DNA methylation QTL analysis identifies new regulators of human longevity

Abstract

Human longevity is a complex trait influenced by both genetic and environmental factors, whose interaction is mediated by epigenetic mechanisms like DNA methylation. Here, we generated genome-wide whole-blood methylome data from 267 individuals, of which 71 were long-lived (90-104 years), by applying reduced representation bisulfite sequencing. We followed a stringent two-stage analysis procedure using discovery and replication samples to detect differentially methylated sites (DMSs) between young and long-lived study participants. Additionally, we performed a DNA methylation quantitative trait loci analysis to identify DMSs that underlie the longevity phenotype. We combined the DMSs results with gene expression data as an indicator of functional relevance. This approach yielded 21 new candidate genes, the majority of which are involved in neurophysiological processes or cancer. Notably, two candidates (PVRL2, ERCC1) are located on chromosome 19q, in close proximity to the well-known longevity- and Alzheimer's disease-associated loci APOE and TOMM40. We propose this region as a longevity hub, operating on both a genetic (APOE, TOMM40) and an epigenetic (PVRL2, ERCC1) level. We hypothesize that the heritable methylation and associated gene expression changes reported here are overall advantageous for the LLI and may prevent/postpone age-related diseases and facilitate survival into very old age.

Figure 1

Flowchart summarizing the study design and statistical approach. DMS, differential methylated site; GO, gene ontology; LLI, long-lived individuals; mQTL, methylation quantitative trait locus.

Figure 2

Depiction of age-related methylation at the CpG site chr6.150040098 annotated to large tumor suppressor kinase 1 (LATS1) and associated gene expression changes in the discovery and replication sample, respectively. (A) methylation difference between young and old (i.e. long-lived, ≥91 years) individuals in the discovery sample (extreme group comparison); (B) methylation changes over the age range 26–102 years in the replication sample; (C) positive correlation between methylation and gene expression; (D) gene expression difference between young and old (i.e. long-lived, ≥91 years) individuals in the discovery sample (extreme group comparison); (E) allele-specific methylation at rs10872646 with additive coding (0, homozygous for the major allele; 1, heterozygous; 2, homozygous for the minor allele).

Figure 3

The protein–protein interaction network with the 21 candidate genes and their top 13 interaction partners as input variables analyzed using the STRING database. The top 13 interaction partners were selected based on the InterMine database. The colored edges represent the types of evidence used in the interaction predictions by the STRING software: experimental evidence (dark pink), text mining evidence (light green), database evidence (light blue), protein homology evidence (purple), and co-expression evidence (black). The 21 candidate genes are indicated by bold circles.

Figure 4

The 21 candidate genes clustered according to their functions and/or associations with diseases or other phenotypes.

Figure 5

Genomic locations of PVRL2, TOMM40, APOE and ERCC1 with some adjacent genes on chromosome 19q.