Research Article: Mendelian Randomization Analysis Identifies CpG Sites as Putative Mediators for Genetic Influences on Cardiovascular Disease Risk

Date Published: October 05, 2017

Publisher: Elsevier

Author(s): Tom G. Richardson, Jie Zheng, George Davey Smith, Nicholas J. Timpson, Tom R. Gaunt, Caroline L. Relton, Gibran Hemani.


The extent to which genetic influences on cardiovascular disease risk are mediated by changes in DNA methylation levels has not been systematically explored. We developed an analytical framework that integrates genetic fine mapping and Mendelian randomization with epigenome-wide association studies to evaluate the causal relationships between methylation levels and 14 cardiovascular disease traits. We identified ten genetic loci known to influence proximal DNA methylation which were also associated with cardiovascular traits after multiple-testing correction. Bivariate fine mapping provided evidence that the individual variants responsible for the observed effects on cardiovascular traits at the ADCY3 and ADIPOQ loci were potentially mediated through changes in DNA methylation, although we highlight that we are unable to reliably separate causality from horizontal pleiotropy. Estimates of causal effects were replicated with results from large-scale consortia. Genetic variants and CpG sites identified in this study were enriched for histone mark peaks in relevant tissue types and gene promoter regions. Integrating our results with expression quantitative trait loci data, we provide evidence that variation at these regulatory regions is likely to also influence gene expression levels at these loci.

Partial Text

Approximately 88% of trait-associated variants detected by genome-wide association studies (GWASs) reside in non-coding regions of the genome and might act through gene regulation.1 Recent studies have incorporated data on genetic variants associated with gene expression (expression quantitative trait loci [eQTLs]) into results from GWASs of complex traits to help identify the putative causal variant in a genomic region, as well as provide evidence suggesting which genes might be influenced by this variant.2, 3, 4, 5 This direction of inquiry can be extended to other “-omic” data types to gain further insights into the mechanistic pathway between genetic variant and causally associated trait. In this study, we introduce an alternative analytical framework to integrate genetic predictors of DNA methylation levels with complex traits to evaluate bi-directional causal relationships.

We have designed a framework for evaluating the putative causal influences of DNA methylation on complex traits and disease via MR. For observed effects on cardiovascular traits that appear to be caused by methylation, we used bivariate fine mapping and JLIM to evaluate whether the putative causal variant influencing methylation was the same causal variant responsible for influencing the trait. The bivariate fine mapping suggested that cardiovascular traits might be influenced by altered DNA methylation levels at the ABO, ADCY3, ADIPOQ, APOA1, APOB, and IL6R regions. However, JLIM supported findings only at the ADCY3 and ADIPOQ loci. This provides compelling evidence that DNA methylation might play a mediatory role for the effects at these loci. 2SMR analyses provided evidence that DNA methylation levels influenced gene expression at these loci, suggesting that functional effects for the causal variants induce a coordinated system of effects. This was important to demonstrate, given that having only single valid instruments available for CpGs meant that we were unable to robustly show that variants were not influencing methylation and traits through horizontal pleiotropy. This limitation has also been encountered by other attempts to evaluate the relationship between DNA methylation and complex traits.16 Nevertheless, the ability to indicate putative mediating molecular phenotypes between genetic factors and complex traits is particularly attractive for therapeutic evaluation of drug targets.




Leave a Reply

Your email address will not be published.