Fernández‑Sanlés et al. Clin Epigenet (2021) 13:86 https://doi.org/10.1186/s13148-021-01078-6 RESEARCH Open Access DNA methylation biomarkers of myocardial infarction and cardiovascular disease Alba Fernández‑Sanlés1,2,3, Sergi Sayols‑Baixeras1,4,5, Isaac Subirana1,6, Mariano Sentí2, S. Pérez‑Fernández1,4, Manuel de Castro Moura7, Manel Esteller7,8,9,10, Jaume Marrugat1,4 and Roberto Elosua1,4,11* Abstract Background: The epigenetic landscape underlying cardiovascular disease (CVD) is not completely understood and the clinical value of the identified biomarkers is still limited. We aimed to identify differentially methylated loci associ‑ ated with acute myocardial infarction (AMI) and assess their validity as predictive and causal biomarkers. Results: We designed a case–control, two-stage, epigenome-wide association study on AMI ­(ndiscovery = 391, ­nvalidation = 204). DNA methylation was assessed using the Infinium MethylationEPIC BeadChip. We performed a fixed- effects meta-analysis of the two samples. 34 CpGs were associated with AMI. Only 12 of them were available in two independent cohort studies (n ~ 1800 and n ~ 2500) with incident coronary and cardiovascular disease (CHD and CVD, respectively). The Infinium HumanMethylation450 BeadChip was used in those two studies. Four of the 12 CpGs were validated in association with incident CHD: AHRR-mapping cg05575921, PTCD2-mapping cg25769469, intergenic cg21566642 and MPO-mapping cg04988978. We then assessed whether methylation risk scores based on those CpGs improved the predictive capacity of the Framingham risk function, but they did not. Finally, we aimed to study the causality of those associations using a Mendelian randomization approach but only one of the CpGs had a genetic influence and therefore the results were not conclusive. Conclusions: We have identified 34 CpGs related to AMI. These loci highlight the relevance of smoking, lipid metab‑ olism, and inflammation in the biological mechanisms related to AMI. Four were additionally associated with incident CHD and CVD but did not provide additional predictive information. Keywords: DNA methylation, Epigenome-wide association study, Predictive biomarkers, Myocardial infarction, Cardiovascular disease Introduction proportion of CHD events occurs in individuals classified Cardiovascular disease (CVD) and more specifically as having moderate or low risk [4]. Additionally, the use coronary heart disease (CHD) remains the number one of currently available drugs to control classical cardio- cause of death and disease burden worldwide [1, 2]. At vascular risk factors (CVRFs) does not prevent all CHD the individual level, prevention is based on the estima- events, underlining the need to identify new strategies tion of cardiovascular risk [3]. However, the sensitivity for reducing this residual cardiovascular risk [5]. Thus, of cardiovascular risk estimation is low and a significant information encoded in biological mechanisms should be unravelled to find new predictive biomarkers and poten- tial therapeutic targets. Among these biomarkers, DNA *Correspondence: relosua@imim.es methylation marks arise as emerging candidates. 1 Cardiovascular Epidemiology and Genetics Research Group, REGICOR DNA methylation is an epigenetic mechanism consist- Study Group, IMIM (Hospital del Mar Medical Research Institute), Dr Aiguader 88, 08003 Barcelona, Catalonia, Spain ing on chemical modifications of cytosines, mostly fol- Full list of author information is available at the end of the article lowed by guanines (CpGs) [6]. Epigenome-wide association © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. The Creative Commons Public Domain Dedication waiver (http://​creat​iveco​ mmons.​org/​publi​cdoma​in/​zero/1.​0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Fernández‑Sanlés et al. Clin Epigenet (2021) 13:86 Page 2 of 11 studies (EWASs) make it possible to find DNA methylation Table 1 Descriptive characteristics of the populations used biomarkers of different traits and outcomes. In fact, DNA in the two-stage EWAS on acute myocardial infarction (AMI): methylation pattern is associated with multiple chronic dis- REGICOR-1 and REGICOR-2 eases [7], including CVD and CHD [8–13]. However, the Variables Discovery: Validation: clinical value of the identified biomarkers is still limited, REGICOR-1 REGICOR-2 and the epigenetic landscape underlying CVD is not com- N = 391 NA‡ N = 204 NA‡ pletely understood. The most common technology to assess DNA methylation AMI cases/controls, n (%)‡ is based on commercial arrays, which do not cover the whole AMI cases 196 (50.1) 0 101 (49.5) 0 methylome. Moreover, most current knowledge on the rela- Controls 195 (49.9) 0 103 (50.5) 0 tion between DNA methylation and cardiovascular risk Age 63.2 (6.94) 0 61.7 (6.90) 0 comes from studies based on the Infinium HumanMeth- AMI cases 63.0 (6.96) 0 61.6 (6.83) 0 ylation450 BeadChip (Illumina, CA, USA; from now on, Controls 63.3 (6.94) 0 61.7 (7.01) 0 450 k) [14] – which has been replaced by the Infinium Meth- Sex, female, n (%) 201 (51.4) 0 100 (49.0) 0 ylationEPIC BeadChip (Illumina, CA, USA; from now on, AMI cases 100 (51.0) 0 49 (48.5) 0 EPIC). Compared to the 450 k, EPIC interrogates 413,745 Controls 101 (51.8) 0 51 (49.5) 0 more methylation sites (but excludes 42,859) increasing the Smokers, n (%) 93 (24.4) 10 67 (33.3) 3 genomic coverage. Moreover, EPIC is enriched with func- AMI cases 76 (40.6) 9 45 (45.0) 1 tional sites analyses such as enhancers, DNase hypersensitive Controls 17 (8.76) 1 22 (21.8) 2 sites, and miRNA promoter regions [15]. Thus, the new chip BMI, kg/m2*‡ 28.5 (4.75) 52 27.5 (4.87) 23 has the potential to identify novel DNA methylation-based AMI cases 28.1 (4.34) 52 28.2 (5.87) 23 biomarkers of cardiovascular events. Controls 28.8 (5.02) 0 26.9 (3.90) 0 We hypothesized that DNA methylation is associated Hypercholesterolaemia, n (%)‡ 179 (53.0) 53 93 (51.4) 23 with MI risk, and that some of these epigenetic marks AMI cases 93 (64.1) 51 51 (65.4) 23 could be predictive of future risk, and have causal effects Controls 86 (44.6) 2 42 (40.8) 0 on cardiovascular outcomes. Thus, this study had three HTN, n (%)‡ 212 (57.1) 20 105 (54.1) 10 aims: 1) to unravel genomic methylation loci associated AMI cases 120 (68.2) 20 62 (68.1) 10 with myocardial infarction (MI), 2) to assess their predic- Controls 92 (47.2) 0 43 (41.7) 0 tive capacity of cardiovascular risk, and 3) to decipher the Diabetes, n (%) 87 (24.7) 39 44 (23.8) 19 causality of those associations. AMI cases 56 (35.7) 39 27 (32.9) 19 Controls 31 (15.9) 0 17 (16.5) 0 