Coronary artery disease (CAD) is the leading cause of death worldwide, counting for trillions of health care expenditure [1]. This situation will continue to deteriorate globally as risk factors continue increasing, such as hypertension, dyslipidemia, diabetes, and obesity. Despite advances in the understanding of causative portions of genetic variants, life styles, and environment factors, existing knowledge cannot fully explain the complex pathophysiology of gene-environment interactions underlying CAD. Epigenetic modification mechanisms of the genome might reveal the clue of this cooperation [2–4].

DNA methylation, methyl group added to the 5-carbon position of a cytosine (5-methylcytosine, 5-mC) within the whole genome that occur mostly at CpG sites [5], is one of the best understood epigenetic mechanisms thus far [6]. DNA methylation has been suggested to regulate gene expression and implicated in numerous biological and pathological processes including CAD [7–11]. Since a large portion of 5-mC is found in repeat sequences and transposable elements, such as long interspersed nuclear element (LINE-1) and ALU [12–14], methylation of these repetitive elements has thus been used as a surrogate for the global methylation of the genome [15]. It could be directly detected using bisulfite sequencing or other methods primarily based on the digestion of genomic DNA by restriction enzymes, like Luminometric Methylation Assay, LUMA, and the [³H]-methyl acceptance-based method [16, 17].

Though variations of global DNA methylation in peripheral blood leukocytes (PBLs) have been exploited in CAD, the observations remain contradictive. Baccarelli et al. found that blood LINE-1 hypomethylation was associated with baseline ischemic heart diseases and stroke in the Boston-area Normative Aging Study [18]. Haley L et al. reported LINE-1 hypomethylation was correlated to higher levels of low-density lipoprotein cholesterol (LDL-C) in a Samoan islander population [19]. In a prospective study conducted by Guarrera et al. [20], global DNA hypomethylation measured in LINE-1 repeats was associated with cardiovascular disease and myocardial infarction risk in men, being more pronounced in cases with shorter time to disease. In contrast, the association between an elevated global DNA methylation and CAD has also been observed in other studies [21, 22]. So far, these studies were based on different populations using different detection methods, part of which did not include PBL counts and classifications as probable confounders. These inconsistent findings indicated further study is necessary.

In recent years, 5-hydroxmethylcytosine (5-hmC) has been discovered as the oxidation product of 5-mC [23]. Though 5-hmC has the relatively low abundance in all cell types, it is a stable modification of genomic DNA [24]. Furthermore, it appears to be required for stem cells to maintain their pluripotency [25] and significantly reduced in many cancer types [26–28]. CAD-associated changes in global DNA hydroxymethylation are not yet known.

So, we adopted the latest liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS) method to compare global 5-mC (n = 535) and 5-hmC (n = 100) of PBLs and to evaluate intra-individual variations from different cell subtypes of PBLs between CAD patients and controls.

DNA methylation plays a vital role in development, aging, and pathological process of certain diseases [29]. Early studies reported the relationship between CAD and global DNA methylation measured in LINE-1 and ALU repeats, and the results included both positive and negative correlations [18–22]. Actually, when using these repeat sequences, there is uncertainty about its comparability and the extent to which it reflects global DNA methylation contents [30]. Moreover, methods using enzyme digestion and isotope labeling or chemiluminescent, in which the methylation status of cytosine is typically assessed by reactions with sodium bisulfite modification or methylation-sensitive restriction enzymes, a complete conversion/restriction-enzymatic-reaction is difficult to be achieved [22, 31, 32]. In addition, these assays could not distinguish 5-mC from 5-hmC, and most studies did not consider the confusion induced by PBL counts and classifications [33]. So, the controversial results with different markers and methods of global DNA methylation urge us to comprehensively investigate the association between the global 5-mC contents of different blood cell sources and CAD.

