Low-penetrance Genes and Their Involvement in Colorectal Cancer Susceptibility¹

Colorectal cancer is an important cause of death from cancer in the Western countries. In the Netherlands, colorectal cancer is the second cause of death from malignant disease in women (after breast cancer) and the third cause of death in men (after lung and prostate cancer; Ref. 1). The genesis of colorectal cancer involves a series of steps in which environmental and/or endogenous carcinogens induce or promote cancer development. These steps include the activation of oncogenes, such as ras, and inactivation of tumor suppressor genes, such as APC, Tp53, and DCC, and genes involved in DNA mismatch repair (2, 3, 4, 5).

Colorectal cancer is a multifactorial disease, i.e., there are many factors contributing to its development. These include on the one hand dietary and lifestyle habits and on the other hand genetic predispositions.

Epidemiological studies indicate that diets high in red meat, diets low in vegetables and fiber, obesity, and smoking are associated with an increased colorectal cancer risk (6, 7, 8, 9, 10, 11). Diets high in calcium and folate and regular physical activity are associated with a reduced risk (8, 9, 10, 12). However, some of these associations are still controversial.

Genetic syndromes predisposing to colorectal cancer include the polyposis syndromes (FAP³ , Peutz Jeghers syndrome, and juvenile polyposis) and HNPCC. These syndromes account for only ∼3% of all cases (13, 14, 15, 16) and are not responsible for the 2-fold increased risk in first-degree relatives of sporadic colorectal cancer patients. This increased risk in relatives suggests a mild genetic predisposition, i.e., the involvement of low-penetrance genes or gene variants. Candidates for this are genes involved in metabolic pathways (17, 18) or in methylation, those modifying the colonic microenvironment, oncogenes, tumor suppressor genes, and genes involved in immune response. Low-penetrance variants in high-penetrance genes (e.g., APC, MLH1, or MSH2) might also be important in sporadic and in familial colorectal cancer.

The phenotypes of FAP and HNPCC, with regard to the colonic disease, vary considerably, not only between families but also within families (19, 20). The cause of this variation is not known but might at least, in part, be because of additional genetic factors, i.e., modifier genes (21). Such modifier genes might be the same genes as the low-penetrance genes involved in sporadic colorectal cancer (19).

The allelic differences in low-penetrance genes may account for the wide interindividual differences in the sensitivity to cancer-inducing or cancer-promoting compounds (18, 22). The increased cancer risk for the individual carrying a variant in one of these genes is estimated to be small, but the high frequency in the population of some of these variants suggests that the population attributable risk can be high (23).

This report focuses on low-penetrance genes and gene variants involved in colorectal cancer susceptibility other than possible low-penetrance variants in the APC gene and the mismatch repair genes. The external variables taken into account were ethnicity, gender, and tumor localization whenever possible. Pooled analyses were performed on all polymorphisms reported in more than one study. In addition, the sample size required to detect an association with colorectal cancer susceptibility with sufficient power was addressed.

Published studies were traced using the PubMed databases from 1980 to 2001 (until September), using the search terms colorectal, adenoma, cancer, risk, and polymorphism(s) to identify candidate genes. For each specific candidate gene, a separate search was performed. For example, the terms HRAS1, colorectal, adenoma, cancer, and risk were used for HRAS1. In addition, the bibliographies of studies identified by the electronic searches were used as source. Studies eligible for our pooled analysis were those that compared genotype or allele frequencies of candidate genes in colorectal adenoma and cancer cases with healthy controls using genomic DNA. Only studies describing primary data or data that superseded earlier work were taken into account.

For a better insight in the possible effects of the various genes on colorectal cancer susceptibility, a pooled analysis for each polymorphism was performed. The studies and the results of the pooled analyses are shown in Table 1. The allele frequencies were based on the control individuals. The raw number of cases and controls from comparable studies were analyzed together. The genotype-specific ORs and 95% CIs were calculated for all studies combined, without adjustment for external variables. This can result in values that differ from those in the original article. Whenever possible, a distinction was made between heterozygous and homozygous carriers of the variant allele. ORs from colonoscopy- or sigmoidoscopy-based studies may not be interpretable as relative risks because the indication for colonoscopy is often a positive family history for colorectal cancer. In these patients, even when no polyps have been found with colonoscopy, the risk of colorectal cancer can still be increased. Therefore, caution should be taken by extrapolating the findings in these studies to the general population. No adjustment has been made for these studies in the pooled analysis. The colonoscopy or sigmoidoscopy based studies are marked in the tables.

