Using Lymphoblastoid Cell Lines to Investigate Altered DNA Methylation in a Subtype of ASD

Background: Lymphoblastoid cell lines (LCL) have previously been used as an experimental model to examine differences in DNA methylation between monozygotic twins and sib pairs discordant for diagnosis of autism spectrum disorder (ASD).  The rationale for using twins and siblings in epigenetic studies is that the identical or similar genotypes of the twins/siblings would increase the probability of identifying differentially methylated genes that were related to the differential diagnosis of the siblings rather than to disparate genotypes.

Objectives: The goals of this study were to: 1) investigate changes in DNA methylation in LCL from individuals with a specific subtype of ASD and their respective non-autistic siblings, 2) relate the differentially methylated genes to ASD-relevant pathways and functions, and 3) explore the potential for class determination (case-control) biomarkers based on methylation signatures.

Methods:

Selection of cases and controls by subtyping analyses:  To reduce the confounding effects of clinical heterogeneity on the methylation analyses, we subtyped individuals with ASD by performing cluster analyses of ADI-R scores as previously described (Autism Research 2:67-77, 2009), limiting the cases for this study to those with severe language impairment which was further verified by low scores on the Peabody Picture Vocabulary Test.  To further reduce heterogeneity among the 21 sib pairs included in this study, we selected sibling controls of the same gender as the cases.

Analysis of differential DNA methylation in cases and controls:  Affymetrix Human Promoter 1.0R GeneChips were used to analyze differentially methylated promoter regions in the DNA of cases and controls.  For each sample, DNA enriched for methylated regions by methyl-C immunoprecipitation (MeDIP) procedures and input (total) DNA were hybridized on separate chips.  Partek GS Software was used for data analysis of the promoter tiling arrays.

Pathway/functional analyses of differentially methylated genes:  Ingenuity Pathway Analysis network prediction software was used to identify functions and canonical pathways represented by the gene sets encompassing the most significant differentially methylated genes.  This information is being used to select relevant candidate genes for functional validation.

Results: The most significant differentially methylated genes were identified by both paired and unpaired t-tests, with P≤ 0.0001 and a minimum MAT score of ± 5.0.  Pathway analyses of the genes from the t-tests show enrichment in gene networks involved in mTOR signaling, Reelin signaling, semaphorin signaling, and axon guidance.  Moreover, principal components analyses of the samples based upon the lists of differentially methylated annotated promoter regions show relatively good separation of cases and controls. Application of class prediction algorithms to the sets of differentially methylated genes demonstrated the ability to correctly identify cases and controls with >90% accuracy based on a limited number (e.g., 18-35) of genes/regions.

Conclusions: Significantly differentially methylated genes in LCL from autistic vs. non-autistic individuals suggest pathways that may be aberrantly regulated in ASD.  In addition, limited panels of methylation-enriched promoter regions can discriminate between autistic cases and controls with reasonably high sensitivity and specificity.  These results suggest that these differentially methylated regions may be useful as “biomarkers” for the subtype of ASD with language-impairment.