Strategies for the Identification of Mismatch Repair Gene Mutation Carriers
The early identification of individuals with Lynch syndrome and their at-risk family members is crucial for the implementation of effective cancer prevention strategies. Strategies for identifying carriers include the evaluation of personal and family cancer history, molecular diagnostic testing of tumors, clinical prediction models, and the sina qua non, germline DNA mutational analysis.
Clinical Criteria
Historically, assessment of family cancer histories has been the cornerstone for the diagnosis of Lynch syndrome. Because there are phenotypic variations and an extensive spectrum of associated cancers, review of personal and family cancer history involving first-, second-, and third-degree relatives is often necessary. Clinical criteria have been developed to identify individuals at risk and have been modified because of increasing knowledge related to phenotype and molecular tumor testing (Table 2). The Amsterdam criteria, developed in 1991 to identify families with autosomal dominantly inherited CRC without polyposis, required (1) three or more CRC cases in which 2 of the affected individuals are first-degree relatives of the third, (2) CRC occurring in 2 generations, and (3) one CRC diagnosed before age 50 years. The Amsterdam criteria were instrumental in identifying families with Lynch syndrome, but they have limited sensitivity and specificity because 40% of families with known mutations do not fulfill the Amsterdam criteria, and 50% who do meet these criteria do not have a detectable DNA MMR defect. The latter group, categorized as Familial CRC Type X, can be distinguished from Lynch syndrome because the CRC tumors do not have features of MSI. In addition, the risk for CRC is lower, and there may be no increased risk for extracolonic cancers compared with Lynch syndrome. On the basis of expert consensus, it may be reasonable in Familial CRC Type X to offer CRC screening initiated 5–10 years before the age of earliest CRC diagnosis, with frequency determined by initial findings but no less often than every 5 years. In a recent study of families with CRC and abnormal immunohistochemistry (IHC) or MSI-high tumors (referred to as Lynch-like syndrome) but no germline MMR mutation, lower cancer risks were observed compared with Lynch syndrome families with identified gene mutations. However, cancer risks were higher than in families with sporadic CRC, and families with Lynch-like syndrome may benefit from tailored screening and surveillance strategies. Nevertheless, additional research and consensus recommendations are necessary.
With the advent of molecular tumor testing, the Bethesda guidelines were developed in 1997 and updated in 2004 to select those patients with CRC who should undergo MSI testing. The guidelines improved on the Amsterdam criteria's sensitivity by including the expanded spectrum of Lynch syndrome tumors and incorporated the MSI tumor testing results in the identification of carriers. Despite this, up to 28% of MMR gene mutation carriers can be missed when limiting genetic testing to those who meet the Bethesda guidelines.
Reliance solely on the existing clinical criteria presents several problems. Studies suggest health care providers do not systematically apply the complex, cumbersome clinical criteria in their routine evaluation of CRC patients, and lack of time needed to take a comprehensive family history is also a barrier to identifying patients who may benefit from genetic evaluation. As a result, many at-risk individuals and family members go without appropriate cancer preventive care.
Molecular Tumor Testing
The use of molecular testing of CRC tumors in the risk assessment paradigm to identify MMR gene mutation carriers has allowed us to move beyond family history. Routine evaluation of MSI or loss of expression of MLH1, MSH2, MSH6, and/or PMS2 MMR proteins by IHC in all newly diagnosed CRC cases is advocated (Table 2).
MMR-deficient phenotypes are found in more than 90% of colorectal tumors in patients with germline mutations. MSI is assessed by using a panel of 5–10 microsatellite markers where tumors are MSI-high when 30% or more of the microsatellite sequences are mutated, MSI-low when 1 sequence is mutated, and stable if none is mutated. In addition, the expression of protein products of the MMR genes on IHC staining frequently shows loss of staining for the antibodies to affected proteins in tumors from patients with Lynch syndrome.