Where metabolic polymorphisms were assumed to be associated with a specific phenotype, a distinction was made between phenotype- and genotype-based studies (e.g., CYP2D6 and NAT2). For the genotype studies of CYP2D6 and NAT2, genotypes were combined according to phenotypic classes. Also, where possible, separate analyses were performed for the three major ethnicity subgroups (white, African-American, and Asian; Table 1) for gender (Table 2) and for tumor localization (Table 3). A difference in interpretation was made between statistically significant results for polymorphisms reported in one study and that obtained from a pooled analysis of several because all of the single studies were small, and the results have to be replicated. Therefore, the polymorphisms reported in one study are shown in a separate table (Table 4) and are not described in the results section. This separation is present also in the “Discussion.”

Finally, sample sizes required to detect an association with colorectal cancer susceptibility with sufficient power were calculated as described previously (24).

The polymorphisms that were reported in colorectal adenoma and/or cancer patients and controls in more than one study each are described separately and summarized in Tables 1, 2, and 3. The polymorphisms reported in only one study are shown in Table 4.

One of the strongest dietary associations with colorectal cancer susceptibility reported is that with diets high in meat (25, 26). For this association, HAAs formed during the cooking of meat may be mediators (27). Like many other chemical carcinogens, these substances/compounds require metabolic activation to bind to DNA and contribute to cancer causation. In the case of HAAs, the enzymes CYP1A2, NAT1, and NAT2 mediate this activation (28, 29, 30). On the other hand, HAAs can be detoxified by GST enzymes (31). Polymorphisms in these genes may affect the enzyme activity or inducibility. The GST enzymes are also involved in the detoxification of PAHs (32). PAHs are carcinogens present in cigarette smoke. PAHs are activated by CYP1A1 (33).

Certain substrates, including almost all carcinogens, are metabolically activated by the CYP enzymes, which results in the formation of chemically reactive mutagenic electrophiles (18). Most prescribed drugs are substrates for one or more CYP isoenzymes. Individual CYP isoenzymes have a unique substrate specificity, although a certain overlap between the enzymes is present (18).

The CYP1A1 gene encodes the strongly inducible aryl hydrocarbon hydroxylase enzyme responsible for the activation of PAHs (33). Differences in xenobiotic metabolic activity between individuals can be >200-fold, even within one family (34). Two polymorphisms have been examined in relation to colorectal cancer, namely the m1 (MspI RFLP) and m2 (A462G, exon 7) polymorphisms. These polymorphisms are in linkage disequilibrium (18). The functional significance of the m1 polymorphism is unknown, whereas the m2 polymorphism gives an increased enzymatic activity (35). This might result in a higher concentration of activated PAH metabolites.

Pooled analysis for both polymorphisms revealed no association, regardless whether the studies were analyzed separately or combined (36, 37, 38, 39, 40, 41, 42). For the combination of the two polymorphisms, examined in one study, no association with an increased colorectal adenoma or cancer risk was found (37).

The CYP2D6 variant allele is the result of a deletion of a 17.5-kb region, including the entire CYP2D6 gene (43). In the Caucasian population, 5% is homozygous for this mutation (18, 44), referred to as poor metabolizers (18). The CYP2D6 gene mutation is characterized by the inability of homozygous carriers to metabolize specific drugs (e.g., debrisoquine, sparteine, and bufuralol; Ref. 18). The CYP2D6 gene mutation was examined in one phenotype (45) and two genotype studies(40, 46). These studies found no association with colorectal cancer either separately or when combined.

CYP2E1 is an ethanol inducible enzyme, involved in the activation of N-nitrosamines (47). Variant allele carriers (in particular homozygous individuals) of a polymorphism (G1259C) in the CYP2E1 gene have a higher enzyme level (48). This could lead to a faster activation of carcinogens, which may induce colorectal cancer. For this polymorphism and another polymorphism in intron 6, no association was found with colorectal cancer (40, 42).

GSTs belong to a superfamily of detoxification enzymes that provide critical defenses against a large variety of chemical carcinogens and environmental toxins (49, 50). HAAs and PAHs are considered carcinogens that potentially cause colorectal cancer in humans and that are detoxified by GSTs (31, 32, 51). Foods that are known to induce GST synthesis are thought to be protective against colorectal cancer (52).

Deletion variants that are associated with a lack of enzyme function occur at the GSTM1 and GSTT1 gene locus (53). Individuals homozygous for null deletions in the GSTM1 and/or GSTT1 genes may have an impaired ability to metabolically eliminate carcinogens and may therefore be at increased cancer risk. Because GSTM1 is expressed at low levels in the colon, it seems, however, unlikely that the GSTM1 null genotype predisposes to colorectal cancer (54). GSTP1 is a more obvious candidate for a colorectal cancer susceptibility gene as it is present at high levels in the colon (31, 54). For the GSTT1 enzyme, there are no data available concerning the expression in the colon.

