Exercise effects on DNA methylation in EVL, CDKN2A (p14, ARF), and ESR1 in colon tissue from healthy men and women

ABSTRACT

Physical activity reduces risk of colon cancer by 20–30%. Aberrant methylation patterns are common epigenetic alterations in colorectal adenomas, and cancers and play a role in cancer initiation and progression. Alterations identified in normal colon tissue represent apotential ‘field cancerization’ process, where normal colon is primed for carcinogenesis. Here, we investigate methylation patterns in three genes –Ena/VASP-like (EVL), (CDKN2A (p14, ARF)), and Oestrogen Receptor-1 (ESR1)– in normal colon tissue collected at baseline and 12 months from 202 sedentary men and women, 40–75 years, enrolled in a randomized controlled trial testing an exercise intervention vs. control (http://clinicaltrials.gov/show/NCT00668161). Participants were randomized to moderate-to-vigorous intensity exercise, 60 minutes/day, 6 days/week for 12 months, or usual lifestyle. Sigmoid colon biopsies were obtained at baseline and 12-months, DNA extracted, and bisulphite converted. Droplet digital methylation-specific PCR was performed for EVL, p14ARF, and ESR1. Generalized estimating equations modification of linear regression was used to model relationships between intervention effects and gene methylation levels, adjusting for possible confounders.

There were no statistically significant differences between methylation patterns at 12-months between exercisers and controls. ESR1 methylation patterns differed by sex: women −10.58% (exercisers) +11.10% (controls); men +5.54% (exercisers), −8.16% (controls) (P=0.05), adjusting for BMI and age. There were no statistically significant changes in methylation patterns in any gene stratified by change in VO₂max or minutes/week of exercise.

While no statistically significant differences were found in gene methylation patterns comparing exercises vs. controls, 12-month exercise effects on ESR1 methylation differed by sex, warranting further study.

KEYWORDS:

• Methylation
• colon cancer
• droplet digital PCR
• exercise
• randomized controlled trial

Introduction

Physically active persons have a 20%–30% reduced risk of colon cancer compared with sedentary individuals, and the association is stronger in men [1]. Physical activity may alter colon cancer risk through diverse mechanisms, such as maintenance of normal weight, leading to lower inflammation levels, altered levels of adipokines, or direct effects on insulin resistance [2,3]. Less commonly assessed potential mechanisms include direct effects on the colon itself, and little is known about the effects of exercise at the cellular and molecular levels in humans.

Observational data suggest that low levels of physical activity are associated with DNA aberrant methylation of specific genes in blood [4–6], but there is a lack of information on either the direct effects of exercise on DNA methylation in target tissues or, if present, on the amount of exercise necessary to effect these changes.

It is well established that the accumulation of genetic and epigenetic alterations in colon epithelial cells plays a role in the initiation and progression of colon polyps to colorectal cancer (CRC). Other common molecular features in CRC include chromosomal and microsatellite instability [7] and epigenetic instability [8], where CpG island methylation alters gene expression [9]. DNA methylation is a heritable, reversible chemical/structural modification that can regulate gene expression in the absence of underlying changes to DNA sequence, and that involves the addition of methyl groups to cytosines in CpG dinucleotides to form 5-methylcytosine (5mC). Aberrant DNA methylation is a common epigenetic alteration found in colorectal adenomas and cancers, that is thought to play a role in cancer initiation and progression[10]. Aberrant DNA methylation of a number of genes has been identified in the normal colon of people at increased risk for colon cancer and appears to reflect a ‘field cancerization’ process in which the normal colon is primed for carcinogenesis. [11] Studies suggest that this methylation process may be one of the earliest molecular events that initiate colorectal cancer formation [12].

