点击显示 收起
1 From the Nutritional Sciences Research Division, King's College London, London, United Kingdom
2 Address reprint requests and correspondence to M Pufulete, Nutritional Sciences Division, King's College London, Franklin Wilkins Building, 150 Stamford Street, London SE1 9NH, United Kingdom. E-mail: maria.pufulete{at}kcl.ac.uk.
ABSTRACT
Background: Greater promoter methylation in some tumor-suppressor genes underlies most sporadic colorectal cancers and increases with age in the colon.
Objective: We tested the hypothesis that biomarkers of folate and vitamin B-12 status are associated with estrogen receptor (ER) and mutL homolog 1 (MLH1) promoter methylation in subjects with and without neoplasia.
Design: Biopsies of normal-appearing colorectal mucosa from 156 subjects with and without colorectal neoplasia (disease free, n = 76; cancer, n = 28; adenoma, n = 35; hyperplastic polyps, n = 17) were obtained at colonoscopy and used to evaluate methylation in 7 CpG sites in the ER promoter and 13 CpG sites in the MLH1 promoter. Blood samples were obtained for the measurement of serum and red cell folate, serum vitamin B-12, and plasma homocysteine concentrations. Methylation indexes were generated to reflect an average methylation value across all CpG dinucleotides in both ER and MLH1.
Results: The methylation indexes for ER and MLH1 generally were significantly (P < 0.05) higher in subjects with neoplasia than in disease-free subjects. The ER methylation index correlated negatively with serum vitamin B-12 (r = –0.239, P = 0.003) and positively with plasma homocysteine (r = 0.188, P = 0.021). Disease status (P < 0.005), age (P < 0.001), and serum vitamin B-12 concentrations (P = 0.006) were independent determinants of ER promoter methylation. Serum and red cell folate concentrations had no influence on ER promoter methylation.
Conclusion: Serum vitamin B-12 but not folate status may be associated with ER promoter methylation in normal-appearing colorectal mucosa.
Key Words: Folate • vitamin B-12 • homocysteine • DNA methylation • estrogen receptor • ER • mutL homolog 1 • MLH1 • colorectal cancer • colorectal adenoma • hyperplastic polyps
INTRODUCTION
Low dietary folate intake (1) and low blood folate concentrations (2) may increase the risk of colorectal cancer. Folate is a methyl donor in the remethylation of homocysteine to methionine, a reaction catalyzed by the vitamin B-12–dependent enzyme methionine synthase, which generates methyl groups for DNA methylation. It has been proposed that folate or vitamin B-12 deficiency may increase the risk of cancer by altering DNA methylation (3, 4).
Aberrant methylation in DNA—global hypomethylation accompanied by an increase in methylation in CpG islands in the promoter regions of tumor-suppressor genes—is a common epigenetic phenomenon that occurs early in colorectal cancer (5, 6). In normal mucosa, both global hypomethylation (7) and an increase in CpG island methylation in genes such as estrogen receptor (ER), insulin- like growth factor 2 (IGF-2) (8–10), human N33 gene, myoblast determination protein 1 (10, 11), and mutL homolog 1 (MLH1) (10, 12) occur as a function of age in persons with or without neoplasia.
Age-related methylation with possible inactivation of tumor suppressor genes has been suggested as a mechanism for the strong relation between colorectal cancer and age (13). Age-related changes in the metabolism of folate and vitamin B-12 may explain the link between age and colorectal cancer risk. Older persons have both lower dietary intakes of folate and vitamin B-12 (14) and a greater prevalence of vitamin B-12 deficiency that is due to lower absorption of food-bound vitamin B-12. The hyperhomocysteinemia that occurs with aging can be partly explained by suboptimal folate and vitamin B-12 status (15). Lower folate and vitamin B-12 intakes may have a more pronounced effect on aging colonic mucosa; a folate-deplete diet reduced colonic folate concentrations in older but not younger rats (16), and a vitamin B-12–deficient diet caused DNA hypomethylation in rat colon (17). Aging has also been shown to decrease the activity of methionine synthase in the liver (18).
Some human studies suggest associations between global DNA methylation in the colon and folate or vitamin B-12 status in subjects with or without neoplasia (19, 20). Colorectal cancer subjects with low folate and high alcohol intakes had a greater prevalence of CpG island methylation in their tumor tissue in several genes than did those with high folate and low alcohol intakes (21), although the use of diseased tissue makes it difficult to separate this relation from confounding effects of the disease.
