Genetic variants in 3′-UTRs of methylenetetrahydrofolate reductase (MTHFR) predict colorectal cancer susceptibility in Koreans

Polymorphisms in the methylenetetrahydrofolate reductase (MTHFR) play important roles in tumor development, progression, and metastasis. Moreover, recent studies have reported that a number of 3′-UTR polymorphisms potentially bind to specific microRNAs in a variety of cancers. The aim of this study was to investigate the association of four MTHFR polymorphisms, 2572C>A [rs4846049], 4869C>G [rs1537514], 5488C>T [rs3737967], and 6685T>C [rs4846048] with colorectal cancer (CRC) in Koreans. A total of 850 participants (450 CRC patients and 400 controls) were enrolled in the study. The genotyping of MTHFR 3′-UTR polymorphisms was performed by polymerase chain reaction-restriction fragment length polymorphism analysis or TaqMan allelic discrimination assay. We found that MTHFR 2572C>A, 4869C>G, and 5488C>T genotypes were substantially associated with CRC susceptibility. Of the potentially susceptible polymorphisms, MTHFR 2572C>A was associated with increased homocysteine and decreased folate levels in the plasma based on MTHFR 677CC. Our study provides the evidences for 3′-UTR variants in MTHFR gene as potential biomarkers for use in CRC prevention.

to its function in cell homeostasis, FA has been hypothesized to play a role in carcinogenesis, especially in development of CRC 5 . Several mechanisms could underlie FA deficiency-mediated CRC, including DNA strand breaks, aberrant DNA methylation, and impaired DNA repair. Thus, FA has been proposed as a possible candidate nutrient for CRC prevention 6 . Genetic variants in FA metabolism-related genes may modulate levels of this vitamin and influence risk of carcinogenesis. Furthermore, previous meta-analyses of numerous epidemiologic studies reported that FA was a determinant of CRC risk 7 .
The effect of several polymorphic genes involved in FA metabolism, including methylenetetrahydrofolate reductase (MTHFR) on CRC susceptibility and progression has been investigated 8 MTHFR catalyzes the reduction of 5,10-methylenetetrahydrofolate to 5-methyltetrahydrofolate; the latter is the methyl donor for the conversion of homocysteine (Hcy) to methionine, whereas the former, and its derivatives, are essential cofactors for both thymidylate and de novo purine synthesis 9,10 . During de novo purine synthesis, thymidylate synthase with the FA binding site catalyzes the conversion of deoxyuridine monophosphate (dUMP) to deoxythymidine monophosphate (dTMP). This conversion is indispensable for the production of thymine, a nucleotide needed for DNA synthesis and repair 11,12 . Decreased MTHFR activity and expression lead to an accumulation of Hcy and/or deficiency of FA 13 .
There are two well-studied polymorphisms of the MTHFR gene, i.e., MTHFR 677C> T and 1298A> C. Despite the existence of a large body of data for studies of associations between these polymorphisms and CRC, findings of genetic associations have been inconsistent for a variety of conditions 8 . Recent studies have shown some clinical impacts of polymorphisms in the 3′ -UTR of certain genes, which may potentially bind to specific microRNAs (miRNAs) in various cancers [14][15][16][17] . However, variants in the MTHFR 3′ -UTR have not been extensively studied.
In the present study, four single nucleotide polymorphisms (SNPs) in the MTHFR 3′ -UTR were identified by a database search; these are MTHFR 2572C> A (rs4846049), 4869C> G (rs1537514), 5488C> T (rs3737967), and 6685T> C (rs4846048). There were 17 MTHFR 3′ -UTR SNPs with minor allele frequencies >5% in the global population. The 13 SNPs in the MTHFR 3′ -UTR were excluded from this study due to following reasons: (1) lack of information for validation in the Asian population (rs35134728 and rs55780505), (2) below 5% of minor allele frequency in the Asian population (rs2184226, rs868014, rs2077360, and rs4845884), and (3) failure of genotyping conditions (rs1537516, rs1537515, rs2184227, rs3737966, rs3820192, rs11559040, and rs72640221). The minor allele frequencies of the four SNPs were all >5% in the Asian population. Little is known about their genetic associations with CRC. Therefore, we investigated whether these polymorphisms of the MTHFR 3′ -UTR correlate with CRC susceptibility in Koreans.