Case and control study would give us some clues for association analysis, on conditions that we could exclude necessary confounders for 5-mC content detection. Hence, we compared global 5-mC contents between CAD cases and controls. Importantly, global 5-mC contents were significantly decreased in cases. However, DNA methylation is not only a heritable epigenetic element but also a dynamic process related to demographical, clinical, and environmental factors. Therefore, we need to survey controls for defining the confounding factors in DNA methylation analysis. Interestingly, in controls, after forward stepwise multivariate linear regression, only age, TC, and PBL classifications were independent factors associated with 5-mC contents, which could partially explain 6.8% individual variation. Thus, age, TC, and PBL classification were identified as important cofounders for case-control study of 5-mC contents. Then, after these confounders were adjusted one by one in unconditional logistic regression, we could conclude 5-mC contents as an independent protective factor for CAD. Moreover, the reverse correlation between risk factors (age, TC, PBL classification) and 5-mC contents was coincident with the decrease of 5-mC contents as a protective factor in CAD cases.

Actually, the longitudinal decline in genome methylation over time has been reported [14], and DNA methylation biomarkers showed high correlation with chronological age and with age-acceleration effects associated with pathological conditions, morbidity, and mortality [34, 35]. Accumulated data supported the idea that DNA methylation showed reduced stringency in maintenance over the lifespan, resulting in an increase in inter-individual variability along with the overall decrease in DNA methylation [36]. Our previous study [37] also observed a slight negative correlation between 5-mC and chronological age.

Meanwhile, the heterogeneous composition of different blood cells within whole blood samples also has an influence on the determined global methylation [38–41]. But whether the variation DNA methylation was the result of the PBL classification, or the cause, or the company, need further investigation. On the other aspect, people more and more realize the important role of circulating immune cells in the inflammation process of atherosclerosis [42]. Not only the cell amount but also the cell classification that is more important decides the outcome of CAD [39]. Therefore, associations of global methylation patterns with certain health-associated conditions can partly be the results of a cell composition effect. Thus, the PBLs were divided into three categories: neutrophils, lymphocytes, and monocytes. Then, we analyzed the 5-mC contents separately to confirm that the changed DNA methylation between cases and controls was not only due to altered PBL classifications. Subgroup analysis results were in accordance to and confirmed the previous finding in groups, while the 5-mC contents of PBLs were significantly decreased in CAD and correlated with those of neutrophils, lymphocytes, and monocytes in controls. However, the correlation of global DNA methylation between PBLs and neutrophils vanished in CAD subgroup, even with a significant increase of neutrophil percentage. But significant correlations of global DNA methylation between PBLs and monocytes were present with the most dominant reduction in global DNA methylation of patients’ monocytes. Based on this result, combining the negative correlation between TC on 5-mC in 270 controls, we hypothesized that monocytes in CAD might be more susceptible to exposures of CAD risk factors (i.e., TC) in epigenetic pathway, compared to other blood cell subtypes. Neutrophils may respond to these exposures via alternative ways [43], such as increased amount.

Recently, it is reported that circulating immune cells can be primed by external stimuli, such as lipid particles, one of the most differences in blood between CAD cases and controls, to obtain a long-term epigenetic memory [44]. A study of the TG-lowering drug fenofibrate suggested that a 3-week daily treatment was not sufficient to reverse lipid-associated DNA methylation changes despite the short half-life of various blood cell types [45]. This suggests these changes did not arise in the circulation but occurred already in hematopoietic stem cells, where lipid priming has previously been started. We thought that monocytes might have been epigenetically altered in microenvironment with CAD risk factors, even before they migrated and settled under endothelia, and further facilitated the CAD process.

Migration of monocytes and formation of lipid-laden macrophage is the trigger of atherosclerosis plaque [38]. However, we were not able to gather the corresponding tissue samples in plaque lesions in CAD patients. Thus, in vitro cell model provides a proper alternative. As a classic model system for investigating the possible molecular mechanisms in foam cell development, THP-1 human monocyte cell line was used as the in vitro model. In agreement with the decreased global 5-mdC contents in CAD patients, lower genomic DNA methylation was also found in THP-1-derived foam cells compared to THP-1 monocytes, which suggests demethylation process might be involved into the foam stage of atheromatous plaque. The lower global 5-mC contents in monocytes of CAD patients, together with a decrease in the process of THP-1 foam cell formation, might imply a persistent decline in monocytes/macrophage during progression of atherosclerosis. Although our observational cross-section study could not infer causal directions, we could conclude monocytes with decreased 5-mC as a character of atherosclerosis pathology, and speculate monocytes were prone to external stimulus and differentiated hereafter to carry a reduced 5-mC. Further investigation is indispensable. In coincidence with previous reports, we also found the reduced global DNA methylation was accompanied with decreased expression of DNMT1 [46, 47].

