The genes GSTM1 (chromosome 1p13.3) and GSTT1 (22q11.2) code for cytosolic enzymes GST-mu and GST-theta respectively, which are involved in phase 2 metabolism. Both genes may be deleted. There is geographical and ethnic variation in genotype frequencies for both genes. In developed countries colorectal cancer (CRC) is the second most common cancer. CRC has been inconsistently associated with polycyclic aromatic hydrocarbons (PAHs) in diet and tobacco. Because GST enzymes are involved in PAH metabolism, it has been postulated that genotype may modify CRC risk associated with PAH exposure. No consistent associations between GSTM1 or GSTT1 genotype and CRC have been observed. However, most studies have methodological limitations. Few studies have investigated gene-environment interactions. No interactions between GSTM1 or GSTT1 genotype and smoking and CRC risk have been reported. One polyp study suggests an interaction between GSTM1 genotype and smoking. Two studies suggest increased disease risk in subjects with high meat intake and GST non-null genotype, contrary to the underlying hypothesis. One study suggests a strong inverse relationship between CRC and broccoli consumption, particularly in subjects GSTM1 null. These finding require confirmation. Methods for determining GSTM1 and GSTT1 genotype are well established. Population testing is not currently justified.

Four GST isoenzyme classes have been identified - alpha, mu, pi and theta (1). Here we consider the two types most investigated in relation to colorectal cancer – GSTM-mu and GST-theta . These are summarised in Table 1.

The GSTM1 and GSTT1 genes code for the cytosolic enzymes GSTM-mu and GST-theta respectively. These enzymes are involved in the conjugation reactions in phase 2 metabolism of xenobiotics (1), catalysing reactions between glutathione and a variety of electrophilic compounds (2). It is thought that most GST substrates are xenobiotics or products of oxidative stress, including some environmental carcinogens (1). In particular, the enzymes detoxify the carcinogenic polycyclic aromatic hydrocarbons (PAHs) present in diet and tobacco smoke (3). They also conjugate isothiocyanates, which are potent inducers of enzymes that detoxify environmental mutagens (4), to glutathione thereby diverting the isothiocyanates from the enzyme induction pathway to excretion (5). It has been postulated that the GST enzymes and the genes encoding these may be involved in susceptibility to cancer (6).

The genes coding for the enzymes GST-mu and GST-theta are polymorphic. There are three alleles at the GSTM1 locus, located on chromosome 1p13.3: GSTM1 null – a deletion, GSTM1a and GSTM1b (6). GSTM1a and b differ by a substitution in one base pair. There is no evidence of functional differences between them (6). The GSTT1 locus is located on chromosome 22q11.2 and is, in some instances, deleted (6). For both GSTM1 and GSTT1, the hypothesised consequence of the null genotype is reduced, or no, conjugation activity. Evidence is lacking on whether heterozygosity in either GSTM1 or GSTT1 affects gene function.

We searched Medline and EMBASE using the MeSH heading "glutathione transferase" and the textwords "GST" and "glutathione S transferase" for papers published between 1993 and 1998. We also searched the CDC Office of Genomics and Disease Prevention Medical Literature Search and reviewed reference lists in published articles. We identified relevant papers and critically appraised them. This section includes studies reporting genotype frequencies in a variety of groups of individuals without cancer.

The frequency of individuals who are homozygous for the GSTM1 null genotype is summarised in Table 2 (7-78), and those homozygous for the GSTT1 null genotype in Table 3 (8, 10, 11, 14, 29, 32, 35-37, 39, 47-49, 51-53, 55, 58, 59, 63, 65, 68, 73-75, 77, 78, 80-85). Many of the series were control groups in case-control studies of cancer. However, few could be described as truly population-based; therefore selection or participation biases may account for some of the variation between studies. Some of the studies have small numbers of participants. Ethnicity is not always straightforward to establish: this limits the generalisability from, for example, one "white" population to another (86).

GSTM1

In African populations, the frequency of the GSTM1 null genotype ranges from 23%-48%; in Asian populations from 33%-63%; and in European populations from 39%-62%.