Our pooled analysis for GSTM1 revealed no association with colorectal adenoma or cancer (38, 40, 42, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71).

Two polymorphisms have been described in the GSTP1 gene, one in codon 105 and one in codon 114. These codons are only ∼1-kb apart, and the polymorphisms exhibit strong linkage disequilibrium with each other (62). The codon 114 variant allele is only found in combination with the codon 105 variant allele. The codon 105 polymorphism modifies the enzyme’s specific activity (72). No association with colorectal cancer was observed for both polymorphisms (62, 68, 70, 73, 74).

The GSTT1 null genotype frequencies show a large variation among racial groups. Our pooled analysis showed that the GSTT1 null genotype was associated with a small increase in colorectal cancer risk (40, 57, 58, 60, 61, 64, 66, 68, 69, 70, 71). This overall association was caused by three strong studies and five studies with a modest nonsignificant association. Subgroup analyses for gender and tumor localization revealed no association with colorectal cancer, but these analyses were based on relatively small numbers of cases and controls as compared with the overall pooled analysis.

One study examined both GSTM1 and GSTM3 polymorphisms (70). The combined high-risk genotype was associated with an increased risk (OR = 2.4, 95% CI: 1.32–4.3). Two other studies examined the GSTM1 and GSTP1 (codon 105) polymorphisms (62, 75), and five studies examined the GSTM1 and GSTT1 polymorphisms (58, 61, 66, 68, 75). Pooled analyses showed no associations. In one study, the GSTP1 (codon 105) and GSTT1 polymorphisms were studied (75). Again, no association was observed. There are no studies available that tested the combination of polymorphisms in three or four genes.

Epithelial cells of the colonic mucosa express both NAT1 and NAT2 activity (76, 77), although the human colonic epithelium appears to contain 100–200-fold more NAT1 than NAT2 (76, 78, 79). Furthermore, the ratio of NAT1:NAT2 activities show large interindividual variations (77). Polymorphic forms of the NAT genes have the potential to affect an individual’s response to carcinogens, thereby influencing cancer risk. There is linkage disequilibrium between the NAT1*10 allele and the NAT2*4 allele (80).

NAT1*4 is the most common NAT1 allele and is presumed to be the wild type, whereas NAT1*10, present in ∼30% in populations of European ancestry, is associated with increased NAT1 activity (81). NAT1 fast acetylators are defined as carriers of at least one copy of the NAT1*10 allele. There may be some misclassification of NAT1*10 alleles because a common test to detect the NAT1*10 allele does not distinguish between NAT1*10 and NAT1*14 or other NAT1 alleles (82). In the overall pooled analysis, no association was found between colorectal cancer and the NAT1 fast genotype (64, 81, 83, 84, 86). Two studies were excluded in the pooled analyses because of lack of data on the NAT1*10 allele (86) or of raw numbers of genotypes for cases and controls (40). Tumor localization was taken into account in the one study that showed an association (81). Only for distal tumors, the association remained significant.

Different point mutations are present in the NAT2 gene. These mutations (variants) cause defective NAT2 alleles. NAT2*4 is the wild-type allele. In the Caucasian population, ∼60% of the individuals have two defective NAT2 alleles. The NAT2 presumed phenotypes are classified as fast (homozygous and heterozygous NAT2*4 allele carriers) and slow (homozygous variant allele carriers) acetylators. Pooled analysis revealed an association between colorectal cancer and fast acetylatorship [Refs. 87, 88, 89; excluding one study that did not provide the raw numbers (90)]. The pooled analysis for genotype studies [Refs. 65, 81, 84, 85, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101; excluding one study that did not provide the raw numbers (40)] detected neither an association between colorectal adenoma or cancer and presumed fast acetylatorship overall, nor in subgroup analyses for ethnicity, gender, and tumor localization. Thus, whereas the phenotype studies showed a positive association between fast acetylatorship and colorectal cancer, the genotype studies revealed no association. A possible explanation for these conflicting results is that some of the negative studies have been carried out in Japan, with a high proportion of fast acetylators and a low incidence of colorectal cancer. However, when Asian studies were excluded in the pooled analysis, still no association was detected between the fast NAT2 genotype and colorectal cancer (OR = 1.05, 95% CI: 0.96–1.14). Furthermore, there is some evidence that phenotype study methods may be influenced by disease or surgery (102, 103) by the specificity of substrates used and by overlapping activity of NAT1. The drug tests are also under the influence of liver and renal functions (104). Apart from these explanations for the conflicting results, it may well be that the result for the combined phenotype studies is falsely positive; as for the genotype studies, pooled analysis was performed on a large sample size (3554 cancer patients and 4070 controls), whereas the pooled analysis for the phenotype studies was based on a small sample size (201 cancer patients and 223 controls).

