DNA methylome profiling of all-cause mortality in comparison with age-associated methylation patterns

Abstract

Background: Multiple epigenome-wide association studies have been performed to identify DNA methylation patterns regulated by aging or correlated with risk of death. However, the inter-relatedness of the epigenetic basis of aging and mortality has not been well investigated.

Methods: Using genome-wide DNA methylation data from the Lothian Birth Cohorts, we conducted a genome-wide association analysis of all-cause mortality and compared this with age-associated methylation patterns reported on the same samples.

Results: Survival analysis using the Cox regression model identified 2552 CpG sites with genome-wide significance (false discovery rate < 0.05) for all-cause mortality. CpGs whose methylation levels are associated with increased mortality appear more distributed from the gene body to the intergenic regions whereas CpGs whose methylation levels are associated with decreased mortality is more concentrated at the promoter regions. In comparison with reported CpGs displaying significant age-dependent methylation patterns in the same samples, we observed a limited but highly significant overlap between mortality-associated and age-associated CpGs (p value 2.52e-06). Most importantly, the overlapping CpGs are dominated by those whose overall age-related methylation patterns reduce the risk of death.

Conclusion: All-cause mortality is significantly associated with altered methylation at multiple genomic sites with differential distribution in gene regions for CpGs correlated with increased or decreased risk of death. The age-dependent methylation changes could reflect an active response to the aging process that contributes to maintain individual survival.

Keywords: Aging; DNA methylation; Epigenome-wide association study; Mortality; Old cohorts.

Fig. 1

Illustration of results from EWAS on mortality presented by plotting − log₁₀(p value) against CpG’s chromosomal location (a, Manhattan plot) and CpG’s regression coefficient in the Cox model (b, volcano plot)

Fig. 2

Star plot based on CpGs showing positive (red) and negative (blue) associations with mortality. The arms of the stars represent gene regions where CpGs are located. a Proportions of CpGs distributed over different regions for positive (red) and negative (blue) mortality-associated CpGs, as well as for all the CpGs on the array (black). b The absolute proportions of positive (red) and negative (blue) mortality-associated CpGs at each gene region

Fig. 3

Scatter plots for mortality-associated CpGs plotted against their coefficients for age from the linear model in the EWAS on aging (a) and for aging-associated CpGs plotted against their coefficients from the Cox model in the EWAS on mortality (b). The colored large dots are 178 CpGs with genome-wide significance in both the EWAS on aging (FWER < 0.05) and the EWAS on mortality (FDR < 0.05)