The main aim of the present study was to determine the relation between folate and vitamin B-12 status and CpG island methylation in normal colonic mucosa. A secondary aim was to determine whether the degree of methylation differs between subjects with and subjects without neoplasia. We chose to investigate ER and MLH1 because an increase in methylation in their promoters leads to diminished or absent protein product in colorectal tumors (8, 22) and because both genes display age-dependent increases in methylation in normal-appearing mucosa of patients with and patients without neoplasia (8, 10, 12). ER mediates the function of the steroid hormone 17ß-estradiol, a critical regulator of growth and differentiation. MLH1 is involved in the mismatch repair pathway, which recognizes and binds errors that occur during DNA replication.
SUBJECTS AND METHODS
Subjects
Subjects were patients referred for colonoscopy at the Department of Colorectal Surgery, King's College Hospital (London, United Kingdom), between 2000 and 2001. Exclusion criteria included a strong family history of colorectal cancer or adenomatous polyposis coli, inflammatory bowel disease, current or past gluten-sensitive enteropathy, and clinical or laboratory (or both) evidence of intestinal malabsorption, pregnancy, alcoholism, and the use of medication known to antagonize the metabolism of folate.
Subjects with colorectal neoplasms were characterized as follows: cancer (tumor but no distant metastases); adenoma (1 adenoma, that was tubular, tubulovillous, villous and serrated at current colonoscopy); hyperplastic (1 hyperplastic polyp at current colonoscopy but no history of adenoma). All tumor tissues were histologically confirmed. Subjects without tumors met the inclusion criteria if they showed no abnormality on full colonoscopy.
Weight and height were recorded, and information on smoking habits, physical activity levels, and current medication and supplement use was gathered before colonoscopy. One week after the colonoscopy, subjects completed a short, previously validated food-frequency questionnaire (23) to assess habitual intakes of alcohol and folate.
Written informed consent was obtained from all subjects. The protocol for the study was approved by the Research Ethics Committees at King's College Hospital and King's College London.
Collection and handling of blood and tissue samples
Fasting venous blood samples were obtained before colonoscopy for the determination of full blood count; the measurement of serum and red cell folate, serum vitamin B-12, and plasma homocysteine concentrations; the performance of liver function tests, and the identification of genetic polymorphisms in methyl pathway enzymes. Blood samples for a full blood count and measurement of red cell folate concentrations were collected in evacuated tubes containing EDTA (Vacutainer; Becton Dickinson, Rutherford, NJ); full blood counts were measured within 24 h of collection. Blood samples for serum and vitamin B-12 measurements were collected in plain evacuated tubes. Blood samples for the measurement of the plasma homocysteine concentration were collected in EDTA Vacutainers, which were chilled on ice and spun within 2 h of collection.
All assays were performed within 2 mo of collection of blood samples. During colonoscopy, 3 tissue samples of normal-appearing mucosa were removed from the rectum (12 cm from the anal verge), immediately snap-frozen in liquid nitrogen, and stored at –70°C.
Laboratory methods
Serum folate and vitamin B-12 concentrations were measured by competitive protein-binding enzyme immunoassays on an Immuno 1 analyzer (Bayer Diagnostics, Newbury, United Kingdom). Samples in EDTA for measurement of red blood cell folate were treated with a lysing reagent, and whole blood folate was measured as for serum folate. Plasma homocysteine was measured by using an Imx homocysteine assay (Abbott Laboratories, Abbott Park, IL). The between-run CV for both serum folate and vitamin B-12 was 4.6%, that for erythrocyte folate was 6.7%, and that for plasma homocysteine was 5.2%.
Global DNA methylation was measured by the in vitro methyl acceptance method with the use of [3H-methyl] S-adenosylmethionine as a methyl donor in the presence of the Sss I prokaryotic methylase enzyme (24). In this assay, the number of radiolabeled methyl groups incorporated into DNA is inversely proportional to DNA methylation status. DNA from pooled whole blood was used as a quality control. Nonmethylated phage DNA was used as a positive control, and DNA from the pooled blood sample above was methylated in vitro and used as a negative control. All data were within the range spanned by the positive and negative controls. The within-run and between-run CVs for this assay were 5.3% and 8.9%, respectively.
