Physical activity is associated with a reduced risk of many chronic diseases, including cardiovascular disease, type 2 diabetes, and cancer (1,2). Sedentary behavior (sitting or lying while awake, with a low-energy expenditure) is also associated with an increased risk of cardiovascular disease, diabetes, and cancer (3,4). Importantly, sedentary behavior is different to physical inactivity, which is defined as an insufficient level of moderate- to vigorous-intensity physical activity (5). The contribution of sedentary behavior to chronic disease risk appears to be largely independent of moderate-to-vigorous physical activity (6). Understanding the biological mechanisms through which physical activity and sedentary behavior affect health can provide insight into the causal nature of associations and inform the development of mechanism-tailored prevention strategies.
One mechanism through which physical activity may influence health is by modulating gene expression via DNA methylation, an epigenetic mechanism (7). DNA methylation is the addition of a methyl group to a cytosine base neighboring a guanine base (CpG site) (8) and plays a critical role in normal human development (9). The epigenome changes over time (10), with ageing (10), and has been linked to genetic, environmental, and lifestyle factors (11,12). Aberrant DNA methylation has been associated with the development of inactivity-associated diseases, including cancer, Alzheimer’s disease, and type 2 diabetes (9,13). Thus, DNA methylation may present a mechanistic pathway between physical activity, sedentary behavior, and health.
Short- and long-term exercise interventions can result in DNA methylation changes in skeletal muscle, heart and brain tissues, and blood (7). Observational studies have mostly focused on DNA methylation in blood from healthy adults, some reporting associations between self-reported physical activity and methylation of cardiovascular and metabolic disease-related genes (14,15), microRNA-encoding genes (16), and global methylation (17,18). However, other observational studies observed no associations between physical activity and blood DNA methylation of specific genes, global methylation, or methylation-based ageing measures (19–22), possibly (partly) due to the generally limited sample sizes (n < 300).
To our knowledge, no genomewide study of DNA methylation (also called epigenome-wide association studies [EWAS]) and physical activity has been conducted to date. The genomewide scanning of thousands of CpG sites in an EWAS can identify novel loci and potential mechanistic pathways. There has also been little research on sedentary behavior and DNA methylation. One study of 248 elderly individuals reported associations between sedentary behavior and methylation-based age acceleration measures, which were nonsignificant after adjustment for multiple testing (22).
We performed an EWAS of physical activity and television viewing time using data from the Melbourne Collaborative Cohort Study (MCCS). Cross-sectional associations between physical activity and television viewing time and DNA methylation in peripheral blood at individual CpG sites across the genome were evaluated using 5513 samples taken at baseline and 1249 at follow-up, a median of 11 yr later. We also performed a longitudinal analysis to assess whether changes in leisure-time physical activity between these time points were associated with methylation changes.
Study Design and Participants
The MCCS is a prospective cohort study (23). Between 1990 and 1994, 41,513 residents of the Melbourne metropolitan area were recruited (59% women; mean age, 55 yr), and 28,240 participants participated in follow-up measurements between 2003 and 2007. The study was approved by the Cancer Council Victoria’s Human Research Ethics Committee, and all participants provided informed consent.
The current study was based on baseline (N = 5513) and follow-up data (N = 1249) from MCCS participants included in one of seven previously conducted nested case–control studies of blood DNA methylation and cancer (Fig. 1). All subjects selected as cases were free of cancer when relevant blood and questionnaire data were collected. Controls were selected using incidence density sampling and matched to cases on sex, year of birth (for colorectal cancer study, year of baseline attendance), country of birth (Australia/NZ/UK/other, Greece, and Italy), sample type (dried blood spots, peripheral blood mononuclear cells, and buffy coats), and smoking status (only lung cancer study).
Blood samples were stored as peripheral blood mononuclear cells, buffy coat samples, or dried blood spots on Guthrie cards. Samples from each study (seven case–control studies and longitudinal reanalysis at follow-up for a subset of controls; Fig. 1) were assayed in nonoverlapping periods. A detailed description of methods used for DNA extraction, DNA methylation assessment, and data processing is provided in the supplemental text appendix (see Text, Supplemental Digital Content 1, Description of methods for DNA extraction, DNA methylation assessment and data processing, https://links.lww.com/MSS/B426). Briefly, DNA methylation was measured using the Illumina Infinium HumanMethylation450K BeadChip (HM450K) array (Illumina, Inc., San Diego, CA), and methylation M-values were calculated for each CpG site.
