The MTHFR C677T and ΔDNMT3B C-149T polymorphisms confer different risks for right- and left-sided colorectal cancer
Etiological risk factors for proximal (right-sided) colon cancers may be different to those of distal colon and rectal (left-sided) cancers if these tumors develop along distinct pathways. The CpG Island Methylator Phenotype (CIMP+) occurs in approximately 15% of colorectal cancers (CRC) and predominantly in the proximal colon. CIMP+ tumors have frequent methylation of gene promoter regions and increased tissue folate levels. The aim here was to determine whether polymorphisms in 2 genes involved in cellular methyl group metabolism were associated with different risks for right- and left-sided CRC. This population-based case–control study involved 859 incident cases of CRC and 973 sex and age-matched controls. Information on dietary folate and alcohol intake was obtained from food frequency questionnaires and information on the anatomical site of tumors from pathology reports. DNA was collected using FTA cards and genotyping performed for the MTHFR C677T and ΔDNMT3B C-149T polymorphisms. The MTHFR 677 T allele was associated with increased risk for proximal colon cancer (adjusted odds ratio, AOR = 1.29) but decreased risk for distal cancers (AOR = 0.87). The increased risk for proximal cancers was especially pronounced in older individuals (AOR = 1.49) and those with a low folate diet (AOR = 1.67) or high alcohol consumption (AOR = 1.90). The ΔDNMT3B-149 TT genotype was protective against proximal colon cancers (AOR = 0.65), but showed no association with the risk of distal colon and rectal cancers (AOR = 1.02). Epidemiological studies on dietary and genetic risk factors for CRC should take into account these may confer different risks for right- and left-sided tumors. © 2009 UICC
There is increasing evidence that etiological factors contributing to the risk of proximal colon cancers are different to those of distal colon and rectal cancers.1–3 This is likely to be due to the existence of at least 2 distinct pathways for colorectal tumorigenesis. The most common pathway accounts for approximately 80% of all colorectal cancers (CRC) and is characterized by high frequencies of chromosomal instability (CIN) and APC, p53 and KRAS gene mutations.4 These tumors progress via the adenoma-carcinoma sequence and arise typically in the left-sided colon (distal colon and rectum) of younger males. The CpG island methylator phenotype (CIMP) represents a second pathway of CRC development, which is characterized by frequent hypermethylation of gene promoter regions and activating mutations in the BRAF oncogene.5–9 CIMP+ CRCs occur predominantly in the right-sided (proximal) colon of older women and are thought to arise from serrated adenomas.4 Depending on the markers used to define this phenotype, it has been estimated that approximately 15–25% of CRC are CIMP+ and that 70–90% of CIMP+ tumors arise in the proximal colon.5–10
The microsatellite instability (MSI+) phenotype occurs in approximately 10–15% of CRC and was initially thought to represent an additional pathway for CRC development.11 However, it is now clear the MSI+ phenotype can arise in both the CIN and CIMP pathways and is unlikely to be the initial driving force for the development of sporadic CRC. In unselected CRC series, the majority of MSI+ CRCs arise in the proximal colon11 and share the CIMP+ phenotype.6–9
Because of the frequent hypermethylation of gene promoter regions found in CIMP+ tumors, the etiology of this phenotype could be expected to involve features of folate and methyl group metabolism. These include dietary intake levels for folate, methionine and alcohol. Common genetic variants in folate metabolizing enzymes and DNA methyltransferases could also be associated with the risk of developing CIMP+ CRC. The common C677T polymorphism in the key enzyme methylenetetrahydrofolate reductase (MTHFR) has been widely investigated as a possible risk factor for CRC due to the markedly reduced activity of this variant. A meta-analysis of 25 studies found that individuals with the MTHFR 677TT genotype were at moderately reduced risk of CRC compared with individuals with the 677CC genotype.12 Evidence for genotype-diet interactions was presented in another review of this polymorphism.13 Interestingly, however, the MTHFR 677TT genotype was associated with an increased risk for MSI+ CRC14–17 and for CIMP+ CRC arising in females7 or in the proximal colon.18 In view of the close overlap between MSI+ and CIMP+ tumors,6–9 these observations suggest the MTHFR 677TT genotype may be associated with opposite risks for the 2 major CRC pathways represented by CIN and CIMP, or between MSI+ and MSI- tumors as reported by Hubner et al.17
Polymorphisms in DNA methyltransferases have received considerably less attention as potential low risk genetic factors for CRC than folate metabolizing enzymes. The most extensively studied variants have been in the DNMT3b gene responsible for de novo methylation of DNA. The combined GG/GT genotypes for the DNMT3b G-579T polymorphism were associated with decreased risks of colon cancer in younger and male individuals from a Korean population.19 The TT genotype for the variant initially described as DNMT3b C-149T was linked to an increased risk for CRC polyps amongst individuals with low intake of folate and methionine.20 This polymorphism was recently reported by Reeves et al.21 to be -149 bp from the transcription start site of the ΔDNMT3b isoform (position 1570351, accession no. NT_028392) and not the regular DNMT3b gene as described in earlier studies.
