Evidence to support a role for the mismatch repair genes human mutL homolog 1 (hMLH1) and human mutS homolog 2 (hMSH2) in the etiology of colorectal cancer has come from linkage analysis, segregation studies, and molecular biologic analysis. More recently, carriers of potentially pathogenic mutations in the hMLH1/hMSH2 genes have consistently been shown to be at a greatly increased risk of developing colorectal cancer compared with the general population. When considered together, the available evidence shows a strong, consistent, and biologically plausible association between mismatch repair gene mutations and colorectal cancer. The penetrance of mutations in hMLH1/hMSH2 is incomplete and is significantly higher in males (approximately 80%) than in females (approximately 40%). To date, evidence for gene-gene or gene-environment interactions is limited, although preliminary studies have revealed a number of avenues that merit exploration. Population screening for mutation carriers is not currently a feasible option, and mutation analysis remains restricted to either relatives of mutation carriers or colorectal cancer cases selected on the basis of phenotype.

Introduction

The mismatch repair genes human mutL homolog 1 (hMLH1) and human mutS homolog 2 (hMSH2) are integral components of the DNA mismatch repair pathway. So far, over 200 allelic variants have been identified for each gene, and the majority of these have been reported to be pathogenic in terms of colorectal cancer. The primary objectives of this review are to describe what is known about hMLH1 and hMSH2 and their variants in different populations and to examine the evidence implicating these genes as risk factors in the development of colorectal cancer. Relevant Internet sites are listed in Appendix 1.

General Methodology

Search strategy
The MEDLINE (National Library of Medicine), EMBASE (Excerpta Medica), and CANCERLIT (National Cancer Institute) databases were searched for papers published before December 31, 2001, using the keywords hMSH2 and hMLH1. Relevant papers were identified, critically appraised, and entered into a Reference Manager (ISI ResearchSoft, Berkeley, California) database. In addition, PubMed was searched via Reference Manager, by author name, for papers from research groups that had published several times on this subject. Finally, the database thus created was cross-referenced with papers cited in the International Collaborative Group on Hereditary Nonpolyposis Colorectal Cancer database of mutations (1). For the "Gene variants" section, we considered a total of 109 papers, which were identified by the above strategy and fulfilled the following selection criteria: 1) complete mutation analysis had been performed on more than five patients with colorectal cancer and 2) there was sufficient detail on the molecular nature of the genetic alteration. Details on all gene variants described in these published papers are given in the first supplementary table, which is posted on the website of the Human Genome Epidemiology Network, as well as on the Journal's website.

For the "Associations" section, the above strategy led to the identification of eight studies that had conducted an analysis of the risk of developing colorectal cancer among carriers of mismatch repair gene mutations and 77 papers that included results of complete mutation analysis performed on more than five colorectal cancer patients selected on the basis of family history, microsatellite instability (MSI), or age of onset. These studies are summarized in table 2 and the second supplementary table, respectively. Many papers included information relevant to both gene variants and associations.

click to view

Table 2  Findings of risk analysis studies of colorectal cancer.

Classification of gene variants
For the purposes of this review, we classified gene variants into one of four categories. These categories are loosely based on the definitions given below, modified according to clinical observations.

1. Pathogenic mutation—generally frameshifts, nonsense mutations, and splice variants
2. Probable pathogenic mutation—generally nonconservative amino acid changes
3. Probable polymorphism—generally conservative changes, often observed in controls
4. Definite polymorphism—synonymous variant

hMSH2
The hMSH2 gene is located at chromosome 2p21, an area initially identified as an important candidate region for genes involved in hereditary nonpolyposis colorectal cancer (HNPCC) syndrome by genetic linkage analysis within large affected families (2, 3).

