Use of Molecular Tumor Characteristics to Prioritize Mismatch Repair Gene Testing in Early-Onset Colorectal Cancer
http://www.100md.com
《临床肿瘤学》
the Genetic Epidemiology Laboratory, Department of Pathology
Centre for Molecular, Environmental, Genetic, and Analytical Epidemiology, The University of Melbourne, Parkville
Genetic Technologies Limited, Kew
Peter MacCallum Cancer Centre, East Melbourne
The Australian Red Cross Blood Service, Melbourne
Department of Anatomical Pathology, Western Hospital, Footscray
Department of Colorectal Medicine and Genetics, The Royal Melbourne Hospital, Melbourne
Cancer Epidemiology Centre, The Cancer Council, Carlton, Victoria
ACCFR Molecular Cancer Epidemiology Laboratory, Queensland Institute of Medical Research, Queensland, Australia
Department of Pathology, McGill University, Quebec, Canada
ABSTRACT
PURPOSE: The relationships between mismatch repair (MMR) protein expression, microsatellite instability (MSI), family history, and germline MMR gene mutation status have not been studied on a population basis.
METHODS: We studied 131 unselected patients with colorectal cancer diagnosed younger than age 45 years. For the 105 available tumors, MLH1, MSH2, MSH6, and PMS2 protein expression using immunohistochemistry (IHC) and MSI were measured. Germline DNA was screened for hMLH1, hMSH2, hMSH6, and hPMS2 mutations for the following patients: all from families fulfilling the Amsterdam Criteria for hereditary nonpolyposis colorectal cancer (HNPCC); all with tumors that were high MSI, low MSI, or that lacked expression of any MMR protein; and a random sample of 23 with MS-stable tumors expressing all MMR proteins.
RESULTS: Germline mutations were found in 18 patients (nine hMLH1, four hMSH2, four hMSH6, and one hPMS2); all tumors exhibited loss of MMR protein expression, all but one were high MSI or low MSI, and nine were from a family fulfilling Amsterdam Criteria. Sensitivities of IHC testing, MSI (high or low), and Amsterdam Criteria for MMR gene mutation were 100%, 94%, and 50%, respectively. Corresponding positive predictive values were 69%, 50%, and 75%.
CONCLUSIONS: Tumor IHC analysis of four MMR proteins and MSI testing provide a highly sensitive strategy for identifying MMR gene mutation–carrying, early-onset colorectal cancer patients, half of whom would have been missed using Amsterdam Criteria alone. Tumor-based approaches for triaging early-onset colorectal cancer patients for MMR gene mutation testing, irrespective of family history, appear to be an efficient screening strategy for HNPCC.
INTRODUCTION
Germline mutations in the DNA mismatch repair (MMR) genes hMLH1and hMSH2, and to a lesser extent hMSH6 and hPMS2, are associated with a high lifetime risk of colorectal cancer and an increased risk of extracolonic cancers.1-3 Recognition of the role of these genes in hereditary cancer came about partly because colorectal cancer in multiple-case families often displayed replication errors (microsatellite instability [MSI]) —a consequence of MMR gene dysfunction.4 To detect and repair mismatches occurring during cell replication, the proteins MSH2 and MSH6 bind to form a heterodimer. Likewise, MLH1 binds with PMS2 to form a heterodimer.
Given that genetic testing for germline mutations is technically challenging and costly, criteria have been developed to help identify the individuals with colorectal cancer who carry a MMR gene germline mutation. These criteria include, but are not limited to, tumor loss of MMR protein expression, tumor MSI, and a strong family history of colorectal cancer or other specific cancers.5,6
To justify the use of these criteria to target individuals for germline testing, their performance in terms of sensitivity, specificity, positive predictive value, and negative predictive value needs to be determined. A recent meta-analysis of clinic-based studies of multiple-case and early-onset families concluded that using the Amsterdam Criteria,6 which are based solely on family cancer history, and a previous version of the Bethesda Criteria,7 which extends the Amsterdam Criteria to include tumor MSI testing, "was inadequately sensitive and specific for the identification of carriers." 8
Several studies have selected patients from family cancer clinics to examine relationships between MMR gene germline mutation, tumor MMR protein expression as measured by immunohistochemistry (IHC), tumor MSI, and family history of colorectal cancer.9-16 One of the largest studies included 48 colorectal cancer patients from families referred to a genetics program.17 All 14 identified carriers of germline mutations in hMLH1 or hMSH2 had high MSI colorectal cancer (sensitivity of high MSI, 100%) but there were an additional 14 high MSI colorectal cancers in individuals for whom no germline mutation was detected (positive predictive value, 50%). IHC testing for MLH1 and MSH2 had corresponding sensitivity and predictive values of 57% and 44%, respectively. For Amsterdam Criteria, the corresponding values were only 71% and 50%, respectively. To the best of our knowledge, the performance of these criteria has not been evaluated on a population basis for individuals with colorectal cancer unselected for family history, yet at increased probability of being carriers because of the presence of early-onset disease.
In this article, we studied a population-based sample of colorectal cancer patients diagnosed at age younger than 45 years and unselected for family cancer history.18 We tested for the expression of MLH1, MSH2, MSH6, and PMS2 proteins in the tumors using IHC, conducted MSI testing, and performed germline mutation detection in hMLH1, hMSH2, hMSH6, and hPMS2.
METHODS
The Victorian Colorectal Cancer Family Study is a population-based, case-family study of early-onset colorectal cancer conducted in Australia during 1993 to 1997.18 Approval for the study was obtained from the ethics committees of The University of Melbourne and The Cancer Council of Victoria. All participants provided written informed consent for participation in the study.
Participants
Eligible patients comprised adult men and women living in the Melbourne metropolitan area who were younger than age 45 years when diagnosed with a histologically confirmed, first primary adenocarcinoma of the colon or rectum (International Classification of Diseases, 9th revision, rubrics 153 and 154, respectively) during the period from July 1, 1992, to September 30, 1996. Patients were identified through the population-based Victorian Cancer Registry that receives statutory notifications from all hospitals and pathology laboratories. A letter requesting permission to approach the patients was sent to the attending doctors of a random selection of 222 patients, followed by letters to the patients seeking participation. Attrition due to death (12.6%), refusal (doctor, 10.4%; patient, 12.6%), or because the patient moved and was not located (5.4%) resulted in 131 consenting to participate (59.0% of those eligible). The median time between diagnosis and interview was 9 months.