Almost all previous studies on genetic risk factors for CRC, including the more recent genome-wide association studies, have not considered the possibility of risk factor differences according to tumor location in the large bowel. Accurate information on tumor subsite has not been available for most study cohorts to date. The aim of this population-based case–control study was to determine whether the MTHFR C677T and ΔDNMT3b C-149T polymorphisms were associated with different risks for right- and left-sided CRC. Information on dietary folate and alcohol intake was obtained by food frequency questionnaire and allowed us to investigate for possible interactions with the risk of CRC associated with the MTHFR and ΔDNMT3b polymorphisms.
Material and methods
Eligible cases were histologically confirmed incident cases of CRC, aged 40–79 years, resident in Western Australia and notified to the Western Australian Cancer Registry between June 2005 and August 2007. A population-based screening program for hereditary nonpolyposis CRC operating in Western Australia22 allowed a small number of familial cases to be excluded from the study. Permission was sought from clinicians to contact their patient for entry into the study. Invitations to the study were sent by post to all patients whose clinician did not object. Those who consented were then sent an FTA card for DNA collection (Whatman International Ltd., Maidstone, UK) and 2 questionnaires. The first was a lifestyle questionnaire that obtained data on area of residence, ethnicity, smoking history, dietary supplement use, weight, and height. The second was a Food Frequency Questionnaire (FFQ) from the Cancer Council of Victoria (http://www.cancervic.org.au/about-our-research/cancer_epidemiology_centre/nutritional_assessment_services) modified to address usual consumption 10 years earlier (1995–1997). A total of 1,719 eligible cases were identified, but 98 died before or soon after invitation and 39 were excluded because permission was denied by the clinician. Of the remaining 1,582 cases invited to participate, 1,034 (65%) gave consent and 918 (58%) returned the dietary questionnaires. Nonparticipating cases were more likely to be female and to be in the youngest and oldest age groups.
Controls were persons without a diagnosis of CRC and were randomly selected from the West Australian electoral roll. Registration on the electoral roll is compulsory for Australian citizens. Controls were frequency matched for age and sex based on the approximate distribution of CRC in Western Australia for 2002, as reported by the West Australian Cancer Registry. The participation rate amongst the 2,198 eligible controls that were approached was 1,021 (46%). Ethics approval for the study was obtained from the Human Research Ethics Committee at the University of Western Australia and the Confidentiality of Health Information Committee, West Australian Department of Health.
Demographic and socioeconomic information for cases and controls was obtained by questionnaire and is shown in Table I. Socioeconomic status was based upon the participants' address at the time of diagnosis or enrolment in the study and was determined from the Socioeconomic Index for Areas (SEIFA) produced by the Australian Bureau of Statistics. Comparisons were made across quintiles of the SEIFA disadvantage index based upon the distribution amongst controls.
