Case-Control Study of Candidate Gene Methylation and Adenomatous Polyp Formation

Abstract

Purpose

Colorectal cancer (CRC) is one of the most common and preventable forms of cancer, but remains the second leading cause of cancer-related death. Colorectal adenomas are precursor lesions that develop in 70–90% of CRC cases. Identification of peripheral biomarkers for adenomas would help to enhance screening efforts. This exploratory study examined the methylation status of 20 candidate markers in peripheral blood leukocytes and their association with adenoma formation.

Methods

Patients recruited from a local endoscopy clinic provided informed consent, and completed an interview to ascertain demographic, lifestyle, and adenoma risk factors. Cases were individuals with a histopathologically confirmed adenoma, and controls included patients with a normal colonoscopy, or those with histopathological findings not requiring heightened surveillance (normal biopsy, hyperplastic polyp). Methylation-specific polymerase chain reaction was used to characterize candidate gene promoter methylation. Odds ratios and 95% confidence intervals (OR, 95% CI) were calculated using unconditional multivariable logistic regression to test the hypothesis that candidate gene methylation differed between cases and controls, after adjustment for confounders.

Results

Complete data were available for 107 participants; 36% had adenomas (men: 40%, women: 31%). Hypomethylation of the MINT1 locus (OR: 5.3, 95% CI: 1.0–28.2), and the PER1 (OR: 2.9, 95% CI: 1.1–7.7) and PER3 (OR: 11.6, 95% CI: 1.6–78.5) clock gene promoters was more common among adenoma cases. While specificity was moderate to high for the three markers (71–97%), sensitivity was relatively low (18–45%).

Conclusion

Follow-up of these epigenetic markers is suggested to further evaluate their utility for adenoma screening or surveillance.

INTRODUCTION

Despite recent decreases in colorectal cancer (CRC) incidence and associated mortality, CRC remains one of the most common and deadly forms of cancer in the United States, with a lifetime risk for diagnosis of 5%.¹ Screening via colonoscopy presents an opportunity to interrupt the disease process (via polypectomy), thus facilitating primary CRC prevention. However, in 2012, the Centers for Disease Control and Prevention reported that CRC screening rates among eligible men and women fell far short of the Healthy People 2020 target of 70.5%.² Indeed, only ~25–60% of eligible adults undergo CRC screening due to socioeconomic, racial, geographic, or other barriers that contribute to a lack of compliance.^3–6 This suggests a need for valid and effective CRC screening methods that are more accessible to patients.

Some procedures that are less invasive than colonoscopy, such as stool-based tests (fecal occult blood test [FOBT] or the fecal immunochemical test [FIT]), may have slightly higher adherence (uptake is around one-third among eligible adults who are not under an intervention designed to improve compliance),⁷ but they are limited by reduced adenoma detection sensitivity (FOBT: 7–23%; FIT: 13–26%) compared to a colonoscopy, which has the ability to detect 73–94% of adenomas.^8–11 A recent meta-analysis of stool-based DNA, an alternative CRC screening strategy, found that these tests may provide a diagnostic benefit for CRC or advanced adenoma detection in high-risk, but not average-risk, populations,¹² plus cost remains a potential barrier.¹³ Interestingly, methylation markers provided better diagnostic performance for adenoma detection than did mutations (67% sensitivity vs. 10% sensitivity, respectively),¹² FOBT, or FIT (≤26%).^10,11 Because blood-based measures can increase compliance with CRC screening and surveillance, genetic and epigenetic biomarkers in circulation are under active investigation.^14,15