The hMSH2 protein product is a component of the DNA mismatch repair pathway, the role of which is well established in bacteria and yeast. hMSH2 can form a heterodimer with one of two other mismatch repair proteins, hMSH6 or hMSH3. This protein complex recognizes and binds any errors that may have occurred during DNA replication, and a larger protein complex is then recruited to excise the incorrect daughter sequence and replace it with the correct sequence, using the parental strand as a template. In Escherichia coli, mutS has been implicated in both short- and long-patch repair systems (4).

hMLH1
The hMLH1 gene is located at chromosome 3p21–23, an area also identified by genetic linkage analysis as an important candidate region within large HNPCC families that are not connected with the chromosome region 2p21–22 (5, 6).

The hMLH1 protein product is another component of the DNA mismatch repair pathway, and it has been shown to form a heterodimer with hMLH3, hPMS2, or hPMS1. The hMLH1 protein has no known enzymatic activity and probably acts as a "molecular matchmaker," in that it recruits other DNA repair proteins to the mismatch repair complex. Again, the bacterial homolog of hMLH1 has been implicated in both short- and long-patch repair (4).

One conclusion generated by early attempts to identify precise genetic alterations in hMLH1 and hMSH2 was that variants in these genes are extremely heterogeneous. All 16 exons of the hMSH2 gene and 19 exons of the hMLH1 gene have been found to contain pathogenic mutations.

At present, there are no standard criteria for classifying variants as pathogenic mutations or polymorphisms, and consequently there is considerable variation in interpretation by different researchers. In general, categorization of alterations is based on the predicted effect on protein, with segregation of the mutation with colorectal cancer in the kindred in question and/or analysis of control subjects for that specific mutation also being considered when possible. However, the functional consequences of many mutations are difficult to predict accurately. It has been suggested that even alterations that do not affect the amino acid sequence could lead to aberrant splicing, and that the position of the mutation may be more significant than the type (7). In vitro functional assays have been developed and applied to the task of determining the pathogenicity of missense mutations (8–10) and may eventually facilitate accurate classification of such changes.

The first supplementary table lists all of the gene variants identified as part of this review, illustrating the extreme range of mutations identified and the fact that the observed spectrum of mutation is not entirely uniform. Figures 1-4 summarize some features of this table. Figures 1 and 2 illustrate the distributions of unique gene variants that have been fully characterized at the molecular level in hMLH1 and hMSH2, respectively, according to their position on the gene. Figures 3 and 4 are designed to show the actual numbers of families in which pathogenic mutations have been identified.

click to view

Figure 1  Distribution of unique gene variants in the mismatch repair gene hMLH1.

click to view

Figure 2  Distribution of unique gene variants in the mismatch repair gene hMSH2.

click to view

Figure 3  Distribution of gene variants in the mismatch repair gene hMLH1 by the number of families affected.

click to view

Figure 4  Distribution of gene variants in the mismatch repair gene hMSH2 by the number of families affected.

It is evident from the above figures that certain specific mutations have been identified in more than one kindred. Indeed, some mutations are found with a relatively high frequency. The most commonly observed mutations are summarized in table 1, which displays all mutations identified in more than four ostensibly independent kindreds.

click to view

Table 1  Commonly observed pathogenic mutations in persons with colorectal cancer

The observed spectrum of gene variants may be largely due to genuine differences in the mutability of specific nucleotides or sequences within the gene, but in some cases variants identified in apparently unrelated kindreds can be traced to a common ancestor. Such "founder effects" have been identified in the Finnish population, where two specific founder mutations in hMLH1 account for the vast majority of families in which mismatch repair gene mutations have been identified (12, 13). Another hMLH1 founder effect is evident in the Danish population (14). The extent to which founder effects are responsible for other frequently detected alterations is not entirely clear from the data currently available, and it is likely that some of the kindreds included in the first supplementary table share a common ancestor. Interestingly, the intervening sequence 5 variant A-T at nucleotide 942+3 has been shown to occur as a founder mutation in Newfoundland (15), but another study found no evidence for a common haplotype in 10 carriers of this variant, of various origins, and concluded that the mutation also arises frequently de novo (16). This example underlines the notion that observations of mutation frequency are the result of both the probability of a mutation at a given nucleotide and the demographic history of the population in question.