Results ‡ NA, missing information; AMI, Acute myocardial infarction; BMI, Body mass Quality control of DNA methylation data, cardiovascular index; Hypercholesterolaemia, defined as self-reported high cholesterol levels or treatment; HTN, Hypertension, defined as self-reported high blood pressure or outcomes and covariates treatment; Diabetes, defined as self-reported diabetes or treatment We finally included 391 individuals (196 cases and 195 † Median (Interquartile Range) controls) in the REGICOR-1 sample (Girona Heart Regis- *Mean (Standard deviation) try; REgistre GIroní del COR), 204 individuals (101 cases and 103 controls) in the REGICOR-2 sample, 1,863 women in the WHI Women’s Health Initiative) sample, and 2,540 validation (p-value < ­10–5), and their Manhattan and Q-Q participants in the FOS (Framingham Offspring Study) plots are shown in the Additional file 2: Table S1, and sample. The main sociodemographic and clinical charac- Additional file 1: Figs. S1 and S2]. In total, we identified teristics of the three populations are shown in Tables 1 and 68 CpGs suggestively related to MI (Additional file 1: 2. Regarding the number of CpGs, we analysed 811,610 Fig. S3). Model 1 provided 56 CpGs, of which three were CpGs in the REGICOR-1 sample, 820,183 CpGs in the also found in both model 2 and 3, and 13 in model 2. One REGICOR-2 sample, 478,369 CpGs in the WHI sample, additional CpG was found in both model 2 and 3, two in and 483,656 CpGs in the FOS sample. Figure 1 illustrates model 2 and nine in model 3. the steps included in this study. Validation and meta‑analysis The association studies Association between DNA methylation and cardiovascular performed in the validation stage included the 68 CpGs outcomes suggestively related to MI. We meta-analysed the results Two‑stage EWAS on acute myocardial infarction of those 68 associations from both stages. We identified Discovery stage The associations from the discovery 34 differentially methylated CpGs related to MI, with sim- stage (REGICOR-1) that were taken to the subsequent ilar effect sizes in all three models for most of the CpGs Fernández‑Sanlés et al. Clin Epigenet (2021) 13:86 Page 3 of 11 Table 2 Descriptive characteristics of the populations used in the follow-up association studies on incident cardiovascular (CVD) and coronary heart disease (CHD) events: Women’s Health Initiative (WHI) and Framingham Offspring Study (FOS) Variables WHI FOS CHD FOS CVD N = 1863 NA N = 2266 NA N = 2123 NA Age 64.2 (7.03) 0 65.7 (8.82) 0 65.3 (8.66) 0 Sex, female, n (%) 1863 (100) 0 1284 (56.7) 0 1206 (56.8) 0 CHD, n (%)‡ 914 (49.1) 0 106 (4.68) 0 _ _ CVD, n (%)‡ 984 (52.8) 0 _ _ 222 (10.5) 0 Smokers, n (%) 174 (9.46) 24 217 (9.62) 11 202 (9.56) 11 BMI, kg/m2*‡ 29.8 (5.97) 11 28.1 (5.39) 8 28.0 (5.33) 6 Total cholesterol, mg/dl* 233 (42.3) 1 189 (36.2) 1 190 (35.8) 1 LDL-C, mg/dl*‡ 152 (38.7) 31 107 (30.8) 2 108 (30.5) 2 HDL-C, mg/dl*‡ 52.2 (13.2) 1 58.3 (18.3) 2 58.7 (18.3) 2 TG, mg/d† 126 [93.0;176] 1 101 [73.0;140] 1 100 [73.0;139] 1 SBP, mmHg*‡ 131 (17.5) 0 126 (16.9) 1 125 (16.7) 1 DBP, mmHg*‡ 76.4 (9.27) 0 72.1 (9.99) 3 72.3 (9.92) 3 HTN treatment, n (%)‡ 672 (45.9) 400 1025 (45.4) 7 927 (43.8) 6 Glucose, mg/dl† 107 (38.3) 1 106 (21.8) 2 105 (21.5) 2 Diabetes, n (%) 319 (17.1) 0 307 (14.3) 126 266 (13.3) 119 ‡ CHD, Coronary heart disease; CVD, Cardiovascular disease; BMI, Body mass index; LDL-C, Cholesterol in low-density lipoprotein; HDL-C, Cholesterol in high- density lipoprotein; TG, Triglycerides; SBP, Systolic blood pressure; DBP, Diastolic blood pressure; HTN: hypertension; Diabetes, defined as previous treatment or glycaemia ≥ 126 mg/dl † Median (Interquartile range) *Mean (Standard deviation) Fig. 1 Flow chart of the steps included in this study Fernández‑Sanlés et al. Clin Epigenet (2021) 13:86 Page 4 of 11 (except cg21566642, cg05575921, cg03636183). The 34 Discussion CpGs were located in 25 different loci (26 genes, with one We have identified 34 methylation loci associated with CpG mapping to two genes) and nine intergenic regions acute MI in a two-stage EWAS, analysing ~ 850,000 CpGs. (Table 3, and Additional file 2: Table S2). All but two of these MI-associated sites (cg05575921 located in AHRR and the intergenic cg21566642) are Follow‑up association studies on incident CHD and CVD newly reported. Of those, 12 CpGs could be studied in events association with incident cases of CHD and CVD, and Out of the 34 identified CpGs associated with MI, only 12 we identified four of them associated with incident CHD were available in the samples with incident cases (whose (three of them also with incident CVD). All four were DNA methylation was profiled with the array 450 k). In also related to traditional CVRFs, supporting their role in total, we validated four CpGs after the meta-analysis of the development of these diseases. However, their clinical the separate association studies in the WHI and the FOS utility as predictive biomarkers or drug targets was not samples (p-value < 0.05/12 = 4.17 × ­10–3): AHRR-mapping proven. cg05575921, PTCD2-mapping cg25769469, intergenic Recently, two EWASs on incident CHD were published cg21566642 and MPO-mapping cg04988978. The four providing different findings from ours. Ward-Caviness CpGs were associated with CHD but cg25769469 was not et al.found nine CpGs associated with incident acute MI related to CVD (Table 4, Additional file 2: Table S3). [9]. Agha et al. reported 52 CpGs related to incident CHD [8]. None of them was replicated in our study. This lack of Association between the identified CpGs and CVRFs concordance could be related to methodological differ- Table 5 shows the associations observed between the ences (incident vs prevalent cases; myocardial infarction identified CpGs and classical CVRFs. The four validated vs CHD; considered confounder variables; characteristics CpGs were related to some CVRF [p-value < 0.05/(4 of the populations), and highlights the complexity of the CpGs*8 CVRF) = 1.56 × ­10–3]. study of these diseases. Association between MRSs and incidence of CHD and CVD CpG sites associated with acute MI events The associations between the methylation risk scores The 34 identified CpGs showed similar effect sizes in the (MRSs) and the incidence of coronary (n = 94) and car- two REGICOR samples and we considered them poten- diovascular (n = 222) events in the FOS population are tially relevant. Similarly, all but three CpGs (AHRR- shown in Additional file 2: Table S4. The median of the mapping cg05575921, F2RL3-mapping cg03636183, and follow-up periods for CVD and CHD incidence were 7.67 the