Identification of Shared Molecular Pathway Involved In Autism by Transcriptional Profiling

Background:  ASD is a heterogeneous disorder of neural development that results from the combined effects of genetic changes interacting with environmental factors. It is highly inheritable, yet only a few genetic factors have been specifically identified. Rare chromosomal disorders, such as copy number variations (CNVs), are among the most common causes of ASD so far, yet each loci accounting for less than 1% of ASD. Given its complexity, focusing on recurrent monogenic forms of ASD with rare CNVs can provide important insights into disease pathogenesis.

Objectives:  With genome-wide transcriptional profiling, we aim to identify genes that are dysregulated concurrently in different recurrent monogenic forms of ASD with rare CNVs: 15q11-q13 duplication (dup(15q)), 16p11.2 deletion (del(16p)), and 22q11.2 duplication (dup(22q)). The shared genetic changes may underlie the common molecular pathways involved in ASD pathogenesis.

Methods: We performed genome-wide measurement of RNA expression using Illumina Human Ref8_v2 microarrays on lymphoblast cells lines from 13 cases (4 del(16p), 6 dup(15q), 3 22q(dup)) and 14 age and gender matched controls. Differential expression was assessed using the limma R package with threshold of p-value < 0.05. We applied Weighted Gene Co-expression Network Analysis (WGCNA) to identify co-expressed gene groups (modules), under the assumption that the genes in the same module were more likely to be functionally related.

Results:  We identified 67 genes differentially expressed (DE) between all autism and controls. GO analysis revealed enrichment for genes categorized by neurological process of cell lines (p-value=1.65E-5) and neurite branching (p-value=3.30E-3). Hierarchical clustering using the 67 genes clearly distinguished different ASD forms from each other and controls. Remarkably, del(16p) and dup(22q) showed very similar gene expression pattern over the DE genes, indicating the convergence of molecular pathways in these two distinct ASD conditions. WGCNA identified four gene co-expression modules that were significantly correlated with autism, one of which was enriched with known autism candidate genes, such as PAK2, and ATRX. Two modules separated the del(16p) and dup(22q) autism groups from the dup(15q) and controls, supporting the notion that del(16p) and dup(22q) may share molecular pathways. One of the top gene ontology categories in these modules was related to alternative splicing, a mechanism commonly employed during normal neural development.

Conclusions: These results support the use of blood-derived lymphoblastoid cells to identify etiological subsets of autism. The DE analysis and network analysis bring the initial evidence of convergent molecular dysregulation exists in autism, especially for the del(16p) and dup(22q). We plan to run more arrays from these recurrent CNV to try to extend and replicate these results. The new samples will be run with current cases together as a new cohort to remove potential confounders, such as batch effects, so as to increase our statistic power.