Materials and methods

PATIENTS AND TUMOUR SAMPLES

For our study, we made use of an ongoing survey of the frequency of MMR germline mutations in early onset colorectal cancer (patientsless than 55 years old at diagnosis) systematically ascertained through three regional UK cancer registries. Formalin fixed, paraffin wax embedded blocks of colorectal cancers and EDTA venous blood samples were obtained from each patient. Our study was undertaken with approval from the relevant local ethics committee.

MOLECULAR ANALYSES

MSI was assessed using a fluorescent polymerase chain reaction (PCR) based assay at the following loci: D1S508, D2S123, D3S1561, D5S346, D11S29, TGFBIIR, D15S970, DCC, D19S565, BAT25, and BAT26. Microsatellite instability was defined as the presence of altered allele sizes in the PCR amplified product of tumour DNA compared with normal DNA. Tumours were designated as MSI positive if altered bands were seen in two or more microsatellites. The full coding sequences and splice junctions of human MLH1 (hMLH1) and hMSH2 were amplified using the PCR. Both PCR primers were end labelled with γ[³²P] ATP using T4 polynucleotide kinase and the amplified fragments were analysed by conformation sensitive gel electrophoresis.⁷ PCR products from samples that showed migration shifts were directly sequenced in forward and reverse directions and analysed on ABI377 DNA sequencers. Mutations were numbered according to accepted conventions.

IMMUNOHISTOCHEMISTRY

For each colorectal cancer, eight sections (3–4 μm thick) were cut and mounted on to glass slides. After dewaxing and rehydration of sections, antigenic site retrieval was accomplished by microwaving each slide for five minutes in 0.01 M citric acid buffer (pH 6.0). Endogenous peroxidase activity was blocked by incubation with 2% hydrogen peroxide for 20 minutes and non-specific binding prevented by incubation with 1% bovine serum albumin (BSA) in phosphate buffered saline (PBS). Sections were subsequently incubated with either monoclonal anti-MSH2 or anti-MLH1 antibodies (Oncogene, Cambridge, Massachusetts, USA) for two hours at room temperature. Antibody binding was detected using the Elite Vectastain ABC kit (Vector Laboratories Ltd, Peterborough, UK), which is based on the biotin–avidin system, using the manufacturer's protocol. The reaction was visualised using a VIP substrate kit for peroxidase (Vector Laboratories Ltd). Sections were then dehydrated and mounted. Normal colorectal tissue adjacent to the carcinoma was used as the positive control. Loss of expression was recorded when nuclear staining was observed in normal tissue but not in adjacent malignant cells.

Results

Table 1 shows the MSI and germline MSH1 and MLH2 status of the 46 colorectal cancers studied. Also shown are the immunohistochemical results for the anti-MSH1 and anti-MLH2 antibodies. The scoring of both antibodies was essentially straightforward. Intact nuclear staining of tumour cells with antibodies to both MSH1 and MLH2 was seen in all of the 23 MSI negative colorectal cancers, concordant with the molecular analysis indicating that these tumours had no evidence of MMR deficiency. In contrast, 22 of the 23 cancers displaying MSI showed no nuclear staining for one of the MMR proteins, permitting the underlying gene inactivation to be inferred (table 1).

The four colorectal cancers harbouring germline mutations in MLH1 were correctly identified. Of the MSI positive colorectal cancer cases that did not harbour a germline MMR mutation there was a preponderance of MLH1 deficiency compared with MSH2 (2.8 : 1). No tumour showed loss of staining with both anti-MLH1 and anti-MSH2 antibodies. These observations of MSH1 and MLH2 staining suggest that immunohistochemistry identifies MMR deficiency with 96% sensitivity (95% confidence limits (CL), 90% to 100%). The negative predictive value of 96% (95% CL, 90% to 100%) reflects the fact that one tumour exhibited MSI but showed staining for both antibodies. Table 2 shows summary statistics for our study and other recently published studies that have evaluated the use of immunohistochemistry as a method of defining the MMR status of colorectal cancers.

Discussion

Colorectal cancers displaying MMR deficiency are characterised by a distinctive morphological and clinical phenotype. Testing cancers for MMR deficiency provides a method of delineating a subset of sporadic cancers with a different clinical course and possible difference in response to chemotherapy.^{5, 6} Furthermore, and more importantly, testing colorectal cancers for this phenotype offers a means of refining the identification of patients harbouring germline mutations. The high risk of cancer conferred by constitutional mutations in the MMR genes (an approximate 70% risk of colorectal cancer and a pronounced increase in the risk of uterine and other adenocarcinomas)¹ necessitates the long term follow up and screening of carriers.

The screening of individuals with colorectal cancer for germline MMR mutations is currently undertaken in cases compatible with a hereditary basis (young onset, right sided, dominant inheritance of colorectal cancer). However, the analysis of constitutional DNA for MSH1 and MLH2 mutations is not straightforward because both genes are relatively large and mutations are scattered throughout each gene; hence, identification of MSI in tumours provides a practical method of identifying those patients in whom mutational analysis is appropriate. The detection of MSI in colorectal cancers classically requires microdissection and access to molecular biological facilities. Thus, it is a relatively complex procedure, expensive, and labour intensive, and is therefore not ideally suited to routine clinical practice. Furthermore, it does not circumvent the issue of screening more than one MMR gene for mutations.

In our study, we have evaluated the usefulness of immunohistochemistry as a method of identifying colorectal cancers with MMR deficiency as a result of the inactivation of the MSH1 or MLH2 gene. We found the sensitivity of this approach to be over 90%, which supports the findings of other recently published studies (table 2). Therefore, immunohistochemistry appears to be a good surrogate test for the detection of MSI caused by MSH2 and MLH1 dysfunction. One caveat to this is that, because of the interdependency between MMR genes, absent or reduced protein expression may be a consequence of the disruption of an interacting MMR gene. The other is that the demonstration of normal protein expression does not entirely preclude gene disruption because mutations that have pathological consequences through RNA decay or other similar mechanisms will go undetected. Therefore, testing cancers for MSI will still be required in cases where there is a high probability of HNPCC but protein expression is present.

Although immunohistochemistry alone cannot be relied upon to distinguish MSI colorectal cancers, it offers an additional method of prioritising mutation analyses suited to routine clinical laboratory practice. Because a small number of MSI positive colorectal cancers are caused by mutations in MMR genes other than MSH1 and MLH2,¹ the usefulness of immunohistochemistry as an initial screening tool will be extended by the use of antibodies against MSH6, PMS1, and PMS2.