Published data from the Americas relate only to studies carried out in the USA; the range of reported frequencies is 23%-62%. In African-Americans and blacks the range is 23-41%, and in whites 35%-62%. In the two studies of subjects of Asian origin, the range was 32%-53%, and in the three studies including Hispanic/Mexican-American subjects the range was 40%-53%.

In two Australian series the frequency is 51%-54%. The highest frequencies have been reported in studies involving small numbers of subjects from parts of the South Pacific – 64%-100% (13). These studies differed from the others in that Southern Blotting rather than PCR methods were used.

GSTT1

The range of frequencies of the GSTT1 null genotype is 16%-64% in Asia, with frequencies of 44% or higher being reported in China, Japan, Korea and the Singapore Chinese. Thus, in some Asian populations, it has been suggested that the frequency of GSTT1 null deletions is similar to that of GSTM1 null. However, in African, African-American and white populations, the frequencies of GSTT1 null are lower than for GSTM1 null. The range of frequencies in three African series is 15%-26%, and in Europe 10%-21%. As was the case for GSTM1, data from the Americas relate only to the USA, where the range of frequencies is 10%-36%. In whites the range is 15%-27%; in African-Americans and blacks 22%-29%; and in Mexican-Americans, based on two studies, 10%-12%. No data on Asian subjects in the US is available. In three groups in Australia, the frequency of GSTT1 null ranged from 9%-19%.

Concordance between genotype and phenotype

Individuals lacking GST-mu or GST-theta activity can be identified using phenotypic assays which classify individuals as active or inactive on the basis of a bimodal distribution. Use of PCR methods informs of the presence or absence of the GSTM1 or GSTT1 alleles. Several studies have investigated concordance between genotype and phenotype; this can be a means of determining whether the appropriate section of DNA coding for the particular phenotype has been identified.

Four studies in Europe and one in the USA have demonstrated concordances between GSTM1 genotype and GST-mu phenotype of 94% or greater (26, 33, 87-89). However, in one study in which genotype and phenotypic status was compared in 63 healthy Zimbabwean volunteers, concordance was lower – 84% (90). This may have been due to presence of (i) other mutations which affect protein expression, (ii) compounds in the diet that may affect protein levels, or (iii) mutations in the regions of the gene which bind to the primers during PCR but which do not affect enzyme activity (90). Genotyping methods developed in populations of European origin may slightly underestimate the proportion of African populations with GSTM1 null genotype (90).

In two small studies (82, 91) and one larger one (83) in northern European populations concordance between GSTT1 genotype and conjugator status (phenotype) in excess of 95% was found.

Worldwide in 1996, there were estimated to be 875,000 new cases of colorectal cancer (92). There is substantial geographical variation in incidence (Figure 1) (93). Epidemiologic evidence suggests that much of the geographical variation reflects variations in environmental or lifestyle exposures, perhaps acting with variations in genetic factors. In developed countries colorectal cancer is the second most common cancer, and in developing countries it ranks sixth most common in men and fifth in women (94). In developed countries, the age standardised rates (30-47 per 100,000 in men and 24-31 in women) are typically about four times higher than in some developing countries (rates below 10 per 100,000 for both sexes) (93). The incidence of colorectal cancer is rising in most populations (95).

In most populations, cancer of the colon is more common than that of the rectum (93). The male:female ratio for colon cancer is around unity; and for rectal cancer 1.5 or greater (93). The incidence of colorectal cancer increases with age (93).

Excluding familial adenomatous polyposis or hereditary non-polyposis colorectal cancer, the risks of colorectal cancer to first degree relatives of index patients with the disease is about twice that of the general population (96, 97). The genetic basis of this familial aggregation has not yet been characterised.

Colorectal adenomatous polyps are thought to be precursors of colorectal cancer. While there is no direct evidence in support of the adenoma-carcinoma sequence, there is considerable indirect evidence from a range of epidemiological, histopathological and molecular genetic studies (98).