Overall, little ethnic or population variation is apparent from the available gene variant data. However, the current biases towards highly penetrant mutations are such that the effect of the identified mutation is likely to transcend any population differences. Clearly, there is a need for accurate and extensive population-based data before any population differences in the spectrum and frequency of mismatch repair gene variants become apparent.

There is no clear evidence to suggest that any specific mismatch repair gene mutation produces a specific phenotype of colorectal cancer, although it has been suggested that some differences exist between the spectrum of extracolonic cancers associated with hMSH2 mutations in comparison with hMLH1 mutations (17, 18).

Colorectal cancer is a major public health problem worldwide, with a current annual incidence approaching 950,000 cases (19). Colorectal cancer is more common in males than in females, and in both sexes the incidence rate increases with advancing age. Incidence rates vary globally and are about four times higher in developed countries than in developing countries (20). While incidence rates do vary according to ethnicity (21), there is compelling evidence that the observed variation between countries is primarily due to the role of environmental factors. This hypothesis is supported by the rising incidence of colorectal cancer in populations undergoing rapid economic development, with associated "westernization" of diet and lifestyle. Further evidence for a strong environmental influence comes from migrant data; despite the relatively low incidence of colorectal cancer in Japan, incidence rates in Hawaiian Japanese are among the highest in the world (22).

Considerable effort and resources have been expended with the aim of elucidating the precise dietary and other variables responsible for the observed environmental influences on colorectal cancer incidence. A report commissioned by the World Cancer Research Fund and the American Institute for Cancer Research concluded that evidence was sufficient to suggest that colorectal cancer risk could be substantially reduced by adhering to a diet high in vegetables and low in meat, together with regular physical activity and avoidance of alcohol (23). Other reviews have reached similar conclusions (24), but some studies have failed to provide evidence to uphold the hypothesis that dietary modification can prevent colorectal cancer. Clinical intervention studies (25, 26) and observational cohort studies (27), as well as studies utilizing animal models (28, 29), have shown no evidence of polyp prevention related to diet. Nonetheless, polyp prevention may not be the best endpoint, so results of further clinical studies with cancer prevention as the endpoint are awaited.

Both epidemiologic evidence and experiments utilizing murine models have suggested that nonsteroidal antiinflammatory drugs have antitumor properties that may prevent colorectal cancer. Sulindac has been shown to inhibit tumor growth in experimental systems and to reduce adenoma counts in humans with familial adenomatous polyposis (30), as has a recent study of the specific cyclooxygenase-2 inhibitor celecoxib (31).

A number of case-control and cohort studies have reported an association between hormone replacement therapy and colorectal cancer, with the majority of these providing evidence in favor of a protective effect (24). Accumulating evidence also implicates obesity as a risk factor for colorectal cancer (32), and a positive association may exist between colorectal cancer and diabetes (33, 34). The weight of evidence also suggests that smoking may be a significant risk factor (35).

Colorectal cancer is a multifactorial condition, and while environmental factors are clearly important in the etiology of the disease, there is a significant input from genetic factors. A recent study of twins provided evidence suggesting that about 35 percent of all colorectal cancer cases have a genetic component (36), and first-degree relatives of colorectal cancer patients are well-recognized to have a 2- to 4-fold increased risk of developing the disease themselves. The genetic factors involved are poorly understood and may include recessive genes, pathogenic mutations of low penetrance, and complex gene-gene and gene-environment interactions.

