), de novo Single Nucleotide Variants (SNVs) or inherited recessive biallelic SNVs; however, most cases remain unexplained.

Objectives:  

We analyzed CNVs and homozygosity in 187 families ascertained through the Homozygosity Mapping Collaborative for Autism (HMCA), an ASD cohort highly enriched for families that are consanguineous (792 individuals). We compared findings in our cohort to data from the Autism Genetic Resource Exchange cohort (AGRE, 740 families, 2985 individuals) and from the Simons Simplex Collection (SSC, 1,027 families, 3881 individuals).

Methods:  

CNV detection, annotation, and analysis were done using a custom build, rule based “Variant Explorer” pipeline. It uses concordant calls between multiple algorithms to maximize specificity, and classified as common/rare based on overlap with 1,251 HapMap controls processed by the same pipeline.

Results:  

In consanguineous families, children affected with ASD are significantly enriched for autosomal homozygous deletions compared to unaffected siblings (17% versus 4%, p<0.001). Most homozygous deletions were small (<50 kb) and only a few were predicted to result in protein inactivation through coding exon disruption. In contrast, affected children were significantly enriched for homozygous deletions in DNA regulatory regions, with regions disrupting ENCODE histone methylation peaks, a rate much higher than predicted by chance (p<0.001).

Conclusions:  

While the importance of regulatory elements has been previously anticipated, such non-coding variants are not as readily identifiable as the disruption of coding sequence. High consanguinity allows us to study biallelic deletions and take advantage of the favorable signal to noise ratio provided by complete loss of coding or regulatory regions. The importance of biallelic deletions is supported by two lines of evidence: 1) a significantly higher rate of biallelic deletions in affected children relative to their unaffected siblings (p<0.001), and 2) a striking enrichment/depletion pattern of intersection between biallelic deletions and ENCODE control regions in affected/unaffected children (p<0.001). While (1) provides unambiguous evidence for the role of biallelic deletions in recessive ASD, the ENCODE analysis provides formal evidence for the mode of action of a significant subset of biallelic deletions.

The biallelic deletions regions identified here represent an important starting point to the understanding of the role of patterned gene activation/regulation in cognitive and social function. ASD in general appears to show an especially important role for gene dosage in causation, given the central importance of heterozygous CNV and SNV all of which affect gene dosage, and the importance of hypomorphic recessive mutation. Upon neuronal depolarization neurons are known to show rapid and reversible changes in the levels of expression of a large number of activity-regulated genes, and this temporally regulated transcriptional program is known to be essential for the functional changes that underlie memory formation and learning. Biallelic noncoding mutations, or heterozygous mutations affecting gene dosage, may well cause disease by disrupting such finely tuned transcriptional programs.