Phenotype Comparison of MLH1 and MSH2 Mutation Carriers in a Cohort of 1,914 Individuals Undergoing Clinical Genetic Testing in the United States

Lynch syndrome, also known as hereditary nonpolyposis colorectal cancer (HNPCC), is the most common inherited colorectal cancer syndrome. It is estimated to account for up to 5% of all colorectal cancers and is caused predominantly by a mutation in one of four DNA mismatch repair genes: MSH2, MLH1, MSH6, and PMS2. Germ-line alterations in the associated cancer susceptibility genes seem to confer a 60% to 80% lifetime risk of developing colorectal cancer in the absence of medical intervention (1-4).

Mismatch repair gene mutations are also associated with a significantly increased risk for certain types of extracolonic malignancies. Lifetime endometrial cancer risk is estimated to be 40% to 60%, and risks for ovarian cancer and tumors of the urinary bladder and renal collecting system are believed to range from 10% to 20% (5-7). Other gastrointestinal tumors such as stomach, hepatobiliary, and pancreatic carcinomas, although overrepresented in Lynch syndrome compared with their prevalence in the general population, seem to occur less commonly. The possibility that cancer risks may vary depending on the type of mismatch repair gene mutation may have significant implications on cancer screening recommendations.

Despite the major advances made in molecular genetics, there remains a limited awareness of the syndrome in the general medical community. Deficiencies in family information and incomplete gene penetrance have contributed to the underrecognition of affected patients and families. Clinicians are challenged by the heterogeneity that exists when assessing the Lynch syndrome phenotype (8). Multiple predictive models have been recently developed to aid in the recognition of mutation carriers (9-11). However, whether there is a difference in genotype-phenotype expression between carriers of the two most common mismatch repair gene mutations, MLH1 and MSH2, has not been resolved. The existing data are limited and derived predominantly from European family registries that invariably are subject to ascertainment bias.

In the present study, we examine the genotype-phenotype differences in Lynch syndrome mutation carriers in a large United States population undergoing clinical genetic testing for mutations in the MLH1 and MSH2 genes. Because different mutation types (point mutations versus large genomic rearrangements) may have different functional consequences on MLH1 and MSH2 genes, correlations between mutation type and phenotypic expression were also explored.

A total of 2,276 unrelated probands submitted blood samples for full gene sequencing of MLH1 and MSH2 to Myriad Genetic Laboratories, Inc. (Salt Lake City, UT), starting in 2000. Information on the proband demographic profile, including age, gender, and ethnicity, as well as personal and family cancer history was obtained from the test order form that was completed by the health care professional ordering the genetic testing. Because of missing information, 362 probands were excluded. Data from 1,914 probands were available for analysis (9).

The total number of relatives per proband included those who were (a) first-degree relatives (FDR) or second-degree relatives (SDR) of the probands; (b) affected with the following Lynch syndrome–related cancers: colorectum, endometrium, stomach, small intestine, pancreas, bile ducts, ovaries, urinary tract (kidney, ureter, bladder), brain (glioblastoma multiforme), or sebaceous glands; and (c) on the affected side of the family. Overall, there were a total of 4,361 relatives with Lynch syndrome–related cancers in the final study sample.

Data collection and analysis occurred independently. Collection of clinical information and molecular analyses occurred at Myriad Genetic Laboratories and an anonymized, electronic data set was provided to investigators at Dana-Farber/Harvard Cancer Center for further statistical analyses. This study was reviewed and approved by the Dana-Farber/Harvard Cancer Center Institutional Review Board.

Mutation analysis was done using methods previously described (9). Full gene sequencing of MLH1 and MSH2 was done on 898 unrelated probands that submitted blood samples to Myriad Genetics starting in 2000. Starting in August 2004, full gene sequencing and large genomic rearrangement analysis of MLH1 and MSH2 genes were done on 1,016 additional blood samples from unrelated probands. Southern blot analysis was used to search for large genomic alterations in the MLH1 and MSH2 genes. Methylation analyses to evaluate the presence of MLH1 or MSH2 germ-line epimutations were not done on the samples provided by the subjects in this study.

Individuals with deleterious mutations or “suspected deleterious” mutations were defined as mutation positive. Mutations leading to truncating or unstable proteins were considered deleterious and included frameshift, nonsense, and splice site mutations, as well as large deletions and rearrangements. All exon deletions and nonsense and frameshift mutations occurring at or before amino acids 733 and 888 of MLH1 and MSH2, respectively, were defined as mutation-positive in this study. “Suspected deleterious” mutations were those genetic variants for which the available evidence indicated a likelihood (but no proof) that the mutation was deleterious. Examples include splice site mutations that occur at the conserved locations of splice acceptors and splice donors. In this study, missense mutations and noncoding intervening sequence mutations were defined as deleterious based on data derived from linkage analysis of high-risk families, functional assays, biochemical evidence, and/or demonstration of abnormal mRNA transcript processing. Mutation-negative probands were individuals with identified missense mutations within intronic regions whose clinical significance is not yet established, as well as those with polymorphisms and unclassified variants.

Statistical analysis was done using SAS statistical software (version 9.1; SAS Institute, Inc.). The variables related to the proband included presence and age(s) of colorectal cancer (none, one, two, or more), endometrial cancer, and/or other Lynch syndrome–related cancer [stomach, small intestine, pancreas, bile ducts, ovaries, urinary tract (kidney, ureter, bladder), brain (glioblastoma multiforme), or sebaceous glands]. Age was treated as a continuous variable and thereafter dichotomized to <50 or ≥50 years. The variables related to the proband family members included number of colorectal cancers, endometrial cancers, and/or other Lynch syndrome–related cancers and corresponding age(s) of diagnosis, as well as relationship to proband (FDR or SDR). To incorporate the effect of genetic distance between FDR and SDR, a weighted sum of cancer diagnoses among relatives was used. Cancer diagnoses in SDR were weighted half that in FDR (12). In probands and family members diagnosed as having the same cancer more than once, the age at diagnosis was defined as the youngest age.

Univariate analyses were used to assess the relationship between mismatch repair genes mutation status or mutation type and potential predictive variables related to the carriers' personal and family history of cancer. Categorical outcome data were reported as frequencies. Comparisons between mismatch repair mutation status groups (MLH1 versus MSH2) and mutation types (point mutations versus large rearrangements) were assessed using χ² tests. Continuous data were reported in mean values and compared between groups using Student's t test. Comparisons are reported as odds ratios (OR) with 95% confidence intervals (95% CI). A two-sided P value of <0.05 was considered statistically significant.

Multivariable logistic regression analysis was used to further assess the associations of clinical features with mutation status and type. The multivariable analysis of type of mutation controlled for the presence of a mutation in the MLH1 or MSH2 gene. To account for all Lynch syndrome–related cancers, it was determined a priori that the presence of endometrial cancer and any other Lynch syndrome–related cancer(s) would be included in the model irrespective of statistical significance in the univariate analysis.

Two hundred eighty-five (15%) patients had detectable deleterious mismatch repair mutations in the MLH1 or MSH2 genes (112 MLH1 and 173 MSH2). The clinical characteristics of the identified mutation carriers appear in Table 1. Additional demographic data pertaining to ancestry and country of origin have been previously reported (9). Ninety-eight percent of the samples and data provided were by individuals residing in the United States. Only 63% (179 of 285) of all mutation carriers fulfilled the Modified Amsterdam Criteria. Conversely, the mismatch repair mutation detection rate was 34% (179 of 539) among all subjects fulfilling the Modified Amsterdam Criteria in the entire study population. MSH2 mutation carriers were more likely to have family histories fulfilling the Amsterdam criteria than MLH1 carriers (OR, 1.7; 95% CI, 1.04-2.8). Mutation carriers reported a total of 936 family members with colorectal cancer or other Lynch syndrome–related cancers.

Colorectal Cancer. The prevalence of colorectal cancer was higher in MLH1 than MSH2 mutation carriers (79% versus 69%, P = 0.08), but the percentage of mutation carriers with two or more colorectal cancer diagnoses was similar (14% versus 12% for MLH1 and MSH2 carriers, respectively; Table 1). Among MLH1 carriers, 55 of 112 (50%) reported colorectal cancer as their only cancer compared with 74 of 173 (43%) of MSH2 carriers (P = 0.3). MLH1 carriers were significantly more likely to be diagnosed with colorectal cancer at age less than 50 years compared with MSH2 carriers (P = 0.01). The mean age of colorectal cancer diagnosis in the entire sample was 42.2 years for MLH1 carriers compared with 44.8 years for MSH2 carriers (Table 2) where male MLH1 carriers had a significantly younger mean age at colorectal cancer diagnosis compared with MSH2 carriers (38 versus 44 years; P < 0.01). This age differential was not appreciated among female carriers.

The mean number of colorectal cancer tumors per family was significantly higher in families with MLH1 than MSH2 mutations (1.8 versus 1.5, respectively; P = 0.04, Table 1). In addition, relatives (FDR and SDR) of MLH1 carriers reported earlier age at colorectal cancer diagnosis compared with relatives of MSH2 carriers (mean age at colorectal cancer diagnosis: 40.5 versus 44.6 years; P = 0.01; Table 2).

Endometrial Cancer. Endometrial cancer was reported in 41% of all mutation carriers (68 of 167) and its prevalence did not statistically differ among females with MLH1 or MSH2 mutations (36% versus 44%, respectively). Of all females undergoing predictive genetic testing, 9% (119 of 1,308) reported endometrial cancer as their only cancer diagnosis; 16% (19 of 119) of these women were found to be MSH2 mutation carriers and 6% (7 of 119) were MLH1 mutation carriers (P = 0.2). The mean age at the time of endometrial cancer diagnosis was also similar for both MLH1 and MSH2 carriers, 45.8 versus 45.5 years, respectively (Table 2). There was no difference in the mean number of endometrial cancers or the mean age of cancer diagnosis among family members.

Other Lynch syndrome–associated cancers. Eighteen percent (51 of 285) of mutation carriers reported other Lynch syndrome–associated cancers. The prevalence of such cancers was significantly higher among MSH2 carriers compared with MLH1 carriers: 24% versus 9%, respectively (OR, 3.2; 95% CI: 1.5-6.6; Tables 1 and 3).

The proportion of carriers with history of sebaceous skin tumors differed significantly between MLH1 and MSH2 mutation carriers, 2% versus 8% (P = 0.05). Eighty-seven percent (13 of 15) of mutation carriers with sebaceous skin tumors had a MSH2 mutation. Most of the urinary tract cancers (11 of 15) and ovarian cancers (11 of 14) were found in subjects with MSH2 mutations. All other cancer types associated with Lynch syndrome were uncommon (Table 3). The mean number of other Lynch syndrome–related tumors per family (FDR + SDR) was lower in MLH1 compared with MSH2 carriers (P < 0.001; Table 1).

Type of Deleterious Mutation: Large Genomic Rearrangements versus Point Mutations. A total of 42 patients had large genomic rearrangements among 1,016 probands who underwent both full gene sequencing and large rearrangement analysis, with 7 in the MLH1 and 35 in the MSH2 genes. The age at diagnosis of colorectal cancer was the most striking difference in phenotype among carriers (48.3 years in those with large rearrangements versus 43.0 years for those with point mutations, P = 0.001; Table 5). Individuals with point mutations were more likely to have colorectal cancer diagnosed at age less than 50 years (Table 4).

The mean number of colorectal and endometrial tumors among FDRs and SDRs was similar between the two mutation types. However, family members of mutation carriers with large rearrangement mutations had a slightly higher mean number of other Lynch syndrome–related cancers (not statistically significant; Table 4). Overall, the mean youngest age of any Lynch syndrome cancer (including colorectal cancer) among FDRs and SDRs did not differ between carriers with point mutations or large rearrangements (Table 5).

Multivariable logistic regression analysis identified three clinical features predictive of MLH1 versus MSH2 mutation: (a) MLH1 mutation carriers had a slightly higher likelihood of having more colorectal cancer diagnosed among family members than MSH2 carriers (OR, 1.3; 95% CI, 1.0-1.7); (b) MSH2 mutation carriers were more likely to have a personal history of nonendometrial extracolonic cancers compared with MLH1 mutation carriers (OR, 3.2; 95% CI, 1.4-7.3); and (c) MSH2 mutation carriers were more likely to have an increased number of other cancers diagnosed among relatives than MLH1 mutation carriers (OR, 2.1; 95% CI, 1.2-3.7). Personal history of endometrial cancer and colorectal cancer diagnosed at a young age were not predictive of which mismatch repair gene was mutated (Table 6). No significant interactions were found between the independent predictors and age or gender. There were also no clinical variables predictive of the type of mutation (Table 6).

Our study provides data on genotype-phenotype relationships associated with germ-line mutations in the mismatch repair genes MLH1 and MSH2 from a large, diverse population of mutation carriers from the United States. The results shed light on important similarities and differences between other studies of mismatch repair gene mutation carriers, which have been conducted primarily in European registry–based populations (Table 7). Because of the large number of mutation carriers, we were also able to evaluate whether the type of mutation affected phenotypic expression and found results contrary to our expectations.

Compared with MSH2 carriers, MLH1 mutation carriers were more likely to have colorectal cancer as the only type of cancer diagnosed, an increased number of colorectal cancer among relatives, and a younger age of cancer diagnosis. The younger age at diagnosis of colorectal cancer was more notable among MLH1 mutation–positive male carriers than in females. Contrary to prior findings in other studies, we found no appreciable difference in endometrial cancer prevalence or age of diagnosis between MLH1 and MSH2 mutation carriers. Similar to our previously reported experience (13) and studies from European registry–based populations, other Lynch syndrome–associated cancers were more prevalent among both MSH2 mutation carriers and their relatives. Overall, we found that a personal history or an increased number of extracolonic cancers other than endometrial among family members were clinical features associated with the presence of a MSH2 gene mutation.

We evaluated the effect of the type of gene alteration on phenotype, expecting that large rearrangements and deletions would lead to a more severe phenotype with more multiple cancers and earlier ages of onset (14). The results unexpectedly showed that individuals who had a rearrangement had older ages at cancer diagnosis but age was not found to be a significant independent predictor of mutation type. Overall, there were few differences between the cancer histories of individuals with point mutations versus large rearrangements, which suggest that the specific mismatch repair gene involved is more important than the type of mutation in determining phenotype.

Our results illustrate some important similarities and differences among reports of genotype-phenotype associations in Lynch syndrome. Reviewing the data from relatively large cohorts of Dutch, Finnish, French, and German MSH2 and MLH1 mutation carriers (7, 15-18), most studies have reported that extracolonic nonendometrial cancers are more prevalent in MSH2 carriers than in MLH1. Although the small numbers of these cancers in this study limited statistical power, ovarian and upper urinary tract cancers were slightly more prevalent among MSH2 mutation carriers, corroborating previous reports by Vasen et al. (7). Our finding that sebaceous skin tumors were significantly associated with MSH2 mutations also supports data from Mangold et al. (19) regarding the increased prevalence of MSH2 mutations among individuals with the Muir-Torre variant of Lynch syndrome.

The endometrial cancer prevalence results have shown more variability. A majority of prior studies, some of which did not distinguish endometrial cancer from other extracolonic cancers, report an increased prevalence of endometrial cancer in MSH2 mutation carriers. Most recently, this association has not been supported (18). Therefore, in combination with our results, it is becoming apparent that endometrial cancer rates and ages of onset do not differ between MLH1 and MSH2 mutation carriers.

Finally, the earlier age of onset of colon cancer that is more prevalent in male MLH1 mutation carriers is an interesting finding. The French study suggested this gender difference although it was not statistically significant (17). Only recently have Goecke and the German HNPCC Consortium reported a similar association in their analysis of 988 related subjects in 281 mutation-positive families (18). The biological basis for this gender difference is not clear and warrants further study.

An important strength of our study is that it represents a diverse population of unrelated individuals who were not selected from family registries. It represents the largest number of unrelated mutation carriers studied to date. The majority of subjects did not meet the strict clinical criteria that have previously been used to select persons suitable for mutation analysis. Previous studies have specifically analyzed mismatch repair gene mutations in selected kindreds with familial clustering of colorectal cancer. Therefore, the ascertainment of the families not only leads to an overestimation of colorectal cancer risk but also hampers the ability to define a reliable phenotypic association to particular germ-line mismatch repair gene mutations. Ethnic homogeneity among certain European cohorts may highlight a founder mutation effect leading to an overrepresentation of certain extracolonic cancers. In addition, environmental factors may contribute to the differences seen in the prevalence of extracolonic cancers. For example, the higher incidence of gastric cancer in the general populations of Finland compared with the Dutch population may suggest an environmental contribution to the differences seen in the gastric cancer incidence between the Dutch and the Finnish families (6, 7).

Nevertheless, potential limitations of the current analysis must be acknowledged. This study is not truly “population-based.” The least biased of all possible samples would involve population screening of patients with sporadic cancer, which represents an enormous, expensive undertaking that may provide very low-yield results with respect to the number of mismatch repair gene mutations detected. In this study, subjects were selected to undergo testing based on some personal or family clinical history that triggered a heightened suspicion for Lynch syndrome. This population is typical of one that is referred to cancer genetics clinics for consultation, counseling, and predictive genetic testing, and best represents a mutation frequency more reflective of the general population at risk.

In addition, the cross-sectional prevalence study design does not afford us the ability to determine whether a prognostic advantage exists between the two mismatch repair gene mutation carriers. Analyzing data provided by surviving mutation carriers at one distinct point in time may influence estimates of the overall prevalence of disease and the comparisons made between the two mutation carrier groups. In turn, the conclusions drawn from this cross-sectional prevalence study may be vulnerable to inferential reasoning. Prospective, population-based studies of a large number of mutation carriers would be necessary to completely resolve such a survival bias and such studies are difficult to undertake due to the rarity of Lynch syndrome.

A third limitation to this study is the inability to confirm the cancer diagnoses reported in probands and family members. Although previous evidence shows that family history reported by probands in FDRs is reliable, the same cannot be said for SDRs. Reporting errors are possible in this study. However, because health care professionals were responsible for the clinical data provided, it is less likely that erroneous diagnoses were documented. Despite these shortcomings, the strong predictive effects as reported in our previous work using this study group (9) that have recently been validated in a population-based sample (20) are indicative that the information is likely to be reliable.

In conclusion, the phenotypic expression of disease in individuals and within families with Lynch syndrome is influenced by whether the genetic mutation is present in MLH1 or MSH2. In the United States, a preponderance of colorectal cancer is more often associated with a mutation in the MLH1 gene with an increased number of colorectal tumors seen among relatives of MLH1 mutation carriers. Extracolonic Lynch syndrome–associated tumors, other than endometrial cancer, predominate in MSH2 carriers with a higher tumor burden noted among family members. Testing for a specific mismatch repair gene based on family history features may provide an efficient approach to predictive genetic testing. The clinical phenotypes described in this study could be used to tailor recommendations regarding cancer screening, surveillance, and prevention for patients and families with Lynch syndrome.

Dr. Syngal reports having received lecture honoraria from Myriad Genetics Laboratories, Inc. No other conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Myriad Genetics, Inc., for sharing the data for analysis.