intergenic cg21566642) showed consistent effect and 7.87 years, respectively. The MRSs were not associ- sizes in the three models. The effect size of those three ated with higher cardiovascular risk independently of the was reduced by half when adjusted for smoking, which classical CVRFs. Consistently, the addition of any of the highlights the important role of this risk factor in the MI MRSs to the Framingham risk function did not improve context. In fact, all three sites are widely described to be its predictive capacity in the FOS cohort (Additional related to smoking [17–19]. file 2: Table S4). Differentially methylated genes were enriched in diverse molecular and physiological pathways, including Causality of the associations between DNA methylation lipid metabolism and metabolic and inflammatory dis- and cardiovascular outcomes eases, underlining their relevance on the pathogenesis Of the four identified CpGs, only cg21566642 showed a of CHD. Interestingly, the SERPINA1 locus also anchors genetic influence; its methylation levels in adolescence genetic variants related to CHD [20], and other identified were associated with rs72617176 and in childhood with loci present with genetic variants associated with body rs139595493. We did not have individual data to test the mass index (DNMT3A, ABTB2, ZBTB16, NISCH, AHRR, first and second Mendelian randomization assumptions, DLEU1), inflammatory biomarkers or blood cell counts but the meQTLs were associated with the CpGs methyla- (AIM2, ITPKB, DNMT3A, LZTFL1, PSMB7, ZBTB16, tion levels at genome-wide significance independently of ACTN1, SERPINA1, MPO, DNAJC5B, CPM, DLEU1, age, sex or ancestry principal components [16]. Only the ZFPM1), blood pressure (PTCD2, PSMB7, SERPINA1, Wald ratio method could be conducted, since it uses a AHRR) and lipids (SERPINA1, NISCH, DLEU1, ZFPM1) single instrumental variable. The results did not support [21]. a causal effect of methylation at cg21566642 on either MI Nonetheless, the case–control design of our initial or CHD (Additional file 2: Table S5). We could not per- discovery sample limits the inference of the biologi- form sensitivity tests for pleiotropic effects or its strength. cal sequence of the epigenetic marks, the related bio- The other three CpGs could not be instrumented. logical mechanisms, and the clinical event. One possible Table 3 CpGs differentially methylated in association with prevalent myocardial infarction in the fixed-effects meta-analyses of the REGICOR case–control samples Genomic information Model 1 Model 2 Model 3 CpG Chr* Position (bp) Gene Coef* SE* P Coef* SE* P Coef* SE* P –1 –9 –1 –8 –1 cg15724772 1 6,373,534 ACOT7 1.17 1.94 × ­10 1.76 × ­10 1.16 2.17 × ­10 8.14 × ­10 1.00 2.30 × ­10 1.38 × ­10–5 –1 –8 –1 –7 –1 cg11919692 1 26,343,868 Intergenic 1.08 1.98 × ­10 4.94 × ­10 1.11 2.19 × ­10 4.00 × ­10 1.04 2.34 × ­10 9.29 × ­10–6 –1 –11 –1 –10 –1 cg10636246 1 159,046,973 AIM2 − 0.96 1.41 × ­10 1.03 × ­10 − 0.98 1.58 × ­10 5.21 × ­10 − 0.86 1.68 × ­10 3.68 × ­10–7 –1 –9 –1 –8 –1 cg08345526 1 226,916,320 ITPKB 1.79 2.93 × ­10 1.02 × ­10 1.85 3.29 × ­10 1.71 × ­10 1.60 3.45 × ­10 3.42 × ­10–6 –1 –9 –1 –8 –1 cg01552966 2 25,528,330 DNMT3A 1.15 1.93 × ­10 2.43 × ­10 1.25 2.22 × ­10 1.78 × ­10 1.04 2.33 × ­10 7.66 × ­10–6 –1 –9 –1 –9 –1 cg25420477 2 70,319,121 Intergenic 1.54 1.63 1.47 Fernández‑Sanlés et al. Clin Epigenet 2.52 × ­10 1.08 × ­10 2.84 × ­10 9.01 × ­10 3.01 × ­10 9.45 × ­10–7 –1 –10 –1 –9 –1 cg06528771 2 102,513,985 Intergenic − 1.71 2.79 × ­10 9.88 × ­10 − 1.89 3.16 × ­10 2.26 × ­10 − 1.81 3.48 × ­10 1.81 × ­10–7 –1 –8 –1 –7 –1 cg12206840 2 136,826,451 Intergenic 1.12 1.98 × ­10 1.31 × ­10 1.12 2.23 × ­10 4.89 × ­10 1.03 2.33 × ­10 9.08 × ­10–6 –1 –9 –1 –2 –1 cg21566642 2 233,284,661 Intergenic − 0.65 1.13 × ­10 7.70 × ­10 − 0.30 1.56 × ­10 5.26 × ­10 − 0.41 1.70 × ­10 1.50 × ­10–2 cg10650214 3 45,914,188 LZTFL1 1.28 1.97 × ­10–1 8.01 × ­10–11 1.26 2.19 × ­10–1 8.05 × ­10–9 1.18 2.31 × ­10–1 2.99 × ­10–7 (2021) 13:86 cg07386061 3 52,492,874 NISCH 0.99 1.79 × ­10–1 2.86 × ­10–8 1.03 2.01 × ­10–1 3.27 × ­10–7 0.86 2.09 × ­10–1 4.01 × ­10–5 cg13373048 3 71,892,575 Intergenic − 2.76 3.57 × ­10–1 1.02 × ­10–14 − 3.24 4.23 × ­10–1 1.93 × ­10–14 − 3.03 4.29 × ­10–1 1.77 × ­10–12 cg03733470 3 196,613,114 SENP5 − 1.20 2.07 × ­10–1 5.94 × ­10–9 − 1.28 2.30 × ­10–1 2.70 × ­10–8 − 1.17 2.46 × ­10–1 2.01 × ­10–6 cg05575921 5 373,378 AHRR − 0.75 1.25 × ­10–1 2.05 × ­10–9 − 0.40 1.91 × 10–1 3.48 × ­10–2 − 0.45 2.04 × ­10–1 2.87 × ­10–2 –1 –9 –1 –8 –1 cg25769469 5 71,643,841 PTCD2 − 1.33 2.20 × ­10 1.33 × ­10 − 1.38 2.53 × ­10 5.40 × ­10 − 1.41 2.80 × ­10 4.92 × ­10–7 –1 –8 –1 –6 1 –1 cg04725636 8 67,004,538 DNAJC5B − 0.80 1.45 × ­10 3.33 × ­10 − 0.72 1.59 × ­10 5.47 × ­10 − 0.75 1.74 × ­10 1.76 × ­10–5 –1 –8 –1 –10 –1 cg12066594 9 127,162,066 PSMB7 − 1.89 3.37 × ­10 2.00 × ­10 − 2.57 4.07 × ­10 2.67 × ­10 − 2.36 4.25 × ­10 2.83 × ­10–8 –1 –8 –1 –7 –1 cg24501073 11 34,210,305 ABTB2 0.99 1.73 × ­10 1.07 × ­10 0.98 1.96 × ­10 6.11 × ­10 0.82 1.98 × ­10 3.07 × ­10–5 –1 –10 –1 –7 –1 cg02042026 11 113,936,782 ZBTB16 1.04 1.70 × ­10 8.50 × ­10 0.99 1.88 × ­10 1.51 × ­10 0.98 2.04 × ­10 1.50 × ­10–6 –1 –8 –1 –6 –1 cg01381203 12 2,985,630 FOXM1;RHNO1 − 0.88 1.62 × ­10 5.17 × ­10 − 0.84 1.76 × ­10 1.88 × ­10 − 0.92 1.95 × ­10 2.44 × ­10–6 –1 –9 –1 –8 –1 cg02175213 12 69,352,407 CPM − 1.08 1.85 × ­10 5.15 × ­10 − 1.10 2.06 × ­10 9.76 × ­10 − 0.82 2.14 × ­10 1.27 × ­10–4 –1 –8 –1 –6 –1 cg06084585 13 50,797,125 DLEU1 − 1.60 2.87 × ­10 2.37 × ­10 − 1.47 3.14 × ­10 2.96 × ­10 − 1.36 3.30 × ­10 3.54 × ­10–5 –1 –9 –1 –9 –1 cg15867698 14 69,438,267 ACTN1 1.13 1.86 × ­10 1.58 × ­10 1.26 2.14 × ­10 3.33 × ­10 1.14 2.25 × ­10 4.04 × ­10–7 –1 –10 –1 –8 –1 cg07252680 14 94,857,224 SERPINA1 − 1.56 2.41 × ­10 1.17 × ­10 − 1.43 2.65 × ­10 6.57 × ­10 − 1.41 2.86 × ­10 7.93 × ­10–7 cg16125725 15 70,101,302 Intergenic − 2.03 3.45 × ­10–1 4.02 × ­10–9 − 2.30 3.97 × ­10–1 6.82 × ­10–9 − 2.20 4.18 × ­10–1 1.39 × ­10–7 cg09967176 15 77,280,925 Intergenic − 1.28 1.78 × ­10–1 6.93 × ­10–13 − 1.40 2.03 × ­10–1 5.21 × ­10–12 − 1.16 2.10 × ­10–1 3.82 × ­10–8 cg17261469 16 87,420,879 FBXO31 0.99 1.73 × ­10–1 9.49 × ­10–9 1.00 1.94 × ­10–1 3.01 × ­10–7 1.00 2.04 × ­10–1 9.92 × ­10–7 cg04983687 16 88,558,223 ZFPM1 0.72 1.29 × ­10–1 2.61 × ­10–8 0.68 1.42 × ­10–1 1.79 × ­10–6 0.44 1.46 × ­10–1 2.49 × ­10–3 –1 –11 –1 –9 –1 cg22508829 17 17,843,585 TOM1L2 1.28 1.96 × ­10 7.97 × ­10 1.35 2.26 × ­10 2.41 × ­10 1.10 2.28 × ­10 1.23 × ­10–6 –1 –10 –1 –8 –1 cg22146772 17 26,793,673 Intergenic 1.23 1.97 × ­10 5.38 × ­10 1.18 2.22 × ­10 8.98 × ­10 