The polyp→carcinoma sequence within the colon has been widely studied, and both genetic and epigenetic alterations have been observed at different stages of CRC development. An extensive literature describes the genetic lesions involved in this process, and an estimated 30% of CRC cases can be attributed to heritable factors.¹⁶ However, highly penetrant genetic markers account for only ~5% of CRC cases.¹⁶ Epigenetic modification represents an important and early event in the adenoma→CRC sequence. Methylation within gene promoter regions can lead to gene silencing and reduced gene expression.^17,18 Promoter methylation of genes responsible for tumor growth suppression, DNA repair, or other critical pathways may reduce their expression and thereby promote tumor development. Aberrant methylation of several CRC-related loci (e.g., MLH1, CDKN2A [p16], MINT1, MINT2, MINT31) corresponds to a specific CRC subtype, the CpG island methylator phenotype (CIMP),¹⁹ which occurs in 0–44% of adenomas,^20,21 and up to 50% of CRCs.²² However, associations between aberrant gene methylation and cancer have not always been consistent, possibly due to differences in the various molecular paths leading to CRC development.²³ For example, MINT1, MINT2, and MINT12 are putative tumor suppressors, but were less methylated in adenomas from individuals with multiple polyps relative to those with at least one polyp and high microsatellite instability (MSI-H) in their adenomas.²³ Also, hypermethylation of MINT31 has been associated with increased CRC risk,²⁴ but also has been associated with longer disease-free survival among CRC patients.²⁵ These findings highlight the need for more research on the relationship between promoter gene methylation and adenoma development or progression.

Despite the inherent advantages in terms of specimen accessibility and patient compliance relative to endoscopy, only a few studies have examined epigenetic markers in peripheral blood leukocytes (PBLs) for early detection of adenomas or CRC. Global hypomethylation of retrotransposable (LINE-1) elements in PBLs has been associated with increased odds of adenoma formation in some studies,^26–29 and other studies targeting candidate genes also have had some success in identifying patients with adenomas or CRC, although results among these investigations have been mixed.^15,3–34 Epigenetic changes in PBLs may reflect those observed in target tissues³⁵ or may serve as biologically relevant indicators of behavioral, environmental, psychosocial, or lifestyle factors that contribute to adenoma formation.^18,36

This exploratory case-control study tested the hypothesis that adenoma case status was associated with aberrant methylation of candidate gene promoters in PBL DNA relative to control colonoscopy patients. The objective of this study was to identify candidate epigenetic markers that can be used to improve adenoma detection and prevention. A panel of twenty candidate gene markers was selected for evaluation following a literature review based on their potential role in adenoma formation or CRC risk: APC, BRCA1, CDKN2AP16, CYP24A, CYP27B1, ER-alpha, IGF2, MGMT, MINT1, MLH1, NGFR, PER1, PER2, PER3, SEPT9, SFRP4, SFRP5, TIMP3, TMEFF2, and WIF1.^{15,18,37–43} Two candidate genes associated with adenomas in this study (PER1, PER3) are involved in the maintenance of endogenous circadian rhythms, as well as biochemical pathways that are important in carcinogenesis (cell cycle control, DNA damage response, apoptosis).^44–46

MATERIALS AND METHODS

The Epigenetics and Diet in the Carcinogenesis Process (EDCaP) study was performed among patients undergoing a colonoscopy procedure at a community endoscopy center in Columbia, SC.⁴⁷ Prior to their clinic visit, 138 participants provided informed consent, including permission to access medical records, in accordance with the University of South Carolina’s Institutional Review Board approval process. Participants completed a questionnaire to provide information on: demographic (sex, marital status, income, race/ethnicity), lifestyle (smoking history, diet, physical activity), and occupational (employment status, job industry, type of shift, history of shiftwork) factors, as well as personal and family medical history (ever being diagnosed with other diseases, personal cancer history, family CRC history).

A total of 256 subjects scheduled for a screening or surveillance colonoscopy were eligible to participate, and 154 were enrolled. Among those subjects, 138 had complete questionnaire data, and 107 of those participants provided biospecimens for epigenetic analysis (Figure 1). Information on cecal intubation, visualization of the appendix, and bowel preparation were abstracted from the patient’s medical record to evaluate colonoscopy quality. Cases were defined as individuals with a complete colonoscopy procedure and at least one histologically confirmed adenoma based on review by a consulting clinical pathology laboratory. Controls included patients with a complete procedure and normal findings, or histopathological findings not requiring heightened surveillance (normal biopsy, hyperplastic polyp).⁴⁸

Figure 1.