Two studies examined colorectal adenoma and cancer risk for the genotype combinations of NAT1 and NAT2 (83, 84). No association was detected.

NQO1 is a polymorphic enzyme involved in the detoxification of potentially mutagenic and carcinogenic quinones (105) and HAAs (106). NQO1 also plays a role in the activation of the potent antioxidant vitamin K (107). The homozygous C609T genotype, which is associated with nondetectable NQO1 activity (108, 109), is found in 2–20% of individuals (110). The heterozygous genotype was shown to have a 3-fold decreased NQO1 activity (111). Our pooled analysis, however, showed no association with colorectal cancer for this polymorphism (112, 113). Another polymorphism (Pro187Ser) generates a null allele and produces no active enzyme (114). One study, which did not provide the raw numbers, found this polymorphism in 19% of colon cancer patients and in 20% of healthy controls (109).

Imbalanced DNA methylation, characterized by genomic hypomethylation (115, 116) and methylation of usually unmethylated CpG sites (117, 118), is observed consistently in colon cancer (119).

The 5,10 MTHFR enzyme and the MTR enzyme play an important role in folate and methionine metabolism and are both important for DNA methylation and synthesis (120).

A common polymorphism (C677T) in the MTHFR gene is associated with decreased enzyme activity (121), lower plasma folate levels, and increased plasma homocysteine levels (122, 123). Enzyme activity is 70% lower in homozygous carriers of the variant allele and 35% in heterozygous carriers than in homozygous carriers of the wild-type allele (121). Pooled analysis did not show an association between adenoma risk and the polymorphism (124, 125, 126, 127, 128, 129, 130, 131). However, for colorectal cancer, a decreased risk was observed for homozygous carriers of the risk allele. Analysis by gender revealed a decreased cancer risk for both subgroups, although the association reached significance only for the males (OR = 0.70, 95% CI: 0.55–0.91; females OR = 0.86, 95% CI: 0.61–1.22).

The lack of association with adenoma risk suggests that the MTHFR enzyme plays a role in a late stage of colorectal tumorigenesis. Homozygous carriers of the C677T polymorphism probably represent a subpopulation with increased folate needs. The homozygous carriers with adequate folate levels seem to be at lower risk for colorectal cancer (132). However, when folate levels are low (because of low intake or depleted by alcohol consumption), both DNA methylation and DNA synthesis (as a result of disturbances in nucleotide synthesis) might be impaired among homozygous carriers, resulting in an increased colorectal cancer risk (132).

The most common polymorphism of this gene (Asp919Gly) alters the amino acid sequence of the protein at a potentially functional site (133). However, no difference in biological function has been reported for this polymorphism. Although decreased risks were found for two studies, one in women with adenomas (134) and one in men with cancer (135), statistical significance was not reached.

The APOE enzyme plays a role in the regulation of cholesterol and bile acid metabolism (136). APOE also effects immunoregulation and cell proliferation (137, 138). The human APOE gene has three common alleles (ε2, ε3, and ε4). Individuals with the ε4 allele have higher levels of serum total and low density lipoprotein cholesterol, whereas individuals with the ε2 allele have lower levels (139, 140). In the pooled analysis, no association with colorectal adenoma or cancer was discovered (40, 141, 142). However, in subgroup analyses for localization, decreased risks were found for proximal adenomas and cancer in carriers of the ε4 allele.

The PLA2G2A gene encodes for secretory phospholipase A2, which is involved in synthesis of prostaglandins. A possible role for PLA2G2A in colorectal cancer susceptibility was revealed by studies in the Min mouse. The Min mouse, with a germ-line mutation in the APC gene, is a model for FAP. Mutations in the PLA2G2A gene dramatically increase the number of intestinal polyps in the Min mouse. Fourteen patients with FAP and 20 patients with sporadic colorectal cancer were screened for PLA2GA2 germ-line and somatic mutations (143). In one sporadic colorectal cancer patient, a frameshift germ-line mutation was detected. The wild-type allele was somatically lost in the tumor of this patient. Because this loss of heterozygosity did not include a flanking microsatellite locus (D1S436), the authors concluded that the homozygous loss of the PLA2G2A gene itself might have contributed to cancer development. This marker was located telomeric to the PLA2G2A gene, the region centromeric to the gene was not examined for loss of heterozygosity. Two polymorphisms have been described in the PLA2G2A gene, one in exon 1 and one in exon 3. Both polymorphisms are unlikely to have any functional effect (144). No association was found in the one study that examined both polymorphisms (144).