Physical Activity and Television Viewing Time
At baseline, physical activity was assessed by three questionnaire items about the frequency (times per week) of moderate and vigorous exercise and walking. Participants were classified as doing some (≥1 time per week) or no (0 time per week) leisure-time physical activity (moderate and/or vigorous exercise). Similarly, based on the frequency of moderate and vigorous exercise and walking, participants were classified as doing some or no total physical activity.
At follow-up, the long version of the International Physical Activity Questionnaire (IPAQ) was used to assess self-reported frequency and duration of moderate and vigorous leisure-time physical activity and walking. The standard IPAQ recall period of 7 d was changed to 3 months. Metabolic equivalent of task (MET)-hours per week was calculated for leisure-time (moderate and vigorous physical activity) and total physical activity (including also walking), by multiplying the hours per week spent in each activity by their MET value and summing these. Moderate and vigorous physical activities were assigned 4 and 8 METs, respectively, according to the IPAQ scoring protocol. The MET value for walking was based on the reported pace, i.e., slow, 2.8 METs; normal/missing, 3.3 METs; brisk, 4.3 METs; and very brisk, 5.0 METs.
At follow-up, self-reported television viewing time was assessed by questionnaire items asking about total time spent watching television on week and weekend days (h·d−1). A minority of participants completed instead the standard IPAQ items of total sitting time on week and weekend days. As reported previously (24), we truncated time spent watching television or sitting to a maximum of 18 h·d−1.
Body mass index (BMI, kg·m−2) was calculated from measured weight and height. Smoking status (current, former, and never), cigarettes per day, and time since quitting were self-reported at baseline; these variables were reassessed at follow-up only for those who reported current or former smoking at baseline. Alcohol consumption (g·d−1) was based on beverage-specific frequency and quantity questions within a self-administered food frequency questionnaire. Socioeconomic position was based on the Socio-Economic Indexes for Areas (SEIFA)–Index of Disadvantage, divided into population-based deciles.
The reliability of methylation measures was assessed by calculating intraclass correlation coefficients, as previously described (25).
Associations between each physical activity measure and DNA methylation M-values at each CpG site were assessed by linear mixed regression. We studied cross-sectional associations at both baseline and follow-up because the baseline sample included more participants (i.e., greater power), whereas physical activity measures at follow-up were continuous and measured with a validated questionnaire (i.e., more information and probably less measurement error). Physical activity measures included “some” or “no” leisure-time physical activity at baseline and MET-hours per week of total and leisure-time physical activity at follow-up (log-transformed to better approximate a normal distribution). We observed a large peak at 0 MET·h·wk−1 in the distribution of both variables. Therefore, we included in the regression model both the continuous variable as well as a dichotomous variable indicating ≥1 or 0 MET·h·wk−1 and evaluated overall significance using a likelihood ratio test. Models were adjusted for age (yr, continuous), sex, country of birth (Australia/NZ/UK/other, Greece, Italy), SEIFA deciles (continuous), smoking status (never; former, quit ≥15 yr ago; former, quit <15 yr ago; current, <20 cigarettes per day; current, ≥20 cigarettes per day), alcohol consumption (ln-transformed, continuous), sample type (dried blood spots, peripheral blood mononuclear cells, buffy coats), and estimated white blood cell composition (proportion of CD4+ T cells, CD8+ T cells, NK cells, B cells, granulocytes, and monocytes, all continuous and calculated using a modified version of the Houseman algorithm) (26) as fixed effects, and study, plate, and chip as random effects. For the follow-up analysis, study (urothelial cell carcinoma study vs longitudinal reanalysis; Fig. 1) was included as a fixed effect; sample type was not included as all urothelial cell carcinoma samples were buffy coats and all other samples were from dried blood spots.
Linear mixed regression was also applied to analyze cross-sectional associations of television viewing time (h·d−1, continuous) with methylation M-values using follow-up data. Models were adjusted for the same factors as in the follow-up physical activity analysis, and additionally for total physical activity (MET·h·wk−1). Because for 149 participants data were available for total sitting time instead of television viewing time, we performed an additional analysis in a larger study sample using quintiles of television viewing or sitting time as explanatory variable (continuous).
