Using data from a group African Americans who are slightly older than our previous group of subjects, we confirm and extend our prior findings, showing that AHRR appears to be the locus whose methylation is significantly affected by nascent smoking, with degree of demethylation strongly associated with level of exposure. In addition, we show a strong correlation between demethylation at cg05575921 and serum cotinine levels. Significant limitations of the current study include the reliance on self-reported data for certain aspects of the study and the lack of self-reported data with respect to smoking at the time of the actual blood sampling.
The findings with respect to AHRR extend the prior findings in 19-year-old African American subjects and indicate that smoking induces a steady yet predictable series of changes in the methylation signature of lymphocytes. In our first group of 19-year-old men, only cg05575921 was significantly changed with an average change of 6%. In this group of slightly older subjects, with a presumably longer smoking history, the average demethylation at cg05575921 was 11%, with two other probes from AHRR achieving at least a trend for genome-wide significance. Taken together with other evidence, this suggests that continued smoking increases the degree of change at AHRR and other genes, even though degree of smoking, on average, remained quite low in this slightly older sample. Some other changes may be notable at genes suggested by others, including MYO1G (herein the fourth-ranked probe), F2RL3 and GFI1 [9, 10, 12]. Indeed, in our analyses of the effects of smoking on DNA methylation in 50-year-old African American smokers, the methylation signatures of a large number of genes are significantly remodeled (Dogan et al., unpublished data). Hence, it may be that as individuals continue to smoke, the degree of differential methylation at these other loci continues to develop to the point that it is detectable at genome-wide levels using similarly powered analyses. This also suggests the possibility of dose–response relationships at other CpG sites in addition to those on AHRR.
The semiquantitative nature of the relationship between serum cotinine levels and AHRR methylation status raises the possibility that DNA methylation could be used as a biomarker for smoking in place of exhaled carbon monoxide or serum cotinine levels when such measures are unavailable. Indeed, for large-scale epidemiological work, DNA demethylation at AHRR might prove useful as an index of smoking if there is stored blood or if other potential assessments are unavailable. For those existing data sets without separate serum samples or quantitative smoking data, this is certainly an attractive possibility. In addition, given the relatively short half-life of exhaled carbon dioxide (3 to 5 hours) [15] and serum cotinine levels (15 hours) [8, 16], the current data suggest that altered DNA methylation could be used to detect otherwise undetectable smoking by individuals such as ‘chippers’, who smoke only periodically [8, 16]. Further research to develop the response profile for AHRR and related loci could result in the development of a versatile assessment tool that could find considerable use in both research and clinical applications.
It is natural to ask why AHRR is the most significant locus. Although not immediately intuitive at first glance, changes in the epigenetic status of AHRR could be expected to be one of the first cellular responses to tobacco smoke exposure, owing to the interaction of AHRR with the aryl hydrocarbon receptor (AHR), which is the induction point for the xenobiotic pathway [17]. This catabolic pathway, which is active both in the liver and in lymphocytes, includes several well-known P450 enzymes, including CYP1A1, and is responsible for the degradation of environment toxins, such as polyaromatic hydrocarbons and dioxins commonly found in cigarettes [18, 19]. Activation of the pathway is initiated by the binding of ligands such as dioxin, which also serve as targets for degradation to the PAH domain of AHR. Following ligand binding, the AHR protein dimerizes with the aryl nuclear receptor translocator (ARNT), which facilitates its translocation to the nucleus and to binding to the promoters of key catabolic genes. AHRR serves as a negative feedback regulator of AHR induction and does so by competing with AHR for binding with ARNT and by sterically competing with AHR at critical gene promoters [20]. Critically, changes in AHRR methylation are known to alter AHRR gene expression [11]. Unfortunately, because AHRR has at least 21 known splice variants and 10 known protein isoforms, the relationship between these toxin exposures, AHRR methylation changes, and AHR pathway activity is likely to be complex. However, given the extant data, it is reasonable to hypothesize that the demethylation seen in smokers is associated with increased AHR activation of the xenobiotic pathway, with the current findings highlighting the need for further understanding of these processes.
A pertinent negative in the current study is the failure to observe significant changes in the DNA methylation signature at nicotinic cholinergic receptors (NChRs). However, it is important to note that in contrast to the situation with respect to AHRR, NChRs are not expressed heavily nor are they functionally coupled in lymphocytes. Furthermore, the genome-wide approaches used in this paper are relatively insensitive to smaller scale, yet more behaviorally relevant smoking associated changes in genes, such as monoamine oxidase A (MAOA), which is only lightly expressed in the lymphocytes [14]. Therefore, examinations of the role of smoking associated changes of NChR methylation in addictive processes should perhaps focus on those cell types in which the genes are heavily expressed and functionally coupled.
A potential problem for any epigenetic study is the presence of confounding genetic vulnerability. However, this is not likely to be a problem for our findings with respect to cg05575921, for several reasons. The nearest polymorphisms, rs6869832 and rs6894195, are relatively uninformative in the African American population (minor allele frequency 0.02); in a previous study of 399 subjects, we genotyped these loci and found no effect on cg05575921 methylation [13]. Still, genetic variation may have an effect on the methylation status at other interesting loci and we encourage the reader to inspect Additional file 1: Table S1 carefully for further details on polymorphisms flanking potentially interesting CpG residues.
An unanticipated finding was the degree of disparity between self-reported smoking status at wave 4 and the serum cotinine levels determined using samples collected 6 months after wave-4 self-reported data collection. Some discrepancy is, of course, understandable. Because the reliability of recall dims with increasing time, and because our yearly examinations only interrogated smoking behavior over the past month, some inaccuracy of self-reporting is to be expected. At the same time, such problems are common in both investigations of adolescent, nascent smoking [6, 7] and in studies of smoking in minority populations [21], highlighting the need for biochemical confirmation of smoking status in studies of tobacco use. In addition, some of the disparity between negative self-report and positive cotinine levels may reflect recent onset in smoking.
Our choice of a 2 ng/ml cutoff level was based on analyses of the shape of the cumulative distribution curve. This level is quite consistent with the optimum cutoff levels developed by Benowitz and colleagues using data from 16,156 subjects from the National Health and Nutrition Examination Study (NHANES) [22]. However, it is possible that a few of our lower ‘positive’ cotinine levels reflected secondhand smoke exposure in the home or from friends who smoked. However, in our opinion, secondhand smoke exposure is unlikely to explain more than one or two false-positives. The lowest cotinine level in the self-reported nonsmokers who had serum cotinine levels of >1.0 ng/dl was 9.3 ng/dl, which is considerably above that expected for secondhand smoke exposure [23]. Accordingly, the finding that one-third of the subjects with positive cotinine levels denied smoking at wave 4 suggests either a surge of smoking initiation at this age, or the possibility that both substantive intermittent, fast-moving changes in smoking behaviors and resulting unreliable self-reporting account for the discrepancies. Given the later onset of smoking in African Americans [24] and the higher rates of discrepant reports in underserved minorities [6, 21], these findings reemphasize the need for repeated measures with shorter lags between assessments and the need for use of biomarkers in both phenomenological and biological examinations of the effects of smoking. In this context, AHRR emerges as a potentially useful adjunct to self-reporting of smoking and may have particular utility in studies of the early phases of smoking.