Flow chart of participant recruitment and participation in the Epigenetics and Diet in the Carcinogenesis Process (EDCaP) study.

A peripheral whole blood sample was collected from each participant using vacutainers containing EDTA anticoagulant; and genomic DNA was isolated using the PUREGENE genomic DNA purification kit (Gentra Systems, Minneapolis, MN) according to the manufacturer’s protocol. Sodium bisulfite conversion of DNA was performed using EpiTect kits (QIAGEN, Germantown, MD, USA). Briefly, DNA (2 µg) was added to the reaction buffer and bisulfite DNA conversion was conducted in a Bio-Rad S1000 thermal cycler (Bio-Rad Laboratories, Hercules, CA, USA) using the manufacturer’s suggested program for the conversion reaction. Converted DNA was eluted from a spin column in TE buffer and either used immediately or stored at −20°C. DNA oligonucleotides for primer pairs were purchased from Integrated DNA Technologies (Coralville, IA, USA, Table S1). To ensure that the polymerase chain reaction (PCR) primers were specific for the detection of methylated DNA, a control human DNA set (EpiTect Control DNA, QIAGEN Sciences, Germantown, MD) containing positive (bisulfite converted, 100% methylated) and negative (bisulfite converted, 0% methylated) control DNA was used. To quantify methylation specific PCR (MSP) results, the QIAxcel system was used (QIAGEN Sciences, Germantown, MD), which utilizes a multiplexed fluorescence detection method with a resolution of 3–5 bp for DNA fragments. The detection sensitivity of QIAxcel system was 0.1 ng/µl DNA in undiluted amplification reactions. The QIAxcel system was equipped with the BioCalculator software that produces a tabular display of a variety of peak properties, including number of peaks as well as the height, width, and area of each peak. The height of the peak was used for separation of the results into three methylation categories: high (peak height >0.30), medium (peak height 0.29-0.05), and low (peak height <0.05), or none (no methylation detected). Results were categorized as either unmethylated or methylated (combination of medium/low, and high methylation categories). The methylation status of each sample was normalized via a two-step process. First, a NanoDrop™ Lite Spectrophotometer (Thermo Fisher Scientific, Waltham, MA) was used to measure concentrations of the DNA templates. Subsequent experiments used the same amount of DNA template for each MSP reaction. To facilitate quantification, equal volumes of the MSP products were loaded on the QIAxcel system for analysis. Second, BRCA1 gene methylation was used as an internal standard since the same amount of DNA template had extremely high methylation in each sample. The measured peak height of BRCA1 methylation was then used to normalize the methylation of the genes of interest for each sample. All methylation assays were performed blinded to case status.

