Ancestry-specific DNA methylation signatures of smoking and associations with atherosclerotic cardiovascular disease risk: findings from the Million Veteran Program

In the largest study of its kind ever conducted, researchers analyzing the genomes and health records of more than 42,000 US Veterans have mapped how cigarette smoking leaves chemical fingerprints on DNA, and how those fingerprints differ across ancestral backgrounds. The findings, drawn from the Department of Veterans Affairs’ Million Veteran Program and published in Genome Medicine, catalog tens of thousands of sites across the epigenome where smoking alters DNA methylation, a chemical modification that regulates gene activity without changing the underlying genetic code. The study also demonstrates that a compact risk score built from a shared set of 31 smoking-responsive methylation sites can modestly improve prediction of atherosclerotic cardiovascular disease beyond what standard smoking status alone provides.

DNA methylation has long served as one of molecular epidemiology’s most informative bridges between environmental exposures and disease. The addition of a methyl group to cytosine bases, typically at CpG dinucleotides, can silence genes, alter enhancer activity, and reshape chromatin architecture, and these marks are known to respond dynamically to external stimuli such as diet, stress, pollutants, and tobacco smoke. Tobacco smoking is widely regarded as one of the most powerful exogenous modifiers of the human epigenome, producing methylation changes that persist for years, and in some cases decades, after cessation. Because of this durability and sensitivity, methylation signatures of smoking have been proposed as objective biomarkers of exposure that could complement self-reported smoking history, which is known to be imperfect and prone to recall bias.

Until now, however, the field’s largest meta-analyses of smoking-related methylation have been constrained by limited ancestral diversity among participants and by sample sizes that, while impressive, could not support well-powered comparisons between populations. Most prior epigenome-wide association studies of smoking drew predominantly from cohorts of European ancestry, leaving open the question of whether the methylation marks discovered in those populations generalize to people of African, Admixed American, Asian, or other ancestries. Genetic ancestry influences patterns of genetic variation that in turn shape methylation, through mechanisms such as meQTLs, or methylation quantitative trait loci, where nearby genetic variants control methylation levels at specific sites. Differences in allele frequencies and linkage disequilibrium across populations therefore have the potential to alter both the detection and the magnitude of exposure-related methylation associations.

The new analysis, led by Dennis Khodasevich of VA Palo Alto Health Care System and Stanford University School of Medicine together with Andres Cardenas and colleagues, exploited the extraordinary scale and diversity of the Million Veteran Program to address these gaps directly. The team performed epigenome-wide association analyses, or EWAS, comparing methylation measured across the genome between current smokers, former smokers, and never smokers. In the overall study population, the researchers identified 40,529 methylation sites significantly associated with current smoking and 4,240 sites associated with former smoking after applying Bonferroni correction, the stringent multiple-testing threshold that guards against false discoveries when testing hundreds of thousands of positions simultaneously. These numbers represent by far the most comprehensive catalog of smoking-associated methylation ever assembled, dwarfing previous efforts and revealing an enormous breadth of low-magnitude signals that smaller studies could not reliably detect.

Crucially, the team then repeated the analyses within ancestry-specific strata, examining European, African, and Admixed American subgroups separately. The results revealed a striking degree of commonality: between 95 and 100 percent of the associations identified in each ancestry-specific analysis were also detected in the overall population, indicating that the fundamental epigenetic response to smoking is broadly shared across human populations. Yet the study also uncovered evidence of ancestry-specific associations and differences in the magnitude of methylation changes between groups, suggesting that while the core smoking response is conserved, genetic background and environmental context modulate its precise epigenomic expression. Across all analyses, a set of 31 CpG sites emerged as differentially methylated regardless of ancestry, forming a robust, universally responsive signature of smoking exposure.

The identification of these 31 common sites enabled one of the study’s most clinically oriented analyses: the construction of a methylation risk score for predicting atherosclerotic cardiovascular disease, or ASCVD. Smoking is a major driver of atherosclerosis, the pathological buildup of cholesterol-laden plaque in arterial walls that underlies coronary artery disease, peripheral arterial disease, and acute ischemic stroke. The researchers tested whether aggregating methylation values across the 31 shared CpG sites into a single quantitative score could improve prediction of time to ASCVD events beyond conventional three-category smoking status, which distinguishes current, former, and never smokers. The methylation risk score achieved a concordance statistic of 0.673, compared with 0.668 for smoking status alone, a modest but meaningful improvement that illustrates how epigenetic information can capture gradations of biological exposure and residual physiological effect that coarse behavioral categories miss.