The laboratory assays used to identify genetic polymorphisms in methyl pathway enzymes [methylenetetrahydrofolate reductase (MTHFR C677T and A1298C), methionine synthase (MS A2756G), and cystathionine ß-synthase (CBS 844ins68)] were described in detail elsewhere (19). All of the above analyses were performed at the time the original study (19) was conducted (2000–2001).
DNA was extracted from tissue samples in 2001, and aliquots were stored at –70°C; these were used for the measurement of ER and MLH1 promoter methylation in the present study. Agarose gel electrophoresis was used to determine DNA size (>20 kb in all cases) and degradation (one sample was degraded, but that had no effect on subsequent measurements). DNA concentration was measured in all samples by using an ultraviolet spectrophotometer (ND-1000; Nano Drop Technologies, Wilmington, DC). All samples had ratios of A260 to A280 of >1.7.
Bisulfite modification of DNA (500 ng/sample) was carried out by using the EZ DNA Methylation-Gold Kit (Zymo Research, Orange, CA) according to the manufacturer's protocol. This converts unmethylated cytosines, by a process of deamination, to uracil, which leaves the methylated cytosines unchanged. The converted DNA was stored at –70°C until use. Normal human blood donor DNA was used as a negative control. The same DNA was methylated in vitro by using Sss1 (CpG) methylase (New England Biolabs, Beverly, MA), and it served as a positive control.
Regions of interest
The polymerase chain reaction (PCR) assays were designed to amplify a part of the gene promoter containing 7 CpG sites in the ER promoter (Figure 1) and 13 CpG sites in the MLH1 promoter (Figure 2). Because of the large number of sites that were analyzed, 2 sequencing primers were necessary for the MLH1 assay. PCR primer sequences and sequencing primer sequences for ER and MLH1 were designed and validated by Biotage AB (Uppsala, Sweden). Primers targeted CpG-free regions to ensure that the PCR product would proportionally represent the methylation characteristics of the template DNA. Sequencing primers were designed to cover as few CpG sites as possible. The targeted CpG sites were chosen to correlate with those assessed by methylation-specific PCR in earlier studies (8, 12). Cytosine and thymine are incorporated during pyrosequencing if the template CpG is methylated or unmethylated, respectively. Therefore the proportion of cytosine to thymine (C:T) is stoichiometrically proportional to the degree of methylation at that CpG site in the template DNA (Figure 1B and Figure 2B, 2C). A non-CpG cytosine was used as an internal control for the completion of the bisulfite treatment (Figure 2C).
FIGURE 1.. A) Bisulfite-converted sequence for the estrogen receptor (ER) gene promoter (ENST00000206249 Ensembl ID). Polymerase chain reaction (PCR) primers are shown in bold, the sequencing primer is shown in italics and underlined, and the CpG sites analyzed are shown in bold and with gray shading. The position t shaded in gray can be used as control for the completion of bisulfite treatment. A control for the completion of bisulfite treatment was not included in the ER assay; this enabled the analysis of more CpG sites. However, the assay was repeated on a subset of samples excluding the last 2 CpG sites by changing the sequence to analyze (to aygtygyggtygtygttaat) and nucleotide dispensation order. This confirmed completion of bisulfite treatment (data not shown). B) Representative pyrogram for the ER gene promoter.
FIGURE 2.. A) Bisulfite-converted sequence for the mutL homolog 1 (MLH1) gene promoter (ENST00000231790 Ensembl ID). Polymerase chain reaction (PCR) primers are shown in bold, the sequencing primers are shown in italics and underlined, and the CpG sites analyzed are shown in bold and with gray shading. The position t shaded in gray was used as control for the completion of bisulfite treatment. The CpG site shown in bold showed a very low degree of methylation in the positive control. A BLAST search was performed on the original sequence used for design, which showed that only some submitted sequences have a C in this position (and is therefore a CpG site), whereas others have a T. This finding suggested that the position is probably a C/T single-nucleotide polymorphism, and it was therefore excluded from the assay. B) A representative pyrogram for sequencing primer 1. C) Representative pyrogram for sequencing primer 2. The third position indicates the control for the completion of bisulfite treatment.