Exposure of meats to pyrolysis temperatures produces heterocyclic amines and PAHs (99, 100). The World Cancer Research Fund (WCRF) and American Institute of Cancer Research (AICR) panel recently concluded that consumption of red meat "probably" and intake of heavily cooked meats "possibly" increase the risk of colorectal cancer (101). In some studies raised risks of colorectal cancer have been associated with consumption of broiled or grilled meats and browning of the meat surface (102, 103). In a recent study an increase in risk associated with higher levels of both a white meat and an overall meat mutagen index in men was found (104). However, in other studies no association with consumption of broiled or grilled meats or browning of the meat surface was observed (105, 106).

High intakes of alcohol may be associated with increased risk of colorectal lesions (98). With regard to dietary factors which may be protective, the WCRF/AICR panel concluded that there is "convincing" evidence that consumption of vegetables and "possible" evidence that the consumption of non-starch polysaccharides/fibre, starch and carotenoids decrease risk (101). In eight out of twelve studies of colon cancer and all five studies of rectal cancer, high levels of consumption of cruciferous vegetables were associated with decreased risks (101). Cruciferous vegetables may have anti-carcinogenic properties as they contain isothiocyanates which induce enzymes that detoxify environmental mutagens (4, 5).

In addition to diet, the other major environmental source of exposure to PAHs is tobacco smoke. Most studies show a positive association between smoking and colorectal adenomas, but the association between smoking and colorectal cancer is less clear (107). However, in four recent large cohort studies, smoking has been associated with colorectal cancer after a long latent period (108-111).

There is consistent evidence from observational studies that higher levels of physical activity are associated with a reduced risk of colon cancer (112).

While the evidence from observational studies suggests that regular use of aspirin or other non-steroidal-anti-inflammatory drugs reduces the risk for colorectal cancer, no protective effect was found in the intervention trial in which US male physicians were given 325mg aspirin on alternate days or placebo for, on average, five years (113).

The studies appraised in this section were identified using the search strategy described earlier with the addition of MeSH headings and textwords relevant to colorectal cancer or polyps.

GSTM1 & Colorectal Cancer

The eight available case-control studies of GSTM1 and colorectal cancer (14, 23, 49, 54, 58, 69, 77, 78), and one of colorectal adenomas (114) are summarised in Table 4 and discussed below in order of publication. In four, exposure to environmental and lifestyle factors was assessed (14, 58, 69, 114); this is discussed in section 6.

The results of the colorectal cancer studies are inconsistent, three suggesting no association (49, 58, 77), three a slightly lower risk in those with the GSTM1 null genotype (23, 69, 78), and two an increased risk associated with this genotype (14, 54). The study of colorectal adenomas suggests a slightly lower risk in those with GSTM1 null genotype (114).

In the first reported study of colorectal cancer and GSTM1, Zhong et al. (54) found a significantly raised relative risk (RR) associated with the GSTM1 null genotype among 196 cases from an Edinburgh hospital and 225 controls from Sheffield, Edinburgh and Potters Bar (RR=1.8, 95% confidence interval 1.2-2.6). This is the only study in which a statistically significant association was observed. The risk was especially elevated for those with a proximal tumour (RR=3.4; 1.9-6.0).

Chenevix-Trench et al. (78) investigated 132 patients with colorectal adenocarcinoma and 200 controls in Australia. 100 of the controls were "unselected" and no further information on them was presented; the other 100 were geriatric patients without cancer or a family history of cancer. The RR of colorectal cancer associated with the GSTM1 null genotype was 0.9 (0.6-1.4). When the analysis was restricted to cases with a proximal tumour, the RR was also 0.9 (0.4-1.8). The proportions of cases aged less than 70 and over 70 years who were GSTM1 null were not significantly different. The authors acknowledge that their study had fewer cases, a smaller proportion of cases with proximal tumours, and a higher proportion of controls carrying the null genotype than the study of Zhong et al. (54); and hence there may have been inadequate statistical power to detect a relationship of the type observed in the earlier study.

In a Japanese study of 103 consecutive colorectal adenocarcinoma patients, and 126 subjects with no gastrointestinal symptoms or current or previous diagnosis of cancer visiting local medical clinics for regular medical check-ups, a RR of 1.5 (0.9-2.6) associated with the GSTM1 null genotype was observed (14). For proximal cases, the RR was 1.2 (0.6-2.3); and for distal cases the RR was 2.0 (1.0-3.9).