The proto-oncogene HRAS1 encodes a protein involved in mitogenic signal transduction and differentiation (145). The HRAS1 gene is highly polymorphic in the human population (146). It encompasses four exons with a variable number of tandem repeats region at the 3′ end (147), with four common alleles and dozens of variants. Each variant is derived from the common allele nearest in size to it (148). Allele-specific effects have been observed (149). The available studies for the HRAS1 polymorphism were heterogeneous; the number of different HRAS1 alleles range from 5 or 6 to >20 according to different authors, and this situation makes it difficult to compare data obtained in different laboratories. However, in most studies there are four common alleles, and the rest of the alleles are listed as rare. Our pooled analysis revealed that the rare HRAS1 alleles were associated with a moderately increased colorectal cancer risk (150, 151, 152, 153, 154).

The L-myc gene encodes a DNA-binding, nucleus-associated protein that is sometimes activated late in tumorigenesis (155). A polymorphism has been described in this gene, with a large (L) allele and a small (S) allele (156). There is no evidence for the functional significance of this polymorphism.

No association with colorectal cancer was observed, regardless whether the studies were analyzed separately or combined (57, 157, 158, 159).

The Tp53 gene plays a role in the protection against replication of damaged DNA (160, 161). Somatic mutations in the Tp53 gene have been found in many tumor types (162, 163), including colorectal cancer (164). Three different Tp53 polymorphisms (i.e., in intron 3, exon 4, and intron 6) have been studied in colorectal cancer patients. All three polymorphisms exhibit strong linkage disequilibrium with each other (165). The exon 4 polymorphism appears to be functionally relevant because the wild-type allele (Arg) has in vitro a weaker affinity for several transcription-activating factors (166). Although the functional significance of the intron 3 and intron 6 polymorphisms is unclear, intronic sequences in Tp53 have been implicated in the regulation of gene expression and in DNA protein interactions (167, 168).

In one study, all three polymorphisms were examined (169) and, in another study, only the exon 4 polymorphism (170). The only association found with polymorphisms individually was with the intron 3 polymorphism (169). For both heterozygous and homozygous carriers of the variant allele, a decreased colorectal cancer risk was detected, although only the OR for heterozygous carriers reached significance (169). The same study examined all three polymorphisms and analyzed the three possible combinations of two polymorphisms (169). For the haplotype composed of two variant alleles, decreased colorectal cancer risks were detected for all three possible combinations, although for only one combination, significance was reached (combination intron 3 and exon 4, OR = 0.49, 95% CI: 0.28–0.85). For the two haplotypes composed each of one variant allele and one wild-type allele, increased risks were found for all three combinations. Two combinations reached significance, i.e., the combination of intron 3 and exon 4 (OR = 1.80, 95% CI: 1.26–2.58) and that of exon 4 and intron 6 (OR = 1.66, 95% CI: 1.14–2.42). When the genotypes were combined for all three polymorphisms, an association was observed for carriers of two variant alleles (OR = 1.85, 95% CI: 1.07–3.22) and three variant alleles (OR = 2.44, 95% CI: 1.41–4.22). No explanation was given for these conflicting results. One possible explanation is that the polymorphisms themselves are nonfunctional but in linkage disequilibrium with other variants. However, the small sample size of the Tp53 studies might also be an explanation.

TNF-α and TNF-β play an important role in the inflammatory response (171, 172). The TNF cytokines are well known for their cytotoxic and antitumor activity. However, some of their properties such as enhanced angiogenesis and up-regulation of adhesion molecules could be advantageous for cancer development. Increased serum TNF-α levels have been described in cancer patients, including those with colon cancer (173, 174).

One polymorphism (TNFa) in the vicinity of the TNF-α gene has 14 different alleles (a1–a14). It has been a matter of controversy whether or not the a2 allele is associated with a change in TNF-α production. The a6 allele was associated with lower TNF-α secretion from activated monocytes (175). Two other polymorphisms in the TNF-α promoter region have been described at position −308 (G to A substitution) and −238 (G to A substitution). The −308 polymorphism is associated with increased TNF-α production (176), whereas the functional significance of the −238 polymorphism is unknown. In the pooled analysis for the TNFa polymorphism, associations with colorectal cancer were detected for the a2, the a5, and the a13 alleles (177, 178). The studies for the −308 and −238 polymorphisms revealed no association with colorectal cancer separately or when combined (179, 180).