1.06 2.28 × ­10 3.31 × ­10–6 –1 –9 –1 –8 –1 cg13064873 17 48,360,726 TMEM92− AS1 − 1.08 1.85 × ­10 5.31 × ­10 − 1.19 2.12 × ­10 1.82 × ­10 − 1.04 2.23 × ­10 3.19 × ­10–6 –1 –10 –1 –10 –1 cg04988978 17 56,359,578 MPO − 1.31 2.07 × ­10 2.33 × ­10 − 1.45 2.37 × ­10 9.71 × ­10 − 1.22 2.43 × ­10 5.41 × ­10–7 –1 –8 –1 –2 –1 cg03636183 19 17,000,585 F2RL3 − 0.74 1.36 × ­10 4.56 × ­10 − 0.35 1.80 × ­10 5.24 × ­10 − 0.39 1.90 × ­10 4.12 × ­10–2 –1 –10 –1 –10 –1 cg19448292 20 35,504,064 C20orf118 1.24 1.93 × ­10 1.17 × ­10 1.38 2.25 × ­10 7.89 × ­10 1.15 2.30 × ­10 6.05 × ­10–7 *Chr, Chromosome; Coef; Coefficient of association; SE, Standard error Model 1 was adjusted for estimated cell counts and two surrogate variables. Model 2 was further adjusted for smoking status. Model 3 was additionally adjusted for diabetes, hypercholesterolemia, and hypertension. The Page 5 of 11 Bonferroni-corrected p threshold was 6.17 × ­10–8 Fernández‑Sanlés et al. Clin Epigenet (2021) 13:86 Page 6 of 11 scenario could be that the identified DNA methylation events in the FOS, nor we observed an improvement in marks occurred before the acute event, as potential bio- the predictive capacity of the Framingham risk func- logical mechanisms involved in MI pathogenesis. This tion when including this score. This highlights the chal- may be the case of the three CpGs that were related to lenge of novel biomarkers to improve cardiovascular risk smoking. Conversely, as blood samples of MI cases were prediction. collected within the initial 24 h after hospitalization, the other possibility could be that methylation at the iden- Causality of the associations between methylation loci tified CpGs had changed as a consequence of the acute and cardiovascular outcomes event or the therapeutic procedures. If the first scenario The four CpGs associated not only with acute MI, but can be proven in further studies, these DNA methylation also incident CHD, may suggest that DNA methylation marks could be potential predictive biomarkers of MI or changes at those loci occur prior to the event. However, new therapeutic targets. If they are found to be post-MI this association does not guarantee whether differen- marks, further studies could evaluate their potential as tial DNA methylation at those loci has a causal effect on biomarkers of prognosis. CHD. Mendelian randomization can be used to ascer- tain this causal relationship. However, this approach could only be undertaken for cg21566642. Although CpG sites consistently related to prevalent and incident a non-causal relationship was suggested, this must be CVD events interpreted with caution as there was a single genetic Twelve of the 34 identified CpGs could be evaluated in instrumental variable, and we cannot discard that the prospective samples and four of them were also related to meQTL is in linkage disequilibrium with the causal vari- incident cases of CHD. cg21566642 maps to an intergenic ant for CHD, reverse causation or horizontal pleiotropy region, and cg05575921, cg04988978 and cg25769469 using this framework [28, 29]. Moreover, cg21566642 annotate to AHRR, MPO and PTCD2, respectively. To showed a genetic influence in childhood and adolescence, our knowledge, these CpGs were not associated with car- while CHD events typically occur during adulthood. diovascular events in previous EWAS reports. cg21566642 and cg05575921 were highly and inversely Strengths and limitations associated with smoking, which is supported by previ- The main strength of our study is that it is the first two- ous EWAS [18, 19]. We have also previously reported stage EWAS on MI to be based on more than 800,000 both CpGs as related to age-independent cardiovascular CpGs across the genome. Moreover, we aimed to vali- risk [13], and they have been related to all-cause mortal- date our findings in prospective samples of CHD and ity in an EWAS [22]. cg05575921 was further associated CVD as a proxy of MI. Also, we aimed to prove the clini- directly with cholesterol in high-density lipoproteins cal relevance of our findings. However, some limitations (HDL-C) and inversely with cholesterol in low-density should be acknowledged. First, two thirds of the CpGs lipoproteins (LDL-C) and triglyceride levels in our study. identified in the initial case–control study could not be This CpG has been related to both CHD prevalence and assessed in the incident studies as the methylation arrays incidence in a candidate gene study [23]. differed in the number of CpGs (EPIC VS 450 k, respec- cg04988978 and cg25769469 annotate to MPO and tively). Second, we used self-reported information about PTCD2, respectively. Both CpGs were associated directly cardiovascular risk factors in the case–control study, as with HDL-C and inversely with triglyceride and glucose an event such as MI modifies risk factor levels during levels. MPO encodes the myeloperoxidase, which pro- the acute phase. Third, we cannot infer causality since motes atherosclerotic lesions by enhancing APOB oxida- changes in methylation could have occurred as a conse- tion within low-density lipoproteins [24] and was causally quence of the acute phase and disease management of associated with incident cardiovascular outcomes [25]. the MI event. We aimed to perform MR studies of the One CpG located within PTCD2 was previously identi- association between the identified CpGs and cardiovas- fied to be associated with hypertension in obstructive cular events, but available methylation Quantitative Trait sleep apnea patients [26], and genetic variants in this Loci (meQTL) datasets are still limited. Last, our study is gene have been related with blood pressure [21]. based on populations of European origin and the results cannot be extrapolated to other populations. MRSs as predictive CVD biomarkers To assess the value of the four identified CpGs as pre- Conclusion dictive biomarkers, we followed the AHA recommenda- Our study provides 34 novel DNA methylation loci tions [27]. However, neither we observed an independent related to MI. The results shed some light on the molec- association between the MRSs and the incidence of CVD ular landscape of MI, highlighting the importance of Fernández‑Sanlés et al. Clin Epigenet Table 4 CpGs differentially methylated in association with incident coronary/cardiovascular disease in the fixed-effects meta-analyses of the samples from the Women’s Health Initiative (WHI) and the Framingham Offspring Study (FOS) Genomic Model 1 Model 2 Model 3 (2021) 13:86 information CpG Chr* Position (bp) Gene Coef* SE* P Coef* SE* P Coef* SE* P Incident coronary heart disease