Several genes have been identified to be more commonly aberrantly methylated in the normal colon of CRC patients compared to healthy individuals, including the alternate reading frame (ARF) protein product of the alternatively spliced cyclin-dependent kinase inhibitor 2A (CDKN2A p14, ARF) locus (hereon p14ARF) [13], EVL, and ESR1. p14ARF inhibits cell proliferation by blocking the Mdm2-mediated degradation of p53, and p14ARF inactivation disrupts the p53/p21 pathway, which contributes to uncontrolled proliferation [14,15]. Methylation of the p14ARF locus has been found in the normal-appearing colorectal mucosa adjacent to colorectal cancer, and this mucosa was more likely to carry methylated p14ARF than the mucosa from healthy people, suggesting that methylated p14ARF may be a field cancerization marker. Another locus implicated as a marker of field cancerization in the colon is EVL/miR-342; the microRNA hsa-miR-342 is encoded in an intronic sequence in the Ena/VASP-like (EVL) gene, which is epigenetically silenced in the majority of colorectal cancers. [16] The EVL gene encodes an actin-associated protein involved in a variety of processes related to cytoskeleton remodelling and cell polarity[17]. An earlier report demonstrated that normal colon mucosa 10 cm away from the colorectal cancer had methylated EVL/miR-342 in almost half of the cases, whereas only 12% of normal colon mucosa from individuals without cancer had methylated EVL/miR-342 [18]. Finally, the oestrogen receptor 1 gene (ESR1), which encodes oestrogen receptor α (ERα), is increasingly methylated as a function of age in human colonic mucosa, [19] of interest as colorectal cancer is an age-related disease. This increase in colon and rectum ESR1 DNA methylation with increasing age suggests it may be involved in the age-related risk of colon cancer [20,21]. Furthermore, ESR1 could be involved in the association of menopausal hormone therapy with reduced risk for colon cancer in women [22].

It is notable that studies of field cancerization in the colon have yielded inconsistent results, which is in part a consequence of the use of different assay technologies and the use of suboptimal assays. Previous studies assessed DNA methylation levels using established techniques that lack precision, sensitivity, and reproducibility. To address this, we developed and successfully demonstrated a technique called MethyLight droplet digital PCR (ddPCR) [10] can accurately and robustly detect low levels of DNA methylation in normal colon tissue. Using this technique, we assessed the effects of a 12-month moderate-to-vigorous aerobic exercise intervention on DNA methylation in three candidate genes associated with field cancerization in colorectal mucosa – p14ARF locus, EVL/miR-342 and ESR1 – in sigmoid colon biopsies in healthy men and women, in a randomized controlled trial (RCT). To our knowledge, this will be the first study to investigate the effects of exercise on methylation of genes in normal human colon tissue.

Materials and methods

This study is ancillary to the A Programme Promoting Exercise and an Active Lifestyle (APPEAL) study (http://clinicaltrials.gov/show/NCT00668161) [23]. The study was approved by the Institutional Review Board of the Fred Hutchinson Cancer Research Center, and all participants signed informed consent, and gave permission for stored samples to be used in future studies

The study population includes 102 men and 100 women who were: aged 40–75 years, recruited principally from gastroenterology practices, screened, randomly assigned to intervention or control, and followed for 12 months. Eligibility criteria included sedentary (<90 minutes/week in the previous 3 months, of moderate-to-vigorous intensity sports, recreational, or walking exercise); consumption of <2 alcoholic beverages per day; no history of invasive cancer or other serious medical conditions; normal response to a maximal exercise tolerance test, normal complete blood count and blood chemistries; no familial polyposis, Gardner’s syndrome, or other known familial colorectal cancer syndrome, no ulcerative colitis or short bowel; and history of having had a colonoscopy within the previous 3 years. They could not be regular users (>2/week) of NSAIDs. Participants were randomized in equal numbers to the intervention or control group, blocked on sex, use of NSAIDs medications (two times/week vs. less), smoking (yes/no), and among women, on menopausal status (pre – or peri – vs. postmenopausal) and current use (yes/no) of hormone replacement therapy. Participants were asked not to change their dietary habits during the trial. Controls were asked not to change their exercise habits during the trial and were offered a 2-month exercise programme at the end of the study.

