1. Introduction

Cardiovascular diseases, including myocardial infarction (MI), remain a leading cause of morbidity and mortality worldwide. Beyond traditional risk factors, emerging evidence suggests that epigenetic modifications such as DNA methylation play a role in the development and aftermath of MI ^[1]^[2]. Epigenome-wide association studies have identified numerous CpG methylation differences associated with incident coronary heart disease (CHD) or MI risk, and significant shifts in DNA methylation profiles have been observed when comparing individuals before versus after an MI event^[1]. Moreover, specific post-MI methylation changes can persist over time: for example one study identified a reproducible “epigenetic fingerprint” of MI comprised of nine CpG sites that remained altered after an incident MI and correlated with metabolic changes ^[2]. These findings support the notion that MI leaves a lasting epigenetic imprint and that DNA methylation marks might serve as biomarkers of cardiovascular health.

Saliva has emerged as a convenient DNA source for epigenetic studies, with the added benefit of non-invasive collection. Notably, saliva DNA methylation profiles closely mirror those of blood leukocytes (correlation r ~0.97 after adjustment)^[3], because saliva contains a substantial proportion of leukocytes (e.g. neutrophils) alongside buccal cells. This makes saliva a viable surrogate for detecting systemic epigenetic signatures of disease. Prior studies have used saliva to reflect epigenetic susceptibility to complex diseases ^[4]. Given that epigenetic marks can encode the influence of environmental exposures (such as diet or smoking) on cardiovascular risk, saliva-based methylation profiling could potentially capture both genetic and lifestyle-related risk information.

This study presents a retrospective, cross-sectional analysis of DNA methylation in two cohorts of middle-aged adults: (1) an MI survivor cohort who had a saliva-based 850K DNA methylation test within 3 years of a heart attack, and (2) a control cohort of individuals with no history of cardiac disease, self-reported excellent health, and similar age. The goal of this study was to identify differentially methylated positions (DMPs) in saliva DNA that distinguish MI survivors from healthy controls. We hypothesise that the MI group would exhibit a distinct methylation profile, reflecting either pre-existing risk factor effects or lasting changes from the cardiac event. By focusing on stringently significant CpG sites (false discovery rate q < 0.05), we aim to highlight potential epigenetic biomarkers and gene targets associated with MI. This work is intended as a technical epigenomic analysis that could inform future research into methylation-based risk assessment for cardiovascular disease.

2. Materials and Methods

Study Design and Cohorts: This study is a retrospective case-control analysis using de-identified data from Muhdo Health’s commercial epigenetic testing database. Cohort 1 (MI focus group) consisted of N = 329 individuals (approximately 60% male) who had suffered a myocardial infarction and provided a saliva sample for DNA methylation profiling within 3 years of the event. Their mean age was 52 years (SD ~5.1). Cohort 2 (healthy control group) included N = 1,438 individuals (52% male, mean age 51 years, SD ~4.8) with no history of heart attack or any diagnosed cardiac condition; these participants self-reported excellent health status. The age and sex distributions were roughly comparable between cohorts to minimise confounding by age or sex. All individuals gave saliva samples collected using standard kits, and DNA was extracted via Eurofins Denmark. Notably, data on smoking status and other lifestyle factors were not analysed due to limitations on cohort numbers; thus, no adjustment for smoking (a known modifier of DNA methylation) or other behavioural risk factors could be made.

DNA Methylation Assessment: Genome-wide DNA methylation was measured using the Illumina Infinium MethylationEPIC BeadChip (850K) array, which assays over 850,000 CpG sites per sample. Standard quality control and normalization procedures were applied to the methylation β-values (proportion of DNA methylated at a given CpG site). Probes located on the sex chromosomes (X or Y) were removed from consideration to avoid sex-related methylation differences, since the cohort sex ratio was slightly imbalanced and we focused on autosomal epigenetic differences. After filtering, methylation β-values for autosomal CpGs were analyzed. Each β-value ranges from 0 (completely unmethylated) to 1 (fully methylated) and can be interpreted as the fraction of cells methylated at that locus in the sample.