cg21566642 2 233,284,661 Intergenic − 1.34 × ­10–1 4.33 × ­10–2 2.00 × ­10–3 − 9.63 × ­10–2 4.83 × ­10–2 4.62 × ­10–2 − 9.66 × ­10–2 5.39 × ­10–2 7.32 × ­10–2 –1 –2 –4 –1 –2 –2 –1 –2 cg05575921 5 373,378 AHRR − 1.80 × ­10 4.73 × ­10 1.38 × ­10 − 1.33 × ­10 5.81 × ­10 2.21 × ­10 − 1.22 × ­10 6.52 × ­10 6.05 × ­10–2 –1 –2 –4 –1 –2 –3 –1 –2 cg25769469 5 71,643,841 PTCD2 − 2.62 × ­10 7.94 × ­10 9.67 × ­10 − 2.52 × ­10 8.02 × ­10 1.70 × ­10 − 2.14 × ­10 8.89 × ­10 1.59 × ­10–2 –1 –2 –3 –1 –2 –3 –1 –2 cg04988978 17 56,359,578 MPO − 2.28 × ­10 7.49 × ­10 2.29 × ­10 − 2.07 × ­10 7.58 × ­10 6.19 × ­10 − 1.06 × ­10 8.35 × ­10 2.03 × ­10–1 Incident cardiovascular disease cg21566642 2 233,284,661 Intergenic − 1.68 × ­10–1 3.97 × ­10–2 2.34 × ­10–5 − 1.31 × ­10–1 4.40 × ­10–2 2.94 × ­10–3 − 1.37 × ­10–1 4.85 × ­10–2 4.87 × ­10–3 cg05575921 5 373,378 AHRR − 2.08 × ­10–1 4.30 × ­10–2 1.30 × ­10–6 − 1.67 × ­10–1 5.24 × ­10–2 1.42 × ­10–3 − 1.61 × ­10–1 5.79 × ­10–2 5.54 × ­10–3 cg25769469 5 71,643,841 PTCD2 − 1.92 × ­10–1 7.27 × ­10–2 8.40 × ­10–3 − 1.77 × ­10–1 7.35 × ­10–2 1.58 × ­10–2 − 1.26 × ­10–1 8.09 × ­10–2 1.20 × ­10–1 cg04988978 17 56,359,578 MPO − 2.04 × ­10–1 6.72 × ­10–2 2.42 × ­10–3 − 1.83 × ­10–1 6.79 × ­10–2 6.99 × ­10–3 − 9.08 × ­10–2 7.36 × ­10–2 2.18 × ­10–1 *Chr, Chromosome; Coef; Coefficient of association; SE, Standard error Model 1 was adjusted for age, estimated cell counts and two surrogate variables (plus ancestry in WHI, plus sex in FOS). Model 2 was further adjusted for smoking status. Model 3 was additionally adjusted for diabetes, hypercholesterolemia, and hypertension. The Bonferroni-corrected p threshold was 4.17 × ­10–3 Page 7 of 11 Fernández‑Sanlés et al. Clin Epigenet (2021) 13:86 Page 8 of 11 Table 5 Associations between the identified CpGs and classical cardiovascular risk factors (CVRFs) in the fixed-effects meta-analyses of the four samples CpG Smoking BMI* HDL-C* LDL-C* TG* Glucose SBP DBP* cg21566642 − 1.61 9.55 × ­10–3 7.69 × ­10–4 − 1.10 × ­10–3 − 4.17 × ­10–4 2.32 × ­10–4 2.30 × ­10–3 7.97 × ­10–3 Coefficient –2 –3 –3 –4 –4 –4 –4 4.22 × ­10 2.65 × ­10 1.21 × ­10 4.09 × ­10 1.95 × ­10 4.20 × ­10 9.02 × ­10 1.67 × ­10–3 SE* –16 –4 –1 –3 –2 –1 –2 < 2.2 × ­10 3.21 × ­10 5.24 × ­10 7.11 × ­10 3.22 × ­10 5.81 × ­10 1.08 × ­10 1.78 × ­10–6 P –2 –3 –3 –4 –4 –3 cg05575921 − 1.61 1.17 × ­10 4.36 × ­10 − 1.35 × ­10 − 7.09 × ­10 1.86 × ­10 2.15 × ­10 7.54 × ­10–3 Coefficient –2 –3 –3 –4 –4 –4 –4 4.22 × ­10 2.46 × ­10 1.11 × ­10 3.77 × ­10 1.79 × ­10 3.87 × ­10 8.23 × ­10 1.52 × ­10–3 SE* –16 –6 –5 –4 –5 –1 –3 < 2.2 × ­10 2.22 × ­10 8.22 × ­10 3.39 × ­10 7.63 × ­10 6.30 × ­10 9.06 × ­10 7.69 × ­10–7 P –2 –2 –3 –4 –3 –3 –3 cg25769469 − 9.01 × ­10 − 1.12 × ­10 6.76 × ­10 2.47 × ­10 − 1.38 × ­10 − 1.52 × ­10 − 2.35 × ­10 − 4.17 × ­10–3 Coefficient –2 –3 –3 –4 –4 –4 –4 5.08 × ­10 2.64 × ­10 1.20 × ­10 4.11 × ­10 1.93 × ­10 4.17 × ­10 9.10 × ­10 1.69 × ­10–3 SE* –2 –5 –8 –1 –13 –4 –3 7.65 × ­10 2.32 × ­10 1.92 × ­10 5.48 × ­10 8.75 × ­10 2.69 × ­10 9.97 × ­10 1.36 × ­10–2 P –2 –2 –3 –4 –3 –3 –3 cg04988978 − 7.29 × ­10 − 1.15 × ­10 5.28 × ­10 3.59 × ­10 − 1.19 × ­10 − 1.88 × ­10 − 2.88 × ­10 − 3.54 × ­10–3 Coefficient –2 –3 –3 –4 –4 –4 –4 5.15 × ­10 2.64 × ­10 1.20 × ­10 4.08 × ­10 1.92 × ­10 4.15 × ­10 9.02 × ­10 1.67 × ­10–3 SE* –1 –5 –5 –1 –10 –6 –3 1.57 × ­10 1.26 × ­10 1.08 × ­10 3.80 × ­10 5.79 × ­10 5.97 × ­10 1.41 × ­10 3.47 × ­10–2 P *BMI, Body mass index; HDL-C, Cholesterol in high-density lipoprotein; LDL-C, Cholesterol in low-density lipoprotein; TG, Triglycerides; SBP, Systolic blood pressure; DBP, Diastolic blood pressure; SE, Standard Error traditional CVRFs and inflammation in the develop- and meets the principles expressed in the Declaration of ment of CHD. Our results question the relevance of DNA Helsinki and the relevant Spanish legislation. methylation as a predictive biomarker. Samples with incident cases of CHD and CVD Methods The WHI sample is a case–control study nested in a Study design and populations cohort. The FOS sample is a prospective cohort study. We designed an EWAS using three populations: the Both samples were available in the database of Genotypes Girona Heart Registry (REGICOR, REgistre GIroní del and Phenotypes (http://​dbgap.​ncbi.​nlm.​nih.​gov; Project COR), the Women’s Health Initiative (WHI), and the Number #9047). The graphical abstract shows the design Framingham Offspring Study (FOS). We first performed and flow-chart of this study. a two-stage EWAS on acute MI using two independent age- and sex-matched case–control studies designed in Assessment of cardiovascular outcomes REGICOR. Then, we validated the results in the other The outcomes assessed were acute MI in REGICOR, and two populations with incident cases of CHD and CVD. incident CHD and CVD in the WHI and FOS samples. Additional details are provided in the Additional file 1: Case–control studies of acute MI in REGICOR Methods. The sample used in the discovery stage (REGICOR-1) involved 416 individuals (208 MI cases and 208 con- Assessment of DNA methylation trols). The sample in the validation stage (REGICOR-2) DNA methylation was assessed genome-wide from comprised 208 individuals (104 cases and 104 controls). peripheral blood with commercial arrays from Illumina Cases were selected from patients who were consecu- (CA, USA). The Infinium MethylationEPIC BeadChip, tively attended for a first acute MI in the reference hos- covering over 850,000 CpGs, was used in the REGICOR pital of the monitored area, in the province of Girona, in samples. The Infinium HumanMethylation450 Bead- the northeast of Spain. Women were overrepresented to Chip, covering over 480,000 CpGs, was used in the WHI achieve their inclusion as 50% of our sample. Controls and FOS samples. A detailed quality control pipeline for were participants in a population-based survey per- the methylation data is available in the Additional file 1: formed in the same monitored area. They were randomly Methods. Methylation status at each CpG was reported selected from those attending the 2009–2013 follow-up by β-values [30]. visit (n = 4980), and matched by age and sex with the MI cases. All participants were of European descent and pro- Covariates vided informed written consent. The study was approved In the REGICOR case–control studies the follow- by the local ethics committee (2015/6199/I; 2018/7855/I) ing covariates were considered: self-reported smok- ing, diabetes, hypercholesterolemia and hypertension Fernández‑Sanlés et al. Clin Epigenet (2021) 13:86 Page 9 of 11 (Additional file 1: Methods). In the WHI and FOS stud- Association