Statistical analyses were performed using SAS® version 9.2 (Cary, NC). Frequency distributions of study variables by case status were examined using the chi-squared test for differences between proportions. Student’s t-test was used for comparisons among cases and controls for normally distributed continuous variables. In instances of non-normally distributed continuous variables, the Wilcoxon–Mann–Whitney test was used to compare cases and controls. Unconditional multiple logistic regression was used to estimate odds ratios (OR) with 95% confidence intervals (CIs), comparing candidate gene methylation among adenoma cases relative to controls, after adjustment for the effects of potential confounding factors. Variables considered for inclusion in adjusted models were known or suspected adenoma or CRC risk factors (age, sex, race, body mass index [BMI], CRC family history, smoking history, work-related factors, diet, vitamin and supplement use, physical activity, sociodemographic characteristics [income, education], personal history of inflammatory bowel disease), and variables that differed among participants and non-participants (fruit and vegetable intake). Final models included variables that were statistically significant (p≤0.05) in the saturated model, or whose inclusion produced ≥10% change in the parameter estimate for the gene of interest. Each candidate gene was selected a priori, thus no correction for multiple testing was performed.⁴⁹ Due to sparse data, individuals who received a colonoscopy due to symptoms (presence of gastrointestinal bleeding, fecal occult blood, iron deficiency, or constipation, n=3) were grouped with patients undergoing a screening colonoscopy. Sensitivity analyses were conducted to determine whether associations were driven by data from people with a right-sided polyp, as right-sided hyperplastic polyps have been linked with the serrated adenoma, MSI-H CRC pathway.⁵⁰ To ensure that precision was due to reduced sample size after exclusion of hyperplastics from the control series, sample-size-adjusted 95% CIs were produced using a method as described by Hébert et al,⁵¹ using the formula antilog[b ± 1.96×SE_b/√n_{all controls}/n_{hyperplastics}], where b = log odds ratio, SE_b = standard error of b, n_{all controls} = number of all controls, and n_{hyperplastics} = number of hyperplastic controls. Sensitivity and specificity, and positive and negative predictive values (PPV, NPV, respectively) were calculated for methylation markers with a statistically significant association with adenoma case status.

RESULTS

Complete data were obtained from 107 of the 154 patients (70%) who consented to participate (Figure 1). The mean age (±SD) among participants was 60±9 years; European Americans (EAs) comprised the majority of subjects (73%), and 46% of the population was female (Table 1). Most participants (96%) had a bowel preparation that was rated as fair or better, and the cecal intubation rate was 100%. No cases of advanced adenoma (>10 mm in size in any dimension, high-grade dysplasia, villous histology), serrated adenomatous polyps, or CRC were identified. The detection rate for any polyp was 65%, and the adenoma detection rate was 36% (male: 40%, female: 31%), with a mean of 0.7±0.5 adenomas per person. All procedures with a right-sided hyperplastic polyp (n=7) were synchronous with adenoma detection in the same patient. Cases were more likely to be smokers compared to controls (68% vs. 45%, p=0.02), and were less likely to ever regularly take vitamin D supplements (5% vs. 29%, p <0.01). Though not statistically significant, cases also had a tendency to be ≥65 years of age (p=0.07), married (p=0.16), to currently drink alcohol (p=0.13), or to ever regularly take a multivitamin (p=0.07), or vitamin C (p=0.15) compared to controls.

Table 1.

Characteristics of EDCaP Study Participants

Variable	Total population (n = 107)	Controls (n = 69)	Cases (n = 38)	Controls vs. Cases p-value1
	n (%)	n (%)	n (%)
Gender				0.33
Male	58 (54)	35 (51)	23 (61)
Female	49 (46)	34 (49)	15 (39)
Race				0.55
European American	78 (73)	49 (71)	29 (76)
African American	29 (27)	20 (29)	9 (24)
Married				0.16
Yes	85 (79)	52 (75)	33 (87)
No	22 (21)	17 (25)	5 (13)
Education				0.62
Up to High School	32 (30)	21 (30)	11 (29)
Some College	28 (26)	16 (23)	12 (32)
College Undergraduate or   Post-Graduate Degree	47 (44)	32 (46)	15 (40)
Income Level				0.86
Under $50,000	33 (33)	21 (33)	12 (33)
≥$50,000 to $100,000	44 (44)	29 (46)	15 (42)
≥$100,000	22 (22)	13 (21)	9 (25)
Age Group (Years)				0.07
<54	28 (26)	20 (29)	8 (21)
55–64	46 (43)	33 (48)	13 (34)
≥65	33 (31)	16 (23)	17 (45)
Body Mass Index (kg/m2)				0.57
Underweight or normal	20 (19)	14 (20)	6 (16)
Overweight or obese	87 (81)	55 (80)	32 (84)
History of Smoking				0.02
Ever	57 (53)	31 (45)	26 (68)
Never	50 (47)	38 (55)	12 (32)
Current alcohol consumption				0.13
Yes	46 (43)	26 (38)	20 (53)
No	61 (57)	43 (62)	18 (47)
Ever regularly taken a multivitamin				0.07
Yes	68 (64)	40 (58)	28 (76)
No	38 (36)	29 (42)	9 (24)
Ever regularly taken a vitamin C supplement				0.15
Yes	31 (29)	17 (25)	14 (38)
No	75 (71)	21 (75)	23 (62)
Ever regularly taken a vitamin				<0.01
D supplement
Yes	22 (21)	20 (29)	2 (5)
No	84 (79)	49 (71)	35 (95)
Hemorrhoids				0.18
Yes	67 (63)	40 (58)	27 (71)
No	40 (37)	29 (42)	11 (29)
Procedure reason				<0.01
Routine screening/Symptoms2	34 (32)	30 (43)	4 (11)
Elevated Risk3	73 (68)	39 (57)	34 (89)
Time of day blood was collected				0.23
Morning	83 (78)	56 (81)	27 (71)
Afternoon	24 (22)	13 (19)	11 (29)
Worked >1 year of shift work over lifetime				0.99
Yes	45 (42)	29 (42)	16 (42)
No	62 (58)	40 (58)	22 (58)
Current Work Shift				0.31
Days	53 (47)	33 (48)	20 (53)
Other	4 (3)	4 (6)	0 (0)
Unemployed/Retired	50 (50)	32 (46)	18 (47)

