Performance of the Revised Bethesda Guidelines for Identification of Colorectal Carcinomas With a High Level of Microsatellite Instability

The molecular pathways of colorectal adenocarcinoma include a subgroup of tumors displaying a high level of microsatellite instability (MSI) and deficient DNA mismatch repair (MMR). Microsatellite instability represents a pattern of mutations that are the result of addition or deletion of bases within nucleotide repeats known as microsatellite regions. Microsatellite loci contain repetitive elements of 1 to 6 nucleotides in length and are most commonly (CA)_n or poly A/T sequences. Microsatellite instability was first reported in colorectal adenocarcinomas of patients with hereditary nonpolyposis colorectal cancer (HNPCC), also previously known as Lynch syndrome, and this status of high-frequency mutagenesis is caused by mutations in the main DNA MMR genes. Hereditary nonpolyposis colorectal cancer is an autosomal-dominant cancer predisposition syndrome, characterized by early onset with average presentation of colorectal cancers (CRCs) by 45 years of age, often multiple or associated with other synchronous or metachronous neoplasms, including endometrial, urothelial, gastric, ovarian, brain, and sebaceous gland tumors.1–3 Hereditary nonpolyposis colorectal cancer previously was divided into Lynch syndrome I, characterized by colon cancer susceptibility, and Lynch syndrome II, in which carcinoma of the endometrium, ovary, and other organs also occur.1 At the molecular level, defects in several DNA MMR proteins have been identified to underlie HNPCC, most often affecting hMLH1 and hMSH2 and less frequently hMSH6, hPMS1, and hPMS2.4–8 

Microsatellite instability is also seen in approximately 15% of sporadic CRCs, usually reflecting loss of expression of hMLH1 associated with gene silencing by hMLH1 promoter methylation.9–11 Importantly, the correlation of loss of one of the main DNA MMR proteins in tumors is characteristically associated with high levels of MSI, affecting multiple loci (MSI-H), which can be detected by using a standard panel of 5 microsatellite markers.12 When at least 2 of the 5 markers show MSI, tumors are classified as MSI-H, and loss of 1 of the MMR proteins, most often hMSH2 or hMLH1, can be detected by immunohistochemical stains. In contrast, tumors with stable microsatellites (MSS) or those that show a low level of MSI (MSI-L) affecting only 1 of the 5 markers, generally have preserved expression of the DNA MMR proteins.13,14 In clinical practice, MSI testing is used as a marker of underlying deficient DNA MMR in tumor tissues, although the MSI test does not distinguish between HNPCC and sporadic CRCs.

Because screening for MSI in all CRC cases would not be cost-effective, clinical criteria, such as age of presentation and relevant family history, have been used in the screening of patients for HNPCC.15,16 Recent guidelines, or criteria, to select CRCs for MSI testing were reported in 2004, as the revised Bethesda guidelines.15,16 Based on these criteria, pathologists can play a major role in identifying MSI-positive CRC and potentially HNPCC. The revised Bethesda guidelines for MSI testing of CRCs are as follows15,16: (1) patient diagnosed with CRC before the age of 50 years; (2) presence of synchronous or metachronous colorectal or other HNPCC-related tumors (stomach, urinary bladder, ureter and renal pelvis, biliary tract, brain [glioblastoma], sebaceous gland adenomas, keratoacanthomas, and small bowel cancer), regardless of age; (3) colorectal cancers diagnosed before the age of 60 years with morphology suggestive of MSI-H (including the presence of tumor-infiltrating lymphocytes, Crohn-like lymphocytic reaction, mucinous or signet-ring cell differentiation, or medullary growth pattern); (4) CRC in a patient with 1 or more first-degree relatives with CRC or other HNPCC-related tumors (one of the cancers must have been diagnosed before the age of 50 years, and adenomas must have been diagnosed before the age of 40 years); and (5) CRC in a patient with 2 or more relatives with CRC or other HNPCC-related tumors, regardless of age. The updated guidelines recommend MSI testing in CRCs of patients younger than 50 years, whereas the previous guidelines recommended testing for CRC in patients younger than 45 years. While in the initial Bethesda guidelines the morphology of the tumors was used only as a guideline in patients younger than 45 years, the updated Bethesda guidelines recommend testing in patients with histopathologic features suggestive of MSI for patients younger than 60 years. There are several potential practical implications of the revised guidelines, including a potentially significant role of pathologists in the identification of cases for MSI testing, with reflex testing based on the identification of cases that fall within the Bethesda guidelines at the time of pathologic examination.