To correct for multiple testing, we applied a statistical significance threshold of P < 1.0 × 10−7. The inflation of test statistics was assessed by generating quantile–quantile (Q–Q) plots and by calculating the genomic inflation factor (lambda) (27).
We performed a sensitivity analysis with additional adjustment for BMI. Further, we performed the baseline analysis separately for case and control participants and compared the regression coefficients for evidence of collider bias because of the sampling of participants based on their subsequent cancer diagnosis (cases).
Linear mixed regression was applied to assess whether a change in doing some versus no leisure-time physical activity between baseline and follow-up was associated with changes in methylation at CpG sites with evidence of cross-sectional association at P < 1.0 × 10−5 with leisure-time physical activity. For total physical activity, the number of participants in some change categories was too limited to perform a longitudinal analysis (see Table, Supplemental Digital Content 2, Number of participants reporting some versus no leisure-time and total physical activity at baseline and follow-up, https://links.lww.com/MSS/B427). We performed two separate longitudinal analyses comparing 1) participants who reported no leisure-time physical activity at baseline and some at follow-up (i.e., increased activity), with those who reported none at both time points, and 2) participants who reported some leisure-time physical activity at baseline and none at follow-up (i.e., decreased activity) with those who reported some at both time points. Because baseline and follow-up samples were assayed at different time points, ComBat was applied to adjust for batch effects (12). Regression models were adjusted for baseline age and change in age (continuous), sex, country of birth, baseline level and change in SEIFA deciles (continuous), smoking status at baseline (as defined previously) and follow-up (current; noncurrent), baseline level and change in log-transformed alcohol consumption (continuous), and baseline methylation M-value as fixed effects, and baseline study as a random effect. A sensitivity analysis was performed without adjustment for baseline methylation M-value, as baseline adjustment may introduce bias in the analysis of change (28).
The “gometh” function of the R package missMethyl was used to map CpG sites with the strongest evidence of association (P < 1.0 × 10−4) to genes and to evaluate the overrepresentation of these genes within the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways.
Statistical analyses were performed using R (version 3.4.2), and P values were two-sided.
The majority of included participants were male (68%), and the mean age was 59.2 yr (SD, 7.5 yr) at baseline and 69.1 yr (SD, 8.1 yr) at follow-up (Table 1). At baseline, 48% of participants reported some leisure-time physical activity and 78% some total physical activity, whereas these figures were 42% and 93%, respectively, at follow-up. At follow-up, participants reported a median of 17.3 MET·h·wk−1 in total physical activity (interquartile range [IQR], 6.9–34.7) and 0.0 MET·h·wk−1 in leisure-time physical activity (IQR, 0.0–12.0). The median reported television viewing time was 2.6 h·d−1 (IQR, 1.9–3.5).
At follow-up, we observed a significant cross-sectional association between more total physical activity and a higher level of methylation at cg10266336 in the TSS200 region (i.e., within 200 base pairs upstream from transcription start site) of SAA2 (P = 6.0 × 10−9). Weaker evidence of association (P < 1.0 × 10−5) was found for 14 and 7 nonoverlapping CpG sites with total and leisure-time physical activity, respectively (Table 2; Manhattan plots in Figure S1, Supplemental Digital Content 3, Results of genomewide studies of DNA methylation in peripheral blood in relation to total and leisure-time physical activity, https://links.lww.com/MSS/B428). For total physical activity, smaller P values were generally observed for the dichotomous variable (≥1 vs 0 MET·h·wk−1) than for the continuous variable; no clear pattern was observed for leisure-time physical activity (Table 2). There was no clear pattern in direction of associations among CpG sites, and in some cases, the direction of associations was opposite for the dichotomous and continuous variable. We performed an additional analysis for the CpG sites with P < 1.0 × 10−5 for associations with total physical activity where we only included the dichotomous variable without adjustment for the continuous variable and observed significant associations for ≥1 versus 0 MET·h·wk−1 in total physical activity with higher methylation levels at cg10266336 (P = 6.9 × 10−10) as well as ch.4.134822993R (intergenic probe; P = 2.9 × 10−8). No evidence of associations was observed with P < 1.0 × 10−5 for some versus no leisure-time physical activity at baseline (see Figure, Supplemental Digital Content 4, Results of genomewide study of DNA methylation in peripheral blood relation to doing some versus no leisure-time physical activity at baseline, https://links.lww.com/MSS/B429).