Participants were randomized to an exercise intervention (N = 51 men, 49 women) or control (N = 51 men, 51 women) arm. The intervention was a 12-month moderate-to-vigorous intensity aerobic exercise programme with a goal of 60 minutes/day, 6 days/week, gradually achieved through 8 weeks’ escalation of dose and duration, and continuing throughout the 12-month intervention. Three days per week, participants exercised on treadmills, stationary bikes, elliptical machines, and rowers at one of the four study facilities with an exercise specialist present. Participants were given Polar (Polar Electro Inc.) heart rate monitors, with a prescription corresponding to 60–85% of their maximal heart rate on their baseline VO₂max test, and they recorded maximal session heart rate on the facility and home daily activity logs. Participants also were asked to exercise at home or gym three additional days/week with the same duration and heart rate goals. Intervention adherence was measured by logging of facility attendance and home exercise (both with type, length, and maximal heart rate logged), quarterly physical activity interviews, quarterly 1-week use of pedometers with logging of steps, and VO₂max treadmill tests at baseline and 12 months. Adherence was calculated weekly from logs such as facility and home sessions completed, total minutes/week, MET-minutes/week, and percentage of goal 360 minutes/week of exercise.

Covariates and biospecimen collection

Fasting blood and spot urine samples, colon and rectal biopsies, anthropometrics, and questionnaire data were collected at baseline and 12-months. Medical history, health habits, family history of colorectal and other cancers, detailed history of colorectal polyps, and reproductive history were assessed using a questionnaire. Diet was assessed using a 120-item food frequency questionnaire (FFQ) [24]. Anthropometric and body composition measures included height, weight DEXA scans (fat, lean, and bone mass), and abdominal CT-scans at the L4-5 (subcutaneous and visceral fat areas). Previously measured fasting blood and tissue biomarkers include colon crypt proliferation markers [Ki67 amount and distribution], colon apoptosis [bax/bcl], and serum insulin, glucose, c-reactive protein, sex hormones [testosterone, oestradiol, oestrone, sex hormone-binding globulin], insulin-like growth factor 1 [IGF-1] and IGF-binding protein 3) [23,25–28].

One-mm-thick sigmoid colon and rectal samples were collected from fasting, saline enema prepped participants, at pre-randomization and at the 12-month time point (i.e., post-intervention) using flexible sigmoidoscopy [23]. Seven biopsies were collected from the sigmoid colon in a 5-cm segment, approximately 30–35 cm from the external anal aperture, by physicians using jumbo biopsy forceps (Olympus FB-50U1-1; Olympus America, Inc., Melville, NY). Two of these biopsy samples were flash-frozen in liquid nitrogen and stored in −80°C freezers. Twelve-month sigmoid colon biopsies were obtained in 197 (97%) [N = 50 men, N = 49 women in the exercise group, and N = 47 men and N = 51 women in the control group], and stored at −70°C.

DNA extraction and sodium bisulphite treatment

DNA was extracted from stored biopsy specimens by the FHCRC’s specimen processing lab, using the Qiagen DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA) and treated with RNase (Life Technologies, Carlsbad, CA). Five hundred nanograms of DNA were bisulphite converted using the EZ DNA Methylation Kit (ZymoResearch) following the manufacturer’s instructions and stored at −80°C. All lab personnel were blinded to participant intervention status, and to whether the samples were from baseline or end-of-study.

Droplet digital methylation specific PCR

MethyLight ddPCR was performed using primer/probe sequences for methylated EVL, p14ARF, and ESR1 as previously described [10,29–31]. For each gene, the quantity of methylation (in pictograms) in clinical samples was determined from a 5-point standard curve derived from the 100% methylated EpiTect Methyl control DNA (Qiagen). All samples were assayed by the standards in the same experimental run. Each DNA sample was partitioned into an average of 15,000 droplets/well and replicated in 8 wells. Droplet counts (positive and negative) were combined from all eight replicates to yield a ‘merged’ value for the final analysis. Probes for reference genes are included in each assay, allowing simultaneous detection and quantification of the methylated gene of interest and the reference gene in the same sample. The methylation-independent C-LESS-C1 control [32] was labelled with VIC to determine the total amount of amplifiable DNA.