The primary aim of our study was to evaluate the performance of the revised Bethesda guidelines to identify microsatellite-unstable CRCs, primarily using reflex testing requested by pathologists based on the first 3 revised guidelines.

The cases for this study were selected from either biopsy specimens or colonic resections with a diagnosis of colorectal adenocarcinoma. The inclusion criteria were in accordance with the recommended revised guidelines for MSI testing, based on the criteria that are usually available to the pathologist, including (1) CRC in patients younger than 50 years and (2) CRC with histology suggestive of a high level of MSI in patients at least 50 years but less than 60 years, and (3) synchronous or metachronous CRC or HNPCC-related tumors, regardless of age.15 Four appendiceal mucinous adenocarcinomas were included. There was only 1 case with synchronous adenocarcinomas, and this 49-year-old patient also fulfilled the age criteria. The histopathologic features suggesting MSI-H include poorly differentiated carcinomas with morphology of so-called medullary-type carcinoma, mucinous features, presence of signet ring cells, prominent tumor-infiltrating lymphocytes, defined as 4 or more tumor-infiltrating lymphocytes per ×400-power field,17 and a Crohn-like reaction associated with the tumors.15,18–21 Seventy-five patients were studied, including 42 women and 33 men (mean age, 40.3 years; range, 17–57 years). Family history of cancer was retrieved from the University of Pittsburgh tumor registry. The demographic and pathologic data, including tumor location, pathologic TNM staging, and histopathologic features, were reviewed from the pathology reports and were confirmed after review of representative tumor sections. The study was approved by the University of Pittsburgh Institutional Review Board.

Genomic DNA was extracted from microdissected nonneoplastic colon and representative tumor areas from 5-μm-thick sections of formalin-fixed and paraffin-embedded tissues.

Polymerase chain reaction amplifications were performed with the National Cancer Institute–recommended 5-oligonucleotide primer set (BAT25, BAT26, D2S123, D5S346, and D17S250).12 Polymerase chain reaction amplifications were carried out in 12.5-μL reaction volumes containing 1× PCR Gold reaction buffer (Applied Biosystems, Foster City, Calif), 25 pmol of both labeled and unlabeled primer, and 0.25 U of Taq polymerase (AmpliTaq Gold; Applied Biosystems). The polymerase chain reaction cycling conditions included the following cycles: 1 cycle at 95°C for 15 minutes; 40 cycles at 94°C for 30 seconds, 55°C for 30 seconds, and 72°C for 1 minute; and 1 extension cycle at 72°C for 10 minutes.

Hematoxylin-eosin–stained sections of colorectal tumors were examined for the presence of histologic features suggestive of MSI-positive CRC, as described. The primary antibodies used were the anti-human mouse monoclonal hMLH1 (1:25; G168-15, PharMingen, San Diego, Calif) and the anti-human mouse monoclonal hMSH2 antibody (1:100; NA27, Oncogene Research Products, San Diego, Calif). Immunohistochemical stains for hMLH1 and hMSH2 were performed using a standard biotin-avidin method, in a Dako Autostainer (DakoCytomation, Carpinteria, Calif). Immunohistochemical stains for hMLH1 and hMSH2 were performed on all cases showing MSI-H status, as determined by the MSI microsatellite marker panel. The slides were deparaffinized with xylene and rehydrated with graded ethanol series. Antigen retrieval was performed in a pressure cooker for 20 minutes, in Tris buffer, pH 8.5. The tissue sections were then blocked with 3% hydrogen peroxide for 5 minutes in a moist chamber. After rinsing with Dako TBS buffer, the sections were incubated with the primary antibody, hMLH1 or hMSH2, diluted in Dako diluent, and incubated for 20 minutes in a moist chamber. The slides were incubated for 20 minutes with rabbit Envision Plus (DakoCytomation), rinsed again, and developed with Dako DAB+ for 5 minutes. After rinsing in tap water, the slides were placed in a 1% solution of copper sulfate for 5 minutes and then counterstained with Harris hematoxylin for 1 minute. The slides were then rinsed, dehydrated, and mounted.