A total of nine and five nonoverlapping CpG sites showed weaker evidence of associations (P < 1.0 × 10−5) with television viewing time and quintiles of television viewing and sitting time, respectively (Table 3; Manhattan plots in Figure, Supplemental Digital Content 5, Results of genomewide studies of DNA methylation in peripheral blood in relation to television viewing time and quintiles of television viewing and sitting time, https://links.lww.com/MSS/B430). For most CpG sites (11 of 14), greater television viewing time was associated with higher methylation levels. The median intraclass correlation coefficient based on follow-up replicates for methylation M-values at CpG sites with P < 1.0 × 10−5 for physical activity and/or television viewing/sitting time variables was 0.07 (IQR, 0.00–0.34). The Q–Q plots indicated no inflation of test statistics (see Figure, Supplemental Digital Content 6, Q–Q plots of genomewide studies of DNA methylation in peripheral blood in relation to tested physical activity and television viewing and sitting time variables, https://links.lww.com/MSS/B431), except for total physical activity (lambda, 1.12). This was not observed in Q–Q plots based on P values for each of the dichotomous and continuous variable (lambda, 0.97 and 1.03, respectively; results not shown).
The regression coefficients obtained from models with and without BMI adjustment were very similar (Pearson’s r > 0.99; see Figure, Supplemental Digital Content 7, Comparison of regression coefficients obtained from linear mixed regression models evaluating associations of tested physical activity and television viewing and sitting time variables with DNA methylation, with and without adjustment for BMI, https://links.lww.com/MSS/B432). We found no evidence for systematic differences between results from the separate analysis of baseline data for cases and controls (see Figure, Supplemental Digital Content 8, Comparison of regression coefficients obtained from linear mixed regression models evaluating associations of some versus no leisure-time physical activity with DNA methylation at baseline in cases versus controls only, https://links.lww.com/MSS/B433).
In total, 121 participants (13%) reported no leisure-time physical activity at baseline, and some leisure-time physical activity at follow-up (i.e., increase), and 264 individuals (27%) reported some leisure-time physical activity at baseline but none at follow-up (i.e., decrease; see Table, Supplemental Digital Content 2, Number of participants reporting some versus no leisure-time and total physical activity at baseline and follow-up, https://links.lww.com/MSS/B427). Of the seven CpG sites with P < 1.0 × 10−5 for cross-sectional associations with leisure-time physical activity at follow-up, associations (P < 0.05) were observed between changes in leisure-time physical activity (increase or decrease) and changes in methylation for cg27283993 (gene body of KLF6), cg05896042 (5′ untranslated region of RAB3D), cg05467523 (TSS200 of SV2A), and cg17757602 (intergenic probe) (Table 4), with the directions of associations consistent with the cross-sectional results (Table 2). The regression coefficients in models without adjustment for baseline methylation were similar, albeit less statistically significant (Table 4).
The pathway analysis indicated an overrepresentation (P < 0.05) of the KEGG pathways “ECM–receptor interaction” and “riboflavin metabolism” (physical activity) and “microRNAs in cancer,” “RNA degradation,” and “p53 signaling pathway” (television viewing or sitting time) (Table 5; included CpG sites in Table, Supplemental Digital Content 9, Overview of CpG sites associated with any of the tested physical activity and television viewing/sitting time variables at P < 1.0 × 10−4, that were included within the pathway analysis, https://links.lww.com/MSS/B434.).
In this first EWAS of physical activity and television viewing time, we found a cross-sectional association between total physical activity and methylation at one CpG site, and weaker evidence of association for an additional 14 CpG sites with total physical activity, 7 CpG sites with leisure-time physical activity, and 9 (nonoverlapping) CpG sites with television viewing time. Changes in leisure-time physical activity between baseline and follow-up were associated with methylation changes at four of the seven CpG sites with weaker evidence of cross-sectional associations with leisure-time physical activity. These findings require replication. Nevertheless, our results are in line with observational and intervention research findings that physical activity may affect blood DNA methylation (7) and suggest that television viewing time may also be associated with DNA methylation, independently of moderate to vigorous physical activity.