Table I. Characteristics of Study Participants
|Total||973|| ||859|| |
| ≤65 yr||500||51||446||52|
| >66 yr||473||49||413||48|
| 0 (most disadvantaged)||191||20||209||24|
| 4 (least disadvantaged)||190||19||160||19|
|Country of birth|
| Australia/New Zealand||647||67||538||63|
| United Kingdom/Ireland||214||22||198||23|
|Highest Education Level|
| High School only||405||42||378||44|
| Vocational training||421||43||342||40|
Data on alcohol consumption from the FFQ were adjusted for energy intake.23 High alcohol intake was classified as ≥30 g/day and no or low alcohol intake as <30 g/day based on previous studies that showed an increase in CRC risk at levels of >30 g/day.24 Dietary folate intake was calculated from the food composition data and was adjusted for energy intake.23 Low and high folate intakes were classified as below and above/equal to the median intake of 305 μg/day, respectively. Vitamin supplements were not included as food items.
Information on tumor pathology (site, grade, stage) for each case was obtained from the diagnostic report and is shown in Table II. Proximal (right-sided) CRC was defined as carcinoma occurring in the cecum, ascending colon, hepatic flexure and transverse colon, whereas distal (left-sided) CRC was defined as arising in the splenic flexure, descending colon, sigmoid colon, rectosigmoid junction and rectum.
Table II. Pathological Features of CRC Cases
| Proximal (right)||280||33|
| Distal (left)||579||67|
|AJCC tumor stage|
Buccal DNA was collected from consenting participants using the FTA card method as recommended by the manufacturer (Whatman International). Fresh material from the buccal scrape was transferred from the applicator onto the indicating circle of the FTA card. Single 2-mm diameter punches were taken from each subject's FTA card and prepared for PCR by washing 3 times with 200 μl FTA purification reagent and twice with 200 μl TE buffer. FTA cards are a robust DNA collection method and were recently shown to produce PCR-quality DNA more reliably than buccal swabs.25 The PCR failure rate in the present study was 4.2% for MTHFR C677T and 5.6% for ΔDNMT3b C-149T. Genotyping was successful in 850 cases and 958 controls for MTHFR C677T and in 828 cases and 949 controls for ΔDNMT3b C-149T. The fluorescence-based single strand conformation polymorphism technique described earlier was used to genotype the MTHFR C677T polymorphism.26 Genotyping for the ΔDNMT3b C-149T polymorphism was performed by restriction fragment length polymorphism analysis using the AvrII restriction enzyme (New England BioLabs, Brisbane) and PCR primers described previously.27 The quality of genotyping methods were evaluated in a series of 80 duplicate punch samples from FTA cards, with 100% concordance obtained.
Statistical analyses were performed using Stata version 9 (StataCorp LP, College Station, TX). Odds ratios were calculated using maximum-likelihood multinomial logistic regression. Each model was age- and sex-adjusted because of the frequency matching. Differences in the coefficient for each risk factor were tested across proximal and distal CRC groups using postestimation commands.28
A total of 1,939 eligible participants gave consent to participate in this study and completed the questionnaires (1,021 controls and 918 cases). Of these, 16 (0.8%) did not consent to DNA testing. Five participants were excluded from the analysis because tumor subsite was not recorded and 34 were excluded because of unreasonable energy values (<3,000 kJ/d or >21,000 kJ/d). Genotyping failed in 52 participants for both polymorphisms (2.7%), in 24 participants for the MTHFR polymorphism only and in 55 participants for the ΔDNMT3b polymorphism only. A total of 1,832 participants were therefore included in the study, with 1,808 (958 controls and 850 cases) having complete data for MTHFR and 1,777 (949 controls and 828 cases) having complete data for ΔDNMT3b.
Cases were well matched to controls for gender, age, socioeconomic status, country of birth, educational level and smoking status (Table I). One third of cases were classified as proximal and two thirds as distal CRC (Table II). Allele frequencies for the MTHFR C677T and ΔDNMT3b C-149T polymorphisms were in Hardy-Weinberg equilibrium in both the case and control groups.