Experimental, clinical and epidemiological investigations have shown that iron can influence carcinogenesis (181). Increased body iron stores have been associated with an increased colorectal adenoma and cancer risk (182, 183, 184). A number of genes is involved in iron metabolism, including the HFE gene and the TFR gene.

Two point mutations (Cys282Tyr and His63Asp) have been detected in the HFE gene in HH. Over 80% of the HH patients are homozygous for the Cys282Tyr mutation (185). Heterozygous carriers, comprising 15% of the American population, have, on average, increased iron stores as compared with noncarriers (186, 187). In a study of 1950 HH heterozygotes (parents of HH patients, genotype unknown) and 1656 controls, an increased colorectal cancer risk was detected in males (OR = 1.28, 95% CI: 1.07–1.53) and an increased colorectal adenoma risk in females (OR = 1.29, 95% CI: 1.08–1.53; Ref. 187). The other studies that examined the Cys282Tyr genotype found no association with colorectal cancer separately or combined (188, 189, 190). The study of HH heterozygotes was not included in the pooled analysis because of the difference in study design. The His63Asp genotype was examined in one of these studies (188). No association was found.

The ALDH2 gene encodes a mitochondrial enzyme responsible for the oxidation of acetaldehyde that is generated in alcohol metabolism. This is of interest as acetaldehyde induces cytotoxicity (191) and DNA damage (192) and enhances folate deficiency. A polymorphism in codon 487, only prevalent in Asians, dramatically diminishes the enzyme activity (193). Homozygous carriers of the variant alleles are highly intolerant to alcohol and, consequently, do almost not drink alcohol. The tolerance of heterozygous carriers is intermediate. In an alcohol-challenge study, heterozygous carriers of the variant allele had blood acetaldehyde concentrations almost six times higher than homozygous carriers of the wild-type allele (194). Our pooled analysis for this polymorphism revealed increased colorectal cancer risks for heterozygous as well as for homozygous carriers of the risk allele (195, 196).

The VDR belongs to the steroid/thyroid receptor family. Apart from the regulation of calcium metabolism, vitamin D₃ plays an essential role in cell proliferation and differentiation in several tissues, including colonic epithelium (197). Several polymorphisms have been described in the VDR gene. The BsmI polymorphism is located in the 3′ region of the gene, and the FokI polymorphism is located in the start codon. These polymorphisms are not in linkage disequilibrium (198). The BsmI polymorphism appears to have phenotypic consequences for calcium and vitamin D metabolism (199, 200, 201). The FokI polymorphism produces receptor variants differing in size and activity (202). For both polymorphisms, no association with colorectal adenoma and cancer was found for the studies separately or when combined (203, 204, 205).

Nine combinations of two polymorphisms in different genes were studied. Two of these combinations, i.e., the combination CYP1A2 and NAT2 (OR = 2.77, 95% CI: 1.51–5.06; Ref. 30) and GSTT1 and NAT2 (OR = 2.33, 95% CI: 1.15–4.72; Ref. 68), showed an association with colorectal cancer risk for the combined high risk genotype. However, another study found no association with colorectal cancer risk for the CYP1A2 and NAT2 combination (206). Unfortunately, the raw numbers were not provided. For the other combinations, i.e., CYP1A1 and GSTM1 (38), GSTM1 and NAT1 (75), GSTM1 and NAT2 (65), GSTP1 and NAT1 (75), GSTT1 and NAT1 (75), GSTM1 and L-myc (57), and the combination HFE and TFR (189), no associations with colorectal cancer were found.

One study examined the combination of polymorphisms in three genes, i.e., CYP1A1, CYP2E1, and GSTM1 (42). The genotype of the three variant alleles was associated with an increased colorectal cancer risk (OR = 4.62, 95% CI: 1.29–16.54).

In two studies, the GSTM1 and GSTT1 polymorphisms were examined in HNPCC patients (64, 207). In the first study, which compared 26 unaffected- and 48 cancer-affected HNPCC mutation carriers, neither an association was observed between the null genotypes (separately or combined null) and the occurrence of cancer, nor with the age at onset of the tumor or localization of the tumor (207). In the second study, only 114 affected HNPCC mutation (MLH1 exon 16 deletion) carriers were examined, and there were no unaffected mutation carriers examined (64). Both of the null genotypes were associated with a younger age at onset. Furthermore, the GSTT1 null genotype and the combined null genotype were associated with a proximal tumor localization.

