An integrative approach to detect epigenetic mechanisms that putatively mediate the influence of lifestyle exposures on disease susceptibility

Abstract

Background: There is mounting evidence that our environment and lifestyle has an impact on epigenetic regulatory mechanisms, such as DNA methylation. It has been suggested that these molecular processes may mediate the effect of risk factors on disease susceptibility, although evidence in this regard has been challenging to uncover. Using genetic variants as surrogate variables, we have used two-sample Mendelian randomization (2SMR) to investigate the potential implications of putative changes to DNA methylation levels on disease susceptibility.

Methods: To illustrate our approach, we identified 412 CpG sites where DNA methylation was associated with prenatal smoking. We then applied 2SMR to investigate potential downstream effects of these putative changes on 643 complex traits using findings from large-scale genome-wide association studies. To strengthen evidence of mediatory mechanisms, we used multiple-trait colocalization to assess whether DNA methylation, nearby gene expression and complex trait variation were all influenced by the same causal genetic variant.

Results: We identified 22 associations that survived multiple testing (P < 1.89 × 10-7). In-depth follow-up analyses of particular note suggested that the associations between DNA methylation at the ASPSCR1 and REST/POL2RB gene regions, both linked with reduced lung function, may be mediated by changes in gene expression. We validated associations between DNA methylation and traits using independent samples from different stages across the life course.

Conclusion: Our approach should prove valuable in prioritizing CpG sites that may mediate the effect of causal risk factors on disease. In-depth evaluations of findings are necessary to robustly disentangle causality from alternative explanations such as horizontal pleiotropy.

Keywords: ALSPAC; ARIES; DNA methylation; Mendelian randomization; mediation; smoking.

Figure 1.

An overview of the proposed two-step epigenetic Mendelian randomization approach to evaluate findings from epigenome-wide association studies. (a) Identify CpGs sites from EWAS where a risk factor is associated with DNA methylation. Prenatal risk factors can add value as they are very unlikely to be associated due to reverse causation. (b) Use independent cis-acting methylation quantitative trait loci (mQTL) as an instrumental variable to proxy for changes in DNA methylation levels at this CpG site, allowing investigation into how these effects may influence complex traits.

Figure 2.

Flowchart outlining the analysis pipeline used in this study along with findings from our applied example and the data resources used. mQTL, methylation quantitative trait loci; DNAm, DNA methylation; ARIES, accessible resource for integrated epigenomics studies; GWAS, genome-wide association studies; eQTL, expression quantitative trait loci; ALSPAC, Avon Longitudinal Study of Parents and Children.

Figure 3.

Illustration of genetic colocalization at the ASPSCR1 locus between gene expression and lung function (left) as well as DNA methylation at CpG site cg06105699 and lung function (right). Overlapping distributions of –log₁₀P-values for the effects of genetic variants at the ASPSCR1 gene region on forced expiratory volume in 1 second (FEV1) (grey), gene expression of ASPSCR1 in whole blood (blue) and DNA methylation at cg06105699 (red). Genes at this region are annotated below the x-axis, where a red cross indicates the position of the associated CpG site. Multiple-trait colocalization provides strong evidence that FEV1, ASPSCR1 expression and DNA methylation at cg06105699 all share the same underlying causal variant at this locus. This supports evidence that changes in DNA methylation at this CpG site may influence lung function via changes in gene expression.