Data Collection
Patients answered a risk factor questionnaire and completed a family pedigree by identifying all first- and second-degree relatives and reporting on their vital status and cancer histories during an in-person interview conducted in their homes. Each case patient was then asked to obtain permission from his or her adult relatives for us to approach them for their participation. Relatives answered the same risk factor questionnaire during a telephone interview, and similarly were asked about the family's cancer history. Verification of all reports of cancers in family members was sought through cancer registries, hospital records, treating clinicians, and death certificates (see Jenkins et al18 for details). Patients were also asked to donate a small blood sample for genetic testing.
Families were categorized as meeting the Amsterdam Criteria if all of the following criteria were met: at least three relatives were diagnosed with cancer of the colon, rectum, endometrium, small bowel, ureter, or renal pelvis; at least one of these was a first-degree relative of at least two others; relevant cancers were diagnosed in at least two successive generations; at least one of those cancers was diagnosed at age younger than 50 years; and familial adenomatous polyposis was excluded. This definition has also been referred to as Amsterdam Criteria II.6
Tumor Collection
Invasive tumor samples of primary colorectal adenocarcinoma were obtained from hospitals and private pathology laboratories for 118 patients (90%). Six did not consent to release tissue to the study and two laboratories had not agreed to release the remaining seven samples at the time of this study. All tumor samples were reviewed by a pathologist who directed the microdissection.
Tumor Expression of MMR Genes
The expression of MLH1, MSH2, MSH6, and PMS2 was assessed by IHC on 3-μm, formalin-fixed, paraffin-embedded sections using standard methods, monoclonal antibodies MLH1 (Pharmingen clone G168-728), MSH2 (Oncogene clone FE11), MSH6 (BD Transduction Laboratories clone 44), and PMS2 (Pharmingen clone A16-4) on a DAKO autostainer.19 Normal colonic epithelium adjacent to tumor and lymphocytes served as built-in positive controls. A GI pathologist scored the tumors as positive when nuclear staining in tumor tissue was present or otherwise as negative when the nuclear stain was absent.
Tumor MSI
MSI testing was performed on invasive tumor cells microdissected from 5-μm sections of paraffin-embedded archival tumor tissue stained with 1% methyl green. DNA extracted from histologically normal cells microdissected from colonic or lymph node tissue, or DNA extracted from peripheral-blood lymphocytes, was used to control the assay as previously described.20,21 Ten microsatellite markers were assessed: three dinucleotide repeats (D5S346, D17S250, and D2S123) and seven mononucleotide repeats (BAT-25, BAT-26, BAT-40, MYB, TGF?RII, IGFIIR, and BAX). For this study, the degree of instability in each tumor was scored as stable (MS stable), low (low MSI), and high (high MSI) when 0 to 1, 2 to 5, and 6 to 10 markers were identified as unstable, respectively.21 An assessment of MSI was not successful for 13 (12%) tumor samples because of technical reasons related to tumor DNA quality, which left 105 tumors tested.
Germline MMR Mutations
All exonic and flanking intronic sequences of the hMLH1, hMSH2, hMSH6, and hPMS2 genes were screened for germline mutations using sequencing approaches, except for exon 4 of hMSH6, which was screened in eight overlapping fragments using denaturing high-performance liquid chromatography. Confirmation of putative mutations identified via denaturing high-performance liquid chromatography or sequencing was sought using an independent polymerase chain reaction for direct automated sequencing. Variants were defined to be deleterious if they could be predicted to produce (or were known to produce) a shortened or truncated protein product, or were missense mutations that have been reported previously to be deleterious. The Multiplex Ligation-Dependent Probe Amplification assay22 (MRC-Holland, Amsterdam, the Netherlands) to detect large genomic alterations in hMLH1 and hMSH2 was performed on samples from 10 patients who had tumors lacking at least one MMR protein expression and for which no previous mutation had been identified by sequencing. Break points were characterized by performing a long-range polymerase chain reaction encompassing the predicted genomic alteration, cloning, and sequencing.
Mutation testing was conducted for all patients with one or more of the following characteristics: a family history that fulfilled the Amsterdam Criteria for hereditary nonpolyposis colorectal cancer (HNPCC); having a tumor that was high MSI, low MSI, or that lacked expression of at least one MMR protein; and presence in a random sample of 23 patients selected from those who had tumors that were MS stable and did not lack expression of any MMR protein.
Statistical Analysis
Sensitivity, specificity, and positive predictive and negative predictive values were estimated,23 and exact 95% CIs of these proportions were calculated using STATA software, release 8.0 (STATA Corp, College Station, TX). Because we did not conduct germline MMR testing on all patients who had tumors that were MS stable and IHC normal, we could not calculate sensitivity and specificity directly. Sensitivity is a/(a + c + ), where a is the observed number of carriers who were positive for the tumor test, c is the observed number of carriers who were negative for the tumor test, and is the number of carriers in the phenotypically negative patients who were not tested for germline mutations. Specificity was estimated as (d – )/(d – + b), where d is the observed number of noncarriers who were negative for the tumor test, and b is the observed number of noncarriers who were positive for the tumor test.
We assumed that the proportion of carriers in these untested, phenotypically negative patients was the same as that in the phenotypically negative patients who were tested for germline mutations. This turned out to be zero; therefore, was zero. We then conducted sensitivity analyses that recalculated sensitivity and specificity by supposing one or a few of the untested phenotypically negative patients were carriers. The probability that there are d = k carriers among m untested phenotypically negative patients, given we observed no carriers in the tested phenotypically negative patient group (n = 23), is [(n + 1)m!(n + m – k)!]/[(n + m + 1)!(m – k)!] (k = 0,1...,m [see Appendix]).
RESULTS
Germline Mutation Carriers
Table 1 indicates that deleterious mutations were detected for 18 patients (17% of the 105 patients for whom tumor material was tested for MSI and MMR protein expression); nine in hMLH1 (50% of all mutations), four in hMSH2 (22%), four in hMSH6 (22%), and one in hPMS2 (6%). A number of other well-documented common variants that have not been classified previously as deleterious were also detected (data not shown). Most carriers were male (83%) compared with half of noncarriers (48%; P = .007). The mean age at diagnosis for carriers (36 years) was younger than that for noncarriers (41 years; P = .0005).
