Identification of Cancer Patients with Lynch Syndrome: Clinically Significant Discordances and Problems in Tissue-Based Mismatch Repair Testing

Defects in the DNA mismatch repair (MMR) system cause replication errors or instability in DNA microsatellites. Germline mutations in MMR genes MLH1, MSH2, MSH6, and PMS2 cause the autosomal dominant hereditary disease Lynch syndrome. More than 90% of tumors from patients with this syndrome show high levels of microsatellite instability (MSI-high; ref. 1). MSI also occurs in 10% to 15% of sporadic tumors, especially colorectal and endometrial carcinomas, due to hypermethylation of the MLH1 promoter (2). Screening for defects in the MMR genes has become important for the identification of patients with Lynch syndrome and has treatment and prognostic significance for patients with sporadic colorectal cancer. Screening for Lynch syndrome can not only effectively prevent subsequent cancers in the screened individual, but it can also prevent cancer development in relatives who also harbor a germline mutation of an MMR gene.

Traditionally, clinical screening methods to help identify families at risk for Lynch syndrome included the use of the Amsterdam and Amsterdam II criteria. Because these criteria have been thought to lack sensitivity, particularly when an extensive family history is unavailable, the revised Bethesda criteria were developed. The Bethesda criteria are more sensitive, focus on the patient medical history, and introduced the category of MSI and tumor morphology into clinical guidelines. However, these criteria do not address patients with endometrial cancer and may fail to identify a large number of mutation carriers that do not meet the clinical criteria (3). To further help identify individuals at risk for Lynch syndrome, several clinical prediction models have been developed to quantify the risk for germline MMR mutations (4–6). These have comparable sensitivities to the Bethesda criteria and can provide quantitative risk assessment (7). However, such prediction models require subsequent genetic testing and are not helpful in identifying patients with sporadic MSI-high colorectal cancers requiring alternative treatment regimens.

Two tissue-based screening methods used to assess tumors for MMR defects include PCR-based MSI testing and immunohistochemistry. In general, a number of large studies have concluded that MSI and immunohistochemistry have comparable sensitivity and specificity in detecting MMR defects (8–12). Because it is generally associated with lower cost, faster turnaround time, and wider availability in smaller pathology laboratories, some advocate that immunohistochemistry could be used as the sole method of screening in certain cancers (9, 13–16).

At our institution, we conduct both PCR-based MSI and immunohistochemistry concurrently. Immunohistochemical–MSI analysis discordances have been acknowledged in the research setting, but the frequency and nature of these problems have not been formally examined in a quality assurance fashion in the clinical laboratory. In surgical pathology practice, correlation reviews (cytology/histology and frozen section final pathology correlations) are commonly used tools for quality assurance (17). This type of quality assurance review has not been previously applied to the tissue-based approach to MMR testing. Therefore, we examined sources of discordances and problems in tissue-based testing to help optimize this approach in our patient population.

After obtaining Institutional Board approval, we obtained the results and clinical records for all cancer patients who had MSI and/or immunohistochemical testing between August 2002, and August 2010, at the University of Texas MD Anderson Cancer Center (Houston, TX; n = 736 patients). For 107 of these patients, only immunohistochemical or MSI analysis had been previously conducted, so these patients are excluded from further consideration. Only those patients with both MSI testing and immunohistochemical results for the same tumor were included (n = 629). Demographic, clinical, and pathologic data were collected. Initial immunohistochemical slides were examined by a group of 14 different gastrointestinal and gynecologic pathologists, whereas initial MSI analyses were evaluated by 7 gastrointestinal pathologists and 5 molecular pathologists. Discordance was defined as a discrepancy between the result of MSI and immunohistochemistry. Problem was defined as a case with indeterminate or questionable immunohistochemical or MSI results. The immunohistochemical slides and the MSI chromatograms for all discordant/problematic cases were rereviewed by 2 pathologists (R.R. Broaddus and A.N. Bartley) and results and patient histories discussed with a genetics counselor (D.S. Saraiya). Any additional testing conducted, including MLH1 methylation, BRAF mutational analysis, and subsequent genetic sequencing and large rearrangement testing for the 4 most common MMR genes, was also documented.

