Testing for Defective DNA Mismatch Repair in Colorectal Carcinoma: A Practical Guide

Colorectal carcinoma (CRC) was traditionally thought of as a monolithic disease process. During the last decade, however, variability in the underlying genetic pathogenesis has been described and has become clinically relevant.1 Colon cancer can now be split into (at least) 2 subsets based on molecular pathogenesis, the first characterized by chromosome instability and the second by genomic (genome-wide) instability.

The chromosomal instability (CIN) pathway accounts for 80% to 85% of all colon cancers. The CIN pathway involves loss of tumor suppressor genes and activation of oncogenes, characteristically identified as loss of heterozygosity, and often manifests overt aneuploidy. Commonly involved genes include the adenomatous polyposis coli (APC) gene, K-ras, the deleted in colon cancer (DCC) locus, and p53. This pathway also underlies familial adenomatous polyposis, which accounts for approximately 0.5% of all colon cancers. This large group probably comprises at least 2 major subsets likely to become better understood in the next several years.2 

Genomic instability results from a loss of DNA mismatch repair (MMR) activity.3 Normal DNA synthesis is associated with a low but finite error rate in terms of which nucleotide is put into the daughter strand, or whenever a base is added or deleted. It is, of course, critical to have a high-fidelity repair system to maintain the integrity of the genetic code. One of these repair systems is a family of proteins that works together as the DNA MMR complex. These genes/proteins have unfortunate alpha-numeric names, with the 4 major genes being MLH1, MSH2, PMS2, and MSH6. If any one of these proteins is absent, DNA MMR function is severely compromised. In the absence of DNA MMR activity, genome-wide mutations quickly accrue over the course of several generations of cell division (schematically represented in Figure 1). DNA microsatellites (mononucleotide, dinucleotide, or tetranucleotide repeat sequences found throughout the genome) make excellent test sites to assay for genetic errors that occur as a result of defective DNA MMR. Accordingly, the standard molecular laboratory test for defective DNA MMR is amplification of several microsatellites to look for abnormal variation in length, which is referred to as microsatellite instability (MSI).

Defective DNA MMR accounts for 15% to 20% of sporadic colon cancers, which typically occur at an older age than sporadic CIN carcinomas, are more frequently present in the proximal colon, and affect women twice as often as men. Lynch syndrome is a familial syndrome due to heritable defective DNA MMR that accounts for 2% of all colon cancers, demonstrates autosomal-dominant inheritance, has no gender bias, and has a broad age distribution with peak in the fifth decade.

Sporadic colon cancer with defective DNA MMR is essentially always due to the same, very specific DNA alteration.4 All of these MSI-positive CRCs result from methylation of the MLH1 gene promoter, which stops transcription of both MLH1 alleles. Therefore, not only do all sporadic CRCs due to defective DNA MMR have MSI, but they also are essentially all due to lack of MLH1 expression.

Based on the above discussion, CRCs can be divided into 4 subtypes (schematically represented in Figure 2): (1) familial colon cancer due to chromosomal stability (familial adenomatous polyposis), (2) sporadic colon cancers due to CIN (ordinary-type colon cancer), (3) familial colon cancer due to defective DNA MMR (Lynch syndrome), and (4) sporadic colon cancer due to defective DNA MMR.

Sporadic colon cancers with defective DNA MMR have a significantly improved prognosis over CIN-pathway CRCs.5–7 In fact, stage III CRCs with defective DNA MMR have a survival curve superimposable on that of stage II CIN cancers. This equates to half as much mortality in stage III CRC with defective DNA MMR through stage II CRC with CIN. Unfortunately, there are few data on whether this is dependent or independent of treatment, and the available data are conflicting.8–10 Currently, treatment decisions are not made based on DNA MMR status, despite in vitro data suggesting that 5-FU–based therapies do not work in these cancers.9,11–15 However, as chemotherapeutic regimens become refined and as panels of prognostic markers are put in place to determine which stage II and III CRC patients should receive therapy, these data will become essential.

The candidate tests for detection of defective DNA MMR in colon cancer include histology, immunohistochemistry, and molecular MSI testing. Characteristic histologic features do exist for CRC with defective DNA MMR.16,17 The most specific morphologic feature is a cribriform/solid (medullary) architecture without anaplastic cytology (Figure 3). Mucinous colon cancers (colloid and signet ring) also have an increased incidence of defective DNA MMR, but are much less specific than the medullary pattern. The suboptimal aspect of using the cribriform/ solid architectural pattern as a test for defective DNA MMR is poor sensitivity. The sensitivity is approximately 50%; half of all CRCs with defective DNA MMR have ordinary-type morphology with gland formation and desmoplasia. Tumor-infiltrating lymphocytes are more sensitive for the detection of defective DNA MMR.18,19 Tumor-infiltrating lymphocytes have to be intimately associated with the colon cancer cells to count for this assessment. This does not include peritumoral or Crohn-like inflammation. Unfortunately, there is a sensitivity-specificity trade-off, which results in approximately 80% sensitivity with 80% specificity when you have approximately 5 tumor-infiltrating lymphocytes per high-power field. Therefore, this is a good but not perfect marker for the detection of MSI-positive colon cancers.