Of the 18 mutations detected, 15 (83%) were predicted to produce a truncated protein product (four were alterations in splice site regions, seven were frameshift mutations, two were nonsense mutations, and two were large deletions). Of the remaining three, one (hMSH2 P622L) was a missense mutation known to be deleterious,24 and another (hMLH1 1846delAAG) was an inframe deletion known to be deleterious.25-27 The third was an inframe deletion, previously unreported (hMLH1 987delCAT [C330]), that we considered deleterious for the following reasons: for the patient with this mutation, the mother was a carrier and had stomach cancer diagnosed at age 43 years and colon cancer diagnosed at age 55 years, the maternal grandfather (deceased and untested) had gastric cancer diagnosed at age 35 years, a maternal uncle (deceased and untested) had cancer of the duodenum diagnosed at age 46, a maternal uncle (deceased and untested) had stomach cancer diagnosed at age 48, and the only carrier in siblings was a brother unaffected at age 29 years.
Table 2 indicates that the sensitivity of IHC was 100%, given that all of the carriers' colorectal cancers lacked expression of at least one MMR protein consistent with the gene mutated in the germline. All but two of the hMLH1, hMSH2, and the hPMS2 mutation carriers' tumors clearly lacked expression of both proteins of the relevant dimer. For the tumor from the carrier of a large genomic deletion encompassing exons 4 and 5 of hMLH1, there was no nuclear staining of MLH1 but high cytoplasmic staining present in biopsy and surgical specimens; however, the staining in normal colon tissue was equivocal. We classified this patient as not expressing MLH1. For the carrier of the large genomic deletion encompassing exon 15 of hMLH1, a biopsy sample had nuclear staining for MLH1 in the tumor (some focal cells were negative) as well as normal colon tissue and was thus classified as expressing MLH1. All tumors from the four hMSH6 mutation carriers lacked expression of MSH6 only. Therefore, five of the 18 MMR gene mutation carriers (28%) expressed MLH1 and MSH2 yet lacked expression of MSH6 or PMS2.
All tumors from the 13 hMLH1 and hMSH2 mutation carriers were high MSI and all tumors from the four hMSH6 mutation carriers were low MSI. We classified the tumor from the hPMS2 mutation carrier as MSI stable, although due to technical difficulties caused by tumor DNA quality, only four of the 10 microsatellite markers were scored for instability. Therefore, 72% of carriers were high MSI (sensitivity, 72%) and 94% were either high MSI or low MSI.
Of the carriers, nine were members of Amsterdam Criteria families (sensitivity, 50%); three (17%) were from patients with a family history of colorectal cancer, but not sufficient to fulfill the Amsterdam Criteria; and six (33%) were from patients with no known family history of colorectal cancer.
IHC
Table 1 shows that, of the 105 tumors tested, 26 (19%) lacked expression of any of the MMR proteins, and of these, 18 had a deleterious germline mutation in one of the four MMR genes (positive predictive value, 69%; Table 2). Of the 79 patients with tumors that expressed all proteins, no mutations were detected in the 33 who were tested.
MSI Testing
Table 1 shows that, of the 105 tumors tested for MSI, 18 (17%), 16 (15%), and 71 (68%) were assessed to be high MSI, low MSI, and MS stable, respectively. Of the 18 patients with a high MSI tumor, 13 were found to carry deleterious germline mutations in an MMR gene (all hMLH1 or hMSH2; positive predictive value, 72%). Of the other five patients with a high MSI tumor in which no germline mutation was detected, three lacked the expression of both MLH1 and PMS2, and two lacked the expression of both hMSH2 and MSH6. One belonged to a family meeting the Amsterdam Criteria, one had some family history of colorectal cancer (a paternal grandmother with cancer of the rectum diagnosed at age 64 years), and three had no reported family history of colorectal cancer.
Of the 16 patients with a low MSI tumor, four were found to carry a deleterious MMR gene mutation; all were hMSH6 carriers. Of the other 12, one lacked MLH1 and PMS2 expression, and one lacked MSH2 and MSH6 expression. Therefore, of 34 patients with a tumor that exhibited high MSI or low MSI, 17 had a germline mutation in an MMR gene (positive predictive value, 50%). One had a family history meeting the Amsterdam Criteria, four had a family history of colorectal cancer not sufficient to fulfill the Amsterdam Criteria, and seven reported no family history.
Of the 71 patients with MS stable tumors, 25 were tested for germline mutations, revealing one hPMS2 mutation whose tumor lacked both PMS2 and MLH1 expression. There was also one tumor that lacked PMS2 expression. Of the 70 MS stable tumors without a known deleterious germline mutation, one had a family history meeting the Amsterdam Criteria, 24 had a family history of colorectal cancer not sufficient to fulfill the Amsterdam Criteria, and 45 reported no family history.
Family History
Of the 105 patients, 12 (11%) had a family history that met the Amsterdam Criteria; of these, nine had a germline mutation in one of the MMR genes (positive predictive value, 75%). Thirty-two patients (30%) had a family history of colorectal cancer not sufficient to fulfill the Amsterdam Criteria; of these, three (9%) had a germline mutation. Sixty-one patients (58%) had no family history of colorectal cancer in three generations; of these, six (10%) had a germline mutation (negative predictive value of family history, 90%).
DISCUSSION
We found that IHC testing of the expression the four MMR proteins and MSI testing are both highly sensitive for MMR gene mutation status in early-onset colorectal cancer patients. Half of the carriers in our sample would have been missed if mutation testing had been confined solely to those whose cancer family history met the Amsterdam Criteria for HNPCC.
We did not test for germline mutations in 46 of our 105 patients who did not belong to an Amsterdam Criteria family or have any molecular tumor characteristics suggestive of MMR dysfunction (specifically, loss of protein expression or MSI). For our initial calculation of sensitivities and specificities, we assumed that these 46 early-onset patients were not carriers; this was supported by the fact that sequencing the MMR genes of 23 randomly chosen patients who did not have these phenotypic characteristics of a carrier failed to identify a single carrier. Given this, it is unlikely that many (if any) carriers were missed; the probability that the number of carriers in the 46 untested patients is 0, 1, and 2 is .33, .22, and .15, respectively. Therefore, we recalculated the sensitivity and specificity assuming that one carrier had been missed in the 46 patients that had no suggestion of being a carrier (2%), and was incorrectly classified as a noncarrier with an MS stable and IHC normal tumor. We found that the high sensitivity and specificity of IHC and MSI decreased little. For example, the sensitivity became 95% for lack of expression of any MMR protein, 89% for high MSI or low MSI, and 42% for Amsterdam Criteria family history.