All immunohistochemical and PCR-based molecular analyses were conducted in Clinical Laboratory Improvement Amendments (CLIA)- and College of American Pathology–approved laboratories. Immunohistochemistry was carried out for DNA MMR gene products MLH1 (G168-15, 1:25; BD Biosciences Pharmingen), MSH2 (FE11, 1:100; Calbiochem), MSH6 (44, 1:300; BD Biosciences Pharmingen), and PMS2 (Alb-4, 1:125; BD Biosciences Pharmingen) as previously described (18). Tumors showing loss of nuclear MLH1, MSH2, PMS2, or MSH6 were classified as negative for the protein expression of each respective marker. Stromal and normal tissues within the tumor served as internal positive controls.

MSI analysis was conducted using 6 National Cancer Institute (Bethesda, MD) recommended microsatellites (19) with the addition of analysis for the mutational inactivation of TGF-β receptor type II (TGFBR2). Therefore, 7 microsatellites comprise the MSI panel. A tumor was designated as MSI-high if it showed allelic shift at 3 or more markers and as MSI-low with allelic shift in one or 2 markers. Those without allelic shifts were designated microsatellite (MS)-stable. A methylation-specific PCR MLH1 promoter methylation assay and BRAF V600E mutational testing by pyrosequencing were conducted on a subset of cases (20, 21).

It is controversial whether MSI analysis or immunohistochemistry represents the “gold standard” when considering the tissue testing approach. Therefore, traditional measures of testing efficacy, such as sensitivity and specificity, are likely not appropriate here. Instead, the measures calculated were concordance (percentage of pairs in agreement) and discordance (percentage of pairs in disagreement). Specifically, we calculated concordance and discordance and their exact 95% confidence intervals (CI) for MSI-high tumors, MS-stable tumors, tumors with loss of an MMR protein by immunohistochemistry, and tumors with retained expression of all 4 MMR proteins.

From August 2002 to August 2010, 629 patients had tumors analyzed by both immunohistochemical and MSI analysis. Twenty-six of these patients had more than one tumor analyzed. Therefore, there are a total of 646 tumor analyses included in this study. The majority of tumors were colorectal adenocarcinomas (88%), followed by endometrial carcinomas (7%; Table 1). Table 2 summarizes the distribution of tumors by MSI status. There were 102 (15.8% of the total) MSI-high carcinomas [78/102 colorectal (76%), 17/102 endometrial (17%), and 7/102 other (7%)] and 55 patients had MSI-low carcinomas [49/55 colorectal (89%), 4/55 endometrial (7%), and 2/55 other (4%)]. The remaining 489 tumors were MS-stable.

Table 3 summarizes the concordances and discordances between immunohistochemical and MSI analysis for the MS-stable and MSI-high tumors (n = 591). MSI-low tumors (n = 55) were specifically excluded from these analyses. The nature of MSI-low tumors is controversial and not well understood, so it is not possible to determine whether immunohistochemical and MSI analyses for individual patients are concordant. Overall, for the MS-stable and MSI-high tumors, the agreement between immunohistochemical and MSI analysis was quite good (97.80% concordance, exact 95% CI, 96.27–98.82). The highest level of agreement occurred in MS-stable tumors with intact expression of all 4 MMR proteins. The lowest level of concordance occurred in MSI-high tumors with loss of at least one MMR protein by immunohistochemistry.

