Novel regional age-associated DNA methylation changes within human common disease-associated loci

Abstract

Background: Advancing age progressively impacts on risk and severity of chronic disease. It also modifies the epigenome, with changes in DNA methylation, due to both random drift and variation within specific functional loci.

Results: In a discovery set of 2238 peripheral-blood genome-wide DNA methylomes aged 19-82 years, we identify 71 age-associated differentially methylated regions within the linkage disequilibrium blocks of the single nucleotide polymorphisms from the NIH genome-wide association study catalogue. This included 52 novel regions, 29 within loci not covered by 450 k or 27 k Illumina array, and with enrichment for DNase-I Hypersensitivity sites across the full range of tissues. These age-associated differentially methylated regions also show marked enrichment for enhancers and poised promoters across multiple cell types. In a replication set of 2084 DNA methylomes, 95.7 % of the age-associated differentially methylated regions showed the same direction of ageing effect, with 80.3 % and 53.5 % replicated to p < 0.05 and p < 1.85 × 10^-8, respectively.

Conclusion: By analysing the functionally enriched disease and trait-associated regions of the human genome, we identify novel epigenetic ageing changes, which could be useful biomarkers or provide mechanistic insights into age-related common diseases.

Keywords: Ageing; Common Disease; DNA methylation; Epigenetics; Human.

Fig. 1

Bidirectional Manhattan plot of age-associated differentially methylated regions (a-DMRs). Loci with a positive correlated with age are displayed in the positive y-axis and those regions where DNA methylation changes negatively correlate with age are displayed in the negative y-axis

Fig. 2

Chromatin segmentation enrichment. a Overlap with a-DMRs and GWAS LD blocks for ENCODE ChromHMM [30] genome segmentation from ENCODE in nine tissues (data via and graph adapted from Epiexplorer [85]). Strong poised promoter enrichment seen. Observed versus expected for the GWAS LD block regions. b Overlap with a-DMRs and LD blocks for combined genome segmentation from ENCODE (ChromHMM [30] and Segway [31]) in six tissues. Strong enrichment for enhancers is evident. Sphere size is proportional to genomic space. TSS predicted promoter region including transcription start site, PF predicted promoter flanking region, E predicted enhancer, WE predicted weak enhancer or open chromatin cis regulatory element, CTCF CTCF enriched element, T predicted transcribed region, R predicted repressed or low activity region

Fig. 3

Fold enrichments for a-DMRs compared to LD block non-overlapped 500-bp windows. All categories are enriched in a-DMRs (χ² p all < 1 × 10^–5) except for transcripts (near identical fraction), SINE and LTR repeats (non-significant) and all repeats and LINE repeats, which are significantly depleted (χ² p = 6.73 × 10^–9 and 2.81 × 10^–3, respectively). CGI CpG Islands, TSS transcription start sites, DNase I HSs DNase I hypersensitivity sites in 125 cell types, TFBS transcription factor binding sites, CTCF and Sp1 from all tissues (ENCODE v3), Repeats All, SINE, LINE, LTR repeats, FANTOM5 Enhancers [36], Dynamic Regions [66] and Conserved (100 Vertebrate) regions [87]

Fig. 4

MEME-ChIP [38] enriched sequence identified in hypomethylated a-DMRs (below) and TOMTOM (v4.10.2) enrichment for the SP1 transcription factor motif (above)

Fig. 5

Selected a-DMRs within (i) genomic location; from top: a-DMRs (purple), gene, DNase I HS clusters, transcription factor ChIP-seq, ChromHMM segmentation, combined segmentation and conservation; and (ii) scatterplot: x-axis = age, y-axis = normalised methylation. a) HPAS2 b) IGFBP4

Fig. 6

DHS fold enrichments for a-DMRs compared to the regions within LD blocks. y-axis indicates log₂ fold enrichment. An enrichment across multiple tissue types is seen