In another study in the UK of 252 colorectal cancer patients, and 577 patients without malignancy or inflammatory pathologies recruited through the same hospital, Deakin et al. (49) found a RR of 1.0 (0.7-1.3) associated with the GSTM1 null genotype. For tumours of the right colon the RR was 0.8 (0.5-1.2), for those of the left colon 1.1 (0.6-1.8) and for the rectum 1.2 (0.8-1.8).

In a study reported only in abstract form, Butler et al. (77) compared the frequency of GSTM1 genotypes between 219 white adults with sporadic colorectal cancer and 200 white blood donors in Australia. The relative risk of colorectal cancer associated with the GSTM1 null genotype was 1.0 (0.7-1.4).

Gertig et al. (58) conducted a case-control study nested within the Physicians’ Health Study (PHS) in the USA. 212 men with colorectal cancer were matched on year of birth and smoking history to men without colorectal cancer. A RR of 1.0 (0.7-1.5) was associated with the GSTM1 null genotype (adjusted for BMI, physical activity and alcohol use). The RRs were not substantially different when the analysis was stratified by age (<=60 years, >60 years). For proximal cancer the adjusted RR was 0.7 (0.4-1.3) and for distal cancer 1.4 (0.8-2.3).

Lee et al. (23) investigated the frequency of GSTM1 polymorphisms among Chinese subjects resident in Singapore. 300 cases of colorectal carcinoma were compared with 183 patients without history of neoplasms recruited from a Clinical Chemistry Department. The RR associated with the GSTM1 null genotype was 0.8 (0.5-1.1). For tumours of the right side the RR was 1.2 (0.6-2.5), for those of the left side 1.0 (0.5-2.1) and for recto-sigmoid tumours 0.7 (0.5-1.0). In individuals with poorly differentiated tumours the frequency of the null genotype was 67%, in moderately differentiated tumors 41% and in well differentiated tumours 43%.

In a large multi-centre case-control study in the USA, Slattery et al. (69) compared 1567 cases and 1889 controls randomly selected from medical care program lists, drivers license lists and social security lists and by random digit dialling. The crude RR for men and women combined of colon cancer associated with the GSTM1 null genotype was 0.9 (0.8-1.1). When the analysis was stratified by age (<67, >=67 years) the RRs were not substantially different. When proximal and distal tumours were considered separately, the crude RRs for both genders combined associated with the GSTM1 null genotype were 1.0 (0.8-1.1) and 0.9 (0.8-1.1) respectively.

In the one study of colorectal adenomatous polyps, from the USA (114), 446 cases were matched on sex, age and date of sigmoidoscopy and centre with 488 controls without colorectal adenomas. The RR associated with the GSTM1 null genotype was 0.9 (0.7-1.1) (adjusted for the matching factors). When the analysis was stratified by ethnic group, the RRs for whites was 1.0 (0.7-1.4), for Hispanics (not black) 0.8 (0.4-1.7), for blacks 0.6 (0.3-1.4) and for Asians and Pacific Islanders 0.4 (0.2-1.1). Cases and controls were identified after sigmoidoscopy. Only the left colon is accessible to sigmoidoscopy, so it is possible that some controls harboured tumours in the rest of their colon – the effect of this would be to bias the RRs towards the null.

Lee et al. (23) investigated the frequency of GSTM1 polymorphisms among Chinese subjects resident in Singapore. 300 cases of colorectal carcinoma were compared with 183 patients without history of neoplasms recruited from a Clinical Chemistry Department. The RR associated with the GSTM1 null genotype was 0.8 (0.5-1.1). For tumours of the right side the RR was 1.2 (0.6-2.5), for those of the left side 1.0 (0.5-2.1) and for recto-sigmoid tumours 0.7 (0.5-1.0). In individuals with poorly differentiated tumours the frequency of the null genotype was 67%, in moderately differentiated tumors 41% and in well differentiated tumours 43%.

