Epigenetic predictor of smoking status in buccal cells

Smoking is the leading cause of accelerated aging and death worldwide. Therefore, identifying and intervening at smoke-responsive DNA methylation markers may be a primary way to reduce mortality risk in the population. Many studies have investigated the association between smoking and DNA methylation in blood samples. Only a few studies have examined saliva and buccal cells in this regard. Here, we determined DNA methylation profiles in buccal cells from a total of 280 individuals. Epigenome-wide association analysis (N = 200) uncovered 61 CpG markers, including novel ones, that were significantly associated with smoke exposure in this tissue type. Functional analysis showed that they were overrepresented in the Wnt signaling pathway and fatty acid metabolism. We confirmed that AHRR, a known smoking marker in blood, is also a top locus in buccal cells. However, cg06036945 (p = 1.76 × 10^-10) and cg04066994 (p = 1.36 × 10^-6) may be more informative for smoking in buccal epithelium than the commonly reported cg05575921. Moreover, unlike in blood, several other loci with a similar effect size were discovered in our study, including DKK3 cg16859537 (p = 3.10 × 10^-10) and CYP1B1 cg02162897 (p = 5.67 × 10^-10). The CYP1B1 and AHRR genes are known to interact through the Ah receptor pathway and play an important role in oxidative stress and mediating toxicity. Finally, logistic regression was employed for variable selection and to derive a novel DNA methylation-based classifier for smoking in buccal cells. Four CpG sites, CYP1B1 cg02162897, DKK3 cg16859537, AXIN1 cg12969952 and PKN1 cg12581991 predicted tobacco use with a cross-validated AUC = 0.8, sensitivity of 65.7 % and specificity of 81.5 %. The model was further validated in an independent set of N = 80 samples. The identified genes and pathways may serve as targets for intervention or risk stratification in respiratory diseases, while the developed models can also be instrumental in monitoring patient compliance with treatment recommendations and in behavioral profiling within forensic contexts.