Our finding of a high sensitivity for IHC and MSI is supported by data from nine studies reporting on a total of 105 patients with colorectal cancer with known germline mutations in hMLH1 or hMSH2 who were subsequently tested for both MSI and IHC for MLH1 and MSH2.9-17 In these clinic-based series there were only two reported carriers of hMLH1 or hMSH2 mutations with MS stable tumors that expressed both MLH1 and MSH2.13,28 Such outlier observations may be due to failure in laboratory technique,29 but it is theoretically possible that mutations may destroy enzymatic activity while preserving the immunoreactivity and stability of proteins. It is also remotely possible that some cancers in carriers were phenocopies (ie, coincidental cases of colorectal cancer not caused by or associated with the germline MMR mutation and therefore not exhibiting loss of MMR protein expression or MSI). Despite our exhaustive screening, it is possible that carriers could have been missed because we did not screen for large genomic deletions in hMSH6 or hPMS2, and sequencing is not infallible.
The advantages and disadvantages of targeting mutation testing based on IHC, MSI, or family history are summarized in Table 3. On the basis of our data, using IHC alone as a basis for targeting mutation testing would result in all MMR gene carriers being tested for mutations. If we had not conducted IHC testing for all four proteins we would have missed detecting evidence for some carriers (eg, by not conducting IHC testing for MSH6 or PMS2, we would have failed to identify as likely carriers the two patients with the large deletions in hMLH1 and all four patients with hMSH6 mutations—almost one third of all carriers). Similarly, Christensen et al9 determined that IHC testing for PMS2 identified 23% more hMLH1 carriers than testing for MLH1 alone. On the basis of the positive predictive value of 69%, for every two carriers detected, one noncarrier would have been tested. Importantly, IHC testing for all four MMR proteins indicates which MMR genes are mutated in the germline, and this can decrease the cost of genetic testing.
If instead, mutation testing was targeted based on tumor MSI status, our data would suggest that although nearly all MMR gene mutation carriers would have been tested, for every carrier detected, one noncarrier would have been tested. Our extended microsatellite panel for MSI testing, which contains seven mononucleotide repeats, successfully differentiated the hMLH1 and hMSH2 mutation carriers (all high MSI) from the hMSH6 mutation carriers (all low MSI). This would not have been possible using the standard National Cancer Institute panel (data not shown), which tests five microsatellites only and is commonly used to determine who should be tested for germline mutations in all of the MMR genes.5 A panel of 10 markers currently is also being used in the world's largest colorectal cancer family study, the Colon Cancer Family Registry.30
Although the MSI testing performed according to the Revised Bethesda guidelines for HNPCC5 would have identified almost all MMR mutation carriers, one would have had to test for germline mutations in all four MMR genes to do so. Using IHC of MMR protein expression to guide mutation testing would have identified all mutation carriers, with approximately the same sensitivity, but with the advantage that by testing for all four proteins, the observed pattern would have identified correctly the specific genes to be tested. In contrast, family history alone was not nearly as sensitive as IHC or MSI for targeting mutation testing. Half were from families that satisfied the Amsterdam Criteria (including the patient diagnosis); one sixth were from patients with a family history of colorectal cancer not meeting the Amsterdam Criteria; and one third were from patients with no known family history of colorectal cancer.
There are practical difficulties with all three approaches. IHC and MSI (with microdissection) require a pathologist, at least to select sections for staining. MSI testing needs to be conducted in a molecular laboratory, is more expensive, and does not provide gene-specific information. Access to normal tissue, which is required for interpretation, can be restricted in retrospective studies. Obtaining a complete and accurate family history is expensive and time consuming.
These extensive molecular and genetic analyses of a population-based series reveal that the traditional approach to mutation testing, focused initially on the family cancer history of opportunistically identified multiple-case families and nonsystematic referral of patients meeting the Bethesda Criteria, lacks the efficiency for identifying mutation carriers that a systematic assessment of all early-onset patients identified through normal clinical practice could provide.31 We would argue that identifying mutation carriers would be better facilitated by a pathologist or an appropriately trained technician who would, as part of the description of the phenotypic features of the tumor, conduct an IHC examination of tumor for loss of any of the four MMR proteins in early-onset colorectal cancers. Given that the IHC findings do not constitute definitive germline findings, it is arguable that the genetic counseling process, including consent processes, could be delayed until IHC testing suggests a likelihood of the existence of a germline mutation. After genetic counseling and informed consent, germline genetic testing could then be directed toward one or two of the MMR genes, rather than all four.
The difficulty with incorporating genetic counseling before IHC testing is the burden of consent required of both doctor and patient at a time when the patient's attention is focused on concerns of self-preservation. Furthermore, it could be argued that no matter how statistically correlated to germline information, IHC measures biologic expression and does not provide definitive germline information, which can only come from mutation testing. There are equally or more cogently genetic diagnoses made by pathologists on operative specimens where such presurgical consent for the reporting of a genetic diagnosis is never required, but has equally relevant implications for a germline diagnosis (eg, surgical pathological diagnosis of familial adenomatous polyposis).
The proposed approach will lead to a more efficient identification of individuals at truly high genetic risk of cancer.31 It is important to recognize that the details of how this regimen might be implemented would depend a great deal on local factors. Appropriate cancer surveillance procedures for relatives of identified carriers would also need to be implemented. To achieve this there is a continuing need for research on the efficacy and optimization of surveillance procedures and the identification of environmental, lifestyle, and genetic factors that modify risk in these genetically at-risk individuals. An approach built around IHC testing alone as the first step is likely to be much more cost effective for detecting mutation carriers than one based on family history alone, or one based on MSI testing alone, although the relative cost efficiencies would need to be determined for each local situation depending on resources, skills, infrastructure, and other considerations.31
In conclusion, for early-onset colorectal cancer, IHC testing of tumors for four MMR proteins is a strong predictor of germline MMR mutation status for particular gene(s). It is more efficient, in terms of the extent of genetic testing, than MSI testing and much more efficient than using family history. Tumor-based strategies for triaging colorectal cancer patients for MMR gene mutation testing and subsequent predictive testing may be an efficient way to identify individuals with hereditary colorectal cancer due to an MMR gene mutation (ie, with HNPCC), as well as their relatives who carry the family-specific mutation. Both sets of individuals are at a high colorectal cancer risk, and their lives may be saved or prolonged by intensive surveillance.