For sporadic CRC with defective DNA MMR, immunohistochemistry is an excellent assay (Figure 4).3 MLH1 immunoperoxidase staining has essentially 100% sensitivity and specificity for defective DNA MMR in the sporadic setting due to the homogeneous underlying molecular pathogenesis (MLH1 promoter methylation). One would have to do this immunostain on all colon cancers to have near-100% sensitivity, but molecular microsatellite analysis would be redundant and therefore unnecessary.

Immunohistochemical reagents are available for MLH1, MSH2, MSH6, and PMS2 expression. The vast majority of all cell types (except those very terminally differentiated) normally demonstrate nuclear expression of each of these DNA MMR enzymes. This provides an extremely useful internal positive control when assaying for aberrant loss of a DNA MMR protein (Figure 4).

Approximately 2% of the 150 000 CRCs diagnosed in the United States each year occur in patients with Lynch syndrome (Figure 5).20–23 This means there are approximately 3000 CRCs per year in Lynch syndrome families. In addition, Lynch syndrome families are at risk for endometrial, upper gastrointestinal, upper urinary tract, ovarian, and pancreaticobiliary carcinomas. Identifying these families allows for sorting of at-risk family members. This autosomal-dominant disease results in 50% of family members having an 80% to 100% lifetime risk of carcinoma development, with the other half having normal cancer risk. Therefore, identification of Lynch syndrome probands can impact decision making (eg, screening, prophylactic surgery) tremendously.

It is reasonable to clarify some terminology at this point. Hereditary nonpolyposis colorectal cancer syndrome (HNPCC) is a clinically defined term based on strong family history of colon cancer. The original HNPCC definition, referred to as the Amsterdam Criteria, includes the following: (1) 3 involved relatives, 2 of whom are first-degree relatives; (2) successive generations involved; and (3) 1 involved patient younger than 50 years. In only 60% of the families with a strong family history of colorectal carcinoma, even to the extent of meeting these criteria, the colorectal carcinoma is due to heritable DNA mismatch repair.24,25 Therefore, 40% of HNPCC families remain unexplained as to their genetic basis. Disease in those HNPCC families with heritable DNA MMR is referred to as Lynch syndrome.21 Approximately 30% of colon cancer patients have at least 1 first- or second-degree relative also with colon cancer.

Immunohistochemical findings in Lynch syndrome cases are much more complex than in sporadic MSI-positive CRCs. While 100% of Lynch syndrome–related carcinomas have MSI, only approximately 95% of them will have abnormal immunohistochemistry findings (loss of expression of a DNA MMR enzyme). The rough distribution of abnormal DNA MMR gene and subsequent loss of protein expression within Lynch syndrome families is as follows: MLH1, 40%; MSH2, 40%; MSH6, 5%; and PMS2, 5%. A small percentage of Lynch syndrome families develop MSI-positive carcinomas with intact staining of all 4 of these enzymes due either to a missense mutation affecting function but not antigenicity, or involvement of another minor protein in the complex.

Based on the above information, recommended testing for Lynch syndrome includes a combination of molecular microsatellite analysis and immunohistochemistry for MLH1, MSH2, PMS2, and MSH6. This allows for high-sensitivity detection of MSI within the tumors and identification of which enzyme is abnormal within that family, in turn guiding the subsequent germline mutation analysis. Both microsatellite analysis and immunohistochemistry are performed on paraffin-embedded tissue. The microsatellite analysis requires both tumoral and nontumoral tissue for comparison.

The rule of thumb is any patient with CRC arising with positive family history and/or young age. A series of National Institutes of Health–sponsored consensus conferences have established the Bethesda Guidelines for selecting patients to test.21 

There are instances in which it can be difficult to differentiate between familial and sporadic defective DNA MMR colon cancer. For example, a 50-year-old with a colon cancer showing high levels of MSI and loss of MLH1 expression can be ambiguous as to germline versus somatic abnormality if the family history is not clear or unavailable. Germline analysis is helpful in these situations, but is also less than 100% sensitive. Recently, a specific B-raf mutation in codon 600 has been identified in a majority of sporadic (methylated) defective DNA mismatch repair CRC cases, which rarely is present in Lynch syndrome carcinomas.26,27 Therefore, B-raf mutational analysis may soon become part of the testing algorithm in this situation.

In summary, 15% to 20% of CRCs are due to defective DNA MMR. Ninety percent of these cases will be caused by silencing of MLH1 expression due to methylation of the gene promotor. These sporadic CRCs with defective DNA MMR have a significantly improved prognosis, but therapy is unchanged. Lynch syndrome proband identification is a very active testing program taking advantage of both microsatellite analysis and immunohistochemistry on paraffin-embedded tissue with subsequent gene mutation analysis (peripheral blood DNA) to sort at-risk family members.