In a large multi-centre case-control study in the USA, Slattery et al. (69) compared 1567 cases and 1889 controls randomly selected from medical care program lists, drivers license lists and social security lists and by random digit dialling. The crude RR for men and women combined of colon cancer associated with the GSTM1 null genotype was 0.9 (0.8-1.1). When the analysis was stratified by age (<67, >=67 years) the RRs were not substantially different. When proximal and distal tumours were considered separately, the crude RRs for both genders combined associated with the GSTM1 null genotype were 1.0 (0.8-1.1) and 0.9 (0.8-1.1) respectively.

In the one study of colorectal adenomatous polyps, from the USA (114), 446 cases were matched on sex, age and date of sigmoidoscopy and centre with 488 controls without colorectal adenomas. The RR associated with the GSTM1 null genotype was 0.9 (0.7-1.1) (adjusted for the matching factors). When the analysis was stratified by ethnic group, the RRs for whites was 1.0 (0.7-1.4), for Hispanics (not black) 0.8 (0.4-1.7), for blacks 0.6 (0.3-1.4) and for Asians and Pacific Islanders 0.4 (0.2-1.1). Cases and controls were identified after sigmoidoscopy. Only the left colon is accessible to sigmoidoscopy, so it is possible that some controls harboured tumours in the rest of their colon – the effect of this would be to bias the RRs towards the null.

GSTT1 & Colorectal cancer

Six of the studies described above also reported GSTT1 genotype (Table 5: 14, 23, 49, 58, 77, 78). Two assessed exposure and are discussed in section 6 (14, 58).

The results of these studies are inconsistent. In two studies (49, 77) the GSTT1 null genotype was associated with a statistically significant increase in the risk of colorectal cancer, while in the other four, no noteworthy associations were apparent.

Chenevix-Trench et al. (78) reported a RR for colorectal cancer of 0.9 (0.4-1.7) associated with the GSTT1 null genotype when both the unselected and geriatric controls were considered together. However, when the analysis was repeated using the different control groups separately the RRs were 0.7 (0.3-1.4) (unselected controls) and 1.5 (0.6-4.3) (geriatric controls). This reflects the different proportions of individuals carrying the GSTT1 null genotype in each of the control groups (19% in the unselected controls and 9% in the geriatric controls) and illustrates the potential for selection bias to distort associations between chronic diseases and genetic polymorphisms. Using the unselected and geriatric control groups combined, the RRs associated with the GSTT1 null genotype were 0.4 (0-1.5) for proximal tumours and 1.0 (0.5-2.1) for distal tumours. In those cases who were diagnosed before the age of 70, 21% were homozygous for the GSTT1 null genotype; in those diagnosed at age 70 or older, 7% carried this genotype.

Deakin et al. (49) reported a RR of 1.9 (1.3-2.7) associated with null genotype. They found increased RRs for each of the tumour sub-sites reported: right-sided tumours 1.5 (0.8-2.7), left-sided tumours 2.3 (1.3-4.2) and tumours of the rectum 1.9 (1.1-3.2).

Katoh et al. (14) found a RR of 1.2 (0.7-2.0) for colorectal cancer and Butler et al. (77) report a RR of 3.4 (2.1-5.4) associated with the GSTT1 null genotype. Neither of these studies presented RRs in relation to tumour sub-site.

Gertig et al. (58) reported a RR of colorectal cancer associated with the GSTT1 null genotype of 0.8 (0.5-1.2), adjusted for BMI, physical activity and alcohol use. For proximal tumours the adjusted RR was 0.9 (0.5-1.7) and for distal tumours it was 0.6 (0.3-1.2). In men aged under 60, the RR associated with GSTT1 null genotype was 0.5 (0.2-1.0) and in those aged 60 years or older, the RR was 0.9 (0.5-1.7). It is not clear whether the age-stratified RRs were adjusted.

Lee et al. (23) stated that the frequency of the GSTT1 null genotype was similar in both cases and controls, and that tumour histology had no effect on the frequency of the null genotype. However, insufficient information was presented for a RR to be calculated.

