Objectives: The goal of this study is to identify large (>300kb), potentially pathogenic CNVs contributing to autism by SNP genotyping in autism trios.
Methods: Genotyping was performed using the Illumina Human 1M Beadchip on 635 autism cases and immediate family members from 600 families. To identify and call CNVs, the PennCNV algorithm implementing a hidden Markov model (HMM) was used. CNVs identified that were greater than 300kb in size and overlapped one or more genes were selected for further analysis by comparison to the Database of Genomic Variants (DGV) and functional classification of genes. CNVs not in the DGV and likely involved in neurological function and development were considered to be potentially pathogenic and investigated further. CNVs of greatest interest were validated using TaqMan Copy Number Assays.
Results: We identified a total of 48 de novo CNVs greater than 300kb using this approach. Of those, 21 were single copy duplications of which only two were not present in the DGV and were shown to be involved in neuronal processes. These CNVs occurred in previously identified autism candidate regions on chromosomes 15q11 and 17p13. Of the 27 single copy deletions, only four were not present in DGV and were involved in neuronal processes. These included previously identified autism related microdeletions at 16p11, deletion of the autism candidate gene ASDL on 22q13, and loss of CHRNA7 on chromosome 15q13. In addition we identified a novel 1.5Mb de novo deletion on chromosome 14q23 in a patient with autism, macrocephaly, and spherocytosis. A deletion in this region has previously been reported in a patient with spherocytosis and mental retardation, and includes novel autism candidate genes MTHFD1, a folate metabolism gene previously implicated in neuropsychiatric disorders such as bipolar disorder and schizophrenia, and PLEKHG3, a gene expressed predominantly in the brain and involved in Rho-GTPase signaling in neurons.
Conclusions: In conclusion, genome-wide SNP analysis is a viable way to identify CNVs related to autism. By filtering only those that are large and affect neuronal genes, we have focused our analysis to potentially pathogenic CNVs. This approach has replicated the presence of previously reported autism CNVs and supports the role of CHRNA7 in autism. Moreover, we have identified two additional potential candidates in MTHFD1 and PLEKHG3. Further analysis will be necessary to better understand the roles of these genes in the etiology of autism.