Statistical Analysis: For each CpG locus, we compared the DNA methylation levels between the MI group and the control group. Given the large sample sizes and the possibility of non-normal distribution of β-values, we employed the non-parametric Mann–Whitney U test to assess differences in the distribution of methylation between groups at each CpG. This test does not assume equal variances or normality. We computed a p-value for each CpG’s group difference. To account for the 850,000 tests performed (one per CpG), we applied multiple-testing correction. Differentially methylated positions (DMPs) were defined as those with p < 2×10^–4; this threshold corresponded to FDR q < ~0.05 in our data (the highest q among reported hits was 0.036). In total, 40 CpG sites passed this significance cutoff. For each significant CpG, we recorded the group mean methylation in the MI and control cohorts, the difference of means, and various distribution percentiles. These top loci were annotated to genes using the Illumina 850K manifest (Supplementary info 1).

3. Results

Significant Differentially Methylated CpG Sites: Out of ~850,000 CpGs tested across the genome, 40 sites showed a significant difference in saliva DNA methylation between MI survivors and controls (Mann–Whitney p < 2×10^–4, q < 0.05). The direction of change was skewed toward higher methylation in the MI group: 34 of the 40 DMPs were hypermethylated in the MI cohort relative to controls, whereas 6 DMPs were hypomethylated in the MI cohort. However, the magnitude of methylation differences was generally modest. The average absolute difference in methylation (Δβ) between groups was on the order of 0.8–1.2% (absolute β difference ~0.008–0.012), and even the largest difference was under 4%. These subtle effect sizes are consistent with other epidemiological DNA methylation studies of complex traits, where small percentage changes can nonetheless be highly significant given large sample sizes.

The two top-ranked CpGs (lowest p-values) were cg27665659 and cg22851420, which had p ≈ 2.0×10^–6 (FDR q ~0.001). CpG cg27665659 is located in/near the AP1S1 gene (adaptor-related protein complex 1 sigma 1 subunit), and cg22851420 is annotated to HPCAL4 (hippocalcin-like protein 4). In the MI group, these loci showed slightly higher methylation on average. Several other CpGs in the top 10 were also modestly hypermethylated in the MI cohort, including cg10676084 in GABRB3 (GABA_A receptor beta-3 gene, p = 1.3×10^–5), cg05266781 in IRX5 (Iroquois homeobox 5, p = 1.4×10^–5), cg23995914 in ZNF518B (zinc finger 518B, p = 1.6×10^–5), cg02164046 in SST (somatostatin, p = 2.0×10^–5), and cg16419235 in PENK (proenkephalin, p = 2.0×10^–5). All of these showed about 0.5–1.5% higher methylation in MI cases compared to controls.

Notably, a few genes were represented by multiple significant CpG sites. Two independent probes in the ANKRD34B gene (ankyrin repeat domain 34B) were differentially methylated (p = 4.8×10^–5 and 9.0×10^–5), both showing higher methylation in the MI group. Similarly, two CpGs in ZYG11A (zyg-11 family member A) reached significance (p ~3.1×10^–5 and 1.0×10^–4). Such concordant hits within the same gene region suggest a consistent locus-specific methylation difference. Another hit was cg21620282 annotated to CHGA (chromogranin A), a gene encoding a stress hormone precursor; this probe had p = 3.1×10^–5 and higher methylation in MI (β ~0.80 vs 0.76).

Among the hypermethylated CpGs (MI > control), the largest mean difference was observed at cg10604476 (annotated to ICAM5, encoding a neural cell adhesion molecule also called telencephalin). This locus had a mean β of 0.413 in the MI group vs 0.374 in controls, a Δβ of +0.039 (approximately 3.9 percentage points higher in MI, p = 4.8×10^–5). the hypomethylated CpGs (MI < control) all showed the MI group having slightly lower methylation in comparison. For example, cg01459453 in the SELP gene (P-selectin) had mean β = 0.745 in MI vs 0.764 in controls (Δβ = –0.019, p = 1.0×10^–4), indicating lower P-selectin gene methylation in the MI cohort. Other hypomethylated sites included cg11649376 in ACSS3 (acetyl-CoA synthetase 3, Δβ ~–0.019), cg01256539 in PRR16 (Δβ ~–0.023), cg26581729 in NPDC1 (neural proliferation differentiation control 1, Δβ ~–0.016), cg23435594 (intergenic or unannotated, Δβ ~–0.021), and cg00277397 in CALN1 (calneuron-1, Δβ ~–0.016). Although these differences are small, they were statistically significant and may reflect down-regulation of methylation at specific loci in the MI group.