The combined MTHFR CT/TT genotypes were associated with an increased risk for proximal CRC but with a decreased risk for distal CRC (Table III). Although these associations failed to reach significance for the individual tumor site groups, the difference between proximal and distal CRC was significant (p = 0.03). The ΔDNMT3b TT genotype was associated with a significantly reduced risk for proximal (AOR = 0.65, 95% CI 0.44–0.98, p = 0.03), but not for distal CRC.
Table III. Associations Between the MTHFR C677T and ΔDNMT3B C-149T Genotypes and Risk of Proximal and Distal CRC After Adjustment for Age and Sex
|MTHFR||958|| ||275|| || ||575|| || || |
| CT||429||45||137||50||1.31 (0.98–1.74)||249||43||0.90 (0.72–1.11)|| |
| TT||101||10||31||11||1.23 (0.78–1.94)||51||9||0.77 (0.53–1.12)|| |
| CT/TT||530||55||168||61||1.29 (0.98–1.70)||300||52||0.87 (0.71–1.07)||0.03|
|DNMT3b||949|| ||273|| || ||555|| || || |
| CT||463||49||139||51||0.91 (0.67–1.23)||275||50||1.05 (0.82–1.34)|| |
| TT||212||22||45||16||0.65 (0.44–0.98)2||122||22||1.02 (0.75–1.37)|| |
| CT/TT||675||71||184||67||0.83 (0.62–1.10)||397||72||1.04 (0.82–1.31)||0.35|
There were no significant associations between dietary folate intake and the risk of proximal (AOR = 1.03, 95% CI 0.79–1.35) or distal CRC (AOR = 1.09, 95% CI 0.88–1.34). High alcohol intake was associated with increased risks for both proximal (AOR = 1.27, 95% CI 0.83–1.95) and distal CRC (AOR = 1.21, 95% CI 0.87–1.66), although neither reached significance.
No significant associations were found for interactions between the 2 polymorphisms and sex, age, dietary folate intake or alcohol consumption. However, stratified analyses suggested the different risks for proximal and distal CRC associated with the MTHFR C677T polymorphism were more pronounced in females, in older individuals and in individuals with low folate or high alcohol intakes (Table IV). Interestingly, the MTHFR C677T polymorphism was not associated with an increased risk for proximal CRC in individuals with high folate intake.
Table IV. Stratified Analyses for Associations Between MTHFR C677T Genotype and Risk of Proximal and Distal CRC
| CT||169||43||61||48||1.36 (0.89–2.09)||80||38||0.80 (0.56–1.14)|| |
| TT||39||10||16||12||1.48 (0.76–2.87)||18||9||0.76 (0.41–1.40)|| |
| CT||260||46||76||52||1.24 (0.84–1.83)||169||46||0.94 (0.71–1.24)|| |
| TT||62||11||15||10||1.03 (0.55–1.96)||33||9||0.77 (0.48–1.23)|| |
| CT/TT||322||57||91||62||1.20 (0.83–1.75)||202||55||0.91 (0.70–1.18)||0.37|
| CT||231||47||58||50||1.13 (0.74–1.73)||141||43||0.82 (0.61–1.10)|| |
| TT||53||11||11||9||0.94 (0.45–1.93)||30||9||0.76 (0.46–1.24)|| |
| CT/TT||284||58||69||59||1.09 (0.72–1.65)||171||52||0.81 (0.61–1.07)||0.23|
| CT||198||43||79||50||1.48 (1.0–2.18)||108||44||0.97 (0.70–1.34)|| |
| TT||48||10||20||12||1.52 (0.84–2.76)||21||8||0.79 (0.45–1.39)|| |
| CT/TT||246||53||99||62||1.49 (1.03–2.15)4||129||52||0.93 (0.68–1.28)||0.06|
| CT||211||44||71||52||1.61 (1.06–2.45)4||119||44||0.99 (0.72–1.36)|| |
| TT||45||9||18||13||1.95 (1.03–3.70)4||25||9||0.95 (0.56–1.64)|| |
| CT/TT||255||53||89||65||1.67 (1.12–2.50)4||144||53||0.98 (0.73–1.34)||0.03|