Methylation analysis with pyrosequencing
PCR and sequencing primers are shown in Figures 1A and 2A. For each gene, 10 pmol of forward and reverse primer and 50 ng of bisulfite-converted DNA template were used. The reaction was carried out in a 5-µL final volume and included 5 µL of 10x PCR buffer (containing 15 mmol MgCl2/L, 1.5 mmol MgCl2/L (25 mmol/L), 200 µmol of each dNTP/L, 1.25U HotStarTaq DNA polymerase (Qiagen, Valencia, CA). and nuclease-free water. The final MgCl2 concentration was 3 mmol/L. The same cycling conditions were used for the MLH1 and ER assays: denaturing at 95°C for 15 min; 45 cycles at 95°C for 20 s, at 54°C for 20 s, and at 72°C for 20 s; extension at 72°C for 5 min, and 4°C hold. The amplification cycle was repeated 45 times to ensure complete consumption of biotinylated primers, which can cause background signals during pyrosequencing. Gel electrophoresis was carried out on all PCR products by using 10 µL of PCR product, which was loaded onto a 1.5% (wt:vol) agarose gel. The gels were then visualized by using an ultraviolet transilluminator (Alpha Imager; Alpha Innotech Corp, San Leandro, CA).
Pyrosequencing was carried out by using the PSQ HS System (Biotage AB). Sample preparation was carried out by using the Vacuum Prep Tool (Biotage AB) according to the manufacturer's protocol. Briefly, 10 µL of PCR product (biotinyated strand only) was immobilized to 2 µL Streptavidin Sepharose HP beads [GE Healthcare (formerly Amersham Biosciences), Piscataway, NJ] by using the Vacuum Prep Tool. The beads were then released onto a PSQ HS 96 plate containing 10 pmol sequencing primer. Annealing was carried out for 2 min at 80°C. The PSQ HS system was used with the nucleotide dispensation order assigned for the assays. All equipment was handled according to standard procedures, and the analysis criteria were specified by factory settings. Within- and between-run CVs for both assays were 1.6% and 5%, respectively.
Statistical analysis
Data were analyzed by using SPSS software (version 13.0; SPSS Inc, Chicago, IL). Between-group differences in categorical variables (ie, sex, smoking status, supplement use, and genotypes) were compared by using chi-square tests or Fisher's exact test. Differences in single continuous variables between groups were compared by using analysis of variance with Dunnett's test to adjust for multiple comparisons (ie, age, dietary folate intake, serum folate, and global DNA methylation) or Kruskall-Wallis tests followed by Mann-Whitney U tests for subgroup comparisons (comparing each disease group with the disease-free group) when the data did not have a normal distribution or when there was a violation of the equality of variance assumption (body mass index, alcohol intake, and serum vitamin B-12, red cell folate, and plasma homocysteine concentrations). For the Mann-Whitney U tests, Bonferroni corrections were applied to adjust the P values for multiple comparisons (we corrected for the 3 comparisons of each disease group with the disease-free group).
A methylation index was generated to reflect an average methylation value across all CpG dincleotides studied at each gene promoter by calculating the mean percentage methylation in all CpG sites in the sequence. The relation between the ER and MLH1 methyation indexes and age, markers of folate status (dietary folate intake; serum and red cell folate, serum vitamin B-12, and plasma homocysteine concentrations; and global DNA methylation) were examined in all subjects by using partial correlation coefficients with adjustment for disease status and in each disease group by using Pearson (for age, dietary folate intake, and serum folate concentration) or Spearman (for serum vitamin B-12, red cell folate, and plasma homocysteine concentrations) correlation coefficients. Between-group differences in ER and MLH1 methylation index were compared by using Kruskall-Wallis tests followed by Mann Whitney U tests for subgroup comparisons (with Bonferroni corrections to adjust for multiple comparisons).
A general linear model was used to test the effect of several variables (ie, disease status, age, sex, body mass index, smoking status, supplement use, alcohol intake, folate intake, serum and red cell folate, serum vitamin B-12, plasma homocysteine, and genotype) on ER and MLH1 methylation indexes. Univariate analysis of variance or covariance was used as appropriate. Estimated marginal means were calculated where necessary.
RESULTS
Data were available from 156 subjects (28 with cancer, 35 with adenoma, 17 with hyperplastic polyps, and 76 who were disease free). The basic characteristics of participants and biomarkers of folate and vitamin B-12 by disease status are shown in Table 1. On average, patients with adenoma and cancer were significantly older and patients with cancer had a significantly lower body mass index than did disease-free patients (P < 0.01 for both). There were significantly (P < 0.01) more smokers in the hyperplastic polyp group than in the disease-free group. Reported supplement use and alcohol intakes were significantly higher in the adenoma group than in the disease-free group (P = 0.01 and P < 0.001, respectively).