between the identified CpGs and CVRFs ies self-reported smoking and glycaemia, total and HDL We analysed whether the methylation levels of the identi- cholesterol, and blood pressure measurements were con- fied CpGs were associated with individual CVRFs in the sidered. Moreover, we inferred the peripheral blood cell four samples using multiple linear regression, and then counts with the FlowSorted.Blood.450 k R package [31]. meta-analysed the results. We defined DNA methylation We also estimated two surrogate variables for unknown as the outcome and adjusted for age and sex in the case of sources of potential technical or biological confounding the REGICOR and the Framingham populations, and for using the sva R package [32]. age and ethnicity in the WHI sample. In the case of the REGICOR samples, the continuous variables were only Statistical analysis available for the control individuals. We meta-analysed All statistical analyses were performed using R version the results from the four populations using a fixed-effects 3.4.0. The codes of the Singularity images used to run the meta-analysis weighted by the inverse of the variance. EWASs in the high performance computing system of the The p value threshold was estimated as 0.05 divided by Hospital del Mar Medical Research Institute are avail- the multiplication of the number of CVRFs and the num- able in the repositories at https://​github.​com/​regic​or/​ ber of CpGs assessed. methy​lation_​ami/. A detailed description of the statistical methods is provided in the Additional file 1: Methods. Methylation risk scores (MRSs) and predictive capacity We developed two weighted MRSs based on the CpGs Association between DNA methylation and cardiovascular identified, each of them using the results from the meta- outcomes analyses of incident CHD or CVD, respectively. We eval- Logistic regression was used in the analyses in the REGI- uated the association between these scores and CHD and COR and WHI samples, while Cox regression was used CVD incidence, respectively, in the FOS sample, using in the FOS sample. We considered the cardiovascular Cox regression. All analyses were adjusted for age, sex, event (acute MI, CHD or CVD) as the outcome and DNA diabetes, smoking, systolic blood pressure, hypertensive methylation as the exposure. treatment, and levels of total cholesterol and HDL-C [36]. We defined three models. Model 1 was adjusted for We also assessed the potential added predictive value of estimated cell counts and two surrogate variables (plus including the MRSs in the Framingham risk function. age and ethnicity in the WHI sample, plus age and sex We evaluated the increase in the discrimination and the in the FOS samples). Model 2 was additionally adjusted reclassification. for smoking. Model 3 was further adjusted for diabetes, hypercholesterolemia and hypertension. Causality of associations between DNA methylation In order to reduce epigenomic inflation, we corrected and cardiovascular outcomes the coefficients, the standard errors and the p values We took a two-sample Mendelian Randomization studies using the bacon R package if necessary [33]. The bacon using the MR-Base platform [37]. We used the MRInstru- R package controls for bias and inflation using a Bayesian ments R package to select the instrumental variables, and method based on the estimation of the empirical null dis- then, the TwoSampleMR R package. First, we considered tribution and was used in previous EWAS [33–35]. We those methylation-level quantitative trait loci (meQTL) used coefficients and standard errors from the regres- from the Accesible Resource for Integrated Epigenomic sion models as the input data and we set a random seed Studies (ARIES) project [16] included in the MR-Base at 123. database [37]. Then, we interrogated their association We selected those associations from the discovery stage with MI and with CHD using summary statistic data (REGICOR-1) with a corrected p-value < ­10–5 for assess- from a meta-analysis of GWAS on CHD [38]. A more ment in the validation stage (REGICOR-2). Moreover, we detailed description of the analysis is included in the performed a fixed-effect meta-analysis of the corrected Additional file 1: Methods. effect sizes observed in both stages, weighted by the inverse of the variance. Thereafter, we studied the associ- Abbreviations ation of the identified CpGs with incident CHD and with 450 k: Infinium HumanMethylation450 BeadChip; MI: Myocardial infarction; CVD events in the WHI and the FOS samples, separately. CHD: Coronary heart disease; CVD: Cardiovascular disease; CVR: Cardiovascular The results from both samples were meta-analysed (for risk; CVRF: Cardiovascular risk factor; EPIC: Infinium MethylationEPIC BeadChip; EWAS: Epigenome-wide association study; FOS: Framingham offspring study; CHD and CVD, separately). We used the Bonferroni cri- MRS: Methylation risk score; REGICOR: REgistre GIroní del COR; WHI: Women’s teria to correct for multiple comparisons (0.05 divided by Health Iniciative. the number of probes analysed in each specific analysis). Fernández‑Sanlés et al. Clin Epigenet (2021) 13:86 Page 10 of 11 Supplementary Information (WHI) program is funded by the NHLBI [Contracts N01WH22110, 24152, 32100-2, 32105-6, 32108-9, 32111-13, 32115, 32118-32119, 32122, 42107-26, The online version contains supplementary material available at https://​doi.​ 42129–32, and 44221]. This manuscript was not prepared in collaboration with org/​10.​1186/​s13148-​021-​01078-6. investigators of the FHS/ WHI, has not been reviewed and/or approved by the FHS/WHI, and does not necessarily reflect the opinions or views of the FHS Additional file 1. Additional material. and WHI investigators or the NHLBI. Additional file 2: Table S1. Discovery stage of the EWAS on acute Availability of data and materials myocardial infarction (REGICOR-1 sample). Model 1 was adjusted for All data generated during this study are included in this published article and estimated cell counts and two surrogate variables. Model 2 was further its Additional files. The REGICOR datasets analysed during the current study adjusted for smoking status. Model 3 was additionally adjusted for are available from the corresponding author on reasonable request. diabetes, hypercholesterolemia and hypertension. Coefficients, standard errors and p-values are given for each model before and after the correc‑ tion of the inflation using the bacon R package. Suggestive significant Declarations associations (p-value<10-5) are in bold. The total number of suggestive significant associations in each