Appendix
Suppose the binomial parameter initially has a uniform prior distribution. If we observe n trials with j successes, the posterior distribution is then
Here D ( a, b) is the Dirichlet distribution with parameters a, b. If this is now taken as the prior for , the marginal probability of observing k successes during a subsequent m trials is just
Note that this is a truncated version of the hypergeometric distribution.
As a special case, if there are no successes in the first n trials, the posterior probability of no successes in the next m is (n + 1)/(n + m + 1); the probability of k successes from the subsequent m is
Authors' Disclosures of Potential Conflicts of Interest
Although all authors completed the disclosure declaration, the following author or immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Acknowledgment
We thank Judith Maskiell and the research interviewers; Gillian Dite and the data management staff; Graham Byrnes for statistical advice; and Sabar Napaki, Deon Venter, and Jane Armes for pathology input. We also thank the men and women who participated in this study.
NOTES
Supported by grants from the National Health and Medical Research Council (Australia) and the Victorian Health Promotion Foundation.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Mitchell RJ, Farrington SM, Dunlop MG, et al: Mismatch repair genes hMLH1 and hMSH2 and colorectal cancer: A HuGE review. Am J Epidemiol 156:885-902, 2002
Aaltonen LA, Peltomaki P, Leach FS, et al: Clues to the pathogenesis of familial colorectal cancer. Science 260:812-816, 1993
Kolodner RD, Tytell JD, Schmeits JL, et al: Germ-line msh6 mutations in colorectal cancer families. Cancer Res 59:5068-5074, 1999
Thompson E, Meldrum CJ, Crooks R, et al: Hereditary non-polyposis colorectal cancer and the role of hPMS2 and hEXO1 mutations. Clin Genet 65:215-225, 2004
Umar A, Boland CR, Terdiman JP, et al: Revised Bethesda Guidelines for hereditary nonpolyposis colorectal cancer (Lynch syndrome) and microsatellite instability. J Natl Cancer Inst 96:261-268, 2004
Vasen HF, Watson P, Mecklin JP, et al: New clinical criteria for hereditary nonpolyposis colorectal cancer (HNPCC, Lynch syndrome) proposed by the International Collaborative group on HNPCC. Gastroenterology 116:1453-1456, 1999
Rodriguez-Bigas MA, Boland CR, Hamilton SR, et al: A National Cancer Institute Workshop on Hereditary Nonpolyposis Colorectal Cancer Syndrome: Meeting highlights and Bethesda guidelines. J Natl Cancer Inst 89:1758-1762, 1997
Kievit W, de Bruin JH, Adang EM, et al: Current clinical selection strategies for identification of hereditary non-polyposis colorectal cancer families are inadequate: A meta-analysis. Clin Genet 65:308-316, 2004
Christensen M, Katballe N, Wikman F, et al: Antibody-based screening for hereditary nonpolyposis colorectal carcinoma compared with microsatellite analysis and sequencing. Cancer 95:2422-2430, 2002
Debniak T, Kurzawski G, Gorski B, et al: Value of pedigree/clinical data, immunohistochemistry and microsatellite instability analyses in reducing the cost of determining hMLH1 and hMSH2 gene mutations in patients with colorectal cancer. Eur J Cancer 36:49-54, 2000
Hoedema R, Monroe T, Bos C, et al: Genetic testing for hereditary nonpolyposis colorectal cancer. Am Surg 69:387-392, 2003
Marcus VA, Madlensky L, Gryfe R, et al: Immunohistochemistry for hMLH1 and hMSH2: a practical test for DNA mismatch repair-deficient tumors. Am J Surg Pathol 23:1248-1255, 1999
Salahshor S, Koelble K, Rubio C, et al: Microsatellite Instability and hMLH1 and hMSH2 expression analysis in familial and sporadic colorectal cancer. Lab Invest 81:535-541, 2001
Scartozzi M, Bianchi F, Rosati S, et al: Mutations of hMLH1 and hMSH2 in patients with suspected hereditary nonpolyposis colorectal cancer: Correlation with microsatellite instability and abnormalities of mismatch repair protein expression. J Clin Oncol 20:1203-1208, 2002
Thibodeau SN, French AJ, Roche PC, et al: Altered expression of hMSH2 and hMLH1 in tumors with microsatellite instability and genetic alterations in mismatch repair genes. Cancer Res 56:4836-4840, 1996
Ward R, Meldrum C, Williams R, et al: Impact of microsatellite testing and mismatch repair protein expression on the clinical interpretation of genetic testing in hereditary non-polyposis colorectal cancer. J Cancer Res Clin Oncol 128:403-411, 2002
Wahlberg SS, Schmeits J, Thomas G, et al: Evaluation of microsatellite instability and immunohistochemistry for the prediction of germ-line MSH2 and MLH1 mutations in hereditary nonpolyposis colon cancer families. Cancer Res 62:3485-3492, 2002
Jenkins MA, Baglietto L, Dite GS, et al: After hMSH2 and hMLH1: What next? Analysis of three-generational, population-based, early-onset colorectal cancer families. Int J Cancer 102:166-171, 2002
Young J, Simms LA, Biden KG, et al: Features of colorectal cancers with high-level microsatellite instability occurring in familial and sporadic settings: Parallel pathways of tumorigenesis. Am J Pathol 159:2107-2116, 2001
Armes JE, Trute L, White D, et al: Distinct molecular pathogeneses of early-onset breast cancers in BRCA1 and BRCA2 mutation carriers: A population-based study. Cancer Res 59:2011-2017, 1999
Southey MC, Young MA, Whitty J, et al: Molecular pathologic analysis enhances the diagnosis and management of Muir-Torre syndrome and gives insight into its underlying molecular pathogenesis. Am J Surg Pathol 25:936-941, 2001
Schouten JP, McElgunn CJ, Waaijer R, et al: Relative quantification of 40 nucleic acid sequences by multiplex ligation-dependent probe amplification. Nucleic Acids Res 30:e57, 2002
Knottnerus JA, van Weel C, Muris JW: Evaluation of diagnostic procedures. BMJ 324:477-480, 2002
Liu B, Parsons RE, Hamilton SR, et al: HMSH2 mutations in hereditary nonpolyposis colorectal cancer kindreds. Cancer Res 54:4590-4594, 1994
Farrington SM, Lin-Goerke J, Ling J, et al: Systematic analysis of hMSH2 and hMLH1 in young colon cancer patients and controls. Am J Hum Genet 63:749-759, 1998
Viel A, Genuardi M, Capozzi E, et al: Characterization of MSH2 and MLH1 mutations in Italian families with hereditary nonpolyposis colorectal cancer. Genes Chromosomes Cancer 18:8-18, 1997
Wijnen J, Khan PM, Vasen H, et al: Majority of hMLH1 mutations responsible for hereditary nonpolyposis colorectal cancer cluster at the exonic region 15-16. Am J Hum Genet 58:300-307, 1996
