Epidemiologic and mechanistic evidence suggests that folate is involved in colorectal neoplasia. Some polymorphic genes involved in folate metabolism-methylenetetrahydrofolate reductase
(MTHFR C677T and A1298C ), methionine synthase ( MTR A2756G ), methionine synthase reductase ( MTRR A66G) , cystathionine ß-synthase ( CBS exon 8, 68-base-pair insertion), and thymidylate synthase ( TS enhancer region and 3' untranslated region)-have been investigated in colorectal neoplasia. For MTHFR C677T and A1298C, the variant allele is associated with reduced enzyme activity in vitro. For the other polymorphisms, functional data are limited and/or inconsistent. Genotype frequencies for all of the polymorphisms show marked ethnic and geographic variation. In most studies, MTHFR C677T (10 studies, >4,000 cases) and 1298CC (four studies, >1,500 cases) are associated with moderately reduced colorectal cancer risk. In four of five genotype-diet interaction studies, C677T subjects who had higher folate levels (or a "high-methyl diet") had the lowest cancer risk. In two studies, 677TT homozygote subjects with the highest alcohol intake had the highest cancer risk. Findings from six studies of MTHFR C677T and adenomatous polyps are inconsistent. There have been only one or two studies of the other polymorphisms; replication is needed. Overall, the roles of folate-pathway genes, folate, and related dietary factors in colorectal neoplasia are complex. Research priorities are suggested.

Abbreviations: CBS, cystathionine ß-synthase; CI, confidence interval; MSI, microsatellite instability; MTHFR, methylenetetrahydrofolate reductase; MTR, methionine synthase; MTRR, methionine synthase reductase; OR, odds ratio; rpt, repeat; TS, thymidylate synthase.

Introduction

Evidence is accumulating for a role of folate in the etiology of colorectal carcinomas and adenomas (1). Many of the genes involved in folate metabolism are polymorphic (2). This paper reviews five polymorphic genes-methylenetetrahydrofolate reductase ( MTHFR ), methionine synthase ( MTR ), methionine synthase reductase
(MTRR ), cystathionine ß-synthase ( CBS ), and thymidylate synthase ( TS )-and their associations with colorectal neoplasia

This section describes polymorphisms in the genes and their functional effects. With the exception of MTHFR , relatively few studies have investigated relations between the polymorphisms and blood levels of folate and related biomarkers in nondiseased persons. In subjects with medical conditions, it is possible that the condition or its treatment, rather than the underlying genotype, influences biomarker levels. Many studies have been small, with limited statistical power. A potential difficulty in interpretation is that any observed difference in biomarker levels by genotype may not be due to the polymorphism under study but to the presence of another polymorphism. Equally, a failure to observe differences in biomarkers by genotype could be due to the presence of another polymorphism with opposing functional effects. So far, there has been little investigation of the effects of combinations of polymorphisms. With regard to MTHFR C677T, only red cell folate measured by microbiologic assay is reliable; results of the radioimmune assay are biased (8). There is differential detection by the assays of various intracellular folates, the distribution of which is related to MTHFR genotype (9). Whether red cell folate results measured by radioimmunoassay are biased for other polymorphisms in the folate-pathway genes is not known.

MTHFR
Several polymorphisms in the MTHFR gene have been reported, and two have been investigated in colorectal neoplasia: 1) C -->T at nucleotide 677, leading to an alanine to valine conversion in the protein (10); and 2) A -->C in exon 7, causing an alanine to glutamate protein change (11,12). These polymorphisms are located 2.1 kb apart. The other polymorphisms-T1059C, T1317C, and G1793A (12-14 )-are not discussed further in this paper.

For C677T , compared with homozygotes for the common variant ( CC ), heterozygotes have 65 percent of their enzyme activity levels in vitro and those who are homozygous variant ( TT ), 30 percent (15). From the microbiologic assay, compared with CC homozygotes, heterozygotes have 10 percent lower and TT homozygotes 18 percent lower red cell folate levels (16). Persons with the TT variant also have lowered plasma folate and vitamin B 12 levels and raised homocysteine levels (17,18). In two studies, the association with homocysteine held only when folate status was low (19, 20); in another, it occurred only when riboflavin status was poor (21). Regarding MTHFR and DNA methylation, one small study found that DNA from subjects with the TT variant had a significantly higher methyl group acceptance capacity than DNA from subjects with the CC variant (22), but this finding was not confirmed in a larger study (23). In 292 subjects (66 percent of whom had coronary atherosclerosis) selected by MTHFR genotype (187 CC, 105 TT), DNA methylation status was affected by genotype among only those with lower plasma folate levels; subjects with the TT variant who had lower plasma folate concentrations had lower methylation levels than all other groups of subjects (24). A few studies have investigated MTHFR and uracil misincorporation, DNA strand breaks, or genetic instability in vivo and in vitro, with inconclusive results (23, 25-27).