There is some overlap between our findings and those reported in a gene ontology meta-analysis (29), which summarized results of observational and intervention studies investigating associations of exercise with DNA methylation in humans. For leisure-time physical activity, we found weak evidence of a cross-sectional and a longitudinal association with methylation at cg27283993, annotated to KLF6, a tumor suppressor gene involved in colon cancer development (30). The meta-analysis included a study (31) of skeletal muscle samples in men, reporting that a 6-month endurance exercise intervention decreased DNA methylation at two CpG sites annotated to KLF6, and increased mRNA expression of KLF6. Further, an animal study reported increased KLF6 expression in adipose tissue after exercise (32). For total physical activity, associations appeared mainly driven by doing some versus no activity. Specifically, doing some versus no total physical activity was associated with methylation at cg10266336, annotated to the SAA2 gene. This gene codes for serum amyloid A-2 protein, which is secreted in response to inflammatory cytokines (33) and is implicated in cardiovascular disease development (34). One intervention study reported upregulated SAA2 transcription in skeletal muscle in response to resistance exercise in postmenopausal women (35). We also observed weaker evidence of total physical activity with methylation at cg00939211, annotated to the SNIP1 gene, coding for smad nuclear interacting protein 1, which influences transforming growth factor-β (TGF-β) and nuclear factor kappa-B (NF-κB) signaling (36,37). TGF-β and NF-κB signaling are known to play an important role in cancer development (38,39) and can be activated in blood by exercise (40,41). Television viewing time showed weak evidence of association with methylation at cg07141142, annotated to the ZMAT3 gene, which influences p53-mediated growth inhibitory signaling that is associated with cancer development (42). Pathway analysis also identified cancer-related pathways including “ECM–receptor interaction,” “microRNAs in cancer,” and “p53 signaling pathway.” Again, there is some consistency with previous results from the gene ontology meta-analysis on exercise and DNA methylation (29) that identified microRNA-associated and tumor suppressor gene networks to be involved. In our analysis, greater television viewing time was mostly associated with higher methylation levels, whereas for physical activity, no clear pattern in direction of associations was observed. Further research is required to replicate our findings and investigate whether and how these may affect gene expression.
An important strength of our study was the large sample size, although it may still have been insufficient to detect associations at the significance threshold. Because ours is the first EWAS of physical activity and television viewing time to date, we also applied a more liberal cutoff point (P < 1.0 × 10−5) for findings that warrant follow-up. Other strengths included the longitudinal analysis and adjustment for several potential confounders.
A limitation of our study included the use of self-reported physical activity data, which may have resulted in measurement error. We had a crude measure of physical activity at baseline and in our longitudinal analysis, which could have limited statistical power. At follow-up, we used data from the comprehensive IPAQ, which has good test–retest reliability but poor to moderate criterion validity (43). The reliability for the identified CpG sites was relatively low, and measurement error may have reduced statistical power. Nevertheless, in a previous study using the same data, we replicated many previously reported BMI-associated CpG sites and identified hundreds of novel associations that were replicated in independent populations (12), which is indirect evidence for the validity and reliability of our data. We observed some inflation in test statistics for total physical activity, whereas none was found for the individual continuous and dichotomous variables. Inflation was still observed after adjusting for 10 methylation-based principal components (results not shown) (27); we cannot explain these findings. Some CpG sites with weaker evidence of associations with leisure-time and total physical activity showed opposite directions of associations for the dichotomous and continuous variable, which may indicate spurious findings and require further investigation.
In conclusion, we observed generally weak evidence of cross-sectional associations between physical activity and television viewing time and DNA methylation in peripheral blood, and that changes in physical activity over approximately a decade may be associated with methylation changes. Studies with accelerometer data and with larger sample sizes will be necessary to replicate our findings. Mechanistic research is needed to investigate whether DNA methylation may be a biological mechanism linking physical activity and sedentary behavior to chronic disease development.
The MCCS was made possible by the contribution of many people, including the original investigators, the teams that recruited the participants and continue working on follow-up, and the many thousands of Melbourne residents who continue to participate in the study.