Liu T, Tannergard P, Hackman P, et al: Missense mutations in hMLH1 associated with colorectal cancer. Hum Genet 105:437-441, 1999
Muller W, Burgart LJ, Krause-Paulus R, et al: The reliability of immunohistochemistry as a prescreening method for the diagnosis of hereditary nonpolyposis colorectal cancer (HNPCC): Results of an international collaborative study. Fam Cancer 1:87-92, 2001
Lindor NM, Burgart LJ, Leontovich O, et al: Immunohistochemistry versus microsatellite instability testing in phenotyping colorectal tumors. J Clin Oncol 20:1043-1048, 2002
Hopper JL. Application of genetics to the prevention of colorectal cancer, in Senn HJ, Morant R (eds): Recent Results in Cancer Research, Vol 166. Berlin, Germany, Springer, 2005, pp 17-33(Melissa C. Southey, Mark )
Centre for Molecular, Environmental, Genetic, and Analytical Epidemiology, The University of Melbourne, Parkville
Genetic Technologies Limited, Kew
Peter MacCallum Cancer Centre, East Melbourne
The Australian Red Cross Blood Service, Melbourne
Department of Anatomical Pathology, Western Hospital, Footscray
Department of Colorectal Medicine and Genetics, The Royal Melbourne Hospital, Melbourne
Cancer Epidemiology Centre, The Cancer Council, Carlton, Victoria
ACCFR Molecular Cancer Epidemiology Laboratory, Queensland Institute of Medical Research, Queensland, Australia
Department of Pathology, McGill University, Quebec, Canada
ABSTRACT
PURPOSE: The relationships between mismatch repair (MMR) protein expression, microsatellite instability (MSI), family history, and germline MMR gene mutation status have not been studied on a population basis.
METHODS: We studied 131 unselected patients with colorectal cancer diagnosed younger than age 45 years. For the 105 available tumors, MLH1, MSH2, MSH6, and PMS2 protein expression using immunohistochemistry (IHC) and MSI were measured. Germline DNA was screened for hMLH1, hMSH2, hMSH6, and hPMS2 mutations for the following patients: all from families fulfilling the Amsterdam Criteria for hereditary nonpolyposis colorectal cancer (HNPCC); all with tumors that were high MSI, low MSI, or that lacked expression of any MMR protein; and a random sample of 23 with MS-stable tumors expressing all MMR proteins.
RESULTS: Germline mutations were found in 18 patients (nine hMLH1, four hMSH2, four hMSH6, and one hPMS2); all tumors exhibited loss of MMR protein expression, all but one were high MSI or low MSI, and nine were from a family fulfilling Amsterdam Criteria. Sensitivities of IHC testing, MSI (high or low), and Amsterdam Criteria for MMR gene mutation were 100%, 94%, and 50%, respectively. Corresponding positive predictive values were 69%, 50%, and 75%.
CONCLUSIONS: Tumor IHC analysis of four MMR proteins and MSI testing provide a highly sensitive strategy for identifying MMR gene mutation–carrying, early-onset colorectal cancer patients, half of whom would have been missed using Amsterdam Criteria alone. Tumor-based approaches for triaging early-onset colorectal cancer patients for MMR gene mutation testing, irrespective of family history, appear to be an efficient screening strategy for HNPCC.
INTRODUCTION
Germline mutations in the DNA mismatch repair (MMR) genes hMLH1and hMSH2, and to a lesser extent hMSH6 and hPMS2, are associated with a high lifetime risk of colorectal cancer and an increased risk of extracolonic cancers.1-3 Recognition of the role of these genes in hereditary cancer came about partly because colorectal cancer in multiple-case families often displayed replication errors (microsatellite instability [MSI]) —a consequence of MMR gene dysfunction.4 To detect and repair mismatches occurring during cell replication, the proteins MSH2 and MSH6 bind to form a heterodimer. Likewise, MLH1 binds with PMS2 to form a heterodimer.
Given that genetic testing for germline mutations is technically challenging and costly, criteria have been developed to help identify the individuals with colorectal cancer who carry a MMR gene germline mutation. These criteria include, but are not limited to, tumor loss of MMR protein expression, tumor MSI, and a strong family history of colorectal cancer or other specific cancers.5,6
To justify the use of these criteria to target individuals for germline testing, their performance in terms of sensitivity, specificity, positive predictive value, and negative predictive value needs to be determined. A recent meta-analysis of clinic-based studies of multiple-case and early-onset families concluded that using the Amsterdam Criteria,6 which are based solely on family cancer history, and a previous version of the Bethesda Criteria,7 which extends the Amsterdam Criteria to include tumor MSI testing, "was inadequately sensitive and specific for the identification of carriers." 8
Several studies have selected patients from family cancer clinics to examine relationships between MMR gene germline mutation, tumor MMR protein expression as measured by immunohistochemistry (IHC), tumor MSI, and family history of colorectal cancer.9-16 One of the largest studies included 48 colorectal cancer patients from families referred to a genetics program.17 All 14 identified carriers of germline mutations in hMLH1 or hMSH2 had high MSI colorectal cancer (sensitivity of high MSI, 100%) but there were an additional 14 high MSI colorectal cancers in individuals for whom no germline mutation was detected (positive predictive value, 50%). IHC testing for MLH1 and MSH2 had corresponding sensitivity and predictive values of 57% and 44%, respectively. For Amsterdam Criteria, the corresponding values were only 71% and 50%, respectively. To the best of our knowledge, the performance of these criteria has not been evaluated on a population basis for individuals with colorectal cancer unselected for family history, yet at increased probability of being carriers because of the presence of early-onset disease.
In this article, we studied a population-based sample of colorectal cancer patients diagnosed at age younger than 45 years and unselected for family cancer history.18 We tested for the expression of MLH1, MSH2, MSH6, and PMS2 proteins in the tumors using IHC, conducted MSI testing, and performed germline mutation detection in hMLH1, hMSH2, hMSH6, and hPMS2.