¹Χ² test for differences in proportions between cases and controls.;

²Three subjects exhibited symptoms (presence of gastrointestinal bleeding, hematochezia, melena, fecal occult blood, iron deficiency, or constipation) and were collapsed into screening category;

³Subject had a previous adenomatous polyp detected (n = 59) and/or a family history of colorectal cancer (n = 20)

After adjustment for potential confounding factors, hypomethylation (i.e., no methylation detected) in the MINT1 (methylated in tumor1; OR: 5.3, 95% CI: 1.0–28.2) locus, and the PER1 (Period 1; OR: 2.9, 95% CI: 1.1–7.7), and PER3 (Period 3; OR: 11.1, 95% CI: 1.6–78.5) clock gene promoters was associated with an increased odds of adenoma detection (Table 2). Figures S1 and S2 (Appendix) present agarose gel images of MSP products for PER1, PER2, PER3 and MINT1, respectively, in a subsample of participants. Sensitivity analyses that were conducted by excluding seven subjects with right-sided hyperplastic polyps yielded the following results for hypomethylation of MINT1 (OR: 5.0, 95% CI: 0.9–27.9), PER1 (OR: 2.8, 95% CI: 1.1–7.5) and PER3 (OR: 9.8, 95% CI: 1.4–70.6). The time of day of sample collection or working at least one year of shift work did not modify the relationship between clock gene (PER1, PER3) promoter methylation and adenoma case status (data not shown). The ORs and corresponding sample-size-adjusted 95% CIs are provided in Table S2. Adjusted ORs for PER1 and PER3 increased in magnitude (OR: 4.7, 95% CI: 1.9–11.6 and OR: 19.9, 95% CI: 2.6–152.1 respectively) and remained significant after exclusion of hyperplastic polyps (n = 32) from the control group; however, MINT1 was no longer statistically significant (OR: 3.1, 95% CI: 0.5–18.0).

Table 2.

Unadjusted and Adjusted Odds Ratios for Leukocyte DNA Promoter Methylation Relative to Adenoma Case Status