This work was supported by the Australian National Health and Medical Research Council (NHMRC) (grant no. 1088405). MCCS cohort recruitment was funded by VicHealth and Cancer Council Victoria. The MCCS was further supported by the Australian NHMRC grant nos. 209057 and 396414 and by infrastructure provided by the Cancer Council Victoria. Cases were ascertained through the Victorian Cancer Registry and the Australian Cancer Database (Australian Institute of Health and Welfare). The nested case–control methylation studies were supported by the NHMRC grant nos. 1011618, 1026892, 1027505, 1050198, 1043616, and 1074383. EHvR was supported by an Endeavour Research Fellowship from the Department of Education and Training of the Australian Government (6059-2017). MCS is an NHMRC Senior Research Fellow (1061177). BML is funded by a National Breast Cancer Foundation Fellowship (ECF-15-012). The study sponsors were not involved in the study design; in the collection, analysis, and interpretation of the data; in the writing of the report; and in the decision to submit the article for publication. The results of the present study do not constitute endorsement by the American College of Sports Medicine.
The authors declared no conflicts of interest. The authors declare that the results of the study are presented clearly and honestly, and without fabrication, falsification, or inappropriate data manipulation.
1. Kyu HH, Bachman VF, Alexander LT, et al. Physical activity and risk of breast cancer, colon cancer, diabetes, ischemic heart disease, and ischemic stroke events: systematic review and dose-response meta-analysis for the Global Burden of Disease Study 2013. BMJ
2. Moore SC, Lee IM, Weiderpass E, et al. Association of leisure-time physical activity with risk of 26 types of cancer in 1.44 million adults. JAMA Intern Med
3. Pandey A, Salahuddin U, Garg S, et al. Continuous dose–response association between sedentary time and risk for cardiovascular disease: a meta-analysis. JAMA Cardiol
4. Biswas A, Oh PI, Faulkner GE, et al. Sedentary time and its association with risk for disease incidence, mortality, and hospitalization in adults: a systematic review and meta-analysis. Ann Intern Med
5. Tremblay MS, Aubert S, Barnes JD, et al. Sedentary Behavior Research Network (SBRN)—terminology consensus project process and outcome. Int J Behav Nutr Phys Act
6. Dunstan DW, Howard B, Healy GN, Owen N. Too much sitting—a health hazard. Diabetes Res Clin Pract
7. Denham J. Exercise and epigenetic inheritance of disease risk. Acta Physiol (Oxf)
8. Rakyan VK, Down TA, Balding DJ, Beck S. Epigenome-wide association studies for common human diseases. Nat Rev Genet
9. Portela A, Esteller M. Epigenetic modifications and human disease. Nat Biotechnol
10. Tan Q, Heijmans BT, Hjelmborg JV, Soerensen M, Christensen K, Christiansen L. Epigenetic drift in the aging genome: a ten-year follow-up in an elderly twin cohort. Int J Epidemiol
11. Bernstein BE, Meissner A, Lander ES. The mammalian epigenome. Cell
12. Geurts YM, Dugue PA, Joo JE, et al. Novel associations between blood DNA methylation and body mass index in middle-aged and older adults. Int J Obes (Lond)
13. Chambers JC, Loh M, Lehne B, et al. Epigenome-wide association of DNA methylation markers in peripheral blood
from Indian Asians and Europeans with incident type 2 diabetes: a nested case–control study. Lancet Diabetes Endocrinol
14. Shaw B, Wc L, Hs T, Al F, Saxton J, Belshaw N. A change in physical activity level affects leukocyte DNA methylation of genes implicated in cardiovascular disease in the elderly: Physiology 2014 (London, UK). Proc Physiol Soc
15. Veenendaal MV, Costello PM, Lillycrop KA, et al. Prenatal famine exposure, health in later life and promoter methylation of four candidate genes. J Dev Orig Health Dis
16. Cordero F, Ferrero G, Polidoro S, et al. Differentially methylated microRNAs in prediagnostic samples of subjects who developed breast cancer in the European Prospective Investigation into Nutrition and Cancer (EPIC-Italy) cohort. Carcinogenesis