Comment on the studies on GSTM1 and GSTT1 and colorectal cancer

It is difficult to assess how far selection and participation biases may account for the inconsistencies in the results. Most studies involved hospital-based case series and most of the control groups were not population-based. This has implications for the generalisability of the study results. The potential problems of selecting controls who do not represent the population from which cases arose is demonstrated by the divergence in RRs obtained for the GSTT1 null genotype in the study by Chenevix-Trench et al. (78) when the different control groups were analysed. Most of the studies were not large – five included fewer than 250 cases. The smaller studies are likely to have limited statistical power, particularly for sub-group analyses. Two of the studies were undertaken in Asian populations – the others were in predominantly white populations. There is little information available in other ethnic groups. It is unclear whether any of the established risk factors for colorectal cancer are associated with GSTM1 or GSTT1 genotype. The studies made little attempt to adjust for potential confounders.

The findings of these studies require confirmation in other populations.

GSTM1 and other cancers

In a recent review, Rebbeck (6) suggests that there is evidence from case-control studies that GSTM1 is involved in the aetiology of both lung and bladder cancer, although not all studies have shown this. While some studies of other cancer sites have shown an association with GSTM1, these findings have not been confirmed.

GSTT1 and other cancers

There have been fewer case-control studies of GSTT1. Statistically significant associations have been reported for astrocytoma, meningioma and myelodysplasia, but these have not been confirmed (6).

Because the GST enzyme family has detoxifying activity it would be expected that, rather than affecting the risk of cancer per se, they would modify risk in relation to exposure to potential carcinogens. The enzymes play a major role in the detoxification of PAHs found in tobacco smoke and in cooked and processed meats. In four studies of colorectal lesions and GSTM1 (14, 58, 69, 114) and two of GSTT1 (14, 58) exposure to tobacco smoke was considered. Meat consumption was considered in relation to GSTM1 in two studies (58, 104), and in relation to GSTT1 in one (58). Consumption of broccoli, the richest source of isothiocyanates, which induce enzymes that detoxify environmental mutagens, was considered in a study of GSTM1 and colorectal adenomas (5). Three studies have considered GST gene-gene interactions and colorectal cancer (23, 58, 69).

The limited statistical power of small studies to detect associations between genotype and disease is particularly important in regard to effect modification. To give adequate statistical power to detect a multiplicative interaction, very large sample sizes (in some circumstances, thousands of cases) may be required (115).

GSTM1 and smoking

Little evidence of interaction between GSTM1 genotype, tobacco exposure and colorectal cancer was found in the three studies (14, 58, 69). However, the one polyp study suggests that GSTM1 genotype may modify the association between smoking and disease (114).

Lin et al. (114) report the effect of cigarette smoking and GSTM1 on adenoma risk. Using the reference group of subjects who were never smokers and were GSTM1 positive, significantly increased adenoma risk were seen in both current smokers who were GSTM1 positive (RR=1.7 (1.0-2.9)) and in current smokers who were GSTM1 null (2.1 (1.1-3.8)). When this analysis was restricted to adenomas >1cm in size, the RRs were 1.3 (0.6-2.9) and 2.5 (1.1-5.5).

Katoh et al. (14) reported that GSTM1 did not influence risk differently in subjects classified by smoking status (smoker or non-smoker) or extent of tobacco exposure (pack-years).

Gertig et al. (58) investigated the joint effect of GSTM1 and cigarette smoking status at entry to the PHS on subsequent risk of colorectal cancer. The RR associated with the GSTM1 null genotype was 1.1 (0.6-2.1) in never smokers; 1.0 (0.6-1.6) in past smokers; and 1.2 (0.3-4.2) in current smokers. There was no significant interaction between pack-years of smoking at baseline and GSTM1 genotype.

In the study of Slattery et al. (69), those who smoked more than a pack per day were at approximately 40% increased risk of colon cancer. No interaction was observed in either men or women between GSTM1 genotype and any of the following categories of tobacco exposure: smoking status, usual number of cigarettes smoked per day, pack-years of cigarettes smoked, age started smoking cigarettes and years since stopped smoking cigarettes.

GSTT1 and smoking

Katoh et al. (14) reported that smoking had no effect on the risk associated with GSTT1 genotypes.

Gertig et al. (58) reported RRs of colorectal cancer associated with the GSTT1 null genotype of 0.8 (0.4-1.8) in those never smokers at the time of enrolment, 0.5 (0.3-1.1) in past smokers, and 1.1 (0.3-4.7) in current smokers. There was no interaction between pack-years of smoking and GSTT1 genotype.