Discordances were identified in 13 of 591 (2.2%) tumors with both MSI and immunohistochemical testing. A detailed summary of these discordances is presented in Table 4. Twelve of 102 (11.8%) MSI-high tumors had discordant immunohistochemical tumor profiles, showing intact immunohistochemical expression of MLH1, MSH2, MSH6, and PMS2 proteins. Seven of 12 tumors were from patients who were younger than 50 years at age of diagnosis, and 8 of 12 tumors were from patients who had a first-degree relative with a Lynch syndrome–associated cancer. Following genetic counseling, 9 patients (8 patients with MSI-high tumors but retained immunohistochemical expression of MLH1, MSH2, MSH6, and PMS2 and the one patient with MS-stable endometrial carcinoma with immunohistochemical loss of MLH1 and PMS2) underwent germline testing; 4 of 9 (44.4%) were identified to have MMR gene alterations. One of these alteration positive patients (patient 3) has an MLH1 mutation that is known to be pathogenic. The remaining identified gene alterations are variants of undetermined significance. Patient 7 had an in-frame duplication of exons 3 to 5 of MLH1. This duplication is not predicted to result in protein truncation, but it is possible that this would affect protein function. Of the 489 MS-stable tumors, only one had loss of an MMR protein by immunohistochemistry. This was an endometrial carcinoma with loss of MLH1 and PMS2 by immunohistochemistry and no evidence of MLH1 methylation. No mutations of MLH1 were detected on subsequent sequencing or large rearrangement analysis. The patient declined genetic testing for PMS2 and is now deceased.

Fifty-three of 55 MSI-low tumors had positive immunohistochemical expression of MLH1, MSH2, MSH6, and PMS2. None of these patients had further genetic testing. One 53-year-old patient with an MSI-low rectal adenocarcinoma had loss of MSH6 by immunohistochemistry; subsequently, a deleterious MSH6 mutation was discovered. A second 68-year-old patient with an MSI-low rectal adenocarcinoma had immunohistochemical loss of MSH2 and MSH6, suggestive of MSH2 mutation. This patient declined genetic testing to confirm this, however.

Problems were identified in 10 patients. These problems do not overlap with the discordances presented in Table 4. The problems were grouped as selection error (n = 2), pathologist error in immunohistochemical interpretation (n = 6), or unusual pattern of immunohistochemical expression (n = 2). Selection error involved 2 different patients, each with 2 separate primary colorectal adenocarcinomas. The proximal tumor in both was MS-stable with intact nuclear expression for all 4 MMR proteins by immunohistochemistry. The distal tumor in both patients was subsequently tested and found to be MSI-high with immunohistochemical loss of MSH2 and MSH6. One of these patients agreed to further genetic testing and was found to have an MSH2 mutation.

Six colorectal adenocarcinomas had pathologist difficulty/error in immunohistochemical interpretation. For 2 of these, our central review changed the immunohistochemical result. The first tumor with pathologist error was from a 32-year-old with an MSI-high rectal adenocarcinoma that was initially interpreted as having positive expression for all 4 MMR proteins. Rereview showed that the tumor had diffuse, strong expression for MLH1, MSH2, and PMS2. Most of the tumor cells, however, were entirely negative for MSH6. A few tumor cells had faint nuclear expression for MSH6 that was much weaker than that of adjacent stromal cells and normal mucosa (Fig. 1). This patient was subsequently identified to have an MSH6 mutation. The second tumor with pathologist error was a rectal adenocarcinoma that was MS-stable and initially interpreted as negative for MLH1 by immunohistochemistry. Re-review revealed that the background stromal cells were also negative for MLH1 and that the immunostaining had, therefore, not worked properly. Immunohistochemistry was repeated using a different tumor block and showed convincing MLH1 protein expression in the tumor (Fig. 2). The remaining 4 tumor analyses in this category involved difficulties with MSH6 immunohistochemistry. Two of these were MSI-high carcinomas (one colorectal, one endometrial) in which the tumors were positive for MLH1, MSH2, and PMS2, but the MSH6 immunohistochemistry did not work (internal control benign tissue also did not stain). Both of these were from patients who had a family history of a Lynch syndrome–associated cancer but neither fulfilled Amsterdam criteria. Neither of these patients consented to further genetic testing. The last 2 tumors in this category were MS-stable colorectal adenocarcinomas which had strong, diffuse nuclear expression for MLH1, MSH2, and PMS2, but the MSH6 expression in the tumor was initially interpreted as only focal or weak. However, upon our rereview, the MSH6 immunohistochemistry was interpreted as positive (retained expression), so these patients did not undergo genetic testing.