Analyses were performed using SPSS (Sigma Stat 3.0, SPSS Inc, Chicago, Ill) software for Windows. Two-by-2 contingency table analyses were performed using a 2-tailed Fisher exact test or χ² test. The variations in tumor size, lymph node counts, and metastatic status were analyzed with the Student t test.

The results of the MSI molecular testing are summarized in Table 1. Seventeen (22.6%) of the 75 CRC cases included by the selection criteria exhibited high-level MSI (MSI-H). Thirteen (19.1%) of the 68 tumors in the age group younger than 50 years were labeled MSI-H, whereas 4 (57.1%) of 7 CRCs in the age group between 50 and less than 60 years were MSI-H (Table 1). Only 1 MSI-L tumor was identified; the patient was a 50-year-old woman with a moderately differentiated adenocarcinoma with associated Crohn-like reaction in the sigmoid colon. The remaining tumors were MSS. The MSI-L case was included in the analysis together with the MSS tumors, since both MSS and MSI-L tumors share similar clinicopathologic features (Table 1). The results obtained with the individual microsatellite markers are summarized in Figure 1. The mononucleotide repeat marker BAT26 and the dinucleotide marker D17S250 were the most sensitive, showing MSI in all cases that were MSI-H.

The mean and median ages of patients with MSI-H tumors were 42 and 46 years, respectively; mean and median ages of patients with MSS tumors were 40 and 41.5 years, respectively (P = .49). Among the 17 MSI-H cases, 8 occurred in women. Of the 68 patients in the age group younger than 50 years, 58 were white, 6 were African American, 1 was Asian, and in 3 cases no information regarding race was available. Of the 7 patients in the age group between 50 and less than 60 years, 5 were white, 1 was African American, and 1 was Asian, reflecting the distribution of the patient population at the University of Pittsburgh Medical Center. One MSI-H tumor occurred in an African American patient, 1 was from an Asian patient, and the remaining tumors were from white patients.

Information regarding tumor size and location in the colon was available in 68 cases (Table 2). Tumors classified as MSI-H were more often proximal, occurring in the cecum, ascending colon, and transverse colon, whereas MSS/MSI-L tumors were frequently distal (Table 2). Interestingly, 4 mucinous appendiceal adenocarcinomas were included in the cases and were all MSS (Table 2). Furthermore, MSI-H tumors were larger than MSS/MSI-L cases (P = .03) (Table 2). Information regarding the total number of identified lymph nodes and the number of positive lymph nodes was available in 57 cases (Table 3). While the total number of identified lymph nodes was higher in MSI-H tumors, the number of positive nodes was on average lower in MSI-H tumors, as compared to MSS/MSI-L tumors (P = .04 and P = .01, respectively) (Table 3). More MSS/MSI-L tumors presented at earlier stages, with 9 of 43 tumors showing no invasion into the muscularis propria (Tis and T1), whereas all 14 MSI-H tumors for which a resection specimen was available were in more advanced stages (T2 and T3) (P = .06).

The histopathologic features were examined in 72 cases in which tissue sections were available for review (Table 4). The histopathologic features were evaluated on routine hematoxylin-eosin stains, as exemplified in Figure 2. The data show a significant association of histologic features, which include prominent tumor lymphocytes, mucinous features, and a Crohn-like reaction with MSI-H status (P < .001, P = .003, and P = .003, respectively), whereas medullary morphology and the presence of signet ring cells did not reach statistical significance (Table 4). Overall, histologic features suggestive of MSI-H phenotype were present in 80% of MSI-H and in 35.1% of MSS/MSI-L tumors (P = .005), and in 8 (66.7%) of 12 MSI-H versus 17 (31.5%) of 54 MSS/MSI-L tumors in patients younger than 50 years (P = .045).

The histologic features suggestive of MSI-H had a positive predictive value of 32% and negative predictive value of 90.2%, sensitivity of 66.7%, and specificity of 68.5% for tumors in patients younger than 50 years.