Stratified effects of clinical and environmental factors.
Unlike other cancers, CRC epidemiology was affected by a variety of identified risk factors in a complex manner 3 . Previous reports identified the following risk factors: aging, male gender, obesity, metabolic syndrome (MetS), hypertension (HTN), diabetes mellitus (DM), deficiency of FA, increased intake of red meat, excessive alcohol consumption, and smoking 3,[18][19][20] . In addition, epidemiologic alterations of CRC pathology and genetic differences correlating with the clinical features of CRC have been previously reported [21][22][23] . Therefore, stratified analyses were useful in elucidating CRC epidemiology resulting from a diversity of confounding variables. Except for FA deficiency, which presented difficulties in establishing a threshold using plasma levels, we conducted stratified analyses of the data according to age, gender, tumor site, tumor size, tumor node metastasis (TNM) stage, presence of MetS, HTN, DM, levels of body mass index (BMI), triglycerides (TG), and high density lipoprotein-cholesterol (HDL-C) to determine whether the 3′ -UTR minor alleles were associated with CRC incidence in specific subsets of the study population. The results for the recessive model were excluded because of a small number of 3′ -UTR minor homozygous genotypes when stratified by a variety of factors. FDR correction was used to eliminate false positive associations from stratified effects. Supplementary Table S3 summarizes the frequencies of MTHFR 3′ -UTR genotypes in each stratified CRC group. The overall results of stratified analyses are shown in Table 4. MTHFR 2572CA + AA presented subset-specific associations in subgroups of patients with the following: Combined effects of 3′-UTR polymorphisms with environmental factors. Because cancer risk is determined by the complex interplay of genetic and environmental factors, we calculated combined gene-environment effects on CRC susceptibility (Table 5 and Supplementary Tables S4-S7). To analyze combined gene-environment effects, we chose the following environmental risk factors which were significantly prevalent in the CRC group: lower plasma FA levels, MetS, HTN, DM, and <40 (male)/50 (female) mg/dL of HDL-C. We used the relative excess odds due to interaction (RERI OR ) to evaluate the additivity of odds between 3′ -UTR genotypes and environmental risk factors 24 . With RERI OR >0, there was an additive interaction between the combined factors. All MTHFR 3′ -UTR minor genotypes Variations of genetic associations for 2572C>A by 677C>T genotypes. Next, we sought to determine whether the polymorphisms of interest within the 3′ -UTR of the MTHFR gene correlated with plasma Hcy and FA concentrations ( Table 6). We analyzed plasma Hcy and FA levels according to studied 3′ -UTR polymorphisms based on MTHFR 677C> T, due to its strong associations with MTHFR activity 13