In addition to the less obvious genetic factors, two autosomally inherited cancer syndromes account for a significant minority of colorectal cancer cases. Familial adenomatous polyposis is a rare syndrome caused by mutations in the adenomatous polyposis coli gene and is characterized by the presence of multiple adenomas. In the HNPCC syndrome, affected kindreds have an unusually high occurrence of colorectal and certain extracolonic cancers, with a relatively early age of onset. HNPCC has traditionally been diagnosed on the basis of family history, and the various criteria used for defining HNPCC are summarized in Appendix 2. For research purposes, the Amsterdam criteria are the most widely used, and by this definition of HNPCC, the syndrome may account for 2–5 percent of all colorectal cancer cases.

It has been established that a large proportion of families diagnosed with HNPCC harbor potentially pathogenic mutations in mismatch repair genes. Of the mutations identified so far, over 90 percent occur in hMLH1 and hMSH2. HNPCC families in which mutations in hMLH1 and hMSH2 are not identified may harbor pathogenic mutations in other mismatch repair genes, such as hMSH6 and hPMS2, or in genes as yet unidentified. Pathogenic mutations in hMLH1 or hMSH2 have also been identified in kindreds that do not meet the traditional criteria for diagnosis of HNPCC. This observation may be due to the inherent misclassification bias involved in diagnosing a condition on the basis of family history alone, particularly in small families.

Evidence implying and supporting a causal role for hMLH1/hMSH2 in colorectal cancer comes from both epidemiologic studies and laboratory-based molecular studies, as summarized in figure 5. Initially, linkage studies revealed that disease expression in a proportion of HNPCC kindreds was linked to either chromosome 2p21 (2, 3, 37) or chromosome 3p21–23 (5, 6, 37, 38).

click to view

Figure 5  Pathways of epidemiologic and biologic research identifying and confirming the causal role of the mismatch repair genes hMLH1 and hMSH2 in colorectal cancer.

The heterogeneity of mutations in mismatch repair genes means that screening for mutations in these genes is a lengthy and complicated process. Consequently, for purely economic, practical, and ethical reasons, mutation analysis has been carried out almost exclusively among colorectal cancer patients, particularly those identified as being at high risk of harboring mutations. Only two studies identified in this review conducted mutation analysis among control subjects. Farrington et al. (45) found that none of 26 Scottish blood donors harbored previously identified mutations, although four variants of unknown significance were found. This was compared with the identification of potentially pathogenic mutations in 14 of 50 colorectal cancer patients diagnosed at less than 30 years of age. Similarly, no pathogenic mutations were reported in an analysis of 73 population controls from Utah (46).

Thus, the practical restrictions on mutation analysis, coupled with the low population prevalence of mismatch repair gene mutations and the fact that such mutations are found only in a minority of colorectal cancer patients, has meant that traditional cohort and case-control study designs have not been feasible. However, despite this lack of conventional epidemiologic evidence, subsequent studies have provided convincing evidence to support the hypothesis that mismatch repair gene mutations cause a subset of colorectal cancer cases.

The most compelling supportive evidence comes from studies which demonstrate that mutation carriers are at greatly increased risk of developing colorectal cancer in comparison with the general population. Such studies are summarized in table 2. Aarnio et al. (47) calculated a standardized incidence ratio of 68 (95 percent confidence interval: 56, 81) for Finnish carriers of hMLH1 or hMSH2 mutations. In the other studies identified in table 2, researchers did not make a formal calculation of the standardized incidence ratio, but approximate estimates utilizing appropriate cancer registry data consistently show that the risk of colorectal cancer in mutation carriers is greatly in excess of the corresponding risk in the general population (see table 2). The relative risk of 8.1 (95 percent confidence interval: 3.5, 15.9) for first-degree relatives of mutation carriers observed by Millar et al. (48) is consistent with a risk that is an order of magnitude greater in mutation carriers than in noncarriers.