17. Luttropp K, Nordfors L, Ekstrom TJ, Lind L. Physical activity is associated with decreased global DNA methylation in Swedish older individuals. Scand J Clin Lab Invest
18. White AJ, Sandler DP, Bolick SC, Xu Z, Taylor JA, DeRoo LA. Recreational and household physical activity at different time points and DNA global methylation. Eur J Cancer
19. Zhang FF, Santella RM, Wolff M, Kappil MA, Markowitz SB, Morabia A. White blood cell global methylation and IL-6 promoter methylation in association with diet and lifestyle risk factors in a cancer-free population. Epigenetics
20. Peplonska B, Bukowska A, Wieczorek E, Przybek M, Zienolddiny S, Reszka E. Rotating night work, lifestyle factors, obesity and promoter methylation in BRCA1 and BRCA2 genes among nurses and midwives. PLoS One
21. Quach A, Levine ME, Tanaka T, et al. Epigenetic clock analysis of diet, exercise, education, and lifestyle factors. Aging (Albany NY)
22. Gale CR, Marioni RE, Cukic I, et al. The epigenetic clock and objectively measured sedentary and walking behavior in older adults: the Lothian Birth Cohort 1936. Clin Epigenetics
23. Milne RL, Fletcher AS, MacInnis RJ, et al. Cohort Profile: The Melbourne Collaborative Cohort Study (Health 2020). Int J Epidemiol
24. Thorp AA, Healy GN, Owen N, et al. Deleterious associations of sitting time and television viewing time with cardiometabolic risk biomarkers: Australian Diabetes, Obesity and Lifestyle (AusDiab) study 2004–2005. Diabetes Care
25. Dugue PA, English DR, MacInnis RJ, et al. Reliability of DNA methylation measures from dried blood spots and mononuclear cells using the HumanMethylation450k BeadArray. Sci Rep
26. Jaffe AE, Irizarry RA. Accounting for cellular heterogeneity is critical in epigenome-wide association studies. Genome Biol
27. Barfield RT, Almli LM, Kilaru V, et al. Accounting for population stratification in DNA methylation studies. Genet Epidemiol
28. Glymour MM, Weuve J, Berkman LF, Kawachi I, Robins JM. When is baseline adjustment useful in analyses of change? An example with education and cognitive change. Am J Epidemiol
29. Brown WM. Exercise-associated DNA methylation change in skeletal muscle and the importance of imprinted genes: a bioinformatics meta-analysis. Br J Sports Med
30. Li D, Yea S, Dolios G, et al. Regulation of Kruppel-like factor 6 tumor suppressor activity by acetylation. Cancer Res
31. Nitert MD, Dayeh T, Volkov P, et al. Impact of an exercise intervention on DNA methylation in skeletal muscle from first-degree relatives of patients with type 2 diabetes. Diabetes
32. Shen Y, Zhou H, Jin W, Lee HJ. Acute exercise regulates adipogenic gene expression in white adipose tissue. Biol Sport
33. De Buck M, Gouwy M, Wang JM, et al. The cytokine-serum amyloid A-chemokine network. Cytokine Growth Factor Rev
34. Getz GS, Krishack PA, Reardon CA. Serum amyloid A and atherosclerosis. Curr Opin Lipidol
35. Buford TW, Cooke MB, Willoughby DS. Resistance exercise-induced changes of inflammatory gene expression within human skeletal muscle. Eur J Appl Physiol
36. Kim RH, Wang D, Tsang M, et al. A novel smad nuclear interacting protein, SNIP1, suppresses p300-dependent TGF-beta signal transduction. Genes Dev
37. Kim RH, Flanders KC, Birkey Reffey S, et al. SNIP1 inhibits NF-kappa B signaling by competing for its binding to the C/H1 domain of CBP/p300 transcriptional co-activators. J Biol Chem
38. Elliott RL, Blobe GC. Role of transforming growth factor Beta in human cancer. J Clin Oncol
39. Hoesel B, Schmid JA. The complexity of NF-κB signaling in inflammation and cancer. Mol Cancer
40. Vider J, Laaksonen DE, Kilk A, et al. Physical exercise induces activation of NF-kappaB in human peripheral blood
lymphocytes. Antioxid Redox Signal
41. Hering S, Jost C, Schulz H, Hellmich B, Schatz H, Pfeiffer H. Circulating transforming growth factor beta1 (TGFbeta1) is elevated by extensive exercise. Eur J Appl Physiol
42. Parikh N, Hilsenbeck S, Creighton CJ, et al. Effects of TP53 mutational status on gene expression patterns across 10 human cancer types. J Pathol
43. Silsbury Z, Goldsmith R, Rushton A. Systematic review of the measurement properties of self-report physical activity questionnaires in healthy adult populations. BMJ Open