Statistical analysis

The concentration and Poisson confidence intervals for each ‘merged’ well for analysed biospecimens were computed using the QuantaSoft software version 1.4.0.99 (BioRad). Coefficients of variation (CVs) ranged from 3.2%-10.3% [10]. DNA methylation results were non-normally distributed and were log-transformed. Missing data at 12-months were imputed using the LOCF (last observation carried forward) method. The primary assessment of the 12-month aerobic exercise intervention effects compared the average change from baseline to follow-up in sigmoid colon DNA methylation in the genes of interest comparing exercisers vs. controls, for all participants, and stratified by sex. The analysis was performed according to the intention-to-treat principle, where data on the level of DNA methylation of the target genes are analysed by treatment assigned at the time of randomization, regardless of the level of participants’ adherence to the exercise intervention. The generalized estimating equations (GEE) modification of linear regression were used to model the relationship between measurements of methylated DNA of the candidate genes and exercise intervention and to account for the correlation within individual data over time, adjusting for possible confounders, including age and baseline body mass index (BMI) (continuous), history of adenomatous polyps, folate, and vitamin B2, B6, and B12 intake, all associated with increased risk of colon cancer. Secondary analyses investigated effect modification by dose of exercise, either as change in VO₂max, or by adherence, measured in minutes/week of exercise. Change in VO₂max was categorized as tertiles (tertile 1 < 5.54% or decreased; tertile 2: VO₂max increased ≥5.54 < 14.76%; tertile 3 VO₂max increased ≥14.76%), and adherence was categorized as tertiles (tertile 1: minutes/week <295.37; tertile 2: minutes/week ≥295.37 < 331.44; tertile 3: minutes/week ≥331.44). Finally, we used Pearson correlations to assess associations between methylation patterns and previously measured covariates including Ki67 amount and distribution], colon apoptosis [bax/bcl], and serum insulin, glucose, c-reactive protein, sex hormones [testosterone, oestradiol, oestrone, sex hormone-binding globulin], insulin-like growth factor 1 [IGF-1] and IGF-binding protein 3) [23,25–28]. Correlation coefficients of 0.2 or greater were considered statistically meaningful Table 1.

Table 1. Baseline characteristics of APPEAL study participants with DNA methylation data.

		Control			Exercise			All participants
		N	Mean	SD	N	Mean	SD	N	Mean	SD
Age (years)	Men	47	56.20	7.04	50	56.00	6.49	97	56.10	6.73
	Women	51	53.69	5.65	49	54.42	7.07	100	54.05	6.36
Weight (kg)	Men	47	96.64	16.05	50	94.68	14.97	97	95.63	15.45
	Women	51	77.94	12.79	49	78.04	17.75	100	77.99	15.34
BMI (kg/m^2)	Men	47	29.98	4.70	50	29.71	3.76	97	29.84	4.22
	Women	51	28.52	4.80	49	28.89	5.48	100	28.70	5.12
Waist to hip ratio	Men	47	0.97	0.06	50	0.96	0.06	97	0.96	0.06
	Women	51	0.78	0.07	49	0.79	0.06	100	0.79	0.07
Waist circumference (cm)	Men	47	104.0	11.70	50	102.8	9.88	97	103.4	10.76
	Women	51	85.25	11.40	49	87.54	13.62	100	86.37	12.52
Fat (%)	Men	47	29.84	6.05	48	31.47	6.50	95	30.66	6.30
	Women	50	43.03	6.80	49	43.28	7.18	99	43.15	6.95
Total lean (g)	Men	47	63,430	7684	48	60,677	7275	95	62,039	7568
	Women	50	40,689	3864	49	40,112	4926	99	40,404	4408
Calories	Men	46	1654	564.0	46	1645	555.7	92	1649	556.8
	Women	50	1583	554.6	49	1543	720.1	99	1563	638.9
Vigorous to moderate physical activity	Men	47	75.39	87.30	50	67.41	115.4	97	71.27	102.3
	Women	51	45.40	80.65	49	43.24	59.89	100	44.34	70.89
Average steps/day	Men	45	6084	3215	49	6004	2794	94	6042	2987
	Women	51	6640	2926	49	5959	2567	100	6306	2763
VO2max	Men	47	30.79	6.08	50	30.21	5.95	97	30.49	5.99
	Women	51	24.81	4.34	49	23.84	5.09	100	24.34	4.72
		N	%		N	%		N	%
Race/Ethnicity Non-Latino White	Men	44	93.6		47	94.0		91	93.8
	Women	47	92.2		42	85.7		89	89.0
Education College degree/higher	Men	27	57.4		33	66.0		60	61.9
	Women	34	66.7		28	57.1		62	62.0
Non-smoker	Men	43	91.5		46	92.0		89	91.8
	Women	50	98.0		47	95.9		97	97.0
First degree relative with colon cancer	Men	17	36.2		14	28.0		31	32.0
	Women	19	37.3		21	42.9		40	40.0
Endoscopy History of adenomatous polyps	Men	27	57.4		29	58.0		56	57.7
	Women	12	23.5		15	30.6		27	27.0