The clinical presentation of colorectal cancer among mutation carriers appears to differ from that found among persons with sporadic cases in several respects, an observation that indirectly supports the hypothesis that mutations in mismatch repair genes account for a distinct subset of colorectal cancer cases. The most obvious clinical characteristic associated with colorectal cancer among mismatch repair gene mutation carriers is familial aggregation. Part a of the second supplementary table, which is available on the website of the Human Genome Epidemiology Network  and the Journal's website, provides details on mutation analysis studies conducted among patients selected on the basis of family history. The results of these studies are summarized in tables 3 and 4. The observed prevalence of potentially pathogenic mutations in individuals meeting the Amsterdam criteria is remarkably consistent across different populations (table 4).

click to view

Table 3  Association between the extent of family history of colorectal cancer and the prevalence of mismatch repair gene mutations

click to view

Table 4  Results of mutation analysis in patients fulfilling the Amsterdam criteria* for colorectal cancer, by geographic origin

MSI is evident in 12–15 percent of sporadic colorectal cancer cases, compared with over 90 percent of cases defined, according to the Amsterdam criteria, as being from HNPCC kindreds (49). MSI is currently thought to result from defective mismatch repair, although evidence to support this hypothesis is limited by two factors. Firstly, the vast majority of studies that examine mutations in MSI-positive patients concentrate on HNPCC families, introducing considerable bias. Secondly, few investigators look systematically for mutations in patients with MSI-negative tumors. Interestingly, when this has been done, there have been a few instances in which tumors from patients with identified mutations in hMLH1 or hMSH2 have not exhibited the MSI phenotype (45, 50, 51). Analysis of all published results from one research group showed that, among kindreds with suspected HNPCC, germline mutations could be detected in 16 out of 22 colorectal cancer patients with MSI-positive tumors, as compared with one out of 37 mutations in MSI-negative patients (52). The presence of mutations in MSI-negative cases may reflect mechanisms of tumorigenesis in people with mismatch repair gene mutations that do not require mutation instability. Mutation analysis studies involving patients selected on the basis of MSI are summarized in the second supplementary table, part b.

The association between early age of colorectal cancer onset and hMLH1/hMSH2 gene mutations is often confounded by the fact that the selection criteria have included family history, but a few studies have performed mutation analysis on patients selected solely on the basis of early age of onset. As is illustrated in table 5, these studies demonstrate a trend towards a higher pathogenic mutation detection rate in individuals diagnosed at a relatively young age, an observation that is consistent with the hypothesis that these genes are involved in colorectal cancer tumorigenesis. Details on the studies considered in table 5 can be found in the second supplementary table, part c.

click to view

Table 5  Association between age at onset of colorectal cancer and mismatch repair gene mutations

Penetrance
While it has become widely accepted that mutations in the mismatch repair genes hMLH1 and hMSH2 play a causal role in a subset of colorectal cancer cases, the precise penetrance of these mutations remains unknown. A number of studies, summarized in table 2, have addressed this issue. Results are presented differently for each study, so direct comparison is difficult. One consistent finding is that risk is higher among male mutation carriers (approximately 80 percent by age 70 years) than among females (approximately 40 percent by age 70 years), an observation with important implications for patient management and surveillance. Observed differences in penetrance between carriers of hMLH1or hMSH2 mutations (53, 54) await confirmation in future studies.

A study by Aarnio et al. (47) classified relatives of clinically defined HNPCC cases as being at a 25 percent, 50 percent, or 100 percent risk of being mutation carriers and calculated the cumulative incidence of colorectal cancer up to age 70 years as being 100 percent and 54 percent for males and females, respectively. A potential source of bias in this particular study is the fact that the majority of the probands had one of the Finnish founder mutations. A similar study carried out in Amsterdam Dutch kindreds calculated risk of colorectal cancer among mutation carriers at age 75 years to be 92 percent in males and 83 percent in females (55).

These studies used family history as a selection criterion, an approach that introduces considerable ascertainment bias. Kindreds identified in this way will inherently have an unusually large number of colorectal cancer cases, and estimates of penetrance obtained in this way are likely to be falsely high. Dunlop et al. (56) used an alternative approach to identify mutation-carrying probands from the Scottish population, performing mutation analysis on colorectal cancer patients with a very early age of onset (<30 years). The cumulative incidence of colorectal cancer among relatives proven to be mutation carriers was found to be 74 percent in males and 30 percent in females at age 70 (56).