For A1298C , enzyme activity in vitro is decreased in homozygotes variants ( CC ) and, to a lesser extent, in heterozygotes compared with those without the variant (11). Studies of A1298C and plasma folate and homocysteine are inconsistent (12, 28-31), which may be due to methodological reasons (e.g., non-population-based study, small sample size), or it may be that there is a relation that depends on the status of folate and/or related nutrients. Enzyme activity in vitro for compound heterozygotes (i.e., heterozygotes for C677T and for A1298C) is unclear (29).

MTR
The A-G polymorphism at position 2756 in the protein binding region of MTR replaces aspartic acid with glycine (32). Most studies suggest that plasma homocysteine level is lower in those with the rarer, G , than the more common, A , allele (18, 33-36). One study found significantly higher plasma folate levels in GG than in AA subjects (34), but this finding was not observed in another study (18). Evidence on red cell folate and on plasma vitamin B 12 and vitamin B 6 is very limited (18, 35, 37).

MTRR
The A66G polymorphism in the MTRR gene results in the substitution of isoleucine with methionine at codon 22 (5). In two studies, subjects homozygous for the common allele ( AA ) had elevated homocysteine levels compared with those who had other genotypes (38, 39); in a third study, genotype was not a significant predictor of homocysteine level (40). No associations were found between genotype and serum folate, vitamin B 6 , or vitamin B 12 in the single known study ( 38 ).

CBS
Many mutations and several polymorphisms in the CBS gene have been reported (41). To our knowledge, the only variant investigated in colorectal neoplasia is the 68-base-pair insertion in the exon 8 coding region. Four studies found lower plasma homocysteine levels in persons carrying the insertion than in those without, although the difference was significant in only one
(35, 36, 39, 42). One study suggested that the effect was modulated by plasma vitamin B 6 concentration (43); another suggested an interaction with MTHFR C677T (35). The one available study that we know of found no associations between genotype and red cell folate or plasma vitamin B 12 level (35).

TS
The TS enhancer region contains a series of 28-base-pair tandem repeats. Two repeats (2 rpt) or three repeats (3 rpt) are most common, with 3 rpt occurring most frequently. More repeats have been observed but are rare (44, 45). In vitro, compared with the double repeat, the triple repeat has been associated with 2.6-fold greater thymidylate synthase expression (46). Among 497 Singapore Chinese, plasma folate levels were significantly lower, and homocysteine levels nonsignificantly higher, in 3 rpt/3 rpt subjects than in those with other genotypes (47). When MTHFR and TS were considered together, plasma folate levels were highest (15.3 nM) in 677CC or 677CT and not 3 rpt/3 rpt subjects, intermediate (13.8 nM) in 677CC or 677CT and 3 rpt/3 rpt subjects, and lowest (11.6 nM) in 677TT subjects (irrespective of TS genotype).

The 3' untranslated region contains a 6-base-pair deletion at base pair 1494, the functional consequences of which are not known (48). The two polymorphisms appear to be in linkage disequilibrium (48).

Refer to the Appendix for Internet sites pertaining to the genes discussed in this review.

This section includes information on studies reporting genotype frequencies in persons without cancer or other diseases. Using appropriate Medical Subject Headings (MeSH) and text words, we searched MEDLINE, EMBASE, and PubMed databases for papers published from 1990 to December 2002. Further relevant articles were identified by hand-searching reference lists in published papers. MTHFR frequencies are from the Human Genome Epidemiology (HuGE) reviews by Botto and Yang (49) and by Robien and Ulrich (50). The A1298C data reported by Robien and Ulrich are augmented with results from less-studied geographic areas and ethnic groups. For the studies tabulated here, Hardy-Weinberg equilibrium of the genotype frequencies was assessed by using the Pearson X ² test.

