Genetic contribution to variation in DNA methylation at maternal smoking sensitive loci in exposed neonates

Epigenome-wide DNA methylation association studies have identified highly replicable genomic loci sensitive to maternal smoking during gestation. The role of inter-individual genetic variation in influencing DNA methylation, leading to the possibility of confounding or bias of such associations, has not been assessed. We investigated whether the DNA methylation levels at the top ten CpG sites previously associated with exposure to maternal smoking during gestation were associated with individual genetic variation at the genome-wide level. Genome-wide association tests between DNA methylation at the top ten candidate CpG and genome-wide SNPs were performed in 736 case and control participants of the California Childhood Leukemia Study. Three of the strongest maternal-smoking sensitive CpG sites in newborns were significantly associated with SNPs located proximal to each gene: <i>cg18146737</i> in the <i>GFI1</i> gene with <i>rs141819830</i> (<i>P</i> = 8.2×10⁻⁴⁴), <i>cg05575921</i> in the <i>AHRR</i> gene with <i>rs148405299</i> (<i>P</i> = 5.3×10⁻¹⁰), and <i>cg12803068</i> in the <i>MYO1G</i> gene with <i>rs61087368</i> (<i>P</i> = 1.3×10⁻¹⁸). For the <i>GFI1</i> CpG <i>cg18146737</i>, the underlying genetic variation at <i>rs141819830</i> confounded the association between maternal smoking and DNA methylation in our data [OR changed from 0.89 (0.75–1.06, <i>P</i> = 0.184) to 0.78 (0.63–0.97, <i>P</i> = 0.024) after including the genotype]. Our results suggest that further studies using DNA methylation at <i>cg18146737, cg05575921</i>, or <i>cg12803068</i> that aim to assess exposure to maternal smoking during gestation should include genotype at the corresponding SNP. New methods are required for adequate and routine inclusion of genotypic influence on DNA methylation in epigenome-wide association studies to control for potential confounding.