Evidence of positive and negative selection associated with DNA methylation

Charlie Hatcher, Gibran Hemani, Santiago Rodrı́guez, Tom R. Gaunt, Daniel J. Lawson, Josine L. MinbioRxiv 2021 · DOI

Abstract

Signatures of negative selection are pervasive amongst complex traits and diseases. However, it is unclear whether such signatures exist for DNA methylation (DNAm), which has been proposed to have a functional role in disease. We estimate polygenicity, SNP-based heritability, and model the joint distribution of effect size and minor allele frequency (MAF) to estimate a selection coefficient ($S$) for 2,000 heritable DNAm sites in 1,774 individuals from the Avon Longitudinal Study of Parents and Children. Additionally, we estimate $S$ for metastable epialleles and DNAm sites associated with aging and mortality, birthweight, and body mass index. Quantification of MAF-dependent genetic architectures reveal evidence of both positive ($S > 0$) and negative selection ($S < 0$) and confirm previous evidence of negative selection for birthweight. These findings highlight the role of DNAm as an intermediary in multiple biological pathways with competing functions.

Introduction

While natural selection is known to prune deleterious mutations from complex traits (negative selection), this study explores whether similar evolutionary forces shape the epigenome. Specifically, it focuses on DNA methylation (DNAm)—chemical "tags" on DNA that regulate gene expression. If DNAm is functional and affects fitness, evolution should leave a signature on the genetic variants that control it (mQTLs). This study demonstrates that DNAm sites experience a tug-of-war between both positive and negative selection, suggesting the epigenome is a functional landscape where evolution actively favors certain regulatory states while suppressing others.

Methods

The researchers utilized BayesNS, a Bayesian mixed linear model, to estimate genetic architecture parameters. The model examines the joint distribution of effect size and Minor Allele Frequency (MAF). In a neutral scenario, effect size is independent of frequency. If rare variants have disproportionately large effects, it signals negative selection ($S < 0$). Conversely, if common variants have large effects, it can signal positive selection ($S > 0$). Unlike previous studies that looked at selection across the entire genome for a single trait, this study applied the model to thousands of individual DNAm sites. The authors also implemented models to account for genetic drift (e.g., the "out-of-Africa" bottleneck) to ensure signals were biological rather than statistical artifacts.

Results and Discussion

The study analyzed data from 1,774 individuals in the ARIES cohort (ALSPAC). Target sites included 2,000 highly heritable DNAm sites, 1,508 "epigenetic supersimilarity" (ESS) sites, and 513 "PhenoAge" sites associated with biological aging.

Selection Coefficients and Heritability

The analysis revealed a dual nature of selection. Unlike many complex traits that primarily show negative selection, DNAm sites show evidence of both positive and negative selection. Highly heritable sites showed a mildly negative average selection coefficient ($\bar{S} = -0.15$), but individual sites reached extreme values ($S > 1$ or $S < -1$). The average SNP-based heritability across the studied DNAm sites was approximately 28.6%, which is higher than previous estimates.

Summary of genetic architecture parameters for different DNAm site categories.

Polygenicity and Functional Enrichment

A striking relationship was found between polygenicity and selection type. Positive selection was associated with "simple" architectures (few variants with large effects), while negative selection was linked to "complex" architectures (many variants with small effects). Positively selected sites were enriched in enhancers and promoters, whereas negatively selected sites were enriched in "poised" promoters—regions kept ready for activation but currently silenced.

Trait-Specific Selection

The model confirmed strong negative selection for DNAm sites associated with birthweight ($p = 3.432 \times 10^{-6}$), consistent with evolutionary pressure to maintain optimal birth size. In contrast, sites associated with PhenoAge and BMI did not show significant enrichment for selection in this specific test.

Fisher’s exact test results for enrichment of selection in trait-associated DNAm sites.

Conclusion and Future Directions

This study provides evidence that DNAm is a functional intermediary subject to complex evolutionary pressures. However, limitations remain. The use of the 450k array covers only 1.5% of the genome and is biased toward promoter regions. Future research using whole-genome bisulfite sequencing or newer EPIC arrays may reveal selection in more remote regulatory regions. Additionally, larger sample sizes are needed to detect "trans" mQTLs and capture the full polygenic tail of negative selection. Finally, expanding this research to diverse global cohorts is essential to determine if these evolutionary signatures are universal or population-specific.

Summary updated

Abstract

Abstract Signatures of negative selection are pervasive amongst complex traits and diseases. However, it is unclear whether such signatures exist for DNA methylation (DNAm) that has been proposed to have a functional role in disease. We estimate polygenicity, SNP-based heritability and model the joint distribution of effect size and minor allele frequency (MAF) to estimate a selection coefficient ( S ) for 2000 heritable DNAm sites in 1774 individuals from the Avon Longitudinal Study of Parents and Children. Additionally, we estimate S for meta stable epi alleles and DNAm sites associated with aging and mortality, birthweight and body mass index. Quantification of MAF-dependent genetic architectures estimated from genotype and DNAm reveal evidence of positive ( S &gt; 0) and negative selection ( S &lt; 0) and confirm previous evidence of negative selection for birthweight. Evidence of both negative and positive selection highlights the role of DNAm as an intermediary in multiple biological pathways with competing function.
BibTeX
@misc{hatcher2021evidence,
  author = {Charlie Hatcher and Gibran Hemani and Santiago Rodrı́guez and Tom R. Gaunt and Daniel J. Lawson and Josine L. Min},
  title = {{Evidence of positive and negative selection associated with DNA methylation}},
  year = {2021},
  doi = {10.1101/2021.11.25.469994},
  howpublished = {bioRxiv},
}
Figure