Differences in smoking associated DNA methylation patterns in South Asians and Europeans

Abstract

Background: DNA methylation is strongly associated with smoking status at multiple sites across the genome. Studies have largely been restricted to European origin individuals yet the greatest increase in smoking is occurring in low income countries, such as the Indian subcontinent. We determined whether there are differences between South Asians and Europeans in smoking related loci, and if a smoking score, combining all smoking related DNA methylation scores, could differentiate smokers from non-smokers.

Results: Illumina HM450k BeadChip arrays were performed on 192 samples from the Southall And Brent REvisited (SABRE) cohort. Differential methylation in smokers was identified in 29 individual CpG sites at 18 unique loci. Interaction between smoking status and ethnic group was identified at the AHRR locus. Ethnic differences in DNA methylation were identified in non-smokers at two further loci, 6p21.33 and GNG12. With the exception of GFI1 and MYO1G these differences were largely unaffected by adjustment for cell composition. A smoking score based on methylation profile was constructed. Current smokers were identified with 100% sensitivity and 97% specificity in Europeans and with 80% sensitivity and 95% specificity in South Asians.

Conclusions: Differences in ethnic groups were identified in both single CpG sites and combined smoking score. The smoking score is a valuable tool for identification of true current smoking behaviour. Explanations for ethnic differences in DNA methylation in association with smoking may provide valuable clues to disease pathways.

Figure 1

Manhattan plot showing association between current and never tobacco smoking and genome-wide DNA methylation. The continuous line marks the P ≤1.1 × 10^-7 significance threshold. CpG sites with corresponding P-values at ≤1.1 × 10^-7 are colour coded to show the direction of difference between smokers and non-smokers. Red CpG sites are hypermethylated in current smokers while blue CpG sites are hypomethylated.

Figure 2

Plot showing interaction between ethnic group and smoking at AHRR cg05575921. Bars show mean DNA methylation levels in each group shown. Error bars represent standard deviations. Ethnic differences were observed between current smokers (t-test: n = 36, P = 2.0 × 10^-3) but not between never smokers (t-test: n = 129, P = 0.44).

Figure 3

Plots of smoking score by reported smoking category in Europeans and South Asians. Box and whisker plots show median and interquartile ranges. Filled black circles show individual data points. Red line indicates threshold score above which individuals were considered to be current smokers.

Figure 4

Plots of associations between smoking score and smoking behaviour in Europeans and South Asians. Filled circles show individual data points coloured by ethnic group (black = European, red = South Asian). R-squared values given are for regression models predicting smoking behaviour using smoking score in each ethnic group.