model is given. Table S2. Meta-analyses Ethics approval and consent to participate of the results from the discovery (REGICOR-1) and the validation stage The study was approved by the local ethics committee (2015/6199/I; (REGICOR-2). Model 1 was adjusted for estimated cell counts and two 2018/7855/I) and meets the principles expressed in the Declaration of Helsinki surrogate variables. Model 2 was further adjusted for smoking status. and the relevant Spanish legislation. Model 3 was additionally adjusted for diabetes, hypercholesterolemia and hypertension. Coefficients, standard errors and p-values of REGICOR-1 Consent for publication are those corrected with the bacon R package. Significant associations Not applicable. (p-value<6.17 × 10-8) are highlighted in bold. The total number of signifi‑ cant associations in each model is given. Table S3. Meta-analysis of the Competing interests results from the follow-up association studies performed in the samples The authors declare that they have no competing interests. with incident cases of cardiovascular (CVD) and coronary heart disease (CHD). Model 1 was adjusted for age, estimated cell counts and two Author details 1 surrogate variables (plus ethnicity in WHI and sex in FOS). Model 2 was Cardiovascular Epidemiology and Genetics Research Group, REGICOR Study further adjusted for smoking status. Model 3 was additionally adjusted for Group, IMIM (Hospital del Mar Medical Research Institute), Dr Aiguader 88, diabetes, hypercholesterolaemia and hypertension. Significant associa‑ 08003 Barcelona, Catalonia, Spain. 2 Pompeu Fabra University (UPF), Barcelona, tions (4.17 × 10-3) found in the fixed-effects meta-analysis are highlighted Catalonia, Spain. 3 Medical Research Council (MRC) Integrative Epidemiology in bold. The total number of significant associations in each model is Unit, University of Bristol, Bristol, UK. 4 CIBER Cardiovascular Diseases (CIBERCV), given. Table S4. Utility of the methylation risk scores (MRS): association Madrid, Spain. 5 Department of Medical Sciences, Molecular Epidemiology, with cardiovascular (CVD) or coronary (CHD) incidence and assessment of Uppsala University, Uppsala, Sweden. 6 CIBER Epidemiology and Public Health their predictive capacity. MRSs were based on the results from model 1 of (CIBERESP), Madrid, Spain. 7 Josep Carreras Leukaemia Research Institute (IJC), incident CHD and CVD. Analyses of the association with the CVD or CHD Badalona, Catalonia, Spain. 8 CIBER Oncology (CIBERONC), Madrid, Spain. 9 incidence were adjusted for age, sex, total cholesterol and HDL-C levels, Catalan Institution for Research and Advanced Studies (ICREA), Barcelona, diabetes, smoking status, systolic blood pressure, and hypertensive treat‑ Catalonia, Spain. 10 Physiological Sciences Department, School of Medicine ment, which are the cardiovascular risk factors considered in the Framing‑ and Health Sciences, University of Barcelona (UB), Barcelona, Catalonia, Spain. 11 ham risk function. Analysis of the improvement in the predictive capacity Medicine Department, Faculty of Medicine, University of Vic-Central Univer‑ of the Framingham function was performed with and without the sity of Catalonia (UVic-UCC), Vic, Catalonia, Spain. corresponding MRS. Table S5. Results of the Wald ratio method applied to determine the causality between the identified CpGs and coronary heart Received: 20 February 2021 Accepted: 13 April 2021 disease or myocardial infarction. Acknowledgements We thank Elaine M. Lilly, PhD, for her critical reading and revision of the English References text. 1. Roth GA, Abate D, Abate KH, et al. Global, regional, and national age-sex- specific mortality for 282 causes of death in 195 countries and territories, Authors’ contributions 1980–2017: a systematic analysis for the Global Burden of Disease Study AF-S, SS-B and RE-L contributed to the conception or design of the work. All 2017. Lancet. 2018;392(10159):1736–88. authors contributed to the acquisition, analysis, or interpretation of data for 2. James SL, Abate D, Abate KH, et al. Global, regional, and national the work. AF-S and RE-L drafted the manuscript. SS-B, IS, MS, SP-F, MC-M, ME, incidence, prevalence, and years lived with disability for 354 diseases JM critically revised the manuscript. All authors read and approved the final and injuries for 195 countries and territories, 1990–2017: a systematic manuscript. analysis for the Global Burden of Disease Study 2017. The Lancet. 2018;392(10159):1789–858. Funding 3. Piepoli MF, Hoes AW, Agewall S, et al. 2016 European Guidelines on The authors disclosed receipt of the following financial support for the cardiovascular disease prevention in clinical practice. Eur Heart J. research, authorship, and/or publication of this article: This work was sup‑ 2016;37(29):2315–81. ported by the Carlos III Health Institute–European Regional Development 4. Marrugat J, Vila J, Baena-Díez JM, Grau M, Sala J, Ramos R, Subirana I, Fitó Fund [Grant Numbers FIS PI18/00017, FIS PI15/00051, PI12/00232, CIBERCV, M, Elosua R. Relative Validity of the 10-Year Cardiovascular Risk Estimate CIBERESP, CIBERONC]; PERIS from Agència de Gestió d’Ajuts Universitaris i in a Population Cohort of the REGICOR Study. Revista Española de Cardi‑ de Recerca [Grant Number SLT002/16/00088]; the Government of Catalonia ología (English Edition). 2011;64(5):385–94. through the Agency for Management of University and Research Grants 5. Patrono C. Fighting residual cardiovascular risk in stable patients with [Grant Numbers 2014SGR240, 2017SGR946]; the Spanish Ministry of Economy atherosclerotic vascular disease: COMPASS in context. Cardiovasc Res. and Competitiveness [Grant Number BES-2014–069718 to AF-S]; and Carlos III 2017;113(14):e61–3. Health Institute-FEDER (grant number IFI14/00007 to SS-B). The Framingham 6. Petronis A. Epigenetics as a unifying principle in the aetiology of complex Heart Study (FHS) is conducted and supported by the National Heart, Lung, traits and diseases. Nature. 2010;465(7299):721–7. and Blood Institute (NHLBI) in collaboration with Boston University [Contract 7. Jin Z, Liu Y. DNA methylation in human diseases. Genes Diseases. No. N01-HC-25195 and HHSN268201500001I]. The Women’s Health Initiative 2018;5(1):1–8. Fernández‑Sanlés et al. Clin Epigenet (2021) 13:86 Page 11 of 11 8. Agha G, Mendelson MM, Ward-Caviness CK, et al. Blood leukocyte DNA 23. Ligthart S, Marzi C, Aslibekyan S, et al. DNA methylation signatures of methylation predicts risk of future myocardial infarction and coronary chronic low-grade inflammation are associated with complex diseases. heart disease. Circulation. 