| CT||218||45||66||47||1.06 (0.71–1.58)||130||43||0.80 (0.59–1.09)|| |
| TT||56||12||13||9||0.78 (0.40–1.54)||26||8||0.63 (0.37–1.05)|| |
| CT/TT||275||57||79||56||1.00 (0.68–1.47)||156||51||0.76 (0.57–1.03)||0.17|
| CT||341||45||105||48||1.18 (0.86–1.63)||178||42||0.82 (0.63–1.04)|| |
| TT||88||12||25||12||1.08 (0.65–1.80)||38||9||0.68 (0.44–1.03)|| |
| CT/TT||429||57||130||60||1.16 (0.85–1.58)||216||51||0.79 (0.62–0.99)||0.04|
| CT||88||44||32||56||1.85 (0.97–3.51)||71||48||1.16 (0.75–1.82)|| |
| TT||13||7||6||11||2.27 (0.76–6.77)||13||9||1.46 (0.64–3.37)|| |
| CT/TT||101||51||38||67||1.90 (1.02–3.55)4||84||57||1.20 (0.78–1.85)||0.12|
Stratified analyses for the ΔDNMT3b C-149T polymorphism showed the TT genotype was most protective against proximal colon tumors in older individuals and in those with low folate intake (Table V). A possible gender difference was apparent for the risk of distal CRC associated with the ΔDNMT3b TT genotype (AOR = 1.59 for women vs. AOR = 0.75 for males), but did not reach statistical significance. In contrast to MTHFR C677T, stratification for alcohol intake did not affect the risk estimates for proximal CRC.
Table V. Stratified Analyses for Associations Between ΔDNMT3b C-149T Genotype and Risk of Proximal and Distal CRC
| CT||194||49||65||51||0.89 (0.56–1.41)||97||48||1.22 (0.80–1.86)|| |
| TT||85||22||21||17||0.70 (0.38–1.27)||56||28||1.59 (0.98–2.59)|| |
| CT/TT||279||71||86||68||0.83 (0.54–1.29)||153||76||1.33 (0.90–1.98)||0.16|
| CT||269||48||74||51||0.90 (0.59–1.36)||178||50||0.96 (0.71–1.31)|| |
| TT||127||23||24||16||0.62 (0.36–1.07)||66||19||0.75 (0.51–1.10)|| |
| CT/TT||396||71||98||67||0.81 (0.55–1.20)||244||69||0.89 (0.67–1.20)||0.51|
| CT||226||46||54||48||0.97 (0.61–1.55)||155||48||1.09 (0.78–1.51)|| |
| TT||114||22||23||20||0.81 (0.45–1.45)||73||23||1.02 (0.69–1.52)|| |
| CT/TT||340||68||77||68||0.92 (0.59–1.42)||228||71||1.06 (0.78–−1.45)||0.81|
| CT||237||51||85||53||0.86 (0.57–1.29)||120||51||0.99 (0.69–1.44)|| |
| TT||98||21||22||14||0.54 (0.31–0.94)4||49||21||0.98 (0.62–1.54)|| |
| CT/TT||335||72||107||67||0.76 (0.52–1.13)||169||72||0.99 (0.70–1.41)||0.37|
| CT||239||50||59||45||0.60 (0.39–0.94)4||124||47||0.97 (0.67–1.40)|| |
| TT||109||23||24||18||0.52 (0.30–0.92)4||71||27||1.23 (0.81–1.88)|| |
| CT/TT||347||73||83||63||0.58 (0.38–0.87)4||195||74||1.05 (0.75–1.49)||0.02|
| CT||224||47||80||57||1.31 (0.85–2.03)||151||52||1.13 (0.81–1.59)|| |
| TT||103||22||21||15||0.77 (0.43–1.40)||51||18||0.82 (0.53–1.27)|| |
| CT/TT||328||69||101||72||1.15 (0.75–1.74)||202||70||1.04 (0.75–1.43)||0.81|
| CT||362||48||109||51||0.93 (0.66–1.31)||203||49||1.10 (0.83–1.46)|| |
| TT||165||22||35||16||0.67 (0.42–1.06)||93||23||1.10 (0.78–1.54)|| |
| CT/TT||527||70||144||67||0.85 (0.61–1.18)||296||72||1.10 (0.84–1.43)||0.34|
| CT||101||51||30||51||0.77 (0.39–1.51)||72||50||0.86 (0.51–1.43)|| |
| TT||47||23||10||17||0.58 (0.23–1.32)||29||20||0.74 (0.40–1.38)|| |
| CT/TT||148||74||40||68||0.70 (0.37–1.33)||101||70||0.82 (0.51–1.33)||0.50|