Note that the identification of families with mismatch repair gene mutations using any phenotypic selection criteria introduces ascertainment bias, and such kindreds may not be representative of all mutation-carrying families in the general population. Thus, there is a considerable need for estimates of penetrance based on systematically collected familial or population data.

Survival
Prior to the identification of mismatch repair genes, several studies suggested that the prognosis for patients with colorectal cancer due to HNPCC was more favorable than that for patients with sporadic colorectal cancer. Whether improved prognosis is specifically a feature of colorectal cancer in patients harboring mismatch repair gene mutations is not yet clear, although preliminary evidence suggests that this may be the case (57, 58).

A possible explanation for this phenomenon may be that the high frequency of mutations characteristic of mismatch repair-deficient tumors actually restricts tumor growth (58). However, kindreds included in survival analysis studies on the basis of a strong family history of colorectal cancer have, by definition, survived to produce a large family group for analysis. Therefore, these kindreds may not be representative of all mutation carriers, and there is a need for survival data from unselected, population-based cohort studies.

It has also been postulated that mismatch repair deficiency may have an effect on response to chemotherapy. Results are not entirely consistent, but several studies suggest an association between hMLH1/hMSH2 deficiency in cell lines and resistance to chemotherapeutic agents (59–62).

While the exact penetrance of specific mutations in hMLH1 and hMSH2 is unknown, it is not complete. Consequently, the age-related risk, pathologic features, and outcomes associated with such mutations are subject to modification by other genetic and environmental factors.

The body of epidemiologic data regarding modification of disease resulting from mismatch repair gene mutations is somewhat limited. The effects of known environmental risk factors for colorectal cancer in mutation carriers are largely unstudied, and much of the suggestive evidence for interactions comes indirectly from studies using MSI-positive or clinically defined HNPCC cases as a surrogate for mutation carriers. Furthermore, the apparent presence of a statistical interaction between mismatch repair gene mutations and other genetic or environmental factors does not necessarily imply the existence of a biologic or causal interaction. Therefore, the studies considered below do not constitute evidence for true interactions involving hMLH1 and hMSH2, although they may prove useful in terms of identifying potential interactions that merit further investigation.

Gene-environment interactions
Reports by Ruschoff et al. (63) and Yamamoto et al. (64) have suggested that treatment of hMLH1- or hMSH2-deficient cell lines with nonsteroidal antiinflammatory drugs leads to a significant reduction in the proportion of cells exhibiting MSI, indicating that this phenotypic manifestation of mismatch repair deficiency may be modified by these drugs.

Slattery et al. (65) have presented evidence suggesting that an interaction may exist between MSI and smoking. Compared with patients with MSI-negative tumors, patients with MSI-positive tumors were more likely to be heavy smokers: Odds ratios were 1.6 (95 percent confidence interval: 1.0, 2.5) in men and 2.2 (95 percent confidence interval: 1.4, 3.5) in women (65). These results are supported by those of another recent study (66), and the implication that smoking is specifically associated with a particular subset of colorectal cancer cases is consistent with the weak associations reported between smoking and sporadic colon cancer. It is possible that mismatch repair deficiency is involved in the observed association between smoking and MSI, but further studies involving known mutation carriers will be required to confirm this hypothesis.

Another recent paper by Slattery et al. (67) showed that the risk of MSI-positive colon cancer may be reduced by estrogens and increased by estrogen withdrawal.

Dietary heterocyclic aromatic amines are another risk factor that requires further evaluation. Wu et al. (66) found that patients with MSI-positive tumors had received a relatively high dietary exposure to heterocyclic aromatic amines, an observation that remained significant after adjustment for smoking and red meat intake. This finding is consistent with laboratory studies, which have shown that rats exposed to particular heterocyclic amines showed the trait of MSI (68).