METHODS
The Victorian Colorectal Cancer Family Study is a population-based, case-family study of early-onset colorectal cancer conducted in Australia during 1993 to 1997.18 Approval for the study was obtained from the ethics committees of The University of Melbourne and The Cancer Council of Victoria. All participants provided written informed consent for participation in the study.
Participants
Eligible patients comprised adult men and women living in the Melbourne metropolitan area who were younger than age 45 years when diagnosed with a histologically confirmed, first primary adenocarcinoma of the colon or rectum (International Classification of Diseases, 9th revision, rubrics 153 and 154, respectively) during the period from July 1, 1992, to September 30, 1996. Patients were identified through the population-based Victorian Cancer Registry that receives statutory notifications from all hospitals and pathology laboratories. A letter requesting permission to approach the patients was sent to the attending doctors of a random selection of 222 patients, followed by letters to the patients seeking participation. Attrition due to death (12.6%), refusal (doctor, 10.4%; patient, 12.6%), or because the patient moved and was not located (5.4%) resulted in 131 consenting to participate (59.0% of those eligible). The median time between diagnosis and interview was 9 months.
Data Collection
Patients answered a risk factor questionnaire and completed a family pedigree by identifying all first- and second-degree relatives and reporting on their vital status and cancer histories during an in-person interview conducted in their homes. Each case patient was then asked to obtain permission from his or her adult relatives for us to approach them for their participation. Relatives answered the same risk factor questionnaire during a telephone interview, and similarly were asked about the family's cancer history. Verification of all reports of cancers in family members was sought through cancer registries, hospital records, treating clinicians, and death certificates (see Jenkins et al18 for details). Patients were also asked to donate a small blood sample for genetic testing.
Families were categorized as meeting the Amsterdam Criteria if all of the following criteria were met: at least three relatives were diagnosed with cancer of the colon, rectum, endometrium, small bowel, ureter, or renal pelvis; at least one of these was a first-degree relative of at least two others; relevant cancers were diagnosed in at least two successive generations; at least one of those cancers was diagnosed at age younger than 50 years; and familial adenomatous polyposis was excluded. This definition has also been referred to as Amsterdam Criteria II.6
Tumor Collection
Invasive tumor samples of primary colorectal adenocarcinoma were obtained from hospitals and private pathology laboratories for 118 patients (90%). Six did not consent to release tissue to the study and two laboratories had not agreed to release the remaining seven samples at the time of this study. All tumor samples were reviewed by a pathologist who directed the microdissection.
Tumor Expression of MMR Genes
The expression of MLH1, MSH2, MSH6, and PMS2 was assessed by IHC on 3-μm, formalin-fixed, paraffin-embedded sections using standard methods, monoclonal antibodies MLH1 (Pharmingen clone G168-728), MSH2 (Oncogene clone FE11), MSH6 (BD Transduction Laboratories clone 44), and PMS2 (Pharmingen clone A16-4) on a DAKO autostainer.19 Normal colonic epithelium adjacent to tumor and lymphocytes served as built-in positive controls. A GI pathologist scored the tumors as positive when nuclear staining in tumor tissue was present or otherwise as negative when the nuclear stain was absent.
Tumor MSI
MSI testing was performed on invasive tumor cells microdissected from 5-μm sections of paraffin-embedded archival tumor tissue stained with 1% methyl green. DNA extracted from histologically normal cells microdissected from colonic or lymph node tissue, or DNA extracted from peripheral-blood lymphocytes, was used to control the assay as previously described.20,21 Ten microsatellite markers were assessed: three dinucleotide repeats (D5S346, D17S250, and D2S123) and seven mononucleotide repeats (BAT-25, BAT-26, BAT-40, MYB, TGF?RII, IGFIIR, and BAX). For this study, the degree of instability in each tumor was scored as stable (MS stable), low (low MSI), and high (high MSI) when 0 to 1, 2 to 5, and 6 to 10 markers were identified as unstable, respectively.21 An assessment of MSI was not successful for 13 (12%) tumor samples because of technical reasons related to tumor DNA quality, which left 105 tumors tested.
Germline MMR Mutations
All exonic and flanking intronic sequences of the hMLH1, hMSH2, hMSH6, and hPMS2 genes were screened for germline mutations using sequencing approaches, except for exon 4 of hMSH6, which was screened in eight overlapping fragments using denaturing high-performance liquid chromatography. Confirmation of putative mutations identified via denaturing high-performance liquid chromatography or sequencing was sought using an independent polymerase chain reaction for direct automated sequencing. Variants were defined to be deleterious if they could be predicted to produce (or were known to produce) a shortened or truncated protein product, or were missense mutations that have been reported previously to be deleterious. The Multiplex Ligation-Dependent Probe Amplification assay22 (MRC-Holland, Amsterdam, the Netherlands) to detect large genomic alterations in hMLH1 and hMSH2 was performed on samples from 10 patients who had tumors lacking at least one MMR protein expression and for which no previous mutation had been identified by sequencing. Break points were characterized by performing a long-range polymerase chain reaction encompassing the predicted genomic alteration, cloning, and sequencing.
Mutation testing was conducted for all patients with one or more of the following characteristics: a family history that fulfilled the Amsterdam Criteria for hereditary nonpolyposis colorectal cancer (HNPCC); having a tumor that was high MSI, low MSI, or that lacked expression of at least one MMR protein; and presence in a random sample of 23 patients selected from those who had tumors that were MS stable and did not lack expression of any MMR protein.
Statistical Analysis
Sensitivity, specificity, and positive predictive and negative predictive values were estimated,23 and exact 95% CIs of these proportions were calculated using STATA software, release 8.0 (STATA Corp, College Station, TX). Because we did not conduct germline MMR testing on all patients who had tumors that were MS stable and IHC normal, we could not calculate sensitivity and specificity directly. Sensitivity is a/(a + c + ), where a is the observed number of carriers who were positive for the tumor test, c is the observed number of carriers who were negative for the tumor test, and is the number of carriers in the phenotypically negative patients who were not tested for germline mutations. Specificity was estimated as (d – )/(d – + b), where d is the observed number of noncarriers who were negative for the tumor test, and b is the observed number of noncarriers who were positive for the tumor test.
We assumed that the proportion of carriers in these untested, phenotypically negative patients was the same as that in the phenotypically negative patients who were tested for germline mutations. This turned out to be zero; therefore, was zero. We then conducted sensitivity analyses that recalculated sensitivity and specificity by supposing one or a few of the untested phenotypically negative patients were carriers. The probability that there are d = k carriers among m untested phenotypically negative patients, given we observed no carriers in the tested phenotypically negative patient group (n = 23), is [(n + 1)m!(n + m – k)!]/[(n + m + 1)!(m – k)!] (k = 0,1...,m [see Appendix]).
