WHAT SHAPES THE CHILD'S EPIGENOME: GENOTYPE, PRENATAL ENVIRONMENT OR BOTH?

Darina Czamara,Goekcen Eraslan,Anna Suarez,Polina Girchenko,Jari Lahti,Marius Lahti,Esa Hamalainen,Eero Kajantie,Hannele Laivuori,Rebecca M. Reynolds,Ivan Kondofersky,Nikola S. Mueller,Fabian J. Theis,Katri Raikkonen,Elisabeth B. Binder

DOI: https://doi.org/10.1016/j.euroneuro.2017.08.363

IF: 5.415

2019-01-01

European Neuropsychopharmacology

Abstract:Prenatal programming describes the process in which environmental events during pregnancy shape and determine the development of the embryo with on-going implications until adulthood. One main trigger of this process is epigenetic changes, for example in DNA methylation. We also know that the genome plays an important role in the regulation of DNA-methylation and de-methylation and a variety of studies have identified meQTLs (methylation quantitative trait loci), i.e. SNPs that are significantly associated with methylation status. We evaluated whether the prenatal environment, genotype and prenatal environment or genotype x prenatal environment interaction was the best predictor of methylation levels in the child's epigenome. We assessed epigenome-wide DNA-methylation levels and genome-wide SNP-genotypes from 817 Finnish cord blood samples from the Prediction and Prevention of Preeclampsia and Intrauterine Growth Restriction (PREDO) cohort. Ten different prenatal environments with regard to prenatal stress (depression score, anxiety score, betamethasone intake), general maternal factors (maternal age, delivery mode, parity) as well as the mother's metabolism (pre-pregnancy BMI, maternal hypertension, gestational diabetes, oral glucose tolerance test) were investigated. Focusing on Variably Methylated Regions (VMRs) and on cis-effects (maximal distance of 1MB between SNP and CpG-site), we tested if environment, SNP-genotype, genotype and environment or genotype x environment interaction best fitted VMR methylation levels. We applied DeepSea, a machine-learning based method, on the top results to identify functionally relevant SNPs. 44% of all VMRs were best explained by SNP genotype alone, 32% by SNP x environment interaction and 24% by additive effects of SNP and environment. Environment alone was only once the best predictor. 8,124 (0.1%) of all investigated SNPs were predicted to be of functional significance. Functionally relevant meQTL-SNPs were located in close proximity to the specific CpG-site and enriched in promoter regions. Functionally relevant SNPs involved in genotype x environment models showed much broader peaks around the CpGs and were enriched for distal intergenic regions. Both SNP groups show different patterns with regard to enrichment for specific transcription factor binding sites. Our results suggest that either genotype, genotype and environment or genotype x environment but not environment alone predict variable methylation in newborn cord blood. Therefore, it is important to include environmental outcomes and genotypic data in epigenetic studies. Furthermore, while meQTLs were enriched in promoter regions, the broader peaks of SNPs involved in interaction models suggest long-distance effects are present in GxE interactions.