GSTM1 and meat intake

In the study of Gertig et al. (58), men who were homozygous for GSTM1 null and consumed more than one serving of red meat per day were at slightly lower risk compared with men who were not homozygous GSTM1 null and consumed less than 0.5 servings per day (RR=0.8 (0.4-2.0)).

Kampmann et al. (104) reported associations between GSTM1 genotype and various measures of meat consumption in the subjects investigated by Slattery et al. (69). There was no evidence that GSTM1 genotype modified the RRs associated with amount of (a) red meat (b) processed meat or (c) poultry consumed; (d) frequency of fried, broiled, baked or barbecued red meat; (e) preferred "doneness" of red meat; frequency of use of (f) red meat or (g) white meat drippings; or (h) red meat mutagen index. GSTM1 genotype modified risks associated with frequency of fried, broiled, baked or barbecued white meat; white meat mutagen index; and total meat mutagen index. Unexpectedly, the strongest positive associations were observed among those who were GSTM1 positive.

GSTT1 and meat intake

In the PHS, men who were GSTT1 null homozygous and consumed more than one serving of red meat daily had a lower risk compared with men who were GSTT1 non-null and consumed less than 0.5 servings daily (RR=0.4 (0.1-1.4)) (58).

GSTM1 and isothiocyanates

Lin et al. (5) postulated that a cancer preventive effect of broccoli would be stronger in GSTM1 null individuals and investigated this in the subjects investigated earlier by Lin et al. (114). Compared to subjects in the lowest quartile of broccoli intake who were GSTM1 null, those in the highest intake quartile who were null had a RR of 0.36 (0.19-0.68) and those in the highest intake quartile who were GSTM1 positive had a RR of 0.74 (0.40-0.99); this interaction was statistically significant (p=0.01).

GSTM1, GSTT1, and other genes

In the PHS, there was no increased risk of colorectal cancer in men who were homozygous null for both GSTM1 and GSTT1 compared to those who were homozygous positive for both GSTM1 and GSTT1 (58). By contrast, Lee et al. (23) reported that 35% of cases with right sided tumours were GSTM1 null and GSTT1 positive compared to 22% of the control series.

Slattery et al. (69) considered the possibility of an interaction between GSTM1 and N-acetyltransferase 2 (NAT2) genotypes. There was a suggestion that women with the combined NAT2 intermediate/rapid and GSTM1 positive genotypes were at increased risk compared to those NAT2 slow/GSTM1 positive genotypes (unadjusted RR=1.5 (1.11-2.05)). This was restricted to women older than 67 and with proximal tumours. However, the association was weaker and not statistically significant in men (unadjusted RR=1.2 (0.89-1.51)). There was no strong evidence of any interaction between NAT2, GSTM1 and smoking in either men or women.

To classify an individual as GSTM1 null or non-null (or GSTT1 null or non-null), the genotyping procedure detects either the absence or the presence of the GSTM1 (or the GSTT1) gene. Therefore after the gene has been amplified by PCR methods, the product needs only to be visualised. This method can not, however, distinguish between the GSTM1*A and GSTM1*B alleles. To do this, a restriction digest must be undertaken. This cleaves the DNA into fragments of characteristic sizes, and the different combinations of these fragments correspond to specific alleles.

To ensure a PCR reaction occurred, a number of quality control procedures should undertaken. Additional "control" primers should be added. These amplify another region of DNA (one that is thought never to be deleted) to confirm that amplification has worked in null individuals. Alongside the samples being amplified, a positive and negative control should be run. The positive control is a sample of DNA known to contain the gene (ie not null); both the band representing the gene in question and the control band should be visible in order for the genotyping to be validated. The negative control allows a check for contamination to be made; if amplification is seen in this control, the samples run at the same time should not be genotyped. In general, the studies present little information on the proportion of subjects for whom genotype could be determined, or reproducibility of genotyping.