RESULTS
Germline Mutation Carriers
Table 1 indicates that deleterious mutations were detected for 18 patients (17% of the 105 patients for whom tumor material was tested for MSI and MMR protein expression); nine in hMLH1 (50% of all mutations), four in hMSH2 (22%), four in hMSH6 (22%), and one in hPMS2 (6%). A number of other well-documented common variants that have not been classified previously as deleterious were also detected (data not shown). Most carriers were male (83%) compared with half of noncarriers (48%; P = .007). The mean age at diagnosis for carriers (36 years) was younger than that for noncarriers (41 years; P = .0005).
Of the 18 mutations detected, 15 (83%) were predicted to produce a truncated protein product (four were alterations in splice site regions, seven were frameshift mutations, two were nonsense mutations, and two were large deletions). Of the remaining three, one (hMSH2 P622L) was a missense mutation known to be deleterious,24 and another (hMLH1 1846delAAG) was an inframe deletion known to be deleterious.25-27 The third was an inframe deletion, previously unreported (hMLH1 987delCAT [C330]), that we considered deleterious for the following reasons: for the patient with this mutation, the mother was a carrier and had stomach cancer diagnosed at age 43 years and colon cancer diagnosed at age 55 years, the maternal grandfather (deceased and untested) had gastric cancer diagnosed at age 35 years, a maternal uncle (deceased and untested) had cancer of the duodenum diagnosed at age 46, a maternal uncle (deceased and untested) had stomach cancer diagnosed at age 48, and the only carrier in siblings was a brother unaffected at age 29 years.
Table 2 indicates that the sensitivity of IHC was 100%, given that all of the carriers' colorectal cancers lacked expression of at least one MMR protein consistent with the gene mutated in the germline. All but two of the hMLH1, hMSH2, and the hPMS2 mutation carriers' tumors clearly lacked expression of both proteins of the relevant dimer. For the tumor from the carrier of a large genomic deletion encompassing exons 4 and 5 of hMLH1, there was no nuclear staining of MLH1 but high cytoplasmic staining present in biopsy and surgical specimens; however, the staining in normal colon tissue was equivocal. We classified this patient as not expressing MLH1. For the carrier of the large genomic deletion encompassing exon 15 of hMLH1, a biopsy sample had nuclear staining for MLH1 in the tumor (some focal cells were negative) as well as normal colon tissue and was thus classified as expressing MLH1. All tumors from the four hMSH6 mutation carriers lacked expression of MSH6 only. Therefore, five of the 18 MMR gene mutation carriers (28%) expressed MLH1 and MSH2 yet lacked expression of MSH6 or PMS2.
All tumors from the 13 hMLH1 and hMSH2 mutation carriers were high MSI and all tumors from the four hMSH6 mutation carriers were low MSI. We classified the tumor from the hPMS2 mutation carrier as MSI stable, although due to technical difficulties caused by tumor DNA quality, only four of the 10 microsatellite markers were scored for instability. Therefore, 72% of carriers were high MSI (sensitivity, 72%) and 94% were either high MSI or low MSI.
Of the carriers, nine were members of Amsterdam Criteria families (sensitivity, 50%); three (17%) were from patients with a family history of colorectal cancer, but not sufficient to fulfill the Amsterdam Criteria; and six (33%) were from patients with no known family history of colorectal cancer.
IHC
Table 1 shows that, of the 105 tumors tested, 26 (19%) lacked expression of any of the MMR proteins, and of these, 18 had a deleterious germline mutation in one of the four MMR genes (positive predictive value, 69%; Table 2). Of the 79 patients with tumors that expressed all proteins, no mutations were detected in the 33 who were tested.
MSI Testing
Table 1 shows that, of the 105 tumors tested for MSI, 18 (17%), 16 (15%), and 71 (68%) were assessed to be high MSI, low MSI, and MS stable, respectively. Of the 18 patients with a high MSI tumor, 13 were found to carry deleterious germline mutations in an MMR gene (all hMLH1 or hMSH2; positive predictive value, 72%). Of the other five patients with a high MSI tumor in which no germline mutation was detected, three lacked the expression of both MLH1 and PMS2, and two lacked the expression of both hMSH2 and MSH6. One belonged to a family meeting the Amsterdam Criteria, one had some family history of colorectal cancer (a paternal grandmother with cancer of the rectum diagnosed at age 64 years), and three had no reported family history of colorectal cancer.
Of the 16 patients with a low MSI tumor, four were found to carry a deleterious MMR gene mutation; all were hMSH6 carriers. Of the other 12, one lacked MLH1 and PMS2 expression, and one lacked MSH2 and MSH6 expression. Therefore, of 34 patients with a tumor that exhibited high MSI or low MSI, 17 had a germline mutation in an MMR gene (positive predictive value, 50%). One had a family history meeting the Amsterdam Criteria, four had a family history of colorectal cancer not sufficient to fulfill the Amsterdam Criteria, and seven reported no family history.
Of the 71 patients with MS stable tumors, 25 were tested for germline mutations, revealing one hPMS2 mutation whose tumor lacked both PMS2 and MLH1 expression. There was also one tumor that lacked PMS2 expression. Of the 70 MS stable tumors without a known deleterious germline mutation, one had a family history meeting the Amsterdam Criteria, 24 had a family history of colorectal cancer not sufficient to fulfill the Amsterdam Criteria, and 45 reported no family history.
Family History
Of the 105 patients, 12 (11%) had a family history that met the Amsterdam Criteria; of these, nine had a germline mutation in one of the MMR genes (positive predictive value, 75%). Thirty-two patients (30%) had a family history of colorectal cancer not sufficient to fulfill the Amsterdam Criteria; of these, three (9%) had a germline mutation. Sixty-one patients (58%) had no family history of colorectal cancer in three generations; of these, six (10%) had a germline mutation (negative predictive value of family history, 90%).
DISCUSSION
We found that IHC testing of the expression the four MMR proteins and MSI testing are both highly sensitive for MMR gene mutation status in early-onset colorectal cancer patients. Half of the carriers in our sample would have been missed if mutation testing had been confined solely to those whose cancer family history met the Amsterdam Criteria for HNPCC.
We did not test for germline mutations in 46 of our 105 patients who did not belong to