Summary of Key Differential Loci: In total, the 40 significant CpGs mapped to over two dozen genes. Many of these genes are not classically known cardiovascular players, suggesting novel or indirect connections to MI. Nonetheless, a few hits stand out for their plausible relevance:

• SELP: One CpG in the SELP gene was hypomethylated in MI survivors. P-selectin is an adhesion molecule expressed on activated endothelial cells and platelets, known to play a critical role in atherosclerosis and thrombosis ^[5]. Lower DNA methylation at a gene locus often correlates with higher gene expression; thus, the MI group’s lower SELP methylation might indicate a propensity for higher P-selectin expression or activity (potentially reflecting a pro-thrombotic or inflammatory state).
• IRX5: A significant hypermethylated CpG was located in IRX5, a cardiac developmental transcription factor. IRX5 is essential for establishing proper ventricular repolarization gradients and electrical conduction in the heart ^[6]. While the exact methylation site found (cg05266781) is not necessarily in a promoter region, the involvement of an IRX5-associated locus is intriguing and could relate to cardiac electrical remodelling or conduction system differences in those who experienced MI.
• AVPR1A (arginine vasopressin receptor 1A): One CpG in the AVPR1A gene (cg10906284) was hypermethylated in MI cases (p = 1.0×10^–4). The AVPR1A receptor mediates vasoconstrictive and pro-hypertensive effects of the hormone vasopressin; overactivation of this receptor pathway has been implicated in hypertension and heart failure pathophysiology. Differential methylation at AVPR1A might reflect alterations in neurohormonal regulation among MI patients (either as a cause or consequence of cardiovascular stress).
• KCNH2 (hERG): We detected a significant CpG (cg04839131) in KCNH2, the gene encoding the hERG potassium channel responsible for the cardiac I_Kr current. While KCNH2 mutations classically cause arrhythmias (Long QT syndrome), it is interesting to observe an epigenetic difference at this gene. It raises the question of whether MI survivors have epigenetically mediated changes in ion channel regulation or if this finding is incidental.
• ICAM5: The ICAM5 gene (Intercellular Adhesion Molecule 5) hit (cg10604476) was the largest effect CpG (as shown in Fig. 1). ICAM5 is actually a neuronal adhesion molecule not normally expressed in the periphery; its role in cardiovascular context is unclear. The methylation difference at ICAM5 might be a surrogate marker for other changes (for instance, reflecting a specific white blood cell subset or activation state) rather than a direct contributor to MI. This illustrates that not every DMP will have an obvious mechanistic link; some may serve as epiphenomenal markers of the post-MI physiological milieu or risk factor exposure.

Overall, the pattern of methylation differences in the MI group suggests a combination of inflammatory/immune activation signals, hormonal and neuronal regulation and cardiac-specific pathways (IRX5, KCNH2), as well as loci of unknown significance. These findings provide candidate genes and pathways for further investigation into how epigenetic modifications might link to the long-term consequences of MI or the underlying susceptibility to MI.

4. Discussion

In this retrospective analysis of saliva DNA methylation, we identified a distinct set of 40 CpG sites that significantly differed between myocardial infarction survivors and age-matched healthy controls. While the absolute differences in methylation at each locus were small (generally under 2–3% in β-value), they were highly statistically significant and overcame stringent multiple-testing correction. Small effect sizes are common in epigenome-wide association studies, yet they can still be biologically meaningful when aggregated or when occurring in key regulatory regions. Our results therefore suggest that MI survivors, even up to three years post-event, carry a subtle but detectable epigenetic “signature” in their saliva DNA.

Some of the differential methylation we observed may reflect long-lasting effects of the MI itself or its treatment. Prior work has shown that the experience of an acute MI can lead to persistent epigenetic changes in blood cells, potentially due to sustained inflammation, stress hormone exposure. For example, the hypomethylation of SELP (P-selectin) in the MI group could be a consequence of chronic inflammatory activation after MI, as circulating P-selectin is often elevated in patients with cardiovascular disease ^[7]. Demethylation at the SELP gene might permit higher expression; indeed, P-selectin protein is known to actively contribute to atherosclerotic lesion growth and thrombosis. This aligns with the idea that MI survivors remain in a pro-inflammatory, pro-coagulant state which could promote recurrence.