Two tumors with unusual immunohistochemical staining patterns were identified. One unusual staining pattern in an MS-stable endometrial adenocarcinoma showed strong nuclear staining for MSH2 and MSH6, but the tumor had definite positive foci and definite negative foci for both MLH1 and PMS2. In the negative foci, internal positive control stromal cells retained expression of both MLH1 and PMS2. Methylation of MLH1 was detected in this tumor, so this was interpreted as a sporadic endometrial carcinoma. The second unusual pattern was an MSI-high colorectal adenocarcinoma with strong expression of MLH1 and MSH2 and complete loss of MSH6 and PMS2 (Fig. 3). These tests were subsequently repeated twice with rereview of each set with the same result. This patient did not assent to genetic testing.

Given the intense scrutiny of the immunohistochemical tests summarized earlier, we next reviewed the results of all 646 tumor analyses for problems in DNA amplification in the MSI analysis to provide better balance to the quality assurance review. Overall, only 0.6% of the microsatellites amplified suboptimally. No tumor had more than one failed microsatellite amplification. Thus, accurate determination of MSI status was possible for all 646 tumor analyses. Only 5 microsatellites, or 1% of the total tested, for the tumors in the MSI-high category failed to amplify. This included one each in 5 different tumors. This did not pose a diagnostic problem as the remaining 6 microsatellites did amplify. For the 55 MSI-low tumors, only 6 of 385 microsatellites (2%) tested failed to amplify (one each in 6 different tumors). Even if this failed microsatellite was actually unstable, this would not be sufficient to redesignate the tumor as MSI-high, as instability in at least 3 of 7 microsatellites is required for a tumor to be MSI-high. For the 489 MS-stable tumors, only 13 of 3,423 microsatellites (0.4%) failed to amplify, one each from 13 different tumors.

Because immunohistochemistry is less expensive, more widely available, relatively simple to carry out, and thought to be comparable with MSI analysis, a number of studies have advocated the use of immunohistochemistry alone in the initial evaluation for MMR defects (14–16, 22). Indeed, we confirm that, in general, concordance between immunohistochemical and MSI analysis is quite high. However, 11.8% of the MSI-high carcinomas had intact immunohistochemical expression of MLH1, MSH2, MSH6, and PMS2. A previous population-based examination of 500 unselected colon cancer patients showed a comparable percentage (9.7%) of MSI-high cancers with intact immunohistochemical expression of MMR proteins (23). Importantly, our study suggests that a significant number of possible patients with Lynch syndrome would not be detected if only immunohistochemistry was used as a screening technique. When family history and the molecular diagnostics tests are considered together, only one (#5) of the 12 MSI-high carcinomas with intact MMR immunohistochemistry summarized in Table 4 is clinically thought to be sporadic. This percentage of likely Lynch syndrome cases is relatively high, especially compared with that of other population-based screening studies (23). This higher percentage may be due to the fact that the vast majority of the patients in our study were initially clinically suspected as having Lynch syndrome due to young age of cancer onset or family history of a Lynch syndrome–related cancer. Reflex testing of all patients with colorectal cancer did not begin at our institution until September, 2009. Previously, Thibodeau and colleagues showed that, in a large series of approximately 1,400 patients with colorectal cancer, there are more immunohistochemical–MSI discrepancies when the patient population screened had moderate to high risk for Lynch syndrome (24). Interestingly, in our current series, we found that 7 of 13 of our immunohistochemical–MSI discordances were prior to September 2009, and 6 of 13 were after this date. So, approximately half of the discordances occurred after commencement of universal screening for colorectal cancer.