Gene-gene interactions
Risk of colorectal cancer among female hMLH1/hMSH2 mutation carriers is approximately half the risk in male mutation carriers (47, 56). In the absence of clear evidence of hormonal influence, the presence of a genetic modifier, X-linked or otherwise, remains a possibility.

The possibility of interaction between mismatch repair genes and other genes known to influence colorectal cancer susceptibility is an area that merits consideration. Initial studies have suggested that genes involved in carcinogen metabolism might modify the phenotypic expression of mismatch repair gene mutations. For example, Moisio et al. (69) demonstrated that a specific polymorphism in the gene encoding the xenobiotic enzyme N-acetyltransferase 1 was associated with a lower age of colorectal cancer onset in Finnish HNPCC kindreds with identified mutations in hMLH1. Similarly, an alteration in cyclin D1 has been associated with earlier age of onset in HNPCC cases; patients who harbor the mutant cyclin D1 allele develop cancer an average of 11 years earlier than patients with two wild-type alleles (70).

Murine studies have demonstrated that MSH2 deficiency accelerates intestinal tumorigenesis in transgenic mice that are heterozygous for a germline mutation in the adenomatous polyposis coli gene (71). Similarly, Toft et al. (72) have used mice mutant for both MSH2 and p53 to demonstrate interaction between these genes. Additionally, in-vitro studies have suggested that interactions may exist between mismatch repair genes and transforming growth factor-ß receptor II (73). While these molecular studies demonstrate that gene-gene interactions may be worth further investigation, the above hypotheses have yet to be tested in human populations for relevance to cancer susceptibility.

The heterogeneity of mutation types found in hMLH1 and hMSH2 has meant that many different techniques have been employed to test for mutations in these genes. A number of techniques are described below, along with their benefits and disadvantages.

In vitro synthesized protein assay
The in vitro synthesized protein assay technique uses an in vitro system to transcribe and translate a large polymerase chain reaction product containing several exons. The translated product is separated on a polyacrylamide gel electrophoresis system, and potential mutations are identified as truncated bands. These may represent a number of mutations that have the effect of altering splicing, therefore producing a translated fragment with certain exons deleted. Out-of-frame deletions or insertions, resulting in frameshifts or splice variants, will also be detected using this method. In vitro synthesized protein assay does not detect missense mutations, in-frame deletions or insertions, large genomic deletions involving numerous exons, promoter mutations, or mutations that silence the gene. The assay also requires the use of mRNA for the production of a cDNA polymerase chain reaction product.

Genomic sequencing
cDNA sequencing also relies on mRNA being available. It will identify all mutation types except large genomic deletions, promoter mutations, and gene silencing mutations. Genomic sequencing detects even fewer changes than cDNA, but it does have the advantage of only requiring genomic DNA. Table 6 shows a comparison of the sensitivity of the two techniques, in vitro synthesized protein assay and genomic sequencing, as described by Farrington et al. (45).

click to view

Table 6  Sensitivity of mutation detection techniques

Table 6 summarizes the available information regarding the sensitivity of the above techniques. The use of various combinations of techniques may enhance sensitivity, but this is usually impractical. Recently, Yan et al. (77) demonstrated that the conversion of chromosomes from the diploid state to the haploid state, by fusion to a recipient rodent cell line, may facilitate improved sensitivity of current mutation detection techniques.

The population prevalence of hMLH1/hMSH2 mutation carriers in the Scottish population aged 15–74 years has been estimated at 1 in 3,139 (78). A recent UK National Screening Committee workshop concluded that there is currently no case to offer population screening in an attempt to identify mutation carriers (Rose et al., UK National Screening Committee, unpublished data). Authors in the United States have reached similar conclusions, agreeing that more information regarding the prevalence and penetrance of mismatch repair gene mutations and more evidence of effective intervention strategies are essential prerequisites for implementing screening outside of the research context (79–84).