Although the original Bethesda guidelines recommended screening CRC cases diagnosed at younger than 50 years for MSI, more than half of Lynch syndrome cases would be missed by limiting tumor MMR testing to this subset of patients. To identify the majority of patients with CRC who may have Lynch syndrome, universal tumor testing, rather than a limited approach to testing only those who meet particular criteria (whether on the basis of personal or family history), has been supported.
The universal approach of testing all nonselected, newly diagnosed CRCs is feasible. The process requires molecular diagnostics laboratories to offer a stepwise series of molecular tests. A common approach is to initially test the tumors for loss of MLH1, MSH2, MSH6, and PMS2 protein expression by IHC testing because it can be more cost-effective in the evaluation of patients with CRC before age 70 years and allows for single-gene germline testing. In those tumors that have loss of MSH2, MSH6, or PMS2 expression, the likelihood of having a germline mutation is extremely high. However, when tumors display loss of MLH1 protein expression, the evaluation is not as straightforward. Most of these cases have inactivation of the MMR system by somatic epigenetic mechanisms rather than germline mutations. Somatic mutations in BRAF V600E, along with hypermethlyation of the 5' untranslated region of the MLH1 gene, silence gene transcription for MLH1, producing an MSI phenotype in the tumor. These epigenetic causes for MMR-deficient tumor phenotypes are frequent in older patients and can account for the 15% of sporadic CRC tumors that show MSI. CRC tumors that display loss of MLH1 expression should also be tested for BRAF mutations and/or MLH1 promoter hypermethylation. The presence of BRAF mutations in tumors of patients with epigenetic alterations in MLH1 can be used to exclude a diagnosis of Lynch syndrome. IHC testing with BRAF analysis can limit unnecessary germline testing and the associated costs for CRC cases that are likely sporadic and not due to MMR deficiency.
Although IHC testing is often helpful, it is not always diagnostic and can provide false-negative results. In some cases, there may be normal staining tumors of truncating-mutation carriers that may be due to the production of an enzymatically inactive but immunologically detectable protein. With missense mutations, there may be a small effect on the protein structure where a stable heterodimer cannot form, and the unstable protein that is present in lower concentrations cannot be recognized properly by the antibody used. This is not uncommon in tumors from MSH6 gene mutation carriers, where up to 28% of tumors may demonstrate normal protein expression on IHC staining. Because of broad variations in penetrance, some studies have emphasized the concomitant use of both IHC and MSI analysis as a pre-selection tool for MSH6 DNA analysis. In addition, studies show a high interobserver variability among pathologists in interpreting results of IHC staining, which may be operator-dependent. As a result, some pathologists suggest that MMR protein immunostaining be restricted to experienced gastrointestinal pathologists in specialized settings to ensure the highest accuracy attainable by molecular tumor testing.
Prediction Models
Prediction models offer an alternative approach to the identification of patients with Lynch syndrome, and the most recently introduced models provide quantifiable estimates of an individual's risk of carrying a MMR gene mutation. These models include MMRPredict, MMRPro, and PREMM1,2,6 (Prediction of Mismatch Repair Gene Mutations in MLH1, MSH2, and MSH6). Validation studies have found that these models outperform the Amsterdam and revised Bethesda criteria in identifying carriers, and it has been suggested that they replace the existing clinical criteria as prescreening tools in the risk assessment process for Lynch syndrome. In addition, the risk estimates provided by the prediction models can be particularly useful when evaluating patients without a cancer diagnosis or for those whose tumors are not available for molecular testing. In such scenarios, if a patient's predicted probability of carrying an MMR gene mutation is ≥5%, direct germline sequencing of MMR genes may be reasonable.
Although the overall purpose of these risk prediction models is similar, they vary in how they were developed and the populations from which they were derived and validated ( Table 2 ). The MMRPredict model was developed by using logistic regression to estimate the overall probability of carrying an MMR gene mutation from CRC patients diagnosed before 55 years of age. The model includes age at CRC diagnosis, presence of multiple CRC diagnoses, gender, location of tumor (proximal vs distal), presence and age(s) of CRC and/or endometrial cancer diagnosis in first-degree relatives, and molecular tumor testing results. It does not include other Lynch syndrome cancers, and its use is limited to the assessment of only individuals with CRC.