In the present study, 12 patients with cancer had MSI-high tumors but intact MMR protein expression by immunohistochemistry. Of these 12, 8 agreed to genetic testing. Genetic testing in these patients is problematic, as the immunohistochemical results cannot help to direct the testing. Seven of these 8 were suspicious for Lynch syndrome based on lack of tumor MLH1 methylation, lack of BRAF mutation, young age of cancer onset, and/or presence of Amsterdam criteria/family history of a Lynch syndrome cancer. Missense mutations could potentially lead to full-length, but nonfunctional, MMR proteins, which would help to explain the immunohistochemical–MSI discordances. However, only 1 of these 8 (12.5%) patients had a pathologic MMR mutation detected (Table 4, #3, MLH1 mutation). The remaining 7 patients either had variants of undetermined significance or no germline variation detected. Tumor 7 was from a patient with an MLH1 in-frame duplication of exons 3 to 5; although this duplication is not thought to cause protein truncation, it could possibly affect protein function. In our previous studies examining patients with endometrial cancer younger than 50 years of age (20), 11 patients had MSI-high endometrial cancers with loss of an MMR protein detected by immunohistochemistry. From these 11 patients, we detected 9 pathologic mutations (81.8%). Three more patients with endometrial cancer had MSI-high tumors but intact MMR protein expression by immunohistochemistry; for these 3 patients, no mutations were identified by genetic testing. Collectively, these results suggest that when immunohistochemistry detects loss of expression of an MMR protein, there is a higher likelihood of ultimately detecting a pathologic MMR gene mutation. Conversely, for tumors that are MSI-high but have retained positive expression of MMR proteins by immunohistochemistry, there could be a lower chance of ultimately discovering a pathologic Lynch syndrome mutation. The reasons for this at this time are not clear. Given time and more experience, it is possible that some of these identified variants of undetermined significance become reclassified as pathologic mutations. Or, mutations in less common genes regulating DNA MMR may be responsible for the high levels of MSI. It is also possible that alternative, not yet identified, genetic mechanisms other than MMR are generating tumor MSI.

In addition to the discordances, our quality assurance review identified several different problems in the tissue-based approach, particularly with immunohistochemistry. Several of these problems are readily correctable, particularly the recognition of lack of internal positive control staining and the testing of all tumors in patients with synchronous primaries. Other problems, such as interpretation of weak or focal immunohistochemical staining (Fig. 1), may require considerable experience to accurately interpret (25). Problems in the MSI analysis, chiefly failure of microsatellite DNA to amplify for adequate testing, were also identified in our study. Of the 646 tumors reviewed, no tumor had more than one microsatellite fail PCR amplification. Given that a panel of 7 microsatellites is used, rather than just one or 2, the failure of one microsatellite to amplify still allows for distinguishing MSI-high from MS-stable.

Although our quality assurance review identified more shortcomings in immunohistochemistry-based testing, it should be emphasized that immunohistochemistry plays an important role in the evaluation of patients with MSI-high carcinomas and can help direct future germline testing. The identification of immunohistochemical loss of MSH2, MSH6, or PMS2 in an MSI-high tumor is virtually diagnostic for Lynch syndrome, even when germline testing fails to reveal a mutation. These patients and their relatives should still be offered cancer prevention services at an earlier age.

Considering the number of possible patients with Lynch syndrome who would be missed by screening by immunohistochemistry alone, we cannot recommend immunohistochemistry as the sole tissue-based screening technique. Rather, for population-based screening of patients with colorectal cancer, it might be preferred to use MSI analysis as an initial screen and then carry out immunohistochemistry in MSI-high cases, as recommended by others (12, 25–27). Alternatively, MSI and immunohistochemistry can be initiated concurrently in patients with moderate or high risk of Lynch syndrome (24). Of course, it is imperative that patients have access to genetic counseling and subsequent germline mutation analysis if these tissue-based tests are informative.

The tissue-based testing of patients with endometrial cancer is more problematic. Endometrial cancers and colon cancers from the same patient can exhibit different levels of MSI, with the endometrial cancers often being MSI-low or MS-stable (28). MSH6 mutation carriers are especially prone to developing endometrial cancer, and these tumors are not typically MSI-high (29, 30). Experience with MSI analysis is even more limited with other cancer types associated with Lynch syndrome, such as cancers of the ureter, ovary, and small intestine. Thus, concurrent MSI analysis and immunohistochemistry should be conducted in patients with these extracolonic malignancies.

D.S. Saraiya is currently employed by Myriad Genetic Laboratories, Inc., as a Regional Medical Specialist (Certified Genetic Counselor). However, all content and writing for the manuscript were completed while she was an employee of MD Anderson Cancer Center. No potential conflicts of interest were disclosed by the other authors.

The study was supported by NIH 2P50 CA098258-06 SPORE in Uterine Cancer (R.R. Broaddus) and NIH CA016672 Cancer Center Support Grant.

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.