One of the major strengths of this study was that the anatomical site of origin in the large bowel for the 859 CRC cases was accurately determined from the pathology report. One third (n = 280) of cases were classified as right-sided and two thirds (n = 579) as left-sided CRC. Both the cases and controls were recruited from population-based samples identified through the Western Australian Cancer Registry and the Western Australian Electoral Roll, respectively. Participation rates of 58% for cases and 46% for controls are typical for this type of study and were similar to rates of 62% and 61%, respectively, reported recently in a comparable UK study of 264 cases and 408 controls.29 As shown in Table I, cases and controls were well matched for gender, age, socioeconomic index, country of birth and education level. However, we cannot exclude that some selection bias may have occurred because of the lower participation rates of females and of younger and older age groups.
One limitation of this study was the potential for misclassification in the assessment of folate. The FFQ used in this study was developed for the ethnically diverse Australian population and has been validated for a large number of nutrients.30 Although it has not been validated specifically for folate consumption, the median folate intake (adjusted for energy) of 305 μg/day was comparable with another Australian survey.31 Folate levels were determined from the Australian Composition tables developed by Foods Standards Australia and New Zealand (http://www.foodstandards.gov.au/monitoringandsurveillance/nuttab2006/). Voluntary folate fortification was introduced in Australia during the time period under investigation; however, relatively few of the recommended foods were fortified (http://www.health.gov.au/internet/main/publishing.nsf/Content/84026385A0EC6DC9CA256F190004C8E1/$File/folate.pdf). Folic acid supplements were taken by 10% of the sample but adjustment for this did not change any of the findings (results not shown).
The MTHFR 677 T allele frequency observed here for controls was 0.33 and compares favorably to the mean allele frequency of 0.32 reported for Caucasian populations in a meta-analysis.12 The ΔDNMT3b-149 T allele frequency of 0.46 observed here for controls is identical to another Australian study21 and similar to the frequency of 0.44 reported for noncancer controls from an American study.27
Because of their roles in cellular methyl group metabolism, the 2 genetic polymorphisms investigated here were hypothesized to show associations with the CIMP+ pathway for CRC. CIMP+ tumors are characterized by frequent methylation of CpG islands5–10 and an elevated concentration of tissue folates.32 A second potential limitation of this study was that tumor site was used as a surrogate indicator for CIMP+. Although the large majority of CIMP+ tumors arise in the proximal colon,5–10 it remains that up to 70% of CRCs that occur at this site are CIMP−. The associations observed here between the 2 polymorphisms and proximal CRC (Tables III–V) may therefore be stronger with the CIMP+ tumor subgroup. On the other hand, intrinsic differences between the proximal and distal colonic mucosa2 could also be very important in determining associations with genetic or dietary risk factors.
