Inheritance of DNA methylation level in healthy human tissues

Abstract

DNA methylation (DNAm) is the covalent modification of DNA by addition of a methyl group primarily at the cytosine directly upstream of a guanine. DNAm level plays a central role in transcriptional regulation and is linked to disease. Therefore, understanding genetic and environmental influences on DNAm level in healthy tissue is an important step in the elucidation of trait and disease etiology. However, at present only a minority of easy to access human tissues and ethnicities have been investigated. Therefore, we studied DNAm level measured in five human tissues: cerebellum, frontal cortex, pons, temporal cortex and colon in either North American or South American samples. We applied a novel statistical approach to estimate the heritability attributable to genomic regions (regional heritability, ĥ²/r,g ) for DNAm level at thousands of individual DNAm sites genome-wide. In all five tissues, DNAm level was significantly associated with the local genomic region for more DNAm sites than expected by chance. Moreover, DNAm level could be predicted from the local sequence variants with an accuracy that scaled with the estimated ĥ²/r,g . Our results inform on molecular mechanisms regulating DNAm level and trait etiology in several ways. Firstly, DNAm level at DNAm sites located in genomic risk regions and measured in a tissue relevant to the disease can be influenced by the local genetic variants. Specifically, we found that genetic variation within a region associated with Fluid Intelligence was also associated with local DNAm level at the proline-rich coiled-coil 1 (PRRC1) gene in healthy temporal cortex tissue. Additionally, we replicated the finding of a Colorectal Cancer risk variant (rs4925386) associated with two DNAm sites in healthy colon tissue. More generally, we showed that DNAm sites located within a susceptibility region and measured in a relevant tissue exhibit a similar overall pattern of estimated ĥ²/r,g to DNAm sites outwith a susceptibility region. Secondly, the propensity for DNAm level to be associated with the local sequence variation differs with respect to CpG dinucleotide density and genic location. Most notably, DNAm sites located in CpG dense regions of the genome are less likely to be heritable than DNAm sites located in CpG sparse regions of the genome. Additionally, within both CpG dense and CpG sparse regions of the genome intergenic DNAm sites are more likely to be heritable than intragenic DNAm sites. Overall, our study suggests that variation in DNAm level at some DNAm sites is at least partially controlled by nuclear genetic variation. Moreover, DNAm level in healthy tissue has the potential to act as an intermediary in trait variation and etiology.