The concordance difference, though small in absolute terms, carries conceptual significance for precision medicine. A person’s smoking history exists on a continuum of intensity, duration, and time since cessation, and the methylation mark integrates this exposure over biological time in ways that self-report cannot. By encoding the cumulative epigenetic consequence of smoking into a risk score, clinicians may eventually gain a more nuanced tool for cardiovascular risk stratification, particularly for patients whose reported smoking history is incomplete or inconsistent. The score’s performance also underscores that methylation is not merely a passive record of exposure; it may reflect downstream molecular pathways, such as inflammatory and endothelial processes, that mediate the link between smoking and arterial disease.

The study’s findings on former smoking are particularly noteworthy. The 4,240 significant sites associated with former smoking, together with what the authors describe as extensive low-magnitude associations, demonstrate that the epigenome retains detectable traces of past tobacco exposure long after smoking ceases. Some methylation changes at well-characterized loci, such as sites in the AHRR and F2RL3 genes that previous literature has repeatedly linked to smoking, are known to partially but incompletely revert after cessation. The persistence of these marks raises the possibility that they contribute to the elevated cardiovascular risk that former smokers continue to carry for years after quitting, although the current study establishes association rather than causation, and disentangling whether methylation changes drive disease or merely track it remains a central challenge for the field.

Methodologically, the work exemplifies the power of biobank-scale epigenomics. The Million Veteran Program, a nationwide cohort of US Veterans who consented to genomic research linked to comprehensive electronic health records, provided the sample size needed to detect effects of small magnitude and to stratify analyses by ancestry without sacrificing statistical power. The researchers acknowledged the Veterans whose participation made the study possible, and the analysis was conducted under VA research infrastructure with appropriate ethical oversight and informed consent from all participants. The consortium-based authorship, credited to the VA Million Veteran Program alongside individual investigators from Stanford, the University of Utah, the University of Pennsylvania, and VA medical centers, reflects the collaborative scale that modern genomic epidemiology increasingly demands.

The implications of the study extend across several domains. For basic epigenomics, the massive catalog of smoking-responsive sites provides a rich resource for investigating the biological pathways that tobacco combustion products perturb, including xenobiotic metabolism, oxidative stress, and immune regulation. For genetic epidemiology, the confirmation that ancestry-specific methylation signals exist alongside broadly shared ones reinforces the importance of diverse cohorts and ancestry-aware analytical designs in biomarker discovery, helping to prevent the inequities that arise when tools calibrated in one population are deployed uncritically in another. For clinical medicine, the methylation risk score, while not yet ready for routine deployment, offers a proof of concept that epigenetic biomarkers can add incremental predictive value to established risk factors for the leading cause of death worldwide.

Future work will need to validate the 31-site score in independent, ancestrally diverse cohorts, refine its calibration, and test whether it improves clinical decision-making in prospective settings. Researchers will also want to probe the ancestry-specific associations more deeply, examining whether genetic variation at methylation quantitative trait loci or differing exposure patterns explain the between-group differences in association magnitude. As epigenome-wide association studies continue to scale up, the smoking methylation signature documented in the Million Veteran Program stands as both a landmark catalog and a template for how large, diverse, deeply phenotyped cohorts can illuminate the molecular conversations between environment, genome, and disease.

Subject of Research: Ancestry-specific DNA methylation signatures of smoking and associations with atherosclerotic cardiovascular disease risk: findings from the Million Veteran Program

Article Title: Ancestry-specific DNA methylation signatures of smoking and associations with atherosclerotic cardiovascular disease risk: findings from the Million Veteran Program

Article References: Khodasevich, D., Hilliard, A. T., Barad, A., Zhou, J., Guarischi-Sousa, R., Clarke, S. L., Pridgen, K. M., Lynch, J. A., Chang, K.-M., Tsao, P. S., Assimes, T. L., Cardenas, A., & VA Million Veteran Program (2026). Ancestry-specific DNA methylation signatures of smoking and associations with atherosclerotic cardiovascular disease risk: findings from the Million Veteran Program. Genome Medicine. https://doi.org/10.1186/s13073-026-01761-4

Image Credits: AI Generated

DOI: 10.1186/s13073-026-01761-4

Keywords: Ancestry-specific, methylation, signatures, smoking, associations, atherosclerotic, cardiovascular, disease, risk, findings, Million, Veteran