Apart from the significant global hypomethylation found in monocytes of CAD patients and in vitro monocyte cell line model, DNA methylation alterations in specific genes in monocyte/macrophage in atherosclerosis have also been reported. Using dietary inducing hypercholesterolemia, mice can lead a hypomethylation state in spleen focus-forming virus proviral integration oncogene (PU.1) and interferon regulatory factor 8 (IRF8) and consequently resulted in a phenotypic alteration in monocytes and macrophages [48]. Monocyte chemotactic protein-1 (MCP-1), related with the migration and gather of monocytes, showed an overexpression due to the DNA hypomethylation induced by homocysteine, which is mediated by NF-κB/DNMT1 pathway [49]. A recently longitudinal study [50] showed ARID5B (transcription coactivator for histone H3K9me2 demethylation) CpG methylation is inversely associated with both ARID5B expression and atherosclerosis in human CD14+ blood monocyte, while the associations vanished in T cell samples from the same subset. These data suggest that whether in the causal or reflective pathway of disease, epigenome features in circulating monocytes are potential “biosensors” that might be useful for detecting early signs of metabolic disorders and elevated CAD risk.

In addition, we found 5-hmC levels are also highly heterogeneous in PBL subtypes (Table 6). 5-hmC is less abundant than 5-mC and is present at higher levels in neurons than in other cell types. Quantification of bulk levels of 5-hmC found that it is relatively rare, with levels varying by tissue from < 0.1 to 0.7% of cytosines globally [51]. The ratio of 5-hmC vs 5-mC is a cell-type specific, and the dynamics of generation and maintenance of 5-mC hydroxylation might also affect 5-mC levels. In PBLs and monocytes, elevated 5-hmC along with decreased 5-mC was found in CAD compared to controls (Fig. 2). It is reasonable because 5-hmC is the oxidized product of 5-mC, which is an intermediate product in DNA demethylation. Additionally, it is widely reported that the levels of 5-hmC inversely correlate with the rate of proliferation [52], and when the abundance and distribution of 5-hmC change, diseases occur. However, what kind of “epigenetic signal” 5-hmC is and how it relates to the 5-mC signal remain to be the main challenges that need to be addressed to date.

Nevertheless, the present study still has some limitations. First, there is significant evidence that many genetic, demographical, clinical, and environmental factors are strong cofounding variables. However, these underappreciated confounding variables all contribute to the overall measured phenotype. Within our present study, only age, sex, history of hypertension, hyperlipidemia, diabetes, blood lipids, fasting blood glucose, PBL counts, and classifications had been taken into account, and data such as smoking habits and BMI failed to be collected, which was one of our limitations. Second, even though we have isolated different subtypes of PBLs to detect 5-mC contents of each subtype and confirmed the association within each PBL subtype in vivo experiments, the sample size is still relatively small. Third, since it was a cross-sectional study, this investigation could not observe in prospective way and thus could not identify the causal effect, and longitudinal studies are more convinced to confirm the temporal relation of epigenetic changes in blood cells in the progress of CAD and to prove TC as causes. Thus, prospective studies are really required for further study.

In summary, the measurement of 5-mC contents by LC-ESI-MS/MS system in a relatively large sample for epigenetic study demonstrated that a significant reduction of 5-mC contents was the independent risk factor of CAD from the molecular epidemiology aspect. The decreased methylation level during the foam cell formation in THP-1 cells and in monocyte subtypes of CAD supported that DNA hypomethylation in PBLs could be a contributor to the pathogenesis of CAD. These suggested a possible epigenetic pathway of monocytes involved in CAD pathogenesis.