Whole-Genome Bisulfite Sequencing Reveals Autism-Associated Hypomethylation and Differentially-Methylated Regions in Umbilical Cord Blood Samples from the Prospective Marbles Study

Background: Autism spectrum disorders (ASD) have complex etiologies, likely involving multiple genetic and environmental insults in perinatal life. Genetic susceptibility can interact with environmental risk factors such as pesticides, air pollution, and persistent organic pollutants. The perinatal period is critical for both nutritional protective factors, such as the methyl donor folate, and interactions with genetic regulators of one-carbon metabolism. The epigenetic layer of DNA methylation, at the interface of genetic and environmental risk and protective factors, holds promise for improved understanding of complex ASD etiologies.

Objectives: Currently, ASD is diagnosed behaviorally at two years of age; however, existing interventions are most effective the earlier they are begun. Biomarkers for ASD at birth are largely unknown and would facilitate earlier diagnosis and more effective treatment. We performed this study to identify DNA methylation biomarkers predictive of ASD diagnosis by age three.

Methods: The MARBLES (Markers of Autism Risk in Babies - Learning Early Signs) prospective study is an enriched risk cohort that enrolls couples who have already had a child with ASD and follows their subsequent pregnancy. We investigated human umbilical cord blood samples from the MARBLES study by whole-genome bisulfite sequencing (WGBS) and expression microarray (n= 26 TD, 26 ASD).

Results: ASD cord blood samples showed significantly lower global percent CpG methylation compared to typically-developing (TD) controls (ASD 76.6% vs TD 77.4%, p = 0.01). ASD-associated hypomethylation was observed across most chromosomes and gene bodies. 20 kb windowing of the genome demonstrated a global shift of hypomethylation, with 81% of windows hypomethylated in ASD samples. Smaller differentially-methylated regions (DMRs) enriched for CpG islands were also identified in ASD cord blood (5,828 total DMRs). 60% of all DMRs were hypomethylated, but all DMRs that were significant after correction for multiple hypothesis testing were hypermethylated. DMRs in ASD cord blood were nearby genes significantly enriched for functions in organic cyclic compound binding (2797 DMRs, Bonferroni-adjusted p = 2.4x10^-11) and abnormal blood cell morphology and development knockout mouse phenotypes (1277 DMRs, Bonferroni-adjusted p = 1.1x10^-11), highlighting the potential influence of the environment and immune system in ASD. Hypomethylated DMRs were also significantly enriched near SFARI ASD candidate risk genes (253 genes, Bonferroni-adjusted p = 7.6x10^-4). Methylation at four of the DMRs with genome-wide significance showed a significant positive association with Autism Diagnostic Observation Schedule (ADOS) severity scores. Methylation at two significant DMRs showed a significant negative association with mRNA levels of nearby genes.

Conclusions: Global hypomethylation in ASD cord blood suggests a methylation deficiency in ASD during perinatal life, which could be a cumulative effect of genetic variants, environmental exposures, and/or shortage in methyl donors. Identified DMRs are relevant to ASD and have potential as a diagnostic tool. In future studies, methylation will be examined in relation to demographic, genetic, environmental, and nutritional information collected in the MARBLES study. These results are expected to improve understanding of perinatal factors in ASD etiology and aid in future preventative and therapeutic treatments.