Results

At baseline, participants were a mean age of 54.1 years (women) and 56.4 years (men); 91.5% non-Hispanic White; with a mean BMI (kg/m²) of 28.7 in women and 29.9 in men [33]. Forty-three per cent had a history of adenomatous polyps, and 36% had a first degree relative with colon cancer[23]. Mean baseline VO₂max was 24.3 ml/kg/min for women and 30.2 for men. From baseline to 12 months, VO₂max increased a mean of 2.5 ml/kg/min (10.5%) in female exercisers and 3.3 ml/kg/min (11%) in male exercisers and decreased in controls (P< 0.001 comparing exercisers to controls). In the exercise group, mean exercise over 12 months was 370 min/week in men (102% of goal) and 295 min/week in women (82% of goal).

Female participants randomized to the exercise intervention had a mean 1.9-kg reduction in total fat mass (measured by DEXA scan), which represents a reduction in per cent body fat from 43.3% to 41.5% vs. a mean gain of 0.2 kg fat in controls (P = 0.001) [33]. Male exercisers experienced an average 3.0-kg loss of total fat mass, a reduction in percentage body fat from 31.5% to 28.8% vs. a slight increase in controls (P < 0.001). Neither exercisers nor controls significantly changed mean total daily caloric intake, intake of fat, fibre, or alcohol, and there were no statistically significant differences for changes in these variables between exercisers and controls.

At baseline, there were no statistically significant differences between methylation levels in ESR1, EVL, or p14ARF, between male and female participants with or without a history of adenomatous polyps (data not shown). There were no meaningful statistically significant differences in methylation patterns at 12 months between exercisers and controls for either men or women (Table 2). However, the distributions of ESR1 methylation patterns differed by sex – ESR methylation in men at 12 months was +5.54% in exercisers and −8.16% in controls; in women values were-10.58% and +11.10%, respectively (P = 0.07); P = 0.05 testing for intervention effects between sexes, adjusting for BMI and age. There were no statistically significant changes in methylation patterns in any gene when stratified by either change in VO₂max, or by minutes/week of exercise (data not shown). Finally, there were no differences in methylation patterns for any gene between exercisers vs. controls comparing (1) women with and without a history of adenomatous polyps and (2) men with and without a history of adenomatous polyps (data not shown).

Table 2. Twelve-month change in DNA methylation patterns.