The PREMM1,2,6 model was developed by using multivariable polytomous logistic regression and provides an overall estimate in having a MMR gene mutation, as well as individual, gene-specific mutation probabilities for MLH1, MSH2, and MSH6. The model was developed from the largest number of unrelated gene mutation carriers (n = 525) compared with MMRPredict (n = 38) and MMRPro (n = 121, validation cohort). The patient-specific variables include gender, presence and age at CRC diagnosis, endometrial cancer and other Lynch syndrome–associated cancers (including cancers of the ovary, stomach, kidney, ureter, bile duct, small bowel, brain [glioblastoma multiforme], pancreas, or sebaceous gland). Variables related to family cancer history include information of first- and second-degree relatives including the number of relatives with CRC, endometrial cancer, or other Lynch syndrome–associated cancers, and the minimum age at diagnosis of each cancer among relatives. It does not incorporate tumor data in risk prediction.
The MMRPro model was developed by using a Bayesian framework and integrates data on MMR mutation prevalence and penetrance informed by published values to estimate the probability of having any gene mutation, as well as gene-specific mutation probabilities. Data for the patient being evaluated and for each first- and second-degree relative include age at diagnosis of CRC, endometrial cancer, and current age or age at last follow-up as well as molecular tumor testing results. The model does not incorporate information pertaining to multiple CRC diagnoses or other extracolonic tumors and requires information on the complete pedigree, including ages of family members unaffected by cancer. Both the MMRPredict and PREMM1,2,6 scores can be generated by using online calculators (http://hnpccpredict.hgu.mrc.ac.uk/ and http://premm.dfci.harvard.edu), and software included in the CancerGene package can be used to calculate MMRPro risk scores (http://astor.som.jhmi.edu/BayesMendel).
The setting in which patients are evaluated for Lynch syndrome may very well determine which prediction model is most appropriate to use. The MMRPro model accounts for the size of the family and the impact of unaffected relatives on risk prediction that requires data entry from the entire pedigree. Its use is likely more feasible in a specialized, high-risk cancer clinic with genetic counselors who can ensure that a complete pedigree is obtained with information on unaffected family members as well as molecular tumor testing results. The PREMM1,2,6 model includes a spectrum of extracolonic cancers associated with Lynch syndrome and evaluates cancer burden among first- and second-degree relatives. In validation studies comparing the models, PREMM1,2,6 has been found easy to use and less time-consuming than MMRPro, which may make it suitable for use in diverse clinical settings where patients are identified for further genetic evaluation or direct predictive testing in certain cases. Last, MMRPredict is best for evaluating young patients with CRC but is likely less accurate in the assessment of families with multiple extracolonic cancers, older individuals with CRC, or patients without a personal history of CRC. Additional collaborative studies are necessary to compare the models' performance in identifying gene mutation carriers among both clinic and population-based cohorts and in diverse patient populations to better promote their systematic use in the routine risk assessment of patients with CRC in clinical practice.
Prediction models for Lynch syndrome may also have applications for CRC risk stratification for patients unaffected by cancer. In a large modeling study of a simulated, population-based cohort of healthy individuals followed through life, the PREMM1,2,6 model was used to assess individuals for Lynch syndrome, taking into account the health risks related to the condition and its management. The study reported that it is cost-effective to screen unaffected individuals between the ages of 25 and 35 years for Lynch syndrome by using the PREMM1,2,6 model and to proceed directly to clinical genetic testing in those subjects with a 5% or higher risk estimate. This approach reduced colorectal and endometrial cancer incidence in mutation carriers by approximately 12.4% and 8.8%, respectively.
DNA Mutational Analyses
Available clinical genetic testing for Lynch syndrome includes full gene sequencing of MLH1, MSH2, MSH6, and PMS2, testing for large rearrangements in these genes, and additional testing for deletions in EPCAM. Although germline genetic testing may be considered the gold standard in identifying MMR gene mutation carriers, its use in screening all CRC patients for Lynch syndrome is not feasible because genetic testing is currently too expensive, time-consuming, and difficult to perform in such a setting. Therefore, family history, molecular tumor testing, and/or risk prediction models remain the primary strategies for the clinical identification of individuals at risk for Lynch syndrome. With advances in DNA sequencing techniques and reductions in the cost of genetic testing, the preferred strategies for identifying patients with Lynch syndrome may indeed change in the future.
Current guidelines recommend that genetic testing for cancer predisposition be offered to at-risk patients when the test will influence their own medical management or that of their at-risk family members. However, the interpretation of genetic tests may sometimes be complicated, and referral to dedicated professionals with expertise in cancer genetics is advised when possible, because results often have important implications on a patient's clinical management for cancer prevention.
The presence of a pathogenic germline mutation identified in 1 of the 4 MMR genes confirms the diagnosis of Lynch syndrome. Once a pathogenic mutation is detected in a family, other at-risk relatives can be tested for the same mutation through single-site, gene-specific DNA mutation analysis; if the mutation is not found, this result can be interpreted as "true negative," and these individuals are considered average-risk and can be managed as such. When genetic testing fails to detect a pathogenic mutation in a family deemed high-risk or reveals a mutation of uncertain pathogenic significance, this result is "uninformative." In such cases, individualized recommendations for specialized cancer screening and surveillance are made on the basis of personal and family cancer history, molecular tumor testing results, and/or risk estimates determined through the prediction models.
A Combined Approach for the Identification of Gene Mutation Carriers
A combination of strategies may likely be the best approach to effectively and efficiently distinguish mutation carriers from noncarriers and thereby limit the shortcomings inherent to each individual strategy. Molecular diagnostic testing of CRC tumors is essential in the evaluation for Lynch syndrome, but family cancer risk assessment should not be overlooked, particularly when genetic testing yields results that are clinically uninformative.
In a recent international study of 10,206 population-based CRC cases, the most sensitive strategy in identifying patients with Lynch syndrome was universal MMR tumor testing regardless of CRC age of diagnosis. However, this approach yielded the highest rates of false-positive results. Selective performance of MMR tumor testing in CRC probands diagnosed at 70 years or younger and in patients older than 70 years who met at least 1 criterion of the revised Bethesda guidelines achieved a similar yield to that of universal strategy. The selective approach incorporating age of CRC diagnosis reduced the number of patients undergoing tumor testing and germline genetic testing by 35% and 29%, respectively. Similarly, IHC plus cancer history assessment by using the PREMM1,2,6 model provided the best combined approach in identifying MMR gene mutation carriers from noncarriers in a study comparing the performance of PREMM1,2,6, MSI, and IHC testing strategies, individually and in different combinations. Higher rates of false-positive results on tumor testing were noted mostly in older patients; with every 10-year increase in the age of CRC diagnosis above 50 years, the performance of IHC and MSI testing decreased, with a considerable decrease in the specificity of IHC testing for loss of MLH1 protein expression in colorectal tumors. In contrast, the performance of PREMM1,2,6 increased with every 10-year increase in age of CRC diagnosis compared with both IHC and/or MSI and improved predictions most when combined with IHC results. In light of these results, a recent approach that incorporates age of CRC diagnosis, molecular tumor results, and cancer history assessment by using PREMM1,2,6, has been proposed in the evaluation of patients with CRC for Lynch syndrome (Figure 1). The algorithm also highlights the importance of identifying individuals without detected MMR gene mutations despite significant personal and family cancer histories (and high prediction scores) or abnormal tumor testing results. Clinical management options still need to be individualized in these patients, and families and additional testing in the future may be considered as novel genes related to familial CRC are discovered.
(Enlarge Image)
Figure 1.
MSS, microsatellite stable. §PREMM1,2,6 score can be calculated at http://premm.dfci.harvard.edu/. ¶Other models (MMRpro, MMRpredict) may also be used with their own specified cut-off scores. *BRAF testing: (+), mutation present; (−), mutation absent/wild type. **Surveillance recommendations based on personal and family history. ‡Gene-specific germline mutational analysis.