View this table:
TABLE 1. Characteristics of subjects by disease status1
Patients with cancer had significantly lower serum folate (P = 0.002), lower serum vitamin B-12 (P = 0.006), and higher plasma homocysteine (P = 0.02) concentrations and higher global DNA hypomethylation in normal-appearing colonic mucosa (P < 0.01) than did patients who were disease free. However, after adjustment for age and MTHFR C677T genotype, plasma homocysteine did not differ significantly between patients with cancer and patients who were disease free (P > 0.05). Patients with adenoma had significantly (P < 0.001) higher global DNA hypomethylation in normal-appearing colonic mucosa than did disease-free patients.
The percentage methylation by CpG site and the average percentage methylation for all sites (methylation index) for ER are shown in Table 2. The median methylation index for ER in the normal-appearing colorectal mucosa differed significantly between patients with adenoma or hyperplastic polyps and disease-free patients (adenoma: 14.6% and 8.1%, respectively; P < 0.001; hyperplastic polyps: 12.3% and 8.1%, respectively; P = 0.02). The median methylation index was marginally higher in patients with cancer than in disease-free subjects group (9.6% and 8.1%, respectively), but the difference between groups was not statistically significant.
View this table:
TABLE 2. Estrogen receptor promoter methylation in normal-appearing colorectal mucosa of subjects with and without colorectal neoplasia1
There were significant positive correlations between ER methylation index and age in all groups (all subjects: r = 0.592, P < 0.001; s: r = 0.626, P < 0.001; cancer patients: r = 0.734, P < 0.001; adenoma patients: r = 0.546, P = 0.01; hyperplastic polyp patients: r = 0.586, P = 0.01). There was a significant negative correlation between ER methylation index and serum vitamin B-12 concentrations in all subjects (r = –0.239, P = 0.003). When analysis was conducted by group, the correlation between ER methylation index and serum vitamin B-12 concentrations was significant in disease-free subjects (r = –0.289, P = 0.01) and adenoma patients (r = –0.481, P = 0.003) but not in cancer (r = –0.217, P = 0.29) or hyperplastic polyp (r = –0.152, P = 0.56) patients, for whom sample sizes were smaller. There was a significant positive correlation between ER methylation index and plasma homocysteine in all subjects (r = 0.188, P = 0.021). When analyzed by group, the correlation between ER methylation index and plasma homocysteine was significant in disease-free subjects (r = 0.286, P = 0.01) and adenoma (r = 0.356, P = 0.04) and hyperplastic polyp (r = 0.755, P < 0.001) patients but not in cancer patients (r = 0.312, P = 0.12). There was a weak correlation between global DNA hypomethylation and ER methylation index in all subjects (r = 0.139, P = 0.09). There were no significant correlations between ER methylation index and serum folate or red cell folate (P > 0.05).
There was heterogeneity in the level of percentage methylation among the 7 sites in all groups (P < 0.001, Friedman test). Analysis of covariance was used to assess the effect of disease status and of markers of folate and vitamin B-12 status on the ER methylation index. The effect of individual factors and their interactions were considered. The model showed disease status as a significant indicator of ER methylation (P < 0.005). The other significant factors were age (P < 0.001) and serum vitamin B-12 concentrations (P = 0.006). Serum and red cell folate and plasma homocysteine concentrations and genotype had no influence on the ER methylation index.
The adjusted means of ER methylation index by disease status are shown in Table 3. When age and vitamin B-12 status were accounted for, there were no significant differences between adenoma and cancer patients disease-free subjects. The adjusted mean difference between patients with hyperplastic polyps and subjects without neoplasia was 4.8% (95% CI: 0.78, 8.90; P = 0.01, Sidak's correction for multiple comparisons), which indicated that factors other than age and vitamin B-12 status influenced ER methylation in this group.