Candidate Gene	Controls (n = 69) n (%)	Cases (n = 38) n (%)	Crude OR (95% CI)	p-value	Adjusted OR (95% CI)	p- value
Hypomethylation*
APCa	44 (64)	23 (61)	0.9 (0.4 – 2.0)	0.74	1.0 (0.4 – 2.5)	0.97
BRCA1b	31 (44)	18 (47)	1.1 (0.5 – 2.4)	0.81	1.5 (0.6 – 3.9)	0.36
MINT1d	4 (6)	7 (18)	3.7 (1.0 – 13.5)	0.05	5.3 (1.0 – 28.2)	0.05
PER1e	20 (29)	17 (45)	2.0 (0.9 – 4.5)	0.10	2.9 (1.1 – 7.7)	0.03
PER3f	2 (3)	6 (16)	6.3 (1.2 – 32.9)	0.03	11.1 (1.6 – 78.5)	0.02
SFRP4e	12 (17)	5 (29)	0.7 (0.2 – 2.2)	0.57	1.1 (0.3 – 3.8)	0.89
SFRP5b	5 (7)	5 (29)	1.9 (0.5 – 7.2)	0.32	3.7 (0.8 – 17.1)	0.09
TIMP3g	42 (61)	25 (66)	1.2 (0.5 – 2.8)	0.61	1.2 (0.4 – 3.0)	0.78
TMEFF2h	33 (48)	21 (55)	1.4 (0.6 – 3.0)	0.46	1.5 (0.6 – 3.4)	0.36
WIF1i	24 (35)	15 (39)	1.2 (0.5 – 2.8)	0.63	1.3 (0.5 – 3.2)	0.56
Hypermethylation
CDKN2A (p16)j	28 (41)	17 (45)	0.8 (0.4 – 1.9)	0.68	0.8 (0.3 – 1.9)	0.54
CYP27B1l	58 (84)	31 (82)	1.2 (0.4 – 3.4)	0.74	0.8 (0.2 – 3.0)	0.72
ER alphaj	31 (45)	20 (53)	0.7 (0.3 – 1.6)	0.45	0.6 (0.3 – 1.5)	0.27
IGF2e	52 (75)	28 (74)	1.1 (0.4 – 2.7)	0.85	0.8 (0.3 – 2.0)	0.57
MGMTh	28 (41)	16 (42)	0.9 (0.4 – 2.1)	0.88	0.9 (0.4 – 2.1)	0.80
MLH1l	63 (91)	34 (89)	1.2 (0.3 – 4.7)	0.76	0.7 (0.2 – 3.1)	0.67
NGFRa	20 (29)	12 (32)	0.9 (0.4 – 2.1)	0.78	0.9 (0.3 – 2.2)	0.75
PER2g	62 (90)	35 (92)	0.8 (0.2 – 3.1)	0.70	0.6 (0.1 – 3.0)	0.56
SEPT9a	32 (46)	18 (47)	1.0 (0.4 – 2.1)	0.92	0.9 (0.4 – 2.1)	0.81

^*- Odds of adenoma status given no methylation detected in the candidate gene. Note: CYP24A was not included in the table due to all subjects being unmethylated in the promoter.

Adjusted for: a – vitamin D and multivitamin use;

^b– vitamin C and vitamin D use and physical activity;

^c– multivitamin, vitamin C, and vitamin D use, physical activity, ever smoking, currently drinking alcohol, being married, and age group;

^d– multivitamin use, vitamin C and D use, and ever smoking;

^e– OR adjusted for vitamin C and D use and physical activity;

^f– multivitamin use, vitamin C use, vitamin D use, physical activity, ever smoking, age, and being married;

^g– multivitamin and vitamin D use, physical activity, and age group;

^h– vitamin D use;

ⁱ– vitamin D use and age group;

^j- vitamin d use, physical activity and age;

^k- vitamin C and D use, physical activity and age;

^l– vitamin D use, current drinking alcohol, and age group. OR: odds ratio; CI: confidence interval

The sensitivity, specificity, PPV, and NPV for hypomethylation of MINT1, PER1, PER3, and their combinations are presented in Table S3 (Appendix). PER1, PER3, or MINT1 hypomethylation correctly identified 16% to 45% of individuals with adenomas. Among patients without adenomas, specificity ranged from 71% to 97%. When evaluated in combination, hypomethylation of these genes yielded sensitivities ranging from 53% to 74%, and specificities ranging from 9% to 36% (Table S3). Among the possible combinations, PER3/MINT1 yielded the highest sensitivity (74%), and the PER1/PER3/MINT1 combination had the highest specificity (36%).

