Maternal Blood DNA Methylation during Pregnancy and Autism Observational Scale for Infants (AOSI) Score at 12-Months in the Early Autism Risk Longitudinal Investigation (EARLI)
Objectives: To test the association between maternal pregnancy whole blood DNA methylation and offspring scores on the 12-month Autism Observational Scale for Infants (AOSI) scale.
Methods: In an ASD-enriched birth cohort, the Early Autism Risk Longitudinal Investigation (EARLI), genome-wide maternal pregnancy blood DNA methylation was measured using two complementary techniques: comprehensive high-throughput arrays for relative methylation (CHARM) and the Illumina Infinium 450k HumanMethylation array. Multivariable linear regression was used to test the association between percent methylation and offspring log transformed 12-month AOSI score (ln(AOSI+1)), adjusting for cell composition, laboratory batch, maternal age, and race. Due to differences in probe density between the methods, CHARM data was used to test for differentially methylated regions and 450k array data was used to identify differentially methylated single CpG sites. As a secondary analysis, we also tested AOSI as a dichotomous variable predicting ASD risk (AOSI < 7 vs. AOSI >=7). We tested for enrichment of locations of DNA methylation differences in gene ontology biological processes as well as autism-associated genes in the Simons Foundation Autism Research Initiative (SFARI) autism-related database.
Results: Paired maternal 450k DNA methylation data from any time during pregnancy and 12-month offspring AOSI scores were available on 79 EARLI families. Infant scores ranged from 0-20 (mean(SD)=5.4(4.8)). Across 450k array single-site models from 485,512 CpGs, we observed a lambda genomic inflation factor of 1.06 and top statistically significant CpG sites associated with offspring AOSI scores overlapped previously identified ASD genetic risk loci. CHARM data from over 2.1 million probes linked with AOSI scores were available from up to 4 repeated visits per mom (V1, n=77; V2, n=76; V3, n=59; V4, n=16), which are now being used to characterize any longitudinal trajectories of ASD-associated maternal DNA methylation differences.
Conclusions: We identified genomic locations of DNA methylation sites measured in maternal pregnancy blood that are associated with ASD-related phenotype at 12-months. These appear to be enriched for genes already implicated in ASD risk through traditional genetic associations and thus may provide further insight about the role of maternal genetic and epigenetic influences on the biology of ASD risk. Further work is needed, including replication in additional populations and model organisms, to clarify ASD mechanisms or the potential role of maternal DNA methylation as a risk biomarker.