2019;140(8):645–57. Genome Biol. 2016;17(1):255. 9. Ward-Caviness CK, Agha G, Chen BH, et al. Analysis of repeated leukocyte 24. Teng N, Maghzal GJ, Talib J, Rashid I, Lau AK, Stocker R. The roles of DNA methylation assessments reveals persistent epigenetic alterations myeloperoxidase in coronary artery disease and its potential implication after an incident myocardial infarction. Clin Epigenetics. 2018;10(1):161. in plaque rupture. Redox Rep. 2017;22(2):51–73. 10. Westerman K, Sebastiani P, Jacques P, Liu S, Demeo D, Ordovás JM. DNA 25. Yao C, Chen G, Song C, et al. Genome-wide mapping of plasma protein methylation modules associate with incident cardiovascular disease and QTLs identifies putatively causal genes and pathways for cardiovascular cumulative risk factor exposure. Clin Epigenet. 2019;11(1):142. disease. Nat Commun. 2018;9(1):3268. 11. Westerman K, Fernández-Sanlés A, Patil P, Sebastiani P, Jacques P, Starr JM, 26. Chen Y-C, Chen T-W, Su M-C, et al. Whole genome DNA methylation Deary I, Liu Q, Liu S, Elosua R, DeMeo DL, Ordovás JM. Epigenomic assess‑ analysis of obstructive sleep apnea: IL1R2, NPR2, AR, SP140 methylation ment of cardiovascular disease risk and interactions with traditional risk and clinical phenotype. Sleep. 2016;39(4):743–55. metrics. J Am Heart Assoc. 2020;9(8):e015299. 27. Hlatky MA, Greenland P, Arnett DK, et al. Criteria for evaluation of novel 12. Fernández-Sanlés A, Sayols-Baixeras S, Subirana I, Degano IRIR, Elosua markers of cardiovascular risk. Circulation. 2009;119(17):2408–16. R. Association between DNA methylation and coronary heart disease 28. Hemani G, Bowden J, Davey SG. Evaluating the potential role of or other atherosclerotic events: a systematic review. Atherosclerosis. pleiotropy in Mendelian randomization studies. Hum Mol Genet. 2017;263:325–33. 2018;27:R195–208. 13. Fernández-Sanlés A, Sayols-Baixeras S, Curcio S, Subirana I, Marrugat J, 29. Richardson TG, Zheng J, Davey Smith G, et al. Mendelian randomization Elosua R. DNA methylation and age-Independent cardiovascular risk, an analysis identifies CpG sites as putative mediators for genetic influences epigenome-Wide approach the REGICOR study (REgistre GIroní del COR). on cardiovascular disease risk. Am J Hum Genet. 2017;101:590–602. Arterioscler Thromb Vasc Biol. 2018;38(3):645–52. 30. Du P, Zhang X, Huang C-C, Jafari N, Kibbe WA, Hou L, Lin SM. Comparison 14. Sandoval J, Heyn H, Moran S, Serra-Musach J, Pujana MA, Bibikova M, of Beta-value and M-value methods for quantifying methylation levels by Esteller M. Validation of a DNA methylation microarray for 450,000 CpG microarray analysis. BMC Bioinf. 2010;11(1):587. sites in the human genome. Epigenetics. 2011;6(6):692–702. 31. Houseman EA, Accomando WP, Koestler DC, Christensen BC, Marsit CJ, 15. Moran S, Arribas C, Esteller M. Validation of a DNA methylation microar‑ Nelson HH, Wiencke JK, Kelsey KT. DNA methylation arrays as surrogate ray for 850,000 CpG sites of the human genome enriched in enhancer measures of cell mixture distribution. BMC Bioinf. 2012;13(1):86. sequences. Epigenomics. 2016;8(3):389–99. 32. Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for 16. Gaunt TR, Shihab HA, Hemani G, Min JL, Woodward G, Lyttleton O, Zheng removing batch effects and other unwanted variation in high-through‑ J, Duggirala A, McArdle WL, Ho K, Ring SM, Evans DM, Davey Smith G, put experiments. Bioinformatics. 2012;28(6):882–3. Relton CL. Systematic identification of genetic influences on methylation 33. van Iterson M, van Zwet EW, Heijmans BT, BIOS Consortium BT. Control‑ across the human life course. Genome Biol. 2016;17(1):61. ling bias and inflation in epigenome- and transcriptome-wide association 17. Sayols-Baixeras S, Lluís-Ganella C, Subirana I, et al. Identification of a new studies using the empirical null distribution. Genome Biol. 2017;18(1):19. locus and validation of previously reported loci showing differential 34. Meeks KAC, Henneman P, Venema A, et al. Epigenome-wide associa‑ methylation associated with smoking The REGICOR study. Epigenetics. tion study in whole blood on type 2 diabetes among sub-Saharan 2015;10(12):1156–65. African individuals: findings from the RODAM study. Int J Epidemiol. 18. Shenker NS, Polidoro S, van Veldhoven K, Sacerdote C, Ricceri F, Birrell MA, 2018;48:58–70. Belvisi MG, Brown R, Vineis P, Flanagan JM. Epigenome-wide association 35. Siemelink MA, van der Laan SW, Haitjema S, et al. Smoking is Associated study in the European Prospective Investigation into Cancer and Nutri‑ to DNA methylation in atherosclerotic carotid lesions. Circ Genomic tion (EPIC-Turin) identifies novel genetic loci associated with smoking. Precis Med. 2018. https://​doi.​org/​10.​1161/​CIRCU​LATIO​NAHA.​107.​699579. Hum Mol Genet. 2013;22(5):843–51. 36. D’Agostino RB, Vasan RS, Pencina MJ, Wolf PA, Cobain M, Massaro JM, 19. Gao X, Jia M, Zhang Y, Breitling LP, Brenner H. DNA methylation changes Kannel WB. General cardiovascular risk profile for use in primary care. of whole blood cells in response to active smoking exposure in Circulation. 2008. https://​doi.​org/​10.​7554/​eLife.​34408. adults: a systematic review of DNA methylation studies. Clin Epigenet. 37. Hemani G, Zheng J, Elsworth B, et al. The MR-Base platform supports sys‑ 2015;7(1):113. tematic causal inference across the human phenome. eLife. 2018. https://​ 20. Van Der Harst P, Verweij N. Identification of 64 novel genetic loci provides doi.​org/​10.​7554/​eLife.​34408. an expanded view on the genetic architecture of coronary artery disease. 38. Nikpay M, Goel A, Won H-H, et al. A comprehensive 1000 Genomes– Circ Res. 2018;122(3):433–43. based genome-wide association meta-analysis of coronary artery 21. Buniello A, MacArthur JAL, Cerezo M, et al. The NHGRI-EBI GWAS Catalog disease. Nat Genet. 2015;47(10):1121–30. of published genome-wide association studies, targeted arrays and sum‑ mary statistics 2019. Nucleic Acids Res. 2019;47(D1):D1005–12. 22. Svane A, Soerensen M, Lund J, Tan Q, Jylhävä J, Wang Y, Pedersen N, Hägg Publisher’s Note S, Debrabant B, Deary I, Christensen K, Christiansen L, Hjelmborg J. DNA Springer Nature remains neutral with regard to jurisdictional claims in pub‑ methylation and all-cause mortality in middle-aged and elderly Danish lished maps and institutional affiliations. Twins. Genes. 2018;9(2):78. Ready to submit your research ? Choose BMC and benefit from: • fast, convenient online submission • thorough peer review by experienced researchers in your field • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions