Abstract

Background Epigenetic programming and epigenetic mechanisms driven by environmental factors are thought to play an important role in human health and ageing. Global DNA methylation has been postulated as an epigenetic marker for epidemiological studies as it is reflective of changes in gene expression linked to disease. How epigenetic mechanisms are affected by psychological, sociological and biological determinants of health still remains unclear. The aim of this study was to investigate the relationship between socio-economic and lifestyle factors and epigenetic status, as measured by global DNA methylation content, in the pSoBid cohort, which is characterized by an extreme socio-economic and health gradient.

Methods DNA was extracted from peripheral blood leukocytes using the Maxwell® 16 System and Maxwell® 16 Blood DNA Purification kit (Promega, UK). Global DNA methylation was assessed using Methylamp™ Global DNA Methylation Quantification Ultra kit (Epigentek, USA). Associations between global DNA methylation and socio-economic and lifestyle factors were investigated in linear regression models.

Results Global DNA hypomethylation was observed in the most socio-economically deprived subjects. Job status demonstrated a similar relationship, with manual workers having 24% lower DNA methylation content than non-manual. Additionally, associations were found between global DNA methylation content and biomarkers of cardiovascular disease (CVD) and inflammation, including fibrinogen and interleukin-6 (IL-6), after adjustment for socio-economic factors.

Conclusions This study has indicated an association between epigenetic status and socio-economic status (SES). This relationship has direct implications for population health and is reflected in further associations between global DNA methylation content and emerging biomarkers of CVD.

Introduction

DNA methylation is an important epigenetic mechanism involved in the regulation of gene expression, genomic stability and imprinting during embryogenesis.1 It may also constitute a novel bio-dosimeter with which to examine the impact of socio-economic and lifestyle factors on disease susceptibility and risk.

Changes in global DNA methylation may be some of the earliest cellular events in disease onset.2 Aberrant DNA methylation has been linked to ageing and a broad range of diseases including cardiovascular and neuronal disorders, as well as cancer.3

DNA methylation is crucial for mammalian development/differentiation and the pattern of this DNA modification is established during embryogenesis, beginning with embryo implantation (global demethylation) followed by de novo re-methylation before organogenesis commences. The methylation pattern becomes imprinted in early life leading to predisposition for specific chronic diseases; however changes can still be made although these are usually minor and can occur throughout adulthood.1 DNA methylation changes have also been linked to the ageing process, with chronological ageing being accompanied by gradual genome hypomethylation, but with an increase of methylation occurring within promoter regions.4 The pattern of methylation within specific genomic regions may be altered in response to diet and lifestyle factors, such as cigarette smoking.4,5 Recent studies have shown that dietary factors, in particular, may be involved in the regulation of global DNA methylation during embryogenesis and this may have further consequences in adult life.1

Insufficient maternal care or diet can be reflected in the methylation status of their offspring. Indeed, an inherited sensitivity to stress, influenced by the mood of the mother during pregnancy, has also been reported.6 These observations suggest that there may be an adaptive mechanism which allows for epigenetic plasticity in response to environmental changes. A broad range of environmental factors may therefore impact on global DNA methylation status and consequently health. This is important to the understanding of the impact of socio-economic drivers of ill health, which may be predominant in communities where there is also a higher prevalence of classical risk factors for disease.

One hypothesis for the increased disease prevalence in these communities is underlying chronic inflammation, a component linked to a wide range of pathologies.7 This, however, may be influenced by epigenetic reprogramming, with consequential transcriptome remodelling in the aetiology of chronic inflammatory diseases.8 We and others have previously reported a strong association between inflammatory biomarkers, such as C-reactive protein and interleukin-6 (IL-6), and DNA methylation in chronic kidney disease patients.9,10

At a mechanistic level, inflammation has a direct impact on the activity of enzymes involved in the establishment, maintenance and change of epigenetic status. In particular, IL-6 is involved in the regulation and localization of HDNMT (human DNA methyltransferase),11,12 influencing expression and activity of this enzyme. Consequently, increased inflammation may lead to aberrant methylation leading to subsequent ill health.13

These observations, in toto, are of particular relevance to Glasgow because of the associated variation in mortality, morbidity and the extreme socio-economic (SE) gradient of health inequality in this city which is not fully explained by conventional risk factors for disease.14

The aim of this study was to enhance our understanding of the relationship between epigenetic status and underlying associations between socio-economic factors and risk factors for ill health, using the Glasgow-based pSoBid cohort.

Rather than using a high resolution analysis of methylation patterns associated with transcriptional changes, we have determined epigenetic status simply by measuring global methylation content and investigated whether we could use this to explain the difference in prevalence of ill health between most and least deprived groups in this cohort. Our working hypothesis was that social deprivation would be associated with genome-wide hypomethylation (i.e. reduced methylation content), which would in turn be associated with enhanced inflammatory status and associated disease risk.

Methods

Study population

Participants comprised a subset (n = 239) of the original pSoBid cohort, for which sufficient quantity and quality DNA was available for analysis,15 as determined by spectral analysis. The design of the psychological, social and biological determinants of ill health (pSoBid) study has been described in detail elsewhere.12 In brief, participants were ranked on the basis of multiple deprivation indicators to define the least and most deprived areas in the NHS Greater Glasgow Health Board area, using criteria based on the Scottish Index for Multiple Deprivation (SIMD). Sampling was stratified to achieve an approximately equal distribution of the 666 participants across males and females and age groups (35–44, 45–54 and 55–64 years) within the most (bottom 5% of SIMD score) and least deprived areas (top 20% of SIMD score). From this original population 239 individual DNA samples were available for analysis. Supplementary Table S1 compares these 239 participants with the 427 who were not included with respect to age, gender, deprivation group, social class, income, education, housing tenure, smoking habit, diet, alcohol consumption, body mass index and waist/hip ratio. Only age group and waist/hip ratio differed between included and excluded participants, there being a lower proportion of 55 to 64-year-olds in the included participants (29% compared with 45%) and a higher mean waist/hip ratio (0.91 compared with 0.89). These samples were blinded with only the provision of an ID number to identify each individual sample, the samples were analysed and data were sent for unblinding to statistical core where statistical analysis was completed. Participants were assessed, with physical examination data, medical history, risk factors, arteriosclerosis indicators (plaque count, carotid intima-media thickness), cognitive functions analysis, morbidity and psychological profiles being collected.

Global DNA methylation determination

DNA was extracted from peripheral blood leukocytes using Maxwell® 16 System and Maxwell® 16 Blood DNA Purification kit (Promega). Quantitative and qualitative DNA analysis was performed using Nanodrop. DNA was analysed for global methylation using Methylamp™ Global DNA Methylation Quantification Ultra kit (Epigentek, USA) according to the manufacturer's instructions. The capture antibody in this kit binds to 5-methylcytosine, thus measuring total DNA methylation level as a percentage of total DNA present in the sample. Hence it measures global DNA content, not specific sites, patterns or methylation types, as other systems do. Samples were run in duplicate and a standard curve was run in triplicate as were the positive and negative controls. Inter-assay precision (%CV) was <8%, intra-assay %CV was <6.5%.