On the other hand, it is also possible that many of these methylation differences preceded the MI and served as risk factors or biomarkers of susceptibility. Epigenetic patterns can encode the impact of lifestyle and environmental exposures over decades. For instance, if an individual in the MI cohort had a history of heavy smoking (a major MI risk factor), they might exhibit DNA hypomethylation at certain loci known to be affected by tobacco smoke. Although we did not measure smoking status, it is noteworthy that none of the classic smoking-associated CpGs (e.g., in the AHRR gene) appeared in our top hits, suggesting that our DMP list is not obviously driven by smoking signatures. Nonetheless, we cannot exclude residual confounding: factors like smoking, diet, stress, or medication use differ between MI patients and healthy people and could partially drive the observed methylation differences. The lack of data on these variables is a limitation of our study. Future analyses would benefit from adjusting for smoking (which broadly reduces DNA methylation) and other lifestyle factors to isolate changes more directly related to the cardiac event or disease process.

It is interesting that several DMPs in our study relate to neuroendocrine and cardiac conduction genes. IRX5 and KCNH2, for example, are crucial for cardiac electrical function – IRX5 helps establish ventricular repolarization gradients and KCNH2 encodes a potassium channel in cardiomyocytes. While our data come from saliva (immune and epithelial cells), one could speculate that epigenetic changes in circulating cells might mirror or even influence systemic factors that impact the heart (e.g., IRX5 expression in cardiac fibroblasts or progenitors, or KCNH2 in circulating progenitor cells). Alternatively, these could be surrogate markers of genetic differences: e.g., certain genetic polymorphisms in IRX5 or KCNH2 might both predispose to heart disease and influence methylation at nearby CpGs (through allele-specific methylation). Without genotype data, this remains speculative. Nonetheless, the presence of such loci suggests an interplay between cardiovascular genetic pathways and the epigenome. It raises the question of whether an epigenetic risk score incorporating these CpGs could add predictive value for future cardiac events, as has been explored in other contexts.

Another noteworthy aspect is that our top list includes genes involved in hormonal signaling and metabolism, such as AVPR1A (vasopressin receptor 1A) and CHGA (Chromogranin A). MI survivors often have chronic neurohormonal activation – for example, elevated sympathetic activity and vasopressin levels are documented in heart failure post-MI^[8]. Hypermethylation of AVPR1A might indicate a compensatory downregulation in response to vasopressin overactivation, or it could be a pre-existing trait (perhaps related to stress responses) that made individuals more vulnerable to MI. Chromogranin A is a marker of neuroendocrine activity and is elevated in acute coronary syndromes; the higher methylation at CHGA CpG we observed might seem counter-intuitive (as it usually implies lower gene expression), but without knowing the exact regulatory context of that CpG (promoter vs gene body), interpretation is difficult. These examples highlight the complexity of linking CpG methylation changes to gene expression and phenotype – methylation can have activating or repressive effects depending on genomic context.

From a technical standpoint, this study demonstrates the feasibility of using saliva-based 850K methylation data in a large-scale case-control design to detect disease-associated epigenetic markers. The strong correlation between saliva and blood DNA methylation profiles implies that many signals we found would likely also be present in blood DNA. Saliva offers a practical alternative for large cohort collection or direct-to-consumer research, and our findings encourage further validation of these CpGs in blood or tissue samples. It is also worth noting that our analysis was retrospective and cross-sectional – thus, we cannot ascertain whether the methylation differences are a cause or consequence of MI. Longitudinal studies (following individuals’ methylation profiles before and after an MI, or tracking healthy people over time for incident MI) would be extremely valuable to determine temporal relationships. Some efforts in this direction have already shown that certain methylation changes precede clinical MI by years, whereas others occur as a result of the acute event. Our results contribute additional candidate loci to this growing body of knowledge.

5. Conclusions

In summary, we identified a set of 40 autosomal CpG sites in saliva DNA that differ significantly between myocardial infarction survivors and healthy controls. These epigenetic differences, detected on the EPIC 850K array, passed stringent significance thresholds and remained evident on average up to three years after the heart attack. The MI-associated methylation changes were modest in magnitude (generally <5% difference in methylation), and most indicated higher methylation in the MI group, with a subset indicating lower methylation. The differentially methylated loci include genes involved in inflammation/thrombosis (SELP/P-selectin), cardiac electrical signalling (IRX5, KCNH2), neurohormonal pathways (AVPR1A), and other cellular processes, suggesting that both cardiovascular-specific and systemic factors are reflected in the epigenome of MI patients. This retrospective study provides a technical demonstration that saliva-based DNA methylation profiles can capture biologically relevant signals related to MI.

Supplementary Info Table 1.