DISCUSSION

Identification of novel genes associated with adenoma risk may lead to a better understanding of carcinogenesis pathways and subsequently improve CRC screening and prevention efforts. Several studies have targeted blood-based epigenetic biomarkers to identify CRC cases, and only a few have used this approach to evaluate the formation of colorectal adenomas.^14,15 In this exploratory case-control study, hypomethylation in the MINT1 locus and the PER1, and PER3 clock gene promoters was more common among adenoma cases relative to controls. Furthermore, associations for PER1 and PER3 strengthened in magnitude and remained significant after adjustment of the 95% CIs for reduced sample size following exclusion of patients with hyperplastic polyps from the control group.

The MINT1 locus is one of several markers used to characterize CIMP status among CRC cases.^19,52 Aberrant DNA methylation is considered an early event in the adenoma→carcinoma sequence,¹⁸ and changes in CIMP markers have been identified in aberrant crypt foci and in colorectal adenomas.^21,53–55 MINT methylation in adenomas or associated CRC tissue has been linked with down regulation of mismatch repair (MLH1) protein expression, microsatellite instability, and BRAF (V600E) mutation relative to normal tissue.⁵⁶ However, prevalence of the CIMP phenotype is variable in adenomas (0–44%)^20,21 as well as normal gastrointestinal tissues.⁵⁷ To our knowledge, the MINT1 locus has not been examined as a circulatory marker for adenoma risk and thus results obtained in the present study require confirmation.

Two of the three candidate genes associated with adenomas in this study (PER1, PER3) are members of the clock gene family. Clock gene expression is involved in the maintenance of endogenous circadian rhythms, as well as the regulation of cellular processes that are considered hallmarks of carcinogenesis (cell cycle control, DNA damage response, apoptosis).^44–46 In colon and other cancer cell lines, PER1 insertion can activate cell cycle checkpoint proteins, sensitize cells to ionizing radiation-induced apoptosis, and exert an antiproliferative effect.⁵⁸ A hypomethylation zone in the PER1 promoter has been associated with uncoupling of PER1 transcription from promoter methylation in cervical cancer cells.¹⁷ Structural or epigenetic variability in E-boxes, particularly the E-box closest to the promoter, may play a role in these processes, and mutations within this element may lead to a reduction of PER1 promoter activity.^17,59 PER1 and PER2 mutations occur in human colorectal tumors,⁶⁰ and PER expression is reduced in colorectal adenomas or tumors relative to normal tissue.^61–66 Polymorphisms in the PER3 gene have been associated with inflammatory bowel disease, an established CRC risk factor,⁶⁷ and with increased odds of adenoma formation.⁶⁶ Aberrant methylation of PER genes has been observed among leukemia and other cancer patients.^68–72 However, to the authors’ knowledge, no studies have examined epigenetic modifications in the PER or other clock genes in relation to adenoma formation.^46,66

Clock genes facilitate the circadian expression of ~5–10% of the mammalian transcriptome, and clock-controlled genes include known tumor suppressors (e.g., p21, Chk2, XPA, ATM) and oncogenes (e.g., β- catenin, c-Myc, WEE-1).⁴⁵ Circadian patterns of clock-controlled gene expression occur via rhythmic epigenetic modification of histones and chromatin.^45,46,73 Histone modification may play a role in tumorigenesis,⁷⁴ although few studies have examined the role of circadian histone modifications in relation to adenoma or CRC development.⁴⁶ A significant proportion of clock-controlled gene regulation in mouse liver is mediated by clock gene interactions with the histone methyltransferase, MLL3 (mixed lineage leukemia 3), a tumor suppressor that has been associated with CRC and other cancers.^73,75 PER complexes recruit histone-modifying proteins that help orchestrate the circadian expression of core clock genes, although it is not known whether this function extends to clock-controlled, cancer-related genes.^46,76