View this table:
TABLE 3. Adjusted means of estrogen receptor methylation index (%) in normal-appearing colorectal mucosa according to disease status1
To avoid potential confounding by disease status, statistical analysis was repeated only in patients without colorectal neoplasia to establish whether age and vitamin B-12 status are determinants of ER methylation index. The analysis of covariance model confirmed the previous analysis, showing that age (P < 0.001) and reduced vitamin B-12 status (P = 0.037) were significantly associated with a rising methylation index. The statistical model was then applied to the individual CpG sites. Age was the only significant determinant of methylation at sites 1, 2, and 7 (P < 0.001), whereas both age (P < 0.001) and serum B-12 concentrations (P = 0.030, 0.019, 0.041, and 0.041 for sites 3–6, respectively) were found to be significant at the other 4 sites.
Investigation of the 13 CpG sites in the MLH1 promoter region showed median methylation levels of <2% (Table 4). The differences between subjects with and without neoplasia were significant (P < 0.001). There was a significant correlation between MLH1 promoter methylation and age in all subjects (r = 2.54, P = 0.002) but no significant correlations with markers of folate or vitamin B-12 status. A detailed statistical analysis was deemed unsuitable because of the low degree of methylation across all sites.
View this table:
TABLE 4. MutL homolog 1 promoter methylation in normal-appearing colorectal mucosa of subjects with and without colorectal neoplasia1
DISCUSSION
This study tested the hypothesis that folate and vitamin B-12 status is associated with CpG island methylation in normal-appearing colorectal mucosa. Dietary folate intakes and red cell folate concentrations did not differ between subjects with and without disease, although serum folate and vitamin B-12 concentrations were significantly lower in the cancer group than in the disease-free group. We identified methylation in 7 CpG sites in the ER promoter and 13 CpG sites in the MLH1 promoter. DNA was extracted from rectal tissue for biopsy that was removed far from any lesions with clean biopsy forceps. We combined the methylation values across all CpG dinucleotides studied into a single methylation index for each gene. The use of such an index has not previously been validated, but the values for each site showed a similar pattern in all groups (see Tables 2 and 4), and, when the statistical analysis was repeated for each CpG site, the results were not altered substantially.
We confirmed the strong relation between ER methylation and age that was first reported by Issa et al (8). We also found a positive correlation between ER methylation and plasma homocysteine concentrations. Serum vitamin B-12 status was an independent determinant of ER methylation in the colon of persons with and persons without neoplasia. This observation is intriguing, particularly because vitamin B-12 deficiency is relatively common in the aging population: it is estimated to affect 10%–15% of people aged >60 y (25), and age is the strongest risk factor for colorectal cancer. It would have been interesting to confirm these findings by using vitamin B-12 intakes, but no dietary data on vitamin B-12 were available.
Compared with evidence for the role of folate, that for the role of vitamin B-12 in colorectal neoplasia is limited. Early studies indicated a positive association between pernicious anemia and the risk of colorectal cancer (26). More recently, Harnack et al (27) showed that the incidence of proximal colon cancer was lowest in those with high vitamin B-12 and folate intakes; this effect was not observed with high folate intake alone. Other studies suggest no relation between dietary vitamin B-12 intake and colorectal cancer risk (28).
An inadequate supply of vitamin B-12 impairs methionine synthase activity, which converts homocysteine to methionine with methyltetrahydrofolate as the methyl donor. Cellular folates accumulate in the methylfolate form but cannot be used, which creates a form of folate deficiency known as the "methylfolate trap" (29). Rats fed vitamin B-12–deplete diets had a lower proportion of tetrahydrofolate and a higher proportion of methyltetrahydrofolate in their colon, which is indicative of the methylfolate trap, and that led to global DNA hypomethylation (17). Piyathilake et al (30) found lower vitamin B-12 concentrations and less hypomethylation in squamous cell lung cancer than in adjacent, normal-appearing mucosa. Pufulete et al (19) reported an inverse relation between serum vitamin B-12 and global DNA hypomethylation in normal-appearing colonic mucosa of subjects with and without neoplasia, although the methyl acceptance assay that was used to measure global DNA methylation is indirect and semiquantitative.
In neoplasia, global hypomethylation and an increase in methylation in CpG-rich promoter regions of tumor-suppressor genes coexist. It has been suggested that a genome-wide decrease in methylation, possibly caused by low methyl group availability, may lead to a reshuffling of methyl groups and may trigger local de novo methylation in susceptible genes (31). The present study suggests that, as ER promoter methylation increases, so does the extent of global hypomethylation. Both ER methylation and global hypomethylation were negatively associated with serum vitamin B-12 concentrations. A limitation of the present study is that, because of limited availability of colorectal tissue samples, we did not measure folate and vitamin B-12 concentrations in the colon.
