Novel DNA methylation signatures of tobacco smoking with trans-ethnic effects

Abstract

Background: Smoking remains one of the leading preventable causes of death. Smoking leaves a strong signature on the blood methylome as shown in multiple studies using the Infinium HumanMethylation450 BeadChip. Here, we explore novel blood methylation smoking signals on the Illumina MethylationEPIC BeadChip (EPIC) array, which also targets novel CpG-sites in enhancers.

Method: A smoking-methylation meta-analysis was carried out using EPIC DNA methylation profiles in 1407 blood samples from four UK population-based cohorts, including the MRC National Survey for Health and Development (NSHD) or 1946 British birth cohort, the National Child Development Study (NCDS) or 1958 birth cohort, the 1970 British Cohort Study (BCS70), and the TwinsUK cohort (TwinsUK). The overall discovery sample included 269 current, 497 former, and 643 never smokers. Replication was pursued in 3425 trans-ethnic samples, including 2325 American Indian individuals participating in the Strong Heart Study (SHS) in 1989-1991 and 1100 African-American participants in the Genetic Epidemiology Network of Arteriopathy Study (GENOA).

Results: Altogether 952 CpG-sites in 500 genes were differentially methylated between smokers and never smokers after Bonferroni correction. There were 526 novel smoking-associated CpG-sites only profiled by the EPIC array, of which 486 (92%) replicated in a meta-analysis of the American Indian and African-American samples. Novel CpG sites mapped both to genes containing previously identified smoking-methylation signals and to 80 novel genes not previously linked to smoking, with the strongest novel signal in SLAMF7. Comparison of former versus never smokers identified that 37 of these sites were persistently differentially methylated after cessation, where 16 represented novel signals only profiled by the EPIC array. We observed a depletion of smoking-associated signals in CpG islands and an enrichment in enhancer regions, consistent with previous results.

Conclusion: This study identified novel smoking-associated signals as possible biomarkers of exposure to smoking and may help improve our understanding of smoking-related disease risk.

Keywords: DNA methylation; Environment; Epigenetics; Lifestyle; SLAMF7; Smoking.

Fig. 1

Methylation association results in current versus never smokers. a Manhattan plot of genome-wide results for methylation association with smoking. Smoking-DMPs are indicated above the Bonferroni-adjusted threshold (red line). b Quantile–quantile (QQ) plot for CpG-site association in current versus never smoker

Fig. 2

Novel smoking-associated DNA methylation signals in SLAMF7. a coMET plot [33] describing the genomic region of epigenome-wide association between smoking and SLAMF7 methylation (top panel) showing the two smoking-DMPs cg00045592 and cg04009575, along with functional annotation of the region (middle panel) where broad ChromHMM regions are displayed using UCSC genome browser color schemes, and pattern of co-methylation at the 12 CpG sites in the EPIC array annotated to SLAMF7 (bottom panel). b Boxplot showing a comparison of DNA methylation levels at cg00045592 between smokers, former smokers and never smokers in the combined TwinsUK, NCDS, NSHD, BCS70 data set. c Boxplot showing a comparison of DNA methylation levels at cg04009575 between smokers, former smokers and never smokers in the Strong Heart Study

Fig. 3

Smoking-DMPs annotation and pathway analysis. a Log fold changes relating to the proportion of probes annotated to particular genomic locations and the proportion of smoking-DMPs. Only annotation categories with statistically significant results based on Fisher’s exact test are shown b. IPA canonical pathway analysis, 20 lowest P values shown