DNA hypomethylation within the context of CRC has been associated with dysregulation of retrotransposable elements (e.g., LINE-1, Alu, SVA), which is considered an early event in CRC development that can facilitate genomic instability, loss of imprinting, oncogene activation, aberrant splicing, and mutagenesis.^18,77 For example, seminal studies identified a somatic LINE-1 insertion in the APC gene, as well as global DNA hypomethylation, as processes associated with CRC.^78,79 More recent studies using PBLs have identified increased odds of global DNA hypomethylation among adenoma or CRC cases relative to controls, although not consistently.^26–29 The PER3 gene contains a variable-number tandem repeat (VNTR) sequence (a component of SVA retrotransposons),^80,81 and individuals with the 5-repeat PER3 VNTR sequence had an increased odds of adenoma formation.⁶⁶ However, the PER3 VNTR did not modify the results obtained for hypomethylation of the PER3 promoter in the present study; thus it is uncertain whether these processes are related. A growing body of evidence indicates that dysregulation of the PERs or other clock genes may have a role in adenoma formation and CRC risk, possibly via epigenetic or retrotransposon pathways, however the specific mechanisms remain to be elucidated.

This study had several noteworthy strengths and limitations. Strengths included the valid ascertainment of adenoma cases and statistical adjustment for known and suspected adenoma risk factors. The sample size limited power and precluded examination of the relationship between methylation patterns and adenoma status within strata of race, reason for colonoscopy, or factors that may influence epigenetic processes. Although each candidate gene was selected a priori following a literature review, this study was exploratory in nature and the potential for spurious or false positive findings cannot be eliminated. Also, methylation among specific white blood cell subsets was not quantified. Methylation patterns do not always vary among T-cell subsets, and the extent to which methylation differs among PBL subtypes for PER1, PER3, or MINT1 is not known. Nonetheless, a lack of standardization for these potential differences represents a study uncertainty.^36,82,83 Other studies that targeted clock genes in PBLs indicated that DNA hypomethylation was associated with breast cancer diagnoses (CLOCK),⁸⁴ or with advanced stage (II-IV) breast cancer (TIMELESS),⁸⁵ whereas promoter hypermethylation was associated with postmenopausal breast cancer (CRY2),⁸⁶ and with chronic myeloid leukemia (PER3)⁶⁸ relative to controls. The mechanism through which PER1 or PER3 gene methylation may impact adenoma formation is unknown and may depend on tissue context, timing of expression relative to other cellular processes, location of epigenetic changes within the gene, or other factors that may disrupt clock gene expression, such as shift work.^45,46 Time of day of sample collection in the present study did not modify the relationship between PER1 or PER3 promoter methylation and adenoma case status (data not shown). However, more detailed genetic and epigenetic studies of the duration, phase and amplitude of PER1 and PER3 expression in conjunction with cancer-related, clock-controlled genes are suggested to help elucidate their role in adenoma formation. Methylation of non-target tissue (i.e., PBLs) can be influenced by psychosocial or environmental stressors, or lifestyle risk factors (e.g., diet, smoking). For example, cancer risk factors that can alter clock gene expression, such as shiftwork or sleep disturbances, also may exert epigenetic effects.^87–91 Global DNA hypomethylation in PBLs was recently observed among night workers relative to those working days, and among the clock genes examined, the largest differences in hypomethylation were observed within PER3 loci.⁸⁹ In the present study, neither the proportion of current shiftworkers nor those who engaged in at least one year of shiftwork differed by adenoma case status (Table 1), or by MINT1, PER1, or PER3 methylation status (data not shown).

In summary, the identification of peripheral epigenetic markers that are associated with adenoma risk may lead to a deeper understanding of colorectal carcinogenesis and facilitate CRC prevention. Results from this study suggest follow-up blood-based monitoring of MINT1, PER1, and PER3 methylation to examine their role in adenoma formation and CRC risk.