Discussion
In the present study, we investigated whether four 3′ -UTR polymorphisms of the MTHFR gene are related to the occurrence of CRC. We found that MTHFR genotypes 2572C> A, 4869C> G, and 5488C> T were substantially associated with CRC susceptibility and displayed significant combined gene-environment effects (RERI OR > 0). Moreover, MTHFR 2572C> A was associated with increased Hcy and decreased FA levels in the plasma based on MTHFR 677CC genotype. To our knowledge, this is the study to provide evidence that 3′ -UTR polymorphisms of MTHFR gene are associated with CRC susceptibility. Numerous studies have investigated the associations of MTHFR genetic polymorphisms with CRC incidence 8,25 . Many reports have demonstrated inconsistent data for the significance of MTHFR 677C> T, although the MTHFR 677T allele was reported to be a potential genetic risk factor for increased CRC  susceptibility, as determined in a recent meta-analysis 25 . There were limited conditions for the association of the MTHFR 677T allele with increased CRC risk: (1) high alcohol intake 26,27 , (2) low FA intake 26,28 , and (3) microsatellite instability (MSI) [29][30][31] . Although there was no similarity when the present study design was compared with numerous previous reports, the subset-specific associations of 2572C> A, 4869C> G, and 5488C> T polymorphisms were observed. We found common subset-specific associations in the subgroup of patients with the following: ≥62 years of age, male, rectal cancer, tumors ≥5 cm, TNM stage I/II disease, and HTN. We hypothesized the reasons for these correlations were high alcohol intake, low FA intake, and microsatellite instability. Korean men showed higher rates of alcohol consumption than   women 32 , FA deficiency increased with age 33 , and FA intake showed stronger associations with decreased risk for the cancer of the rectum than of the colon 34 . HTN may be prevented by FA fortification 35 , and increased prevalence of tumors ≥5 cm were associated with MSI 36 . Finally, CRC with early TNM stage and without lymph node metastases showed a higher frequency of MSI 37 .
The MTHFR 677T allele decreased MTHFR activity correlated with increased plasma Hcy and decreased plasma FA levels 13 . Also, we observed that MTHFR 2572A allele had a tendency of decreased MTHFR mRNA expression in tumor-adjacent tissues. Therefore, it is necessary to gain an understanding of carcinogenic events caused by decreased MTHFR activity to explain why 3′ -UTR polymorphisms may affect CRC susceptibility. Lower MTHFR activity increases Hcy and decrease FA levels in the plasma, inducing development of CRC 38,39 . Plasma FA concentration inversely correlates with Hcy level 40 . Depletion of FA may be considered a risk factor for colorectal carcinogenesis because it induces breaks in human chromosomal DNA 41 . Two plausible mechanisms by which FA deficiency may create such breaks are uracil misincorporation and impaired DNA repair 42,43 . FA deficiency reduces synthesis of deoxythymidylate from deoxyuridylate, and the resultant nucleotide imbalance accelerates the incorporation of uracil into DNA. Uracil in DNA is excised by a repair glycosylase and, in the process, a transient single-strand break develops in the DNA 44 . Simultaneous removal and repair of two adjacent uracil residues on opposite strands can cause a double-strand break, which is difficult to repair and further increases genetic instability. Unrepaired double-strand DNA breaks enhance cellular transformation and increase cancer risk 44 . FA status is also important to modify cell proliferation rates. James et al. reported excessive cell proliferation in livers of FA/methyl-deficient rats 45 . Conversely, FA supplementation has been found to diminish colorectal mucosal proliferation in both animal and human studies. Nensey et al. reported the same phenomenon in an animal model 46 . FA supplementation reduces carcinogen-induced ornithine decarboxylase and tyrosine kinase activities, both of which are indicators of cell proliferation 46 . Biasco et al. 47 reported that FA supplementation significantly decreases rectal mucosal proliferation in patients with long-standing ulcerative colitis, a condition that carries a higher risk of CRC, a predisposition to which is considered to be due in part to reduced availability of FA. Akoglu et al. 48 described a human colon cancer cell line in which dihydrofolate and methyl-THF serve as growth-inhibitory factors.
Epidemiologic studies over the past two or three decades have described an inverse relationship between FA status (assessed by dietary FA intake or measurement of red cell and plasma FA levels) with the risk of cancer of the lungs, oropharynx, esophagus, stomach, colorectum, pancreas, cervix, ovary, prostate, and breast, and the risk of neuroblastoma and leukemia 49,50 . Although the results of epidemiologic and clinical studies are inconsistent, the reports have indicated 20%-40% lower risk of CRC in subjects with the highest dietary intake or blood levels of folate compared with those with the lowest intake or blood levels 49,51,52 . Several intervention studies have shown that FA supplementation can improve or reverse poor prognostic factors of CRC 49,51 , and some epidemiologic studies have shown a beneficial effect of taking multivitamin supplements containing ≥400 μ g FA on CRC risk and mortality [53][54][55] . The data from animal studies generally support a causal association between FA deficiency and CRC risk and an inhibitory effect of modest levels of FA supplementation on colorectal carcinogenesis 51 . The results for previous association studies of plasma Hcy or FA levels with risk of colorectal neoplasia are complicated. Four studies reported increased risk of colorectal neoplasia by higher plasma Hcy or lower plasma FA levels 56-59 , Kato et al. 56 and Pufulete et al. 57 demonstrated associations with decreased plasma FA levels whereas Ulvik et al. 58 and Martinez et al. 59 presented significant relationships with higher plasma Hcy levels. Moreover, Martinez et al. reported that the risk of colorectal neoplasia could be controlled by FA supplementation 59 . Three other studies did not report a significantly elevated risk of colorectal neoplasia with increasing plasma Hcy or decreasing plasma FA levels [60][61][62] . In this study, there were the additive interactions of CRC risk (RERI OR >0) between MTHFR 3′ -UTR minor genotypes and lower plasma FA levels (<5.77 ng/mL: the lowest tertile interval). FA supplementation or fortification may therefore be essential for the prevention of colorectal carcinogenesis in individuals with MTHFR 3′ -UTR minor genotypes.
Genetic variation in the 3′ -UTR region could affect the stability and translation of the mRNA through altered miRNA-binding affinity. In the present study, we could not demonstrate altered miRNA-binding activity depending on MTHFR 3′ -UTR polymorphisms. One study showed altered miR-149-binding activity with polymorphism MTHFR 2572C> A 63 . The MTHFR 2572A allele augmented miR-149-binding activity associated with decreased MTHFR expression. Further studies are needed to directly test for binding activity of miRNA to MTHFR 3′ -UTR polymorphic regions to determine the mechanism by which these polymorphisms may influence cellular proliferation and cancer progression. These studies may have great clinical impact for all diseases related to one-carbon metabolism.
In conclusion, we investigated the relationship of the polymorphisms MTHFR 2572C> A, 4869C> G, 5488C> T, and 6685T> C with CRC susceptibility. We found specific 3′ -UTR polymorphisms positively correlated with CRC susceptibility, depending on a diversity of clinical and environmental risk factors. This study has several limitations. First, the manner in which 3′ -UTR polymorphisms in the MTHFR gene affect development of CRC is still unclear. Second, information regarding additional environmental risk factors in CRC patients is lacking and remains to be investigated. Third, a limited number of 3′ -UTR polymorphisms were studied. Lastly, the population of this study was limited to ethnic Koreans. Although results of our study provide the first evidence for 3′ -UTR variants in the MTHFR gene as Scientific RepoRts | 5:11006 | DOi: 10.1038/srep11006 potential biomarkers for use in CRC prevention, a prospective study on a larger cohort of ethnically diverse patients is warranted to validate these findings.