MTHFR
There is considerable ethnic and geographic variation in the frequency of the C677T variant (49). The TT prevalence ranged from around 1 percent in Black populations in the United States, sub-Saharan Africa, and South America to more than 20 percent in US Hispanics, Colombians, and Amerindians in Brazil. TT genotype frequency in White populations in Europe, North America, and Australia was 8-20 percent. In Europe, there appears to be a trend of increasing frequency of the variant from north to south. Twelve percent of Japanese were TT homozygotes.

For A1298C , the CC prevalence in North American studies, which included mainly White subjects, was 7-12 percent (50). In four Hispanic series ( n < 90), the frequency was 4-5 percent (51-54). In two African-American series, 2 and 4 percent were CC subjects. In Europe, the prevalence of CCranged from 4 to 12 percent in most studies. In two northeast Scotland series of subjects randomly selected from general practitioner registers, the frequencies were 15 percent (95 percent confidence interval (CI): 11.8, 19.2) and 18 percent (95 percent CI: 9.5, 30.4) (55, 56). In Chinese, Japanese, and Hawaiian populations, 1-4 percent were CC (50, 54) subjects. In the single studies in Brazil, Morocco, South Africa, and Turkey and among Israeli Jews, the frequencies were 6 percent (95 percent CI: 2.8, 9.6), 3 percent (95 percent CI not available), 4 percent (95 percent CI: 1.4, 9.9), 6 percent (95 percent CI: 1.7, 14.8), and 13 percent (95 percent CI: 9.7, 16.5), respectively (31, 33, 57-59).

In some series, but not all, a few persons with three or four variant alleles (i.e., 677TT / 1298AC , CT / CC, TT / CC) have been reported (35, 60-64).

For A1298C , the CC prevalence in North American studies, which included mainly White subjects, was 7-12 percent (50). In four Hispanic series ( n < 90), the frequency was 4-5 percent (51-54). In two African-American series, 2 and 4 percent were CC subjects. In Europe, the prevalence of CCranged from 4 to 12 percent in most studies. In two northeast Scotland series of subjects randomly selected from general practitioner registers, the frequencies were 15 percent (95 percent confidence interval (CI): 11.8, 19.2) and 18 percent (95 percent CI: 9.5, 30.4) (55, 56). In Chinese, Japanese, and Hawaiian populations, 1-4 percent were CC (50, 54) subjects. In the single studies in Brazil, Morocco, South Africa, and Turkey and among Israeli Jews, the frequencies were 6 percent (95 percent CI: 2.8, 9.6), 3 percent (95 percent CI not available), 4 percent (95 percent CI: 1.4, 9.9), 6 percent (95 percent CI: 1.7, 14.8), and 13 percent (95 percent CI: 9.7, 16.5), respectively (31, 33, 57-59).

In some series, but not all, a few persons with three or four variant alleles (i.e., 677TT / 1298AC , CT / CC, TT / CC) have been reported (35, 60-64).

MTR
In Japanese, Chinese, and Korean populations, the frequency of the GG genotype was 2-3 percent (18, 32- 37, 54, 65-82; Web table 1). (This information is described in the first of four supplementary tables; each is referred to as "Web table" in the text and is posted on the website of the Human Genome Epidemiology Network (http://www.cdc.gov/genomics/hugenet/default.htm) as well as on the Journal 's website (http://aje.oupjournals.org/). In most European series, approximately 3 percent of the subjects had the GG genotype. Frequencies from all but two North American studies were 1-5 percent. The frequency was 10-11 percent in these two series-one of White children and their mothers in Canada and the other of White persons in Hawaii. In the single African-American population, 6 percent (95 percent CI: 4.3, 8.7) of the subjects had the GG genotype. In three studies, the genotype frequencies were not in Hardy-Weinberg equilibrium (73-75).

Web Table 1  Genotype frequencies of the A2756G polymorphism in MTR

MTRR
The lowest reported prevalence of GG homozygotes was 8-10 percent in Japanese in Hawaii and in Hawaiians (5, 14, 38-40, 54, 83-85; Web table 2). Among 558 subjects in Northern Ireland, 12 percent (95 percent CI: 9.1, 14.6) were GG homozygotes, but this series was not in Hardy-Weinberg equilibrium. In most of the remaining series, the frequency was 19-29 percent. Among 97 African Americans and 96 Hispanics, the frequencies were 42 percent (95 percent CI: 32.3, 52.7) and 50 percent (95 percent CI: 39.6, 60.4), respectively.