		Control							Exercise							P-value*
		Baseline			12 Month			Difference %	Baseline			12 Month			Difference %
Gene	Sex	N	GM	95% CI	N	GM	95% CI		N	GM	95% CI	N	GM	95% CI		P_un	P^[Citation1]	P^[Citation2]
ESR1	Men	47	3.410	(2.840, 4.0927	43	3.131	(2.516, 3.897)	−0.278 –8.16%	50	3.774	(3.124, 4.559)	44	3.985	3.249 4.887	0.211 5.54%	0.62	0.61	0.99
	Women	51	2.760	(2.197, 3.467)	46	3.066	(2.394, 3.926)	0.306 11.10%	48	2.451	(1.959, 3.066)	43	2.191	1.704 2.817	−0.259 –10.58%	0.05	0.06	0.07
	P interaction															0.06	0.05	0.12
EVL	Men	47	0.270	(0.181, 0.405)	43	0.225	(0.158, 0.318)	−0.047 –17.25%	50	0.284	(0.206, 0.392)	44	0.336	0.238 0.474	0.052 18.21%	0.17	0.14	0.58
	Women	51	0.286	(0.203, 0.404)	46	0.340	(0.247, 0.494)	0.063 21.90%	48	0.266	(0.201, 0.352)	43	0.280	0.204 0.383	0.014 5.26%	0.63	0.67	0.60
	P interaction															0.09	0.09	0.29
p14ARF	Men	47	0.016	(0.011, 0.024)	43	0.015	(0.010, 0.021)	−0.001 –7.03%	50	0.021	(0.015, 0.029)	44	0.016	0.011 0.023	−0.004 –21.04%	0.46	0.24	0.46
	Women	51	0.014	(0.010, 0.021)	46	0.017	(0.012, 0.025)	0.003 20.63%	48	0.010	(0.007, 0.013)	43	0.015	0.011 0.021	0.005 54.92%	0.38	0.33	0.38
	P interaction															0.82	0.86	0.82

P_un: P-value based on GEE models comparing the 12-months changes in the DNA methylation between Controls versus participants in the Exercise group within each stratum of gender, unadjusted.

P¹: P-value based on GEE models comparing the 12-months changes in the DNA methylation between Controls versus participants in the Exercise group within each stratum of gender, adjusted for age and baseline BMI (continuous).

P²: P-value based on GEE models comparing the 12-months changes in the DNA methylation between Controls versus participants in the Exercise group within each stratum of gender, adjusted for age, baseline BMI (continuous), history of adenomatous polyps, history of adenomatous polyps, folate, and vitamin B2, B6, and B12 intakes.

P Interaction is the P-value obtained from GEE model comparing the intervention effect on the DNA methylation between men and women.

Baseline associations between methylation levels for the three genes for women and men are shown in Table 3 and Table 4 respectively. Among women, methylation patterns in p14ARF correlated statistically significantly with Bax density in crypt middle, top, and whole crypts, but not in men. ESR1 methylation patterns in men correlated negatively with free testosterone levels (P= 0.23, P = 0.03).

Table 3. Correlations between baseline (log-transformed) DNA methylation patterns and baseline variables in women.

	ESR1			EVL			p14ARF
	N	Rho	P	N	Rho	P	N	Rho	P
Bax density
Crypt bottom	97	0.123	0.23	97	0.084	0.41	97	0.192	0.06
Crypt middle	97	0.126	0.22	97	0.109	0.28	97	0.211	0.04
Crypt top	97	0.126	0.22	97	0.065	0.52	97	0.217	0.03
Crypt whole	97	0.129	0.21	97	0.087	0.39	97	0.215	0.03
Crypt height	97	0.015	0.89	97	0.036	0.74	97	0.067	0.51
Bcl-2 density
Crypt bottom	96	0.027	0.79	96	0.144	0.16	96	0.093	0.37
Crypt middle	96	−0.057	0.58	96	0.040	0.70	96	−0.041	0.70
Crypt top	96	0.112	0.27	96	0.192	0.06	96	0.205	0.05
Crypt whole	96	0.043	0.68	96	0.146	0.15	96	0.111	0.28
Crypt height	96	−0.062	0.55	96	0.054	0.60	96	0.125	0.23
Bax/bcl-2
Crypt bottom	96	−0.111	0.28	96	0.01	0.76	96	−0.130	0.20
Crypt middle	96	−0.162	0.11	96	−0.062	0.55	96	−0.224	0.02
Crypt top	96	−0.001	0.99	96	0.136	0.19	96	0.012	0.91
Crypt whole	96	−0.091	0.38	96	0.049	0.63	96	−0.113	0.28
Crypt height	97	−0.023	0.821	97	0.047	0.647	97	0.104	0.310
Ki67+ cells
Average cell height	97	0.092	0.37	97	0.138	0.18	97	−0.041	0.69
Median cell height	97	0.092	0.37	97	0.138	0.18	97	−0.041	0.69
75^th percentile cell height	97	0.117	0.25	97	0.118	0.25	97	−0.083	0.42
Weighted average cell height	97	0.102	0.32	97	0.109	0.29	97	−0.114	0.26
Average crypt height from KI67 slice	97	0.096	0.34	97	0.126	0.22	97	0.049	0.63
Average percent cell height	97	0.102	0.32	97	0.120	0.24	97	−0.090	0.38
Median percent cell height	97	0.099	0.33	97	0.104	0.31	97	−0.028	0.78
75^th percentile percent cell height	97	0.134	0.19	97	0.092	0.37	97	−0.148	0.14
Cytokeratin-18	98	0.021	0.84	98	0.040	0.69	98	−0.045	0.66
Circulating biomarkers
Insulin	99	0.113	0.27	99	0.125	0.21	99	0.028	0.78
Glucose	99	0.061	0.55	99	0.119	0.24	99	0.133	0.19
HOMA (Homoeostatic Model Assessment)	99	0.104	0.30	99	0.130	0.19	99	0.045	0.66
CRP (C-Reactive Protein)	97	0.053	0.60	97	0.031	0.76	97	−0.114	0.26
Testosterone	99	0.021	0.84	99	0.023	0.82	99	0.112	0.27
Free testosterone	99	0.080	0.43	99	0.088	0.39	99	0.095	0.35
SHBG (Sex Steroid Hormone Binding Globulin)	99	0.006	0.95	99	0.035	0.72	99	0.008	0.93