There are essentially two strategies that could be employed to search for mutations in the context of population screening: searching the entire gene(s) for mutations using the techniques considered above or looking for specific mutations. The latter option is far less expensive and labor-intensive and could be of particular benefit in countries where specific "founder" mutations are prevalent. It may also be possible to apply DNA pooling strategies in this context to enhance efficiency (85). However, this approach is not currently feasible because of the extreme heterogeneity of mismatch repair gene variants and the low allele frequency of individual mutations. The ethical issues inherent in genetic screening, coupled with the poor efficiency and high cost of detecting mutations using current technology, mean that population testing in any form is unlikely to be recommended in the near future.

Another approach to identifying mutation carriers is performing mutation analysis in colorectal cancer patients deemed to be at high risk of harboring mutations, and subsequently performing "cascade screening" of their relatives. The major issue in the context of a cascade screening program is that of how resources can be efficiently targeted towards the identification of kindreds with hMLH1/hMSH2 mutations. This issue is considered in detail in an overview of findings from one research group (52), and the sensitivity and specificity of various clinical criteria are considered by Syngal et al. (86).

Currently, most mutation carriers are identified by referral of patients with a family history of colorectal cancer to cancer genetics services. Another option, under investigation in an ongoing program in Scotland, is to search for mismatch repair gene mutations among persons with early-onset colorectal cancer and subsequently perform cascade screening in the relatives of mutation carriers. The phenotypic features of age at onset, family history, and MSI are commonly used selection criteria in mutation analysis studies, as summarized for reference in the second supplementary table.

At the present time, there is no consensus regarding the most efficient approach to identifying mutation carriers. It is clear, however, that further understanding of the role of mismatch repair genes in colorectal cancer has important scientific and clinical implications.

• Table 1 - Commonly observed pathogenic mutations in persons with colorectal cancer.
• Table 2 - Findings of risk analysis studies of colorectal cancer.
• Table 3 - Association between the extent of family history of colorectal cancer and the prevalence of mismatch repair gene mutations.
• Table 4 - Results of mutation analysis in patients fulfilling the Amsterdam criteria* for colorectal cancer, by geographic origin.
• Table 5 - Association between age at onset of colorectal cancer and mismatch repair gene mutations.
• Table 6 - Sensitivity of mutation detection techniques.
• Supplementary Table 1 - ( 1.7MB) hMLH1/hMSH2 Gene Variants
• Supplementary Table 2 - ( 1.39 MB) Summary of Studies Conducting Analysis of hMLH1/hMSH2 in Colorectal Cancer Cases
• Appendix 1 - Internet Sites
• Appendix 2

Figures

• Figure 1 - Distribution of unique gene variants in the mismatch repair gene hMLH1.
• Figure 2 - Distribution of unique gene variants in the mismatch repair gene hMSH2.
• Figure 3 - Distribution of gene variants in the mismatch repair gene hMLH1 by the number of families affected.
• Figure 4 - Distribution of gene variants in the mismatch repair gene hMSH2 by the number of families affected.
• Figure 5 - Pathways of epidemiologic and biologic research identifying and confirming the causal role of the mismatch repair genes hMLH1 and hMSH2 in colorectal cancer.

List of References

Colorectal Cancer Statistics (last accessed 2/2008)

• International Agency for Research on Cancer
• Surveillance, Epidemiology, and End Results Program

Genetic Information and Database (last accessed 2/2008)

• National Centre for Biotechnology Information
• Online Mendelian Inheritance in Man
• ICG-HNPCC* Database
• Cambridge Public Health Genetics Unit 

Patient Education and Support (last accessed 2/2008)

• World Cancer Research Fund
• Genetic Health
• Medicine Online
• American Cancer Society
• Cancer Research Campaign
• International Union Against Cancer
• UK National Screening Committee

* ICG-HNPCC, International Collaborative Group on Hereditary Nonpolyposis Colorectal Cancer.