Web Table 2 Genotype frequencies of the A66G polymorphism in MTRR

CBS
Homozygosity for the 68-base-pair insertion is rare in all populations (35, 36, 39, 42, 54, 65, 70, 71, 77, 82, 86-98; Web table 3). The highest reported frequency was 3 percent among Blacks from Brazil and Africa. In four other series, the homozygote prevalence also reached 3 percent, but the genotype frequencies were not in Hardy-Weinberg equilibrium (42, 70, 71, 96). In Europe, Australia, and most US populations, the frequency of heterozygotes was 8-19 percent, with most around 13-15 percent. Two Japanese series found no heterozygotes. Heterozygosity occurred in 5 percent (95 percent CI: 1.6, 11.3) of the single Chinese series.

Web Table 3 Genotype frequencies of 68bp insertion at position 844 of CBS

TS
In three studies in the United Kingdom, and in three of mainly White populations in the United States, 19-23 percent of subjects were 2 rpt/2 rpt (44-47, 99-102; Web table 4). The prevalence was 14-20 percent in two African and one African-American series and 17 percent among volunteers born in four southwest Asian countries living in Scotland. Two to 4 percent of two Chinese populations were homozygous variant. In all studies, genotype frequencies were in Hardy-Weinberg equilibrium.

Web Table 4 Genotype frequencies for tandem repeat polymorphism in TS

In a single study of US Whites, 10 percent (95 percent CI: 7.7, 12.5) were homozygotes for the 3' untranslated region deletion (102).

Combinations of genotypes
Most studies reporting frequencies of combinations of genotypes are small (33, 35, 70, 80, 94, 103). In the largest, of almost 1,300 males in the United Kingdom, 8 percent carried the CBS68-base-pair insertion and the MTHFR T allele; 5 percent of subjects had the CBS 68-base-pair insertion and the MTR G allele; and 20 percent carried both the MTR G and MTHFR T alleles (35).

Comments on studies of population frequencies
Few of the studies reviewed here were population based; many relied on convenience samples. Selection and participation biases may therefore explain some of the apparent variations in genotype prevalence. In a few studies, genotype frequencies were not in Hardy-Weinberg equilibrium. Although lack of Hardy-Weinberg equilibrium might indicate that the series were subject to selection or participation biases, there are other reasons why Hardy-Weinberg equilibrium might not hold, including migration or genotyping error (104). Many of the studies are relatively small, so the estimates of genotype frequency lack precision.

In many studies, the ethnic makeup of the participants is not described. Most well characterized are White populations in the United States and western Europe. Other populations, geographic areas, and ethnic groups, particularly in Africa, Asia (other than Japan), and South America, have been less studied. The generalizability from, for example, one "Black African" population to another may be limited since it is not always straightforward to establish ethnicity (105).

An estimated 945,000 new cases of colorectal cancer were diagnosed worldwide in 2000, and 492,000 persons died from the disease (106). Two thirds of incident cases occur in developed countries, where it is the third most common cancer in males and second most common in females (107). There are substantial international variations in incidence (108). Sixty to 70 percent of colorectal cancers arise in the colon (108).

Although most evidence is indirect, the majority of colorectal carcinomas are believed to develop from adenomatous polyps (109). Hyperplastic polyps may be precursors of some right-sided colon cancers (110). Investigation of the first occurrence, and the recurrence, of polyps may reveal factors important in early stages of the neoplastic process.

Fewer than 10 percent of incident colorectal tumors are due to hereditary nonpolyposis colorectal cancer and familial adenomatous polyposis (111). When these syndromes are excluded, there is still familial aggregation of cancers and adenomas (112-114), which is unlikely to be entirely accounted for by familial clustering of environmental factors (115). This information points to the potential importance of genetic susceptibility factors, and the interaction of these with each other and with environmental factors, in the disease causation.

The studies of Japanese migrants to the United States in the 1960s revealed the overwhelming importance of environmental factors in colorectal cancer etiology (116). Established risk factors for the disease are shown in Table 1 (109, 117-125).