Table 4. Correlations between baseline (log-transformed) DNA methylation patterns and baseline variables in men.

	ESR1			EVL			p14ARF
	N	Rho	P	N	Rho	P	N	Rho	P
Bax density
Crypt bottom	94	−0.116	0.26	94	−0.118	0.26	94	−0.112	0.29
Crypt middle	94	−0.140	0.18	94	−0.124	0.24	94	−0.107	0.30
Crypt top	94	−0.138	0.18	94	−0.104	0.32	94	−0.097	0.35
Crypt whole	94	−0.137	0.19	94	−0.118	0.26	94	−0.108	0.30
Crypt height	94	0.091	0.39	94	0.013	0.90	94	−0.039	0.70
Bcl-2 density
Crypt bottom	94	−0.123	0.236	94	−0.094	0.367	94	−0.146	0.16
Crypt middle	94	−0.141	0.176	94	−0.068	0.515	94	−0.190	0.07
Crypt top	94	−0.117	0.261	94	−0.043	0.681	94	−0.053	0.60
Crypt whole	94	−0.140	0.180	94	−0.072	0.492	94	−0.134	0.19
Crypt height	94	0.033	0.750	94	−0.073	0.486	94	−0.113	0.28
Bax/bcl-2
Crypt bottom	92	−0.009	0.93	92	0.030	0.78	92	−0.009	0.93
Crypt middle	92	0.004	0.97	92	0.063	0.55	92	−0.060	0.57
Crypt top	92	0.027	0.79	92	0.061	0.56	92	0.054	0.61
Crypt whole	92	0.012	0.91	92	0.061	0.57	92	0.003	0.98
Crypt height	96	0.066	0.52	96	−0.037	0.72	96	−0.085	0.41
Ki67+ cells
Average cell height	96	0.008	0.94	96	−0.097	0.35	96	−0.119	0.25
Median cell height	96	0.014	0.89	96	−0.116	0.26	96	−0.104	0.31
75^th percentile cell height	96	0.017	0.87	96	−0.074	0.47	96	−0.099	0.34
Weighted average cell height	96	0.028	0.79	96	−0.069	0.50	96	0.010	0.92
Average crypt height from KI67 slice	96	−0.013	0.89	96	−0.080	0.44	96	−0.189	0.07
Average percent cell height	96	0.042	0.69	96	−0.047	0.65	96	0.023	0.82
Median percent cell height	96	0.036	0.73	96	−0.075	0.47	96	0.024	0.81
75^th percentile percent cell height	96	0.044	0.67	96	−0.024	0.82	96	0.051	0.62
Cytokeratin-18	97	0.045	0.66	97	0.043	0.68	97	−0.085	0.41
Circulating biomarkers
Insulin	97	−0.100	0.330	97	−0.088	0.39	97	−0.008	0.95
Glucose	97	0.027	0.795	97	−0.023	0.82	97	−0.044	0.67
HOMA (Homoeostatic Model Assessment)	97	−0.094	0.359	97	−0.089	0.39	97	−0.013	0.89
CRP (C-reactive Protein)	93	−0.070	0.502	93	−0.097	0.36	93	−0.055	0.60
Testosterone	97	−0.196	0.055	97	−0.125	0.22	97	−0.100	0.33
Free testosterone	97	−0.227	0.026	97	−0.097	0.35	97	−0.109	0.29
SHBG (Sex Steroid Hormone Binding Globulin)	97	−0.089	0.387	97	−0.132	0.20	97	−0.065	0.50
DHT (5α-dihydrotestosterone)	97	−0.087	0.40	97	−0.065	0.52	97	−0.085	0.41