A third limitation of our study was that polymorphisms in only 2 of the genes involved in methyl group metabolism were investigated for associations with the risk of proximal or distal CRC. While the common functional polymorphisms in MTHFR and ΔDNMT3b are no doubt very important, genetic variants in other enzymes involved in folate metabolism, DNA methylation and the availability of nucleotides for DNA synthesis could also be associated with the risk of CRC.13 These include polymorphisms in the thymidylate synthase (TS), methionine synthase (MTR), reduced folate carrier (RFC), dihydrofolate reductase (DFE), methylene-tetrahydrofolate dehydrogenase 1 (MTHFD1) and cystathionine β-synthase (CBS) genes. A recent population-based case–control study found the MTHFR A1298C polymorphism, in combination with diet, was associated with CIMP+.33 Although it did not investigate polymorphisms in DNA methyltransferases, the study by Curtin et al. found no evidence for the involvement of several other genes involved in methyl group metabolism including TS, MTR, RFC and MTHFD1.
A meta-analysis of 25 populations involving more than 12,000 cases concluded that the MTHFR 677TT genotype was associated with a moderately reduced risk of CRC (OR = 0.83).12 Amongst the studies that reported tumor site, no evidence was found for different risk associations between the 677TT genotype and colon (OR = 0.78) or rectal (OR = 0.82) cancers. This contrasts with the present results (Table III) and with 5 separate studies showing that the MTHFR C677T polymorphism was associated with an increased risk for MSI+ CRC,14–17, 34 the large majority of which occur in the proximal colon.11 Possible reasons for the discrepancy include the classification of CRC as colon or rectal and whether study populations were from high or low folate intake areas, as discussed below.
The increased risk associated with the MTHFR 677TT genotype was most pronounced for women, older individuals and those with low folate or high alcohol intakes (Table IV). High folate intake appeared to completely reverse the elevated risk of proximal CRC associated with the TT genotype and, similar to distal CRC, may even be protective (AOR = 0.78, 95% CI 0.40–1.54). The modifying effects of folate and alcohol on the risk of CRC associated with the MTHFR C677T polymorphism were first recognized over a decade ago35 and have since been confirmed by many others.13, 29, 33, 34, 36 From these previous reports and results of the present study, we conclude that the MTHFR 677TT genotype is protective against both proximal and distal CRC under conditions of high folate intake. However, when folate intake is low, the MTHFR 677 T allele is associated with an increased risk for proximal, but not distal CRC, particularly in older individuals and those with high alcohol consumption (Table IV).
The mechanism linking the MTHFR C677T polymorphism with risk of CRC is currently unknown. Lower levels of tetrahydrofolate (THF) and 5,10-methylene-THF have been observed in the normal colonic mucosa37 and tumor32 tissues of CRC patients with MTHFR 677 CT or TT genotypes. This is a paradoxical finding since the lower enzymatic activity of the mutant form (T allele) of MTHFR would be expected to result in an increased cellular concentration of its substrate, 5,10-methylene-THF. Nevertheless, tissue levels of 5,10-methylene-THF could influence the development of CRC by altering the rates of production of dTMP required for DNA synthesis and of S-adenosyl methionine required for gene methylation. The MTHFR C677T polymorphism has been associated with an increased frequency of gene promoter methylation in proximal CRC in a Japanese population,18 but no association was observed with gene methylation levels in the normal colonic mucosa38 or tumor tissue32 from Australian CRC patients. The level of folate intake was not evaluated in these studies and may explain the different results observed for the 2 populations. A low folate/high alcohol diet was associated with an increased frequency of gene methylation in CRC,39 but the influence of the MTHFR C677T polymorphism was not reported in this study. Additional work is required to understand how the MTHFR C677T polymorphism confers different risks for proximal and distal CRC under conditions of low folate. This work should include evaluation of tissue folate and gene methylation levels in the proximal and distal colonic mucosa of individuals at risk of developing CRC. Further studies that investigate possible relationships between tissue, plasma and RBC folate with genetic polymorphisms and dietary folate intake are also clearly warranted. The present results suggest that increased dietary folate could reduce the risk of proximal colon cancers in older individuals carrying the MTHFR 677 CT or TT genotypes; however, this finding needs to be replicated in independent studies.