In the first study, also the NAT2 genotype was examined (207). An increased colorectal cancer risk was observed for carriers of the slow NAT2 acetylator genotype, whereas no association was found with age at onset of colorectal cancer. The second study also examined the NAT1 genotype (64). The NAT1*10 genotype was associated with a younger age at onset and with a distal colon tumor localization.

The exon 4 polymorphism in the CCND1 gene was examined in two studies in HNPCC mutation carriers (208, 209). In the first study in 49 affected and 37 unaffected mutation carriers, patients with the AA or AG genotypes were 2.5 times more likely to develop colorectal cancer at any age than were patients with the GG genotype (208). Also an association was observed between variant allele carriers and the age at onset. However, in another study in 146 affected mutation carriers, no association was found with the age at onset (209).

In a study in 43 affected and 24 unaffected HNPCC mutation carriers, the ATM 1853N genotype was associated with a higher incidence of colorectal cancer and other HNPCC-related cancers (OR = 8.90, 95% CI: 1.08–73.44) but not with a younger age at diagnosis (210).

The genotypes of 46 mutation carriers and 31 noncarriers from a FAP kindred were determined at 14 microsatellites surrounding the PLA2GA2 locus on chromosome 1p35–36 (211). The development of extracolonic symptoms was associated with one of the markers. However, this marker (D1S211) was far apart (27 cM) from the PLA2GA2 locus. In another study, a polymorphism in exon 3 in the PLA2GA2 gene was associated with relatively severe colonic FAP (144). Two other studies detected no associations between the PLA2G2A gene and the colonic (212) or extracolonic phenotype (213) of FAP.

Many studies have shown that polymorphisms in a significant number of genes affect colorectal cancer risk. Recently, Houlston et al. (214) described an analysis similar to ours. Whereas they studied 21 polymorphisms in 15 genes, we studied 48 polymorphisms in 35 genes. Houlston et al. (214) concluded in their review that HRAS1-variable number of tandem repeats, MTHFR variants, and APC variants represent the strongest candidates for low-penetrance susceptibility alleles identified to date. We also found an increased risk for the first one and a decreased risk for the second, with similar ORs to the ones they reported. We did not study the APC-I1307K polymorphism.

Houlston et al. (214) did not find an association for the GSTT1 gene. In contrast, our pooled analysis revealed a significant association between this gene and colorectal cancer, although the increase in risk we detected was small to moderate. This discrepancy is probably attributable to the larger sample size that we included in our analysis. Significant associations were found for three other polymorphisms in our pooled analysis, namely the intron 3 polymorphism in the Tp53 gene, the TNFa polymorphism in the TNF-α gene, and the polymorphism in the ALDH2 gene. Houlston et al. (214) only addressed the Tp53 polymorphism and not the other two.

Furthermore, tumor localization and gender were considered separately in our review and studies examining the effects of modifier genes in HNPCC and FAP patients were included.

We examined 30 polymorphisms in 20 different genes, described in more than one study, whenever possible by pooled analysis. This revealed an association with colorectal cancer for seven polymorphisms in seven genes. As mentioned above, increased colorectal cancer risks were found for the polymorphisms in GSTT1, NAT2 (phenotype), HRAS1, TNF-α (a2 allele of the TNFa polymorphism), and ALDH2, with population-attributable risks ranging from 4 to 37%. Decreased colorectal cancer risks were found for MTHFR, Tp53 (intron 3), and TNF-α (a5 and a13 allele of the TNFa polymorphism). For most of these polymorphisms, except for GSTT1 and MTHFR, the pooled analysis was performed on a few cases, ranging from 155 to 399, resulting in low statistical power. Because of this, the associations with colorectal cancer are not firm. Much larger sample sizes are needed to confirm or to refute the described associations. For the MTHFR polymorphism, the pooled analysis was performed on a large sample size (for adenomas 1461 cases and 2088 controls, for cancer 2064 cases and 3229 controls). Therefore, there is strong evidence for the decreased colorectal cancer risk associated with this polymorphism. For the GSTT1 polymorphism, the pooled sample size was intermediate (1490 cases and 2026 controls).

The pooled analysis for 24 polymorphisms in 16 genes, reported in more than one study, namely CYP1A1 (m1 and m2), CYP2D6, CYP2E1 (G1259C and intron 6), GSTM1, GSTP1 (codon 105 and 114), NAT1, NAT2 (genotype), NQO1, MTR, APOE, PLA2G2A (exon 1 and 3), L-myc, Tp53 (exon 4 and intron 6), TNF-α (−308 and −238), HFE (C262Y and H63D), and VDR (BsmI and FokI), revealed no association with colorectal adenoma and/or cancer. For the polymorphisms with large sample sizes, i.e., GSTM1 and NAT2 (genotype), an association with colorectal adenoma and cancer can be excluded. For two polymorphisms, i.e., NAT1 and APOE, with pooled analysis performed on an intermediate sample size, there is probably no association with colorectal adenoma and cancer, although an association cannot be completely excluded. Indeed, subgroup analysis for localization for the APOE polymorphism revealed a decreased proximal colorectal adenoma and cancer risk in carriers of the ε4 allele, although the sample size of this subgroup was small (109 proximal adenomas with 419 controls and 41 proximal carcinomas with 199 controls).