There were no statistically significant changes in methylation patterns in any gene when stratified by either change in VO₂max or by minutes/week of exercise (data not shown).

Discussion

We found no effect of exercise on methylation levels of three genes postulated to play a role in a field cancerization phenomenon in the colon. However, we did observe different methylation patterns in women vs. men for ESR1 (which encodes oestrogen receptor α): methylation decreased at 12 months comparing exercises to controls (P = 0.07 in fully adjusted analysis) in women only and increased in men. Free testosterone levels at baseline correlated with ESR1 methylation levels. This negative association between ESR1 methylation and free testosterone levels suggests peripheral sex steroids may modify DNA methylation. Of note, a small previously published study demonstrated the effect of increased testosterone on methylation of ESR1 in women undergoing cross-sex hormone treatment [34].

We also observed differences between men and women in correlations between p14ARF and Bax (part of the bcl-2 family) staining density; methylation patterns correlated with Bax density in colon crypts in women only. We previously demonstrated that the exercise intervention resulted in greater expression of proteins that promote apoptosis at the bottom of colon crypts in men and decreased expression of proteins that promote apoptosis at the middle and top of colon crypts in women [35]. The pro-apoptotic protein Bax plays a key role in mitochondrial-apoptotic pathways and participates in executing p53-mediated apoptosis [36]. The p14ARF tumour suppressor plays a central role in regulating cell cycle arrest and apoptosis. Of interest is a study demonstrating that the p53-independent activation of the mitochondrial apoptosis pathway by p14ARF is primarily mediated by the pro-apoptotic Bax-homolog Bak [37]. However, it is unclear whether any conclusions can be drawn from the observed correlation between Bax expression and methylation of p14ARF in colon crypts in this study.

The underlying design of the study is based on molecular pathological epidemiology, first described by Ogino and Stampfer in 2010 [38], who defined it as a multidisciplinary investigation of complex interrelationships between environmental, dietary, lifestyle and genetic factors, tumour molecular signatures, and disease pathways and evolution. Here, we examine how the exposure of interest (physical exercise) influences colorectal cancer initiation by examining molecular pathologic marks of tumour initiation (here, methylation patterns of three genes involved in the field cancerization process). A review by Ogino et al. [39] described significant confounding issues in study design, generalizability, and confounding, among others. As this study is ancillary to an RCT, we were able to reduce the impact of confounding factors. However, study limitations, including measuring methylation levels in only three genes, the fact that the majority of participants were non-Latinx Whites, and the relatively small sample size, may impact generalizability.

Strengths of this study include measurement of intervention effects in the context of a well-characterized RCT with excellent study adherence and retention, combined with a highly sensitive and precise method of measuring low levels of DNA methylation.

Disclosure statement

The authors have no conflict of interest to report.

Additional information

Funding

This work was supported by the National Cancer Institute [R01 CA077572]; National Cancer Institute [R21 CA209203]; Breast CancerResearchFund [BCRF-19-107].