We investigated 13 CpG sites in MLH1, 700 bp upstream of the promoter. In tumor tissue, methylation in the proximal region correlates better with transcriptional silencing and the loss of protein (10, 22, 32) than does methylation in the upstream region (22). We chose the latter because methylation in this region has been shown in normal mucosa from subjects with colorectal cancer (12, 33) and has been associated with age in some (10, 12) but not all (34) studies. Nakagawa et al (12) analyzed the entire 700-bp region upstream of MLH1, which covered 51 CpG sites, and found an age-dependent increase in methylation in the normal-appearing mucosa of 123 subjects with sporadic colorectal cancer. We also observed increasing methylation with age in the present study, but median methylation levels were low (<2%) across all groups, which confirmed findings from a recent study (10). We did not perform-in depth statistical analyses because it is not certain that the observed level of methylation is biologically relevant or that the pyrosequencing data are reliable at the quantified level of methylation.
In the present study, ER and MLH1 methylation in subjects with neoplasia was higher than that in disease-free subjects. Subjects with neoplasia also showed more global DNA hypomethylation than did disease-free subjects. The findings in the present study contradict those of Ye et al (35), who found no difference in methylation in 6-O-methylguanine–DNA methyltransferase, MLH1, and cyclin-dependent kinase inhibitor 2A (CDKN2A/p16) in the healthy colorectal mucosa of subjects with or without adenoma. However, in subjects with ulcerative colitis, ER methylation in nonneoplastic mucosa was higher in those with neoplasia than in those without (36–38). In the present study, subjects with cancer, adenoma, and hyperplastic polyps had 19%, 80%, and 52%, respectively, more ER methylation than did disease-free subjects. Cancer subjects did not have the highest ER methylation (as they may be expected to have), although there was wide variation in methylation in those subjects, which, combined with the small sample size, may be responsible for the observed results.
It is surprising that subjects with hyperplastic polyps had high ER methylation in their normal mucosa, which was not explained by age and vitamin B-12 status. Hyperplastic polyps have been classified as nonneoplastic lesions lacking in malignant potential and until recently have largely been ignored in DNA methylation studies. However, reports have linked the condition hyperplastic polyposis, characterized by multiple hyperplastic polyps in the colon, with colorectal cancer (39, 40). There is evidence to suggest that some colon cancers arise from hyperplastic polyps that evolve into a serrated adenoma (41). Three reports have shown extensive DNA methylation in the normal mucosa of subjects with hyperplastic polyposis (42–44). However, none of the 17 subjects with hyperplastic polyps in the present study were diagnosed with hyperplastic polyposis, and yet the extent of ER promoter methylation was similar to that in subjects with adenoma and cancer. Future studies investigating the role of CpG island methylation in sporadic hyperplastic polyps and subsequent risk of colorectal cancer are warranted.
In conclusion, this study suggests that age, serum vitamin B-12 concentrations, and the presence of benign or neoplastic lesions are independent determinants of ER promoter methylation in colorectal mucosa. Markers of folate status and genetic polymorphisms in methyl pathway enzymes appear to have no influence on ER methylation. Future studies involving a larger sample size and a wider panel of genes associated with colorectal neoplasia should investigate a potential association with folate and vitamin B-12 status. To establish a link between methylation and carcinogenesis, future studies also should evaluate the effect of promoter methylation on transcription (which was not possible in the present study because of limited availability of tissue biopsies for RNA extraction) to establish a link between methylation and carcinogenesis. The findings from this study should be confirmed with more sensitive assays of vitamin B-12 status (eg, holotranscobalamin and methylmalonic acid) and global DNA methylation (eg, liquid chromatography–mass spectrometry) because the extent to which the methods used contribute to the observed results is not clear, and discrepancies between studies may be partly due to differences in the methods employed.
ACKNOWLEDGMENTS
We thank Monica Pettersson at Biotage for invaluable assistance with the methylation assays.
The authors' responsibilities were as follows: RA-G: experiment design, patient recruitment, collection of data, statistical analysis, and manuscript writing; JP and RF: collection of data; NH: manuscript writing; and MP: experiment design, patient recruitment, and manuscript writing. None of the authors had a personal or financial conflict of interest.
REFERENCES