Briefly, the plate was coated with 200 ng of DNA (100 ng/µl) per well and dried at 37°C for 40 min followed by incubation at 60°C until the solution evaporated and wells were dried. Simultaneously, the standard curve (range 0.2–20 ng of methylated DNA) and negative control DNA were prepared for each plate. The amount of methylated DNA (pg) was determined according to manufacturer's instruction; additionally %DNA methylation was calculated.

Statistical analysis

Associations between DNA methylation (expressed as a percentage of total DNA by weight) and participant characteristics were investigated in linear regression models adjusted for the stratification variables: age group (35–44, 45–54 and 55–64 years), gender (male and female) and deprivation group (bottom 5% and top 20% of SIMD score). Failure to adjust for these variables would risk uncovering spurious associations, in particular with lifestyle and socio-economic status (SES) factors, which are likely to be closely correlated with deprivation group. DNA methylation content and biomarkers were log transformed to satisfy the assumption of normally distributed residuals. Primary analyses were focused on the associations between methylation and factors reflecting SES and lifestyle. Firstly, the association between methylation and deprivation group was assessed because deprivation was the primary focus of the pSoBid study and drove the sampling strategy. Other factors reflecting SES were added separately to this base model (containing age group, gender and deprivation group), yielding a subset of SES factors that were associated with global DNA methylation. These factors were combined in a single model to assess which factors were independently associated with methylation. Secondly, we assessed the associations between DNA methylation and specific biomarkers associated with biological ageing and cardiovascular disease (CVD) by including DNA methylation as a covariate in linear regression models for each biomarker with adjustment for age, gender and deprivation group. Finally, interactions were tested between SES and age group (when the outcome variable was global DNA methylation) and methylation and age group (when the outcome was biomarker level). This study was exploratory, therefore no adjustment for multiple testing has been made. We detected a negative association between sampling date and DNA methylation content, which was not explained by any of the stratification variables (age group, gender and deprivation group). However, adjusting for sampling date had a negligible effect on the results presented here.

Percentage global DNA methylation and biomarker levels were log transformed for regression analysis. Regression coefficient estimates were therefore multiplicative when transformed back to the original scale, and are presented as the percentage difference in global DNA methylation associated with each patient characteristic where global methylation is the outcome, and percentage difference in biomarker level associated with a standard deviation (SD) increase in log global methylation where methylation is a covariate. Physical activity (inactive, moderately inactive, moderately active, active) and annual income (<£15 000, £16–25 000, £26–35 000, £36–45 000, >£45 000) were converted to scores (1–4 and 1–5, respectively) for inclusion in linear regression models, so that a regression coefficient for each of these characteristics is interpreted as the difference in the response associated with an increase of one level. Self-reported alcohol consumption in units per week was strongly right-skewed and the square root was used in regression models.

Results

Association between global methylation and socio-economic status

In order to determine associations between levels of global DNA methylation and measures of SES and lifestyle factors, linear regression models were applied. Two types of population characteristics were investigated: (i) age group (35–44, 45–54, 55–64 years), gender and deprivation group (most versus least deprived) and (ii) factors reflecting other aspects of SES and lifestyle, adjusted for age group, gender and deprivation group. The median global DNA methylation content within subgroups of SES and lifestyle factors, overall and by age group, is shown in Table 1.

Table 1

Median (interquartile range) %DNA methylation within subgroups of gender, SES and lifestyle, overall and by age group

    Age group
 
 Subgroup n All (n = 239) n = 74, (35–44 years) n = 95, (45–54 years) n = 70, (55–64 years) 
 All 239 3.3 (2.2–4.7) 3.4 (2.2–4.6) 2.8 (2.1–4.4) 3.8 (2.5–5.4) 
Gender Female 122 3.2 (2.2–4.8) 3.1 (2.1–4.7) 2.8 (2.1–4.2) 3.8 (2.7–6.6) 
Male 117 3.4 (2.2–4.6) 3.7 (2.5–4.2) 2.9 (2.3–4.5) 3.9 (2.1–4.8) 
Deprivation group Affluent 115 3.7 (2.3–5.5) 3.5 (2.8–4.5) 3.0 (2.1–5.0) 4.1 (2.6–6.9) 
Deprived 124 3.1 (2.2–4.2) 3.0 (2.0–4.6) 2.8 (2.2–3.8) 3.6 (2.5–4.3) 
Social class Non-manual 151 3.5 (2.4–5.3) 3.5 (2.8–4.8) 2.8 (2.2–4.6) 4.2 (3.4–7.3) 
Manual 78 2.9 (1.9–4.2) 2.9 (1.9–4.5) 2.9 (2.0–4.0) 3.1 (1.8–4.1) 
Household income >£25 000 122 3.4 (2.3–5.1) 3.4 (2.7–4.4) 2.9 (2.2–4.6) 4.7 (2.6–6.7) 
<£25 000 109 3.1 (2.0–4.3) 3.0 (2.0–4.5) 2.8 (2.0–4.0) 3.5 (2.4–4.3) 
Years of education Upper 50% 114 3.4 (2.4–5.1) 3.4 (2.9–4.4) 3.0 (2.2–4.7) 4.5 (2.6–6.7) 
Lower 50% 125 3.1 (2.0–4.3) 3.1 (2.0–4.7) 2.8 (2.0–3.8) 3.4 (2.4–4.3) 
Housing tenure Owner-occupier 151 3.4 (2.2–5.2) 3.5 (2.8–4.9) 2.8 (2.1–4.6) 4.1 (2.4–6.9) 
Tenant 88 3.1 (2.0–4.2) 2.9 (2.0–4.2) 3.0 (2.3–4.0) 3.6 (2.6–4.3) 
Physical activity level Active 110 3.2 (2.3–4.6) 3.4 (2.7–4.7) 2.5 (2.0–3.6) 4.0 (3.2–6.6) 
Inactive 129 3.4 (2.1–4.7) 3.2 (2.0–4.0) 3.5 (2.4–4.7) 3.4 (2.0–4.8) 
Current cigarette smoker No 180 3.4 (2.2–4.8) 3.2 (2.3–4.2) 2.8 (2.1–4.5) 3.8 (2.6–5.6) 
Yes 47 3.2 (2.1–4.4) 3.9 (1.7–4.8) 2.9 (2.3–3.7) 4.0 (2.7–4.7) 
Diet score Upper 50% 117 3.4 (2.1–5.1) 3.4 (2.2–4.8) 2.6 (2.0–4.5) 3.9 (2.6–6.9) 
Lower 50% 122 3.3 (2.3–4.4) 3.4 (2.3–4.2) 3.0 (2.3–4.4) 3.8 (2.3–4.7) 
Excessive alcohol [>14 (F) or 21 (M) U/week] No 192 3.4 (2.2–4.8) 3.4 (2.2–4.8) 2.8 (2.1–4.4) 3.8 (2.6–5.3) 
Yes 47 3.1 (2.2–4.2) 3.0 (2.3–3.9) 2.9 (2.3–4.2) 3.6 (1.9–5.7) 
Obese (BMI >30 kg/m2No 170 3.4 (2.2–4.7) 3.4 (2.1–4.6) 2.9 (2.3–4.5) 3.9 (2.5–5.4) 
Yes 67 3.2 (2.0–4.4) 3.5 (2.9–6.5) 2.6 (1.6–4.0) 3.4 (2.5–5.2) 
Waist/hip ratio Lower 50% 118 3.7 (2.3–4.9) 3.8 (3.0–4.9) 2.7 (2.1–4.4) 3.9 (2.6–6.3) 
Upper 50% 118 2.9 (2.1–4.4) 2.9 (2.0–4.1) 2.8 (2.2–4.3) 3.6 (2.4–5.1) 
    Age group
 
 Subgroup n All (n = 239) n = 74, (35–44 years) n = 95, (45–54 years) n = 70, (55–64 years) 
 All 239 3.3 (2.2–4.7) 3.4 (2.2–4.6) 2.8 (2.1–4.4) 3.8 (2.5–5.4) 
Gender Female 122 3.2 (2.2–4.8) 3.1 (2.1–4.7) 2.8 (2.1–4.2) 3.8 (2.7–6.6) 
Male 117 3.4 (2.2–4.6) 3.7 (2.5–4.2) 2.9 (2.3–4.5) 3.9 (2.1–4.8) 
Deprivation group Affluent 115 3.7 (2.3–5.5) 3.5 (2.8–4.5) 3.0 (2.1–5.0) 4.1 (2.6–6.9) 
Deprived 124 3.1 (2.2–4.2) 3.0 (2.0–4.6) 2.8 (2.2–3.8) 3.6 (2.5–4.3) 
Social class Non-manual 151 3.5 (2.4–5.3) 3.5 (2.8–4.8) 2.8 (2.2–4.6) 4.2 (3.4–7.3) 
Manual 78 2.9 (1.9–4.2) 2.9 (1.9–4.5) 2.9 (2.0–4.0) 3.1 (1.8–4.1) 
Household income >£25 000 122 3.4 (2.3–5.1) 3.4 (2.7–4.4) 2.9 (2.2–4.6) 4.7 (2.6–6.7) 
<£25 000 109 3.1 (2.0–4.3) 3.0 (2.0–4.5) 2.8 (2.0–4.0) 3.5 (2.4–4.3) 
Years of education Upper 50% 114 3.4 (2.4–5.1) 3.4 (2.9–4.4) 3.0 (2.2–4.7) 4.5 (2.6–6.7) 
Lower 50% 125 3.1 (2.0–4.3) 3.1 (2.0–4.7) 2.8 (2.0–3.8) 3.4 (2.4–4.3) 
Housing tenure Owner-occupier 151 3.4 (2.2–5.2) 3.5 (2.8–4.9) 2.8 (2.1–4.6) 4.1 (2.4–6.9) 
Tenant 88 3.1 (2.0–4.2) 2.9 (2.0–4.2) 3.0 (2.3–4.0) 3.6 (2.6–4.3) 
Physical activity level Active 110 3.2 (2.3–4.6) 3.4 (2.7–4.7) 2.5 (2.0–3.6) 4.0 (3.2–6.6) 
Inactive 129 3.4 (2.1–4.7) 3.2 (2.0–4.0) 3.5 (2.4–4.7) 3.4 (2.0–4.8) 
Current cigarette smoker No 180 3.4 (2.2–4.8) 3.2 (2.3–4.2) 2.8 (2.1–4.5) 3.8 (2.6–5.6) 
Yes 47 3.2 (2.1–4.4) 3.9 (1.7–4.8) 2.9 (2.3–3.7) 4.0 (2.7–4.7) 
Diet score Upper 50% 117 3.4 (2.1–5.1) 3.4 (2.2–4.8) 2.6 (2.0–4.5) 3.9 (2.6–6.9) 
Lower 50% 122 3.3 (2.3–4.4) 3.4 (2.3–4.2) 3.0 (2.3–4.4) 3.8 (2.3–4.7) 
Excessive alcohol [>14 (F) or 21 (M) U/week] No 192 3.4 (2.2–4.8) 3.4 (2.2–4.8) 2.8 (2.1–4.4) 3.8 (2.6–5.3) 
Yes 47 3.1 (2.2–4.2) 3.0 (2.3–3.9) 2.9 (2.3–4.2) 3.6 (1.9–5.7) 
Obese (BMI >30 kg/m2No 170 3.4 (2.2–4.7) 3.4 (2.1–4.6) 2.9 (2.3–4.5) 3.9 (2.5–5.4) 
Yes 67 3.2 (2.0–4.4) 3.5 (2.9–6.5) 2.6 (1.6–4.0) 3.4 (2.5–5.2) 
Waist/hip ratio Lower 50% 118 3.7 (2.3–4.9) 3.8 (3.0–4.9) 2.7 (2.1–4.4) 3.9 (2.6–6.3) 
Upper 50% 118 2.9 (2.1–4.4) 2.9 (2.0–4.1) 2.8 (2.2–4.3) 3.6 (2.4–5.1) 

Regression analysis revealed an association between global DNA methylation content and age group, but not gender (Table 2). The association with age group did not appear to reflect a decline of methylation content with increasing age, the lowest average methylation levels being found in the middle age group (45–54 years) and the highest in the oldest age group (55–64), 32% [95% confidence interval (CI) 10 to 59; P = 0.003] higher than in the middle age group. Global DNA methylation content was 17% (95% CI 4 to 29; P = 0.015) lower in the most deprived group than in the least deprived group (P = 0.015, Table 2, Figure 1A). The proportion of variance in log DNA methylation content explained by age group, gender and SES, gauged by adjusted R2, was 4.3%, compared with 2.3% for age group and gender alone.

Figure 1

Predicted DNA methylation (±95% CI) in subgroups of (a) deprivation (affluent/deprived), (b) social class (Non-manual/manual) and (c) years of education (upper 50% / lower 50%), broken down by age group. P-values are given for test of equal geometric mean DNA methylation (indicated by horizontal lines) between subgroups. DNA methylation was predicted from linear regression models of log DNA methylation adjusted to a subject of average deprivation group (b and c only), age and gender

Figure 1

Predicted DNA methylation (±95% CI) in subgroups of (a) deprivation (affluent/deprived), (b) social class (Non-manual/manual) and (c) years of education (upper 50% / lower 50%), broken down by age group. P-values are given for test of equal geometric mean DNA methylation (indicated by horizontal lines) between subgroups. DNA methylation was predicted from linear regression models of log DNA methylation adjusted to a subject of average deprivation group (b and c only), age and gender

Table 2

Percentage difference (95% CI) in DNA methylation associated with characteristics of study participants predicted from linear regression models. Two types of characteristic were investigated: (A) age, gender and deprivation the reference group for which is the 35–44 years age group; and (B) factors reflecting SES and lifestyle, whose association with DNA methylation is being investigated here

  Predicted per cent difference in DNA methylation, adjusted for:
 
  Income, diet score, current smoking
 
 Predictor Per cent difference 95% CI P-value Per cent difference 95% CI P-value 
A Age (years)       
     (45–54) −16.3 −30.4 to 0.5 0.011 −17.5 −31.4 to −0.8 0.002 
     (55–64) 10.7 −9.1 to 34.8 0.011 16.8 −5.0 to 43.6 0.002 
 Gender (male) −2.3 −16.1 to 13.9 0.768 1.3 −13.7 to 18.9 0.872 
 Deprivation group (deprived) −17.3 −29.1 to −3.6 0.015 −11.3 −30.6 to 13.3 0.334 
  Predicted per cent difference in DNA methylation, adjusted for:
 
  Income, diet score, current smoking
 
 Predictor Per cent difference 95% CI P-value Per cent difference 95% CI P-value 
A Age (years)       
     (45–54) −16.3 −30.4 to 0.5 0.011 −17.5 −31.4 to −0.8 0.002 
     (55–64) 10.7 −9.1 to 34.8 0.011 16.8 −5.0 to 43.6 0.002 
 Gender (male) −2.3 −16.1 to 13.9 0.768 1.3 −13.7 to 18.9 0.872 
 Deprivation group (deprived) −17.3 −29.1 to −3.6 0.015 −11.3 −30.6 to 13.3 0.334 
  Age, gender, deprivation group Age, gender, deprivation group, income, diet score, current smoking 
B Social class: Manual −24.3 −37.7 to −8.0 0.005 −27.4 −41.1 to −10.5 0.003 
 Household income score (1–5) 4.0 −3.1 to 11.5 0.279 – – – 
 Years of education 2.4 −0.1 to 5.0 0.062 2.5 −0.1 to 5.3 0.052 
 Housing tenure: Tenant −3.9 −23.5 to 20.8 0.733 −11.6 −32.3 to 15.6 0.367 
 Physical activity level 1.2 −5.3 to 8.2 0.719 −0.8 −7.5 to 6.4 0.827 
 Current cigarette smoker 5.2 −15.0 to 30.3 0.638 – – – 
 Diet score 0.0 −0.2 to 0.1 0.781 – – – 
 Weekly alcohol consumption (sqrt units) −0.4 −3.8 to 3.1 0.820 −1.2 −4.8 to 2.4 0.500 
 Body mass index (kg/m2−0.7 −2.1 to 0.7 0.315 −0.8 −2.2 to 0.7 0.310 
 Waist/hip ratio −12.1 −73.9 to 195.8 0.835 −16.3 −76.7 to 200.5 0.784 
  Age, gender, deprivation group Age, gender, deprivation group, income, diet score, current smoking 
B Social class: Manual −24.3 −37.7 to −8.0 0.005 −27.4 −41.1 to −10.5 0.003 
 Household income score (1–5) 4.0 −3.1 to 11.5 0.279 – – – 
 Years of education 2.4 −0.1 to 5.0 0.062 2.5 −0.1 to 5.3 0.052 
 Housing tenure: Tenant −3.9 −23.5 to 20.8 0.733 −11.6 −32.3 to 15.6 0.367 
 Physical activity level 1.2 −5.3 to 8.2 0.719 −0.8 −7.5 to 6.4 0.827 
 Current cigarette smoker 5.2 −15.0 to 30.3 0.638 – – – 
 Diet score 0.0 −0.2 to 0.1 0.781 – – – 
 Weekly alcohol consumption (sqrt units) −0.4 −3.8 to 3.1 0.820 −1.2 −4.8 to 2.4 0.500 
 Body mass index (kg/m2−0.7 −2.1 to 0.7 0.315 −0.8 −2.2 to 0.7 0.310 
 Waist/hip ratio −12.1 −73.9 to 195.8 0.835 −16.3 −76.7 to 200.5 0.784 

These models were adjusted for age, gender and deprivation group alone (first column), or additionally for income, diet score and current smoking (second column).

Moreover, manual workers had 24% (95% CI 8 to 38; P = 0.005) lower levels of methylated DNA in comparison with non-manual workers, independently of age group, gender and deprivation group (Table 2, Figure 1B). Social class, together with age group, gender and SES, explained 6.6% of the variation (adjusted R2). A trend was observed between the number of years spent in education and global DNA methylation content, with each additional year of education being associated with 2.4% (95% CI −0.1 to 5.0; P = 0.052) greater global DNA methylation content (Table 2, Figure 1C). None of the other SES or lifestyle factors investigated had an influence on overall epigenetic status when considered independently of each other and in combination with age, gender and deprivation status.

To further investigate the role of SES confounding factors, regression models were adjusted to include income, diet score and current smoking status. As previously, a decrease in global DNA methylation content (27%) was noted within the manual workers group. However, the difference in average global DNA methylation content between deprived and affluent participants was substantially reduced, with deprived participants having 11% (95% CI −13 to 31; P = 0.334) lower global DNA methylation content. The attenuation of the deprivation effect was further investigated by adjusting separately for income, diet and smoking status (in addition to age, gender and deprivation), which showed that this attenuation was entirely due to adjusting for income (95% CIs of the deprivation effect: −15% to 28% adjusted for income; 3 to 30% adjusted for diet; 4 to 32% adjusted for smoking status). None of the other factors investigated was associated with global DNA methylation content.

Association between global DNA methylation and biomarkers

Linear regression analyses were performed to investigate associations between biochemical parameters and global DNA methylation content. As before, all analyses were adjusted for age, gender and deprivation group. The list of biomarkers analysed is displayed in Table 3.

Table 3

Percentage difference (95% CI) in biomarker level associated with an increase of one SD in log DNA methylation. These models were adjusted for age, gender and deprivation group alone (first column), or additionally for income, diet score and current smoking (second column)

 Predicted per cent difference in biomarker level, adjusted for:
 
 Age, gender, deprivation group
 
Age, gender, deprivation group, income, diet score, current smoking
 
Outcomes Per cent difference 95% CI P-value Per cent difference 95% CI P-value 
Systolic BP (mmHg) −0.1 −1.7 to 1.7 0.952 −0.3 −2.2 to 1.5 0.725 
Diastolic BP (mmHg) −0.2 −1.9 to 1.5 0.801 −0.3 −2.1 to 1.6 0.772 
Cholesterol (mmol/l) −0.1 −2.9 to 2.7 0.931 0.1 −2.9 to 3.2 0.933 
HDL cholesterol (mmol/l) 0.2 −3.5 to 4.1 0.906 −0.1 −4.0 to 4.1 0.977 
LDL cholesterol (mmol/l) 0.3 −3.7 to 4.5 0.881 1.1 −3.2 to 5.7 0.617 
Triglycerides (mmol/l) −0.5 −7.4 to 6.9 0.888 −1.5 −8.7 to 6.4 0.704 
Glucose (mmol/l) −0.9 −3.0 to 1.3 0.438 −1.8 −4.2 to 0.6 0.144 
Insulin (mU/l) 0.2 −8.4 to 9.7 0.958 1.2 −8.3 to 11.6 0.812 
HOMA – IR 1.3 −8.2 to 11.9 0.795 1.8 −8.6 to 13.5 0.742 
C-reactive protein (mg/l) −10.6 −22.4 to 3.0 0.120 −13.7 −26.0 to 0.7 0.060 
Interleukin 6 (pg/ml) −8.7 −15.5 to −1.4 0.021 −8.5 −16.0 to −0.3 0.042 
Intercellular adhesion molecule 1 (ng/ml) 0.4 −3.0 to 3.9 0.817 −0.2 −3.5 to 3.1 0.890 
Fibrinogen (g/l) −3.1 −5.7 to −0.5 0.020 −3.0 −5.8 to −0.2 0.036 
von Willebrand factor (IU/dl) 0.6 −3.2 to 4.6 0.752 0.1 −4.1 to 4.5 0.956 
D-dimer (ng/ml) −6.5 −13.4 to 0.9 0.083 −4.6 −12.0 to 3.4 0.250 
Mean carotid IMT (mm) −1.7 −4.1 to 0.7 0.159 −1.8 −4.4 to 0.8 0.167 
 Predicted per cent difference in biomarker level, adjusted for:
 
 Age, gender, deprivation group
 
Age, gender, deprivation group, income, diet score, current smoking
 
Outcomes Per cent difference 95% CI P-value Per cent difference 95% CI P-value 
Systolic BP (mmHg) −0.1 −1.7 to 1.7 0.952 −0.3 −2.2 to 1.5 0.725 
Diastolic BP (mmHg) −0.2 −1.9 to 1.5 0.801 −0.3 −2.1 to 1.6 0.772 
Cholesterol (mmol/l) −0.1 −2.9 to 2.7 0.931 0.1 −2.9 to 3.2 0.933 
HDL cholesterol (mmol/l) 0.2 −3.5 to 4.1 0.906 −0.1 −4.0 to 4.1 0.977 
LDL cholesterol (mmol/l) 0.3 −3.7 to 4.5 0.881 1.1 −3.2 to 5.7 0.617 
Triglycerides (mmol/l) −0.5 −7.4 to 6.9 0.888 −1.5 −8.7 to 6.4 0.704 
Glucose (mmol/l) −0.9 −3.0 to 1.3 0.438 −1.8 −4.2 to 0.6 0.144 
Insulin (mU/l) 0.2 −8.4 to 9.7 0.958 1.2 −8.3 to 11.6 0.812 
HOMA – IR 1.3 −8.2 to 11.9 0.795 1.8 −8.6 to 13.5 0.742 
C-reactive protein (mg/l) −10.6 −22.4 to 3.0 0.120 −13.7 −26.0 to 0.7 0.060 
Interleukin 6 (pg/ml) −8.7 −15.5 to −1.4 0.021 −8.5 −16.0 to −0.3 0.042 
Intercellular adhesion molecule 1 (ng/ml) 0.4 −3.0 to 3.9 0.817 −0.2 −3.5 to 3.1 0.890 
Fibrinogen (g/l) −3.1 −5.7 to −0.5 0.020 −3.0 −5.8 to −0.2 0.036 
von Willebrand factor (IU/dl) 0.6 −3.2 to 4.6 0.752 0.1 −4.1 to 4.5 0.956 
D-dimer (ng/ml) −6.5 −13.4 to 0.9 0.083 −4.6 −12.0 to 3.4 0.250 
Mean carotid IMT (mm) −1.7 −4.1 to 0.7 0.159 −1.8 −4.4 to 0.8 0.167 

HDL, high density lipoprotein; LDL, low density lipoprotein; HOMA-IR, homeostatic model assesment-insulin resistance.

IL-6 (pg/ml) and fibrinogen (g/l) levels were inversely associated with global DNA methylation content. A one-SD increase in log DNA methylation was associated with a 9% (95% CI 1 to 15; P = 0.021) decrease in IL-6 levels and a 3% (95% CI 0 to 6; P = 0.020) decrease in fibrinogen levels (Table 3). Combining global DNA methylation content and a model that adjusts for gender, age group and deprivation group increased the adjusted R2 from 20.3% to 21.8% where the outcome was log IL-6 level and from 6.5% to 8.2% where the outcome was log fibrinogen level.

Intriguingly, the greatest decrease in IL-6 (20% decrease for each additional SD of log global DNA methylation) was observed in the oldest age group (aged 55–64 years; P = 0.003) whereas the largest decrease for fibrinogen (7%) was noted in the youngest group (aged 35–44 years; Table 4); however, there was no evidence to support an interaction between global DNA methylation content and age group. When the regression models were additionally adjusted for the lifestyle factors income, diet score and current smoking status, IL-6 levels decreased by 9% (95% CI 0 to 16; P = 0.042) and fibrinogen levels decreased by 3% (95% CI 0 to 6; P = 0.036); both remained associated with a one-SD increase in global DNA methylation (Table 3). Adding these lifestyle factors slightly improved the models of both IL-6 level (adjusted R2 were 19.9 and 21.1% without and with global DNA methylation, respectively) and fibrinogen level (adjusted R2 were 9.2 and 10.7% without and with global DNA methylation, respectively).

Table 4

Percentage difference (95% CI) in biomarker level associated with an increase of one SD in log DNA methylation

Outcomes Predicted per cent difference (95% CI) in biomarker level, adjusted for:
 
Age group (years), gender, deprivation group
 
Age group (years), gender, deprivation group, income, diet score, current smoking
 
35–44 45–54 55–64 35–44 45–54 55–64 
Cholesterol (mmol/l) −2.1 (−7.2 to 3.4) −2.1 (−5.9 to 1.9) 5.8 (0.2 to 11.7) −1.6 (−7.2 to 4.3) −3.0 (−7.0 to 1.2) 8.7 (2.4 to 15.4) 
P = 0.454 P = 0.289 P = 0.043 P = 0.582 P = 0.156 P = 0.006 
LDL cholesterol (mmol/l) −0.4 (−7.9 to 7.7) −4.5 (−9.8 to 1.1) 10.8 (2.4 to 19.8) 1.1 (−7.0 to 10.0) −4.5 (−10.1 to 1.5) 13.4 (4.1 to 23.5) 
P = 0.921 P = 0.116 P = 0.011 P = 0.794 P = 0.139 P = 0.004 
IL-6 (pg/ml) −8.3 (−20.9 to 6.3) −1.8 (−12.0 to 9.6) −20.7 (−31.7 to −7.8) −6.3 (−20.5 to 10.4) −2.9 (−13.9 to 9.5) −20.9 (−33.4 to −6.2) 
P = 0.247 P = 0.741 P = 0.003 P = 0.433 P = 0.626 P = 0.007 
Fibrinogen (g/l) −7.3 (−12.0 to −2.3) −1.4 (−5.1 to 2.4) −2.0 (−6.9 to 3.1) −6.4 (−11.6 to −0.9) −1.4 (−5.3 to 2.7) −2.9 (−8.3 to 2.7) 
P = 0.005 P = 0.451 P = 0.429 P = 0.024 P = 0.501 P = 0.300 
Outcomes Predicted per cent difference (95% CI) in biomarker level, adjusted for:
 
Age group (years), gender, deprivation group
 
Age group (years), gender, deprivation group, income, diet score, current smoking
 
35–44 45–54 55–64 35–44 45–54 55–64 
Cholesterol (mmol/l) −2.1 (−7.2 to 3.4) −2.1 (−5.9 to 1.9) 5.8 (0.2 to 11.7) −1.6 (−7.2 to 4.3) −3.0 (−7.0 to 1.2) 8.7 (2.4 to 15.4) 
P = 0.454 P = 0.289 P = 0.043 P = 0.582 P = 0.156 P = 0.006 
LDL cholesterol (mmol/l) −0.4 (−7.9 to 7.7) −4.5 (−9.8 to 1.1) 10.8 (2.4 to 19.8) 1.1 (−7.0 to 10.0) −4.5 (−10.1 to 1.5) 13.4 (4.1 to 23.5) 
P = 0.921 P = 0.116 P = 0.011 P = 0.794 P = 0.139 P = 0.004 
IL-6 (pg/ml) −8.3 (−20.9 to 6.3) −1.8 (−12.0 to 9.6) −20.7 (−31.7 to −7.8) −6.3 (−20.5 to 10.4) −2.9 (−13.9 to 9.5) −20.9 (−33.4 to −6.2) 
P = 0.247 P = 0.741 P = 0.003 P = 0.433 P = 0.626 P = 0.007 
Fibrinogen (g/l) −7.3 (−12.0 to −2.3) −1.4 (−5.1 to 2.4) −2.0 (−6.9 to 3.1) −6.4 (−11.6 to −0.9) −1.4 (−5.3 to 2.7) −2.9 (−8.3 to 2.7) 
P = 0.005 P = 0.451 P = 0.429 P = 0.024 P = 0.501 P = 0.300 

These models were adjusted for age group, gender and deprivation group alone (first column), or additionally for income, diet score and current smoking (second column).

Analyses were performed to establish which individual factors had an impact on IL-6 and fibrinogen expression. The association between increasing global DNA methylation content and decreasing IL-6 was strongest when analyses were adjusted for age and gender, a one-SD increase in log DNA methylation content being associated with a 13% decrease in IL-6 levels. This association was attenuated (P = 0.007) to 9% with further adjustment for deprivation (Table 5).

Table 5

Percentage change (95% CI) in IL-6 (pg/ml) and fibrinogen (g/l) associated with an increase of one SD in log DNA methylation, predicted from linear regression models with no adjustment and with sequential addition of adjustment covariates (age, gender, deprivation group, income, diet score and current smoking)

Adjustment covariates Estimate (95% CI), P-value P-value for change in estimate 
IL-6 (pg/ml)   
None −9.9 (−17.8 to −1.3), 0.026 – 
Age + gender −12.9 (−20.5 to −4.6), 0.003 0.017 
Age + gender + deprivation −8.8 (−16.2 to −0.7), 0.034 0.007 
Age + gender + deprivation + income −8.7 (−16.2 to −0.5), 0.038 0.828 
Age + gender + deprivation + income + diet −8.5 (−16.0 to −0.3), 0.043 0.681 
Age + gender + deprivation + income + diet + current smoking −8.5 (−16.0 to −0.3), 0.042 0.896 
Fibrinogen (g/l)   
None −3.3 (−6.1, to −0.4), 0.024 – 
Age + gender −4.0 (−6.8, to −1.2), 0.006 0.074 
Age + gender + deprivation −3.4 (−6.1, to −0.5), 0.022 0.019 
Age + gender + deprivation + income −3.1 (−5.8, to −0.3), 0.033 0.337 
Age + gender + deprivation + income + diet −3.0 (−5.8, to −0.1), 0.040 0.475 
Age + gender + deprivation + income + diet + current smoking −3.0 (−5.8, to −0.2), 0.036 0.694 
Adjustment covariates Estimate (95% CI), P-value P-value for change in estimate 
IL-6 (pg/ml)   
None −9.9 (−17.8 to −1.3), 0.026 – 
Age + gender −12.9 (−20.5 to −4.6), 0.003 0.017 
Age + gender + deprivation −8.8 (−16.2 to −0.7), 0.034 0.007 
Age + gender + deprivation + income −8.7 (−16.2 to −0.5), 0.038 0.828 
Age + gender + deprivation + income + diet −8.5 (−16.0 to −0.3), 0.043 0.681 
Age + gender + deprivation + income + diet + current smoking −8.5 (−16.0 to −0.3), 0.042 0.896 
Fibrinogen (g/l)   
None −3.3 (−6.1, to −0.4), 0.024 – 
Age + gender −4.0 (−6.8, to −1.2), 0.006 0.074 
Age + gender + deprivation −3.4 (−6.1, to −0.5), 0.022 0.019 
Age + gender + deprivation + income −3.1 (−5.8, to −0.3), 0.033 0.337 
Age + gender + deprivation + income + diet −3.0 (−5.8, to −0.1), 0.040 0.475 
Age + gender + deprivation + income + diet + current smoking −3.0 (−5.8, to −0.2), 0.036 0.694 

P-values for change in the DNA methylation effect estimate due to each additional adjustment were estimated from 10 000 bootstrap samples. Due to missing data in some covariates, all models were fitted on a reduced data set of 217 participants.

No further significant attenuation resulted from additional adjustment for income, diet and smoking status. Similarly, the strongest association (−4.0%) between fibrinogen and global DNA methylation content was observed when analyses were adjusted for age and gender alone. This association was attenuated (P = 0.019) to −3.4% by adjusting for deprivation status, with no further significant attenuation resulting from additional adjustment for income, diet and smoking status (Table 5).

Additionally, we found tentative evidence that the strength of association between global DNA methylation content and cholesterol levels varied between age groups (interaction P = 0.011 for LDL cholesterol and P = 0.055 for total cholesterol). The association was strongest in the oldest age group (55–64 years), with a one-SD increase in global DNA methylation being associated with an 11% (95% CI 2 to 20; P = 0.011) increase in LDL cholesterol and a 6% (95% CI 0 to 12; P = 0.043) increase in total cholesterol (Table 3). These results are promising but are not sufficient to draw strong conclusions. A larger sample would be required to confirm these preliminary results for cholesterol and LDL cholesterol.

Discussion

We have observed that SES, lifestyle factors and a number of emerging biomarkers for CVD are associated with epigenetic status as measured by global DNA methylation content. Previous investigations seeking to determine such associations have yielded equivocal results, possibly as a result of difficulties in obtaining age-matched samples from SES groups.16 Our study has demonstrated that global DNA hypomethylation was associated with the most deprived group of participants, when compared with the least deprived. This observation is further supported when other markers of SES are examined. Decreased global DNA methylation content was observed in manual workers, as opposed to non-manual workers and a trend was observed with educational status. The extent of DNA hypomethylation in the most deprived group of participants is intriguing. Such global hypomethylation could be reflective of environmental exposures and/or diet during life, or a direct consequence of developmental programming in utero, or a combination of both. Notably, adjusting for diet does not weaken associations with global methylation status, suggesting that in utero programming or environmental factors would be causative for the observed global hypomethylation associated with lower SES. However, it must also be considered that poor reporting of diet may also play a role. This cohort does not present sufficient data points to determine whether fetal programming plays a major role, this would require further investigation in a larger study.

A link with in utero programming is an attractive explanation, as there is some evidence supporting the effect of a poor childhood environment and an increased risk of cardiovascular disorders. Indeed, in utero epigenetic programming has been linked to the development of obesity, arteriosclerosis and diabetes and may be related to maternal diet and/or ‘fuel economy’.17,18 This is pertinent to a Glasgow-based cohort, where persistence of socio-economic deprivation can be invoked to explain specific global DNA hypomethylation, with consequent effects on health in adult life. Such a scenario is consistent with the high prevalence of CVDs observed within Glasgow being associated with socio-economic and lifestyle factors.

Our study supports a relationship between global DNA methylation and emerging biomarkers of CVD and systemic inflammation. We have observed an association between hypomethylation and increased expression of biomarkers of CVD, such as IL-6. Conversely, a relative elevation of global DNA methylation content was associated with higher total cholesterol and LDL levels in the oldest group (55–64 years). This association was further enhanced when the analysis was adjusted for deprivation status, income, diet and smoking status. These findings are supported by the results of previous studies, using the same cohort, whereby higher total cholesterol and LDL levels were shown to be associated with deprivation status and biological age.14,19 Other studies have suggested that DNA hypomethylation may be a result of CVD risk factors,20 whereas yet others suggest it is a causative agent.21,22 Several studies have suggested the mechanism for the latter may be hyper-homocysteinemia.23,24

We also observed that the association of methylation content with age group did not appear to reflect a decline of methylation content with increasing age, which has been reported by other groups,25,26 as the lowest average methylation levels were found in the middle age group (45–54 years) and the highest in the oldest age group (55–64 years). This is intriguing and requires further study. It may reflect the influence of SES confounding factors and survivor effects in this particular cohort. Alternatively, this may be due to other large environmental events that took place at specific points in history, perhaps early childhood for each group; such events could include disease or environmental disasters.

At a mechanistic level, our observations may be explained by lipid- and lipoprotein-induced DNA instability and chromatin remodelling, through changes in genomic DNA methylation content resulting in aberrant gene expression. Recent studies have revealed that free intracellular cholesterol can associate with chromatin as a complex with sphingomyelin, other lipids and an as yet unidentified protein.27 A study conducted by Lund28 concluded that different lipoprotein profiles can affect global DNA methylation status in mice. The results of two studies from the same group suggest that the pro-atherogenic effect of lipoproteins is associated with DNA hypermethylation, before atherosclerosis onset, followed by DNA hypomethylation in the advanced stages of arteriosclerosis, probably as a result of intensive proliferative activity of cells.

It is also possible that these two processes operate in parallel within the same timeline and DNA hypermethylation masks the DNA hypomethylation processes in arteriosclerotic lesions.28,29 Additionally, an increase of LDL fraction can be accompanied by an increase in IL-6 and fibrinogen, further promoting oxidative stress and inflammatory cellular responses.30,31

The relationship between IL-6 and global DNA methylation status observed in our study is consistent with recent studies suggesting a direct link between IL-6 levels and DNA methyltransferase and demethylase activity.13 In particular, IL-6 may be involved in the regulation of HDNMT (human DNA methyltransferase);32 influencing expression and activity of this enzyme would lead to differences in global DNA methylation status and thus correlate with health status.

Additionally, increasing fibrinogen levels were also associated with a decrease in global methylation which became even more apparent when age and gender were considered. However, the overall level of global methylation rose again when other factors (smoking status, diet, income and deprivation) were applied. Taken in context, this alteration could be a result of IL-6 regulation of the fibrinogen promoter; further investigation is necessary to confirm this.

Our observations provide a potentially novel explanation for accelerated age-related disease onset in Glasgow. They present an interesting avenue for the exploration of epigenetic changes in relation to ageing and health outcomes in larger and life-course cohorts. Such an approach will be required to establish if reduced methylation content is associated with SES at birth and leads to health inequalities later in life.

Supplementary Data

Supplementary Data are available at IJE online.

Funding

This work was funded by Glasgow Centre for Population Health, Glasgow, Scotland, UK.

Acknowledgements

No part of this paper has been submitted for publication or published elsewhere. All authors have met the requirements for authorship and have read and approved the manuscript in this form.

Conflict of interest: None declared.

KEY MESSAGE

  • This study presents novel findings in a unique cohort characterised by extremes of the socio-economic scale. The data links global DNA methylation status to socio-economic factors. Additionally, the data suggests a relationship between global methylation status and biomarkers for cardiovascular disease and inflammation potentially providing a link between deprivation and increased risk for cardiovascular disease.

References

1
Thornburg
KL
Shannon
J
Thuillier
P
Turker
MS
In utero life and epigenetic predisposition for disease
Adv Genet
 , 
2010
, vol. 
71
 (pg. 
57
-
78
)
2
Chmielewski
M
Lindholm
B
Stenvinkel
P
Ekstrom
JT
The role of epigenetics in kidney diseases
Prilozi
 , vol. 
l32
 (pg. 
45
-
54
)
3
Portela
A
Esteller
M
Epigenetic modifications and human disease
Nat Biotechnol
 , 
2010
, vol. 
28
 (pg. 
1057
-
68
)
4
Thompson
RF
Atzmon
G
Gheorghe
C
, et al.  . 
Tissue-specific dysregulation of DNA methylation in aging
Aging Cell
 , 
2010
, vol. 
9
 (pg. 
506
-
18
)
5
Mathers
JC
Strathdee
G
Relton
CL
Induction of epigenetic alterations by dietary and other environmental factors
Adv Genet
 , 
2010
, vol. 
71
 (pg. 
3
-
39
)
6
Champagne
FA
Curley
JP
Epigenetic mechanisms mediating the long-term effects of maternal care on development
Neurosci Biobehav Rev
 , 
2009
, vol. 
33
 (pg. 
593
-
600
)
7
Deans
KA
Bezlyak
V
Ford
I
, et al.  . 
Differences in atherosclerosis according to area level socioeconomic deprivation: cross sectional, population based study
BMJ
 , 
2009
, vol. 
339
 pg. 
b4170
 
8
Backdahl
L
Bushell
A
Beck
S
Inflammatory signalling as mediator of epigenetic modulation in tissue-specific chronic inflammation
Int J Biochem Cell Biol
 , 
2009
, vol. 
41
 (pg. 
176
-
84
)
9
Stenvinkel
P
Karimi
M
Johansson
S
, et al.  . 
Impact of inflammation on epigenetic DNA methylation – a novel risk factor for cardiovascular disease?
J Intern Med
 , 
2007
, vol. 
261
 (pg. 
488
-
99
)
10
Carrero
JJ
Stenvinkel
P
Fellstrom
B
, et al.  . 
Telomere attrition is associated with inflammation, low fetuin-A levels and high mortality in prevalent haemodialysis patients
J Intern Med
 , 
2008
, vol. 
263
 (pg. 
302
-
12
)
11
Braconi
C
Huang
N
Patel
T
MicroRNA-dependent regulation of DNA methyltransferase-1 and tumor suppressor gene expression by interleukin-6 in human malignant cholangiocytes
Hepatology
 , 
2010
, vol. 
51
 (pg. 
881
-
90
)
12
Hodge
DR
Cho
E
Copeland
TD
, et al.  . 
IL-6 enhances the nuclear translocation of DNA cytosine-5-methyltransferase 1 (DNMT1) via phosphorylation of the nuclear localization sequence by the AKT kinase
Cancer Genomics Proteomics
 , 
2007
, vol. 
4
 (pg. 
387
-
98
)
13
Hodge
DR
Xiao
W
Clausen
PA
Heidecker
G
Szyf
M
Farrar
WL
Interleukin-6 regulation of the human DNA methyltransferase (HDNMT) gene in human erythroleukemia cells
J Biol Chem
 , 
2001
, vol. 
276
 (pg. 
39508
-
11
)
14
Deans
KA
Bezlyak
V
Ford
I
, et al.  . 
Differences in atherosclerosis according to area level socioeconomic deprivation: cross sectional, population based study
BMJ
 , 
2009
, vol. 
339
 pg. 
b4170
 
15
Velupillai
YN
Packard
CJ
Batty
GD
, et al.  . 
Psychological, social and biological determinants of ill health (pSoBid): study protocol of a population-based study
BMC Public Health
 , 
2008
, vol. 
8
 pg. 
126
 
16
Zhu
ZM
Hou
L
Bollati
V
, et al.  . 
Predictors of global methylation levels in blood DNA of healthy subjects: a combined analysis
Int J Epidemiol
 , 
2012
, vol. 
41
 (pg. 
126
-
39
)
17
Godfrey
KM
Gluckman
PD
Hanson
MA
Developmental origins of metabolic disease: life course and intergenerational perspectives
Trends Endocrinol Metab
 , vol. 
21
 (pg. 
199
-
205
)
18
Burdge
GC
Lillycrop
KA
Jackson
AA
Nutrition in early life, and risk of cancer and metabolic disease: alternative endings in an epigenetic tale?
Br J Nutr
 , 
2009
, vol. 
101
 (pg. 
619
-
30
)
19
Shiels
PG
Improving precision in investigating aging: why telomeres can cause problems
J Gerontol A Biol Sci Med Sci
 , 
2010
, vol. 
65
 (pg. 
789
-
91
)
20
Baccarelli
A
Rienstra
M
Benjamin
EJ
Cardiovascular epigenetics: basic concepts and results from animal and human studies
Circ Cardiovasc Genet
 , 
2010
, vol. 
3
 (pg. 
567
-
73
)
21
Duthie
SJ
Epigenetic modifications and human pathologies: cancer and CVD
Proc Nutr Soc
 , 
2010
, vol. 
70
 (pg. 
47
-
56
)
22
Ordovas
JM
Smith
CE
Epigenetics and cardiovascular disease
Nat Rev Cardiol
 , 
2010
, vol. 
7
 (pg. 
510
-
19
)
23
Cacciapuoti
F
Hyper-homocysteinemia: a novel risk factor or a powerful marker for cardiovascular diseases? Pathogenetic and therapeutical uncertainties
J Thromb Thrombolysis
 , 
2011
, vol. 
32
 (pg. 
82
-
88
)
24
Kim
M
Long
TI
Arakawa
K
Wang
R
Yu
MC
Laird
PW
DNA methylation as a biomarker for cardiovascular disease risk
PLoS One
 , 
2010
, vol. 
5
 pg. 
e9692
 
25
Liu
L
Wylie
RC
Andrews
LG
Tollefsbol
TO
Aging, cancer and nutrition: the DNA methylation connection
Mech Ageing Dev
 , 
2003
, vol. 
124
 (pg. 
989
-
98
)
26
Fraga
MF
Agrelo
R
Esteller
M
Cross-talk between aging and cancer: the epigenetic language
Ann N Y Acad Sci
 , 
2007
, vol. 
1100
 (pg. 
60
-
74
)
27
Zaina
S
Dossing
KBV
Lindholm
MW
Lund
G
Chromatin modification by lipids and lipoprotein components: an initiating event in atherogenesis?
Curr Opin Lipidol
 , 
2005
, vol. 
16
 (pg. 
549
-
53
)
28
Lund
G
Andersson
L
Lauria
M
, et al.  . 
DNA methylation polymorphisms precede any histological sign of atherosclerosis in mice lacking apolipoprotein E
J Biol Chem
 , 
2004
, vol. 
279
 (pg. 
29147
-
54
)
29
Zaina
S
Lindholm
MW
Lund
G
Nutrition and aberrant DNA methylation patterns in atherosclerosis: more than just hyperhomocysteinemia?
J Nutr
 , 
2005
, vol. 
135
 (pg. 
5
-
8
)
30
Sarwar
N
Thompson
AJ
Di Angelantonio
E
Markers of inflammation and risk of coronary heart disease
Dis Markers
 , 
2009
, vol. 
26
 (pg. 
217
-
25
)
31
Marz
W
Scharnagl
H
Winkler
K
, et al.  . 
Low-density lipoprotein triglycerides associated with low-grade systemic inflammation, adhesion molecules, and angiographic coronary artery disease: the Ludwigshafen Risk and Cardiovascular Health study
Circulation
 , 
2004
, vol. 
110
 (pg. 
3068
-
74
)
32
Thaler
R
Agsten
M
Spitzer
S
, et al.  . 
Homocysteine suppresses the expression of the collagen cross-linker lysyl oxidase involving IL-6, Fli1, and epigenetic DNA methylation
J Biol Chem
 , 
2010
, vol. 
286
 (pg. 
5578
-
88
)