The presence of DNA MMR proteins hMLH1 and hMSH2 was evaluated by immunohistochemistry. Among the 17 MSI-H tumors, 10 cases (58.8%) were characterized by loss of hMLH1 and 6 (35.1%) showed loss of hMSH2 (Figures 3 and 4), whereas hMLH1 and hMSH2 had a normal expression in the MSS/MSI-L tumors. One MSI-H tumor expressed both hMLH1 and hMSH2.

The family history of cancer in patients with MSI-H colorectal tumors is summarized in Table 5. All (6 of 6) patients with hMSH2-negative tumors had a positive family history of HNPCC-related cancers, including colorectal, uterine, and brain tumors (namely, astrocytoma) (Table 5). Of the 10 cases with hMLH1-negative tumors, information regarding the family history of cancer was known in 8 cases, and only 3 cases reported a positive family history of carcinoma, including CRC, uterine cancer, and kidney carcinoma (Table 5).

Four patients with MSI-H had complete sequencing of hMLH1 and hMSH2, and 1 also had Southern analysis performed. Patient A had a clinical diagnosis of Muir-Torre and tested positive for a mutation in hMSH2. Patient B met Amsterdam criteria and tested positive for a mutation in hMSH2. Patient C had some family history of cancer, did not meet Amsterdam I or II criteria, and sequencing of hMSH2 and hMLH1 failed to identify a mutation; however, Southern blot analysis of hMSH2 was pursued and identified a deletion in hMSH2. Patient D had no family history of cancer and tested negative for mutations in both genes.

The revised Bethesda guidelines for screening of CRC for MSI allowed the identification of more MSI-H cases than the previous Bethesda recommendations, as only 8 of the 17 MSI-H cases fulfilled the criteria in the previous Bethesda guidelines. At the same time, more MSS cases were identified, since only 37 of the 58 MSS cases fulfilled the initial Bethesda criteria. Three of the 9 cases that matched the revised Bethesda guidelines and not the initial Bethesda recommendations were hMSH2 negative, and all 3 patients were younger than 50 years. Of the 6 patients showing loss of hMLH1 in the tumor, 2 were younger than 50 years. The other 4 patients were between 50 and less than 60 years and were included based on the presence of histologic features suggestive of MSI-H CRC.

The revised Bethesda guidelines for the screening of CRC for MSI testing15 recommend the use of clinical as well as histologic criteria in the selection of cases for molecular testing. While the necessary clinical information may not be available, at least 2 of the criteria recommended in these guidelines are readily available to the pathologist at the time of pathologic evaluation of colon cancer cases, namely, the age of the patient and the histologic features of the tumor. In this study, we evaluated the performance of these 2 parameters to identify MSI-H CRC cases. Therefore, the main criteria for inclusion in the study were the diagnosis of CRC in a patient younger than 50 years regardless of histology, or tumors with histologic features suggestive of an MSI-H tumor in patients 50 to less than 60 years old. Among 75 patients fulfilling these criteria, 22.6% were MSI-H CRC cases. Interestingly, 19.1% of the patients younger than 50 years had MSI-H CRC, and 57% of the cases selected based on the combined criteria of histopathologic features suggestive of MSI-H and between 50 and 60 years of age had MSI-H tumors.

Since the patients aged 50 to less than 60 years were included based on histopathologic features, we excluded these 7 cases from the analysis regarding the ability of histology to predict MSI status and evaluated that association for patients younger than 50 years. Our data reveal findings similar to those of previous studies regarding the sensitivity and specificity of histologic features to predict MSI-H.17,20 While the negative predictive value of histology suggestive of MSI-H was high (90%), the positive predictive value was low (32%), and sensitivity and specificity were relatively low, at 66.7% and 68.5%, respectively. The low specificity can be attributed in part to the inclusion of tumors with mucinous differentiation, which, although the most sensitive criterion, since it identified 9 of 15 MSI-H cases (sensitivity, 60%), it was also the least specific, with 17 of 56 MSS/MSI-L cases showing mucinous differentiation (specificity, 69.6%). There is nonetheless an important role of identification of histologic features for detection of CRC for MSI testing, supported by our results showing that 80% of the MSI-H CRCs had at least 1 of the criteria suggestive of MSI-H status. Interestingly, mucinous appendiceal carcinomas showed preserved expression of MMR proteins and an MSS phenotype (4/4 cases), in keeping with previously reported data.22 While tumors of the appendix are not specifically included or excluded in the revised Bethesda guidelines, our data and previously reported studies suggest that mucinous tumors arising in the appendix in young people could be excluded from reflex MSI testing.