Much of the PCR work on genotyping has used DNA from blood; however work involving DNA from mouthwash samples is now being undertaken (116). This development makes PCR methodology even more appropriate for researchers undertaking molecular epidemiology studies, as it enables subjects to be genotyped without the need for invasive sampling.

GSTM1

In two of the nine studies of GSTM1 and colorectal lesions, no details of the primers used are given (23, 77). Three studies (58, 78, 114) use the same primers to amplify the GSTM1 gene, although they reference different papers for these methods (64, 89, 117). The primer 5’-CTGCCCTACTTGATTGATGGG-3’ anneals to the 5’ region of exon 4 and the primer 5’-CTGGATTGTAGCAGATCATGC-3’ anneals to the 3’ region of exon 5. They amplify a 273 bp product, but use slightly different amplification cycles. Katoh et al. (14) used the method outlined by Bell et al. (67). The primers are 5’-GAACTCCCTGAAAAGCTAAAGC-3’ and 5’-GTTGGGCTCAAATATACGGTGG-3’; and they amplify a 215 bp product. The amplification cycles are undertaken at similar temperatures to Comstock et al. (117) and Brockmöller et al. (89), but the time for each stage of the cycle is considerably shorter, and there are fewer cycles in total.

Deakin et al. (49) used the methods of Warwick et al. (84) and Fryer et al. (118) where three primers (5’-GCTTCACGTGTTATGAAGGTTC-3’, 5’-TTGGGAAGGCGTCCAAGCGC-3’, 5’-TTGGGAAGGCGTCCAAGCAG-3’) are used to amplify DNA in intron 6 and exon 7 and a restriction digest differentiates alleles GSTM1*A and GSTM1*B (118). Slattery et al. (69) use the method outlined by Zhong et al. (54) where three primers (P1 5’-CGCCATCTTGTGCTACATTGCCCG-3’, P2 5’-ATCTTCTCCTCTTCTGTCTC-3’, P3 5’-TTCTGGATTGTAGCAGATCA-3’) are combined in a single PCR. Primers P1 and P3 amplify a 230 bp product specific to GSTM1; primers P1 and P2 anneal to either GSTM1 or GSTM4 and amplify a 157 bp product, thereby acting as the control primers.

In another two studies, explicit mention is made of the use of control primers: Chenevix-Trench et al. (78) used primers for exon 1 of coagulation factor XIII and Katoh et al. (14) used primers for beta-globin. In the methodology of Warwick et al (84) – used by Deakin et al. (49) beta-globin is again used as the control primer. In two studies (14, 58) the use of positive and negative controls samples is reported.

GSTT1

In two of the six studies (23, 77) no details on the methods used are given. In the other four (14, 49, 58, 78) the genotyping methods outlined by Pemble et al. (91) were employed. The primers used for amplification in this method are TTCCTTACTGGTCCTCACATCTC and TCACCGGATCATGGCCAGCA. In two studies the use of control primers is described: Chenevix-Trench et al. (78) used primers for GSTP1 and Katoh et al. (14) used primers for beta-globin. In two studies (14, 58) the use of positive and negative controls samples is mentioned.

To date, there is insufficient evidence implicating GSTM1 or GSTT1 in the aetiology of colorectal neoplasms to make population testing an issue.

List of References

• Table 1
• Table 2
• Table 3
• Table 4
• Table 5
• Figure 1

Data on Disease Frequency

• IARC- Cancer Mondial (last accessed 5/2007)
• SEER   (last accessed 5/2007)

General Information on Cancer

• Cancer Research UK (last accessed 5/2007)
• American Association of Cancer Research (last accessed 5/2007)
• National Cancer Institute (last accessed 5/2007)
• International Union against Cancer (last accessed 5/2007)

Gene Specific Information

GSTM1

• Human Gene Mutation Database (last accessed 5/2007)
• OMIM (last accessed 5/2007)
• GenAtlas (last accessed 5/2007)
• UniGene (last accessed 5/2007)
• GeneCards (last accessed 5/2007)

GSTT1

• Human Gene Mutation Database (last accessed 5/2007)
• OMIM (last accessed 5/2007)
• GenAtlas (last accessed 5/2007)
• UniGene (last accessed 5/2007)
• GeneCard (last accessed 5/2007)