For all other polymorphisms with small sample sizes ranging from 27 to 862 cases, an association with colorectal adenoma and cancer is still unknown.

Eighteen polymorphisms in 15 genes, i.e., CCND1 (215), CYP1A2 (30), CYP1B1 (37), CYP2C9 (216), DCC (217), EphB2 (218), ER (PvuII and XbaI; Ref. 205), GSTM3 (70), mEPHX (exon 3 and exon 4; Ref. 219), TFR (189), TGFB1 (220), TNF-β (175), TSER (221), uPAR (222), and XRCC1 (codon 194 and codon 399; Ref. 223) are each described in only one study, all with very small sample sizes ranging from 31 to 206 cases (Table 4). For 11 polymorphisms in 10 of these genes, an association with colorectal adenoma and/or cancer was found, whereas no association was found for the other seven polymorphisms in five genes. To replicate positive studies for polymorphisms described in only one study, a sample size of roughly four times that of the initial study is needed to conclude from a positive study that the original effect is likely to be an artifact (24).

Five studies examined the role of certain polymorphisms in HNPCC mutation carriers (64, 207, 208, 209, 210). Although associations were found, some results were conflicting, and the sample sizes were too small to draw firm conclusions.

The PLA2G2A gene region (chromosome 1p35–36) was examined in FAP patients. Although a modifier gene was likely to exist in the chromosome region 1p35–36, it was doubtful whether the PLA2G2A gene itself was this gene.

Because the products of several genes interact (almost half of the reviewed genes are metabolic pathway genes), interactions between the genes with respect to cancer risk are likely. For polymorphisms not associated with colorectal adenoma or cancer when studied separately, an association with colorectal adenoma or cancer is still possible in combination with other polymorphisms. Nine combinations of two polymorphisms in different genes were described. For two of these combinations, an association with colorectal cancer was shown with the combined high-risk genotypes of CYP1A2 and NAT2 (30) and of GSTT1 and NAT2 (68). However, all studies were performed on small sample sizes, resulting in low statistical power. One study examined the combination of polymorphisms in three genes, i.e., CYP1A1, CYP2E1, and GSTM1 (42). The genotype of the three variant alleles was found to be associated with an increased colorectal cancer risk but again the sample size was small.

With regard to the pooled analyses, some notes have to be made. Most important is the possibility of publication bias. In general, negative findings might be more difficult to get published. Another important issue is the pooling of the results of the different ethnicity groups because in many studies the ethnicity of subjects is not reported. Linkage disequilibrium for certain variants often differ between populations (224). The overall risk for all samples may therefore be invalid, when the variant itself is nonfunctional, but in linkage disequilibrium with some other functional variant. In most studies, no distinction is made between gender and localization of the tumor. Several lines of evidence suggest that proximal and distal colorectal cancers may differ at least partly in their etiology (225, 226). Thus, it is possible that different genes play a role in proximal as compared with distal colorectal cancer. This may also be true for women versus men. In our pooled analysis, external variables like dietary or lifestyle factors are not taken into account. It is possible or even likely that some of the candidate low-penetrance genes only contribute to colorectal cancer in combination with certain dietary and/or lifestyle factors.

In conclusion, truly reliable judgments on these associations should for most of the candidate low-penetrance genes be based on much larger numbers of affected and unaffected subjects than we were able to study with available data from the literature. Actually, before a study is performed, the sample size required to detect an association with sufficient power should be calculated. It is important to realize then that the genetic effect or OR is often overestimated, resulting in too small samples after all.

Only simultaneous genotyping and combined analysis of different polymorphisms in large numbers of patients and controls, stratified by ethnicity, gender, and tumor localization, and taking relevant dietary and lifestyle habits into account, will make it possible to describe the exact relations between polymorphisms and colorectal cancer susceptibility with an adequate power.

The ability to identify these genes and to understand their interactions with other relevant environmental, endogenous, and genetic factors is important. It will help to identify high-risk individuals for entry into surveillance programs and to reveal causative factors, allowing more effective prevention strategies (93), thereby reducing the morbidity and mortality of colorectal cancer (227).