Table 1  Results of studies of progesterone receptor gene PROGINS polymorphisms  and epithelial ovarian cancer, 1995-2003

Although diet appears to be important in colorectal cancer (120), it has been difficult to identify the specific components involved. Observational epidemiologic evidence shows that a high vegetable intake is related to decreased risk (120), although recent work suggests that the relation is complex (124, 125). Vegetables, particularly green, leafy vegetables, are a major source of folate. The majority of prospective and case-control studies of serum folate, red cell folate, or reported dietary or total folate intake are compatible with inverse associations with colon cancer and adenomas (17, 54, 76, 125-146). There is no consistent association between rectal cancer and folate intake (126, 131, 133-135, 137, 138). One small trial of folic acid supplementation in persons from whom polyps had been removed observed a reduced recurrence rate in the supplemented group (147). Some studies are compatible with a positive association between alcohol intake, which adversely affects folate metabolism (148), and colorectal neoplasia (109). A "low-methyl diet," comprising high alcohol intake and low folate and methionine (and/or vitamins B 6 and B 12 ) intakes, has been associated with increased colon cancer risk (126, 130, 132, 140).

Internet sites providing data and information on colorectal neoplasia are contained in the Appendix.

The observed association of the MTHFR homozygous variant genotypes with reduced carcinoma risk was the opposite of what might have been expected a priori. This finding has led investigators to reconsider the folate metabolism pathway, putting a greater emphasis on the functions of folate and MTHFR in DNA synthesis. The evidence is compatible with interactions between MTHFR genotype and folate, alcohol, and/or related nutrients in relation to colorectal cancer. Evidence on polymorphisms other than MTHFR C677T is extremely limited. The associations observed between MTR, CBS, MTRR, and TS genotypes and colorectal neoplasia are tentative at best and require replication. The few studies of combinations of polymorphisms suggest the possibility of gene-gene interactions; again, further investigation is needed to confirm initial findings. Altogether, the evidence suggests that the roles of folate-metabolizing genes, folate, and related dietary factors in colorectal neoplasia are complex. Methodologies are currently lacking for specification of hypotheses, clarification of functional effects, and statistical analysis relating to such complex gene-environment pathways. This area of research must be a priority if advancements in understanding of disease etiology are to be achieved. Table 6 lists other areas for further research.

Table 6  Research priorities.

Tables

• Table 1 - Environmental factors associated with colorectal cancer
• Table 2 - Studies of the MTHFR * C677T genotype and colorectal carcinoma, with relative risks and 95% confidence intervals
• Table 3 - Studies of the MTHFR * C677T genotype and adenomatous polyps, with relative risks and 95% confidence intervals
• Table 4 - Studies of the MTHFR * C677T genotype and hyperplastic polyps, with relative risks and 95% confidence intervals
• Table 5 - Studies of the MTHFR * A1928C polymorphism, other folate pathway genes, and colorectal neoplasia
• Table 6 - Research priorities

Supplementary Tables

• WebTable 1 - Genotype frequencies of the A2756G polymorphism in MTR
• WebTable 2 - Genotype frequencies of the A66G polymorphism in MTRR
• WebTable 3 - Genotype frequencies of 68bp insertion at position 844 of CBS
• WebTable 4 - Genotype frequencies for tandem repeat polymorphism in TS

Figures

• Figure 1 - The roles of the methylenetetrahydrofolate reductase, methionine synthase, methionine synthase reductase, cystathionine ß-synthase, and thymidylate synthase genes in the metabolism of folate

List of References

Data on cancer incidence, survival, and mortality

• International Agency for Research on Cancer (IARC)-Cancer Mondial (last accessed 5/2007)
• Surveillance, Epidemiology, and End Results (SEER) Program (last accessed 5/2007)
• National Program of Cancer Registries (NPCR) (last accessed 5/2007)

Information on cancer

• Cancer Research UK (last accessed 5/2007)
• National Cancer Institute-cancer.gov (last accessed 5/2007)
• American Cancer Society (last accessed 5/2007)

Genetics information

• Human Genome Epidemiology Network (HuGENet) (last accessed 5/2007)
• Public Health Genetics Unit (last accessed 5/2007)
• Online Mendelian Inheritance in Man (OMIM) (last accessed 5/2007)
• GenAtlas (last accessed 5/2007)
• GeneCards (last accessed 5/2007)
• National Center for Biotechnology Information (last accessed 5/2007)
• UK Human Genome Mapping Project