The presence of tumor-infiltrating lymphocytes proved to be a very specific criterion, as only 1 of 57 of the MSS cases showed a significant number of tumor-infiltrating lymphocytes (specificity, 98.2%); however, this feature has a low sensitivity, since only 5 of 15 MSI-H tumors showed tumor-infiltrating lymphocytes (sensitivity, 33.3%). The other histopathologic criteria had intermediate sensitivities and specificities. The positive predictive value for MSI-H status of the combined age and histologic criteria was 31%. This low predictive value is not surprising, since the revised Bethesda criteria were designed to be more sensitive, therefore more inclusive, at the expense of specificity. Proximal tumors were more likely to show an MSI-H phenotype (10 [66.7%] of 15) than the distal tumors (8 [15.1%] of 53) (P = .01), which concurs with results from other studies.17,20 

In addition to the criteria discussed above, awareness of associated pathology may be available and could direct certain patients for MSI testing. Three patients in our series had associated pathology suggestive of HNPCC. Patient 10 of the MSI-H group presented with synchronous cecal adenocarcinomas (Figure 2) with characteristic histology and had a diagnosis of Muir-Torre syndrome with associated facial sebaceous adenomas. Patient 11 had a prior history of endometrial adenocarcinoma. Patient 17 presented at the age of 57 years with a tubulovillous adenoma with high-grade dysplasia and had a prior right hemicolectomy for colorectal adenocarcinoma 15 years prior to the metachronous tubulovillous adenoma.

Our studies confirm reported data regarding differences in clinicopathologic features of MSI-H CRC. Our data show that while the number of identified lymph nodes in colectomy specimens from patients with MSI-H tumors was higher, the absolute number of nodes with metastatic CRC was significantly lower in MSI-H tumors. Together with the observation that the MSI-H tumors were on average larger, these findings suggest a lower ability of tumor survival in metastatic foci of MSI-H CRC, which could contribute to the overall better 5-year survival of patients with MSI-H tumors reported previously.23,24 Despite limited available information regarding genetic/ germline gene mutation testing in our group of patients, it is noteworthy that 3 of the 4 patients with MSI-H tumors that were tested for germline gene mutation had hMSH2 mutations identified.

In our study, the updated Bethesda guidelines for screening of CRC for MSI allowed the identification of 9 additional cases, as compared to the 8 cases that fulfilled the criteria in the previous Bethesda recommendations. Importantly, 3 of the 9 cases that matched the updated Bethesda guidelines and not the previous recommendations were hMSH2 negative, and 6 showed loss of hMLH1 in the tumor. The identification of additional hMSH2-negative CRCs has significant implications, since CRC associated with loss of hMSH2 is nearly, if not always, associated with HNPCC. Four patients identified by the updated guidelines were between 50 and 59 years of age and were included based on the presence of histologic features suggestive of MSI-H CRC, emphasizing the importance of histologic criteria to select cases for MSI testing.

In summary, the updated Bethesda guidelines for MSI testing of colorectal tumors define an important role of the pathologist in identifying cases for routine testing based on the tumor histopathologic features suggestive of MSI-H status in patients younger than 60 years, in addition to patients with CRC younger than 50 years, regardless of the tumor histopathologic features. These criteria contrast with recommended testing of CRC patients aged 45 years or younger in the previous Bethesda guidelines.15 Our data argue against including appendiceal adenocarcinomas with mucinous features for routine MSI testing, since none of the cases tested in our series had MSI-H status.22 In the absence of expression by the tumor of one of the MMR proteins and/or the identification of an MSI-H phenotype, the patient should be referred for genetic counseling and potential genetic testing to rule out germline mutations associated with HNPCC or to confirm that the tumor represents a sporadic type of colorectal adenocarcinoma.15,25