In comparison with MTHFR, considerably fewer studies have investigated whether polymorphisms in ΔDNMT3b are risk factors for CRC. The expression of DNMT3b has been demonstrated in an animal model to promote tumor development via gene-specific methylation that leads to transcriptional silencing.40 The most frequently studied genetic variant in ΔDNMT3b, C-149T, has been associated with a 2-fold increased promoter activity in in vitro studies.27 However, this is at odds with findings of a lower DNA methylation frequency in ovarian tumors41 and in the normal colonic mucosa38 of individuals with the CT or TT genotypes. Jung et al. found the TT genotype was associated with a trend for increased risk of colorectal polyps amongst individuals with low folate and methionine intake,20 but stratification according to site was not performed.
The current study is the first to evaluate the risk for colorectal carcinoma associated with the ΔDNMT3b C-149T polymorphism. The results showed a protective effect for the TT genotype on risk of proximal CRC, but no association with the risk of distal CRC (Table III). The protective effect against proximal CRC was seen across all gender, age, folate and alcohol subgroups, but reached significance only for older individuals and those with low folate intake (Table V). The protective effect of the ΔDNMT3b-149TT genotype against proximal CRC (AOR = 0.65, 95% CI 0.44–0.98, p = 0.03) may be due to its association with lower gene methylation levels in normal colonic mucosa, as observed in an earlier study by our group.38 Since frequent gene methylation is a hallmark of CIMP+ and these tumors occur predominantly in the proximal colon,5–10 this would explain why the protective effect does not extend to distal CRC. Moreover, gene methylation in the normal colonic mucosa increases with age,38, 42 thus explaining why the protective effect for the ΔDNMT3b-149TT genotype against proximal CRC was strongest in older individuals (AOR = 0.52, 95% CI 0.30–0.92, p = 0.02).
Although not reaching statistical significance, there was some evidence for a sex difference in the risk of distal CRC associated with the ΔDNMT3b-149TT genotype (Table V; AOR = 1.59 vs. AOR = 0.75 for females and males, respectively). This observation may be related to the finding of a gender difference for risk of CRC associated with another common polymorphism (G39179T) in DNMT3b.19 The modifying effects of sex and age on risk of CRC associated with DNMT3b polymorphisms could also be related to the observation that DNMT3b mRNA expression levels in human liver tissue were higher in older and female individuals.43 Further work in this area should include investigation of gene methylation levels in normal colonic mucosa in relation to DNMT3b expression, age, gender, site in the large bowel and polymorphisms in ΔDNMT3b.
In summary, our results show that 2 frequent polymorphisms in genes involved in methyl group metabolism, MTHFR and ΔDNMT3b, were associated with different risks for right- and left-sided CRC. These differences may in part be explained by the CIMP pathway occurring predominantly in the proximal colon. Site differences in risk for CRC are unlikely to be restricted to variants in genes involved in methyl group metabolism and the CIMP pathway. Indeed, site and sex differences have also been reported for the risk of CRC associated with polymorphisms in the TP5344 and VEGF45 genes, suggesting the CIN pathway is similarly influenced by such factors. These findings emphasize the importance of stratifying for age, sex, tumor site and diet when investigating genetic risk factors for CRC.
The authors gratefully acknowledge the people who generously gave their time to participate in this research, the clinicians who gave permission to approach their patients, the WA Cancer Registry staff for assistance with case ascertainment and recruitment, members of the WABOHS team (Ms. Clare Tran, Ms. Beatriz Cuesta-Briand, Mr. Medhi Tabatabaei, Mr. Terry Boyle, Ms. Jenny Landrigan) for collection of data used in this study, Ms. Sophia Ang and Ms. Kathryn Li for assistance with the genotyping and Dr. Kazuyuki Kawakami for helpful discussions. A/Prof Lin Fritschi was supported by an NHMRC Senior Research Fellowship.