Methods
Ethics statement. The study protocol was approved by the Institutional Review Board of CHA Bundang Medical Center. All study subjects provided written informed consent to participate in the study. All the methods applied in the study were carried out in accordance with the approved guidelines.

Study population.
A case-control study of 850 individuals was conducted. Four hundred and fifty patients diagnosed with CRC at CHA Bundang Medical Center, Seongnam, South Korea were enrolled from June 2005 to January 2010. This study only included CRC patients who had undergone surgical resection with a curative intent and who had histologically confirmed adenocarcinoma. The response rate of CRC patients, with an initial number of 598 individuals, who gave written informed consent was 75.3%. Within the CRC cohort, 155 consecutive patients with proximal colon cancer (i.e., from the cecum to the splenic flexure), 101 consecutive patients with distal colon cancer (i.e., descending and sigmoid colon), 8 consecutive patients with mixed colon cancer, and 186 consecutive patients with rectal cancer underwent primary surgery. Information concerning the date of diagnosis and pathological stage was obtained retrospectively. Tumor staging was performed according to the sixth edition of the American Joint Committee on Cancer (AJCC) staging manual. The control group consisted of 400 individuals randomly selected following a health screening. Of the initial 612 normal visitors, 476 individuals gave written informed consent (response rate: 77.8%), and 76 participants were excluded from the study due to a history of thrombotic diseases or cancer. Patients with a high baseline blood pressure (systolic ≥140 mm Hg or diastolic ≥90 mm Hg) on more than one occasion or a history of treatment with antihypertensive medication were classified as having HTN. Patients with high fasting plasma glucose ( ≥126 mg/dl) and those who took oral hypoglycemic agents or who had a history of insulin treatment were classified as having DM. Individuals were diagnosed with MetS if they possessed three or more of the following five risk factors: BMI ≥25.0 kg/m 2 ; TG ≥150 mg/dl; HDL-C <40 mg/dl for men or <50 mg/ dl for women; blood pressure ≥130/85 mm Hg or currently taking anti-hypertension medication; and fasting blood sugar ≥100 mg/dl or currently taking hypoglycemic medication.
Phenotype measurements. Anthropometric measurements included BMI. Systolic and diastolic blood pressures of subjects were measured in the seated position after 10 min of rest. For measurements of physiological parameters, 3-ml samples of blood were obtained after fasting overnight. Plasma glucose was determined in duplicate using the hexokinase method adapted for an automated analyzer (TBA 200FR NEO, Toshiba Medical Systems, Tokyo, Japan). Levels of TG and HDL-C were determined by enzymatic colorimetric methods using commercial reagent sets (Toshiba Medical Systems). The concentration of Hcy in the plasma was measured by fluorescence polarization immunoassay (FPIA) with the IMx automated analyzer (Abbott Laboratories, Chicago, IL, USA). The plasma concentration of FA was determined using a radioassay kit (ACS:180, Bayer, Tarrytown, NY, USA).

Statistical analysis.
To analyze baseline characteristics, chi-square tests were used for categorical data, and Student's t-tests were used for continuous data to compare patient and control baseline data. Association of MTHFR 3′ -UTR polymorphisms with CRC incidence was calculated using adjusted odds ratios (AORs) and 95% confidence intervals (CIs) from multivariate logistic regression adjusted for age, gender, HTN, DM, BMI, TG, and HDL-C. These parameters were selected as adjustment variables because they were directly or potentially associated with CRC 3,18-20 . To evaluate the association data by the Benjamini-Hochberg method, we calculated FDR-corrected P values according to the number of genetic markers and stratified groups 64 . RERI OR was used to calculate additive interactions between genotypes and environmental risk factors 24 . When RERI OR >0, there was an additive interaction between combined factors. Correlation of MTHFR 3′ -UTR polymorphisms with plasma Hcy and FA levels was calculated using regression coefficients and t values from multivariate linear regression adjusted for age, gender, HTN, DM, BMI, TG, and HDL-C. The statistical significances of MTHFR mRNA expression levels according to studied polymorphisms and tissue differences were calculated by Mann-Whitney, Kruskal-Wallis, and Wilcoxon tests. The haplotypes for multiple loci were estimated using the expectation maximization algorithm with SNPAlyze (version 5.1; DYNACOM Co, Ltd, Yokohama, Japan), and those with frequencies of <1% were excluded from statistical analysis. Analyses were performed using GraphPad Prism 4.0 (GraphPad Software Inc., San Diego, CA, USA) and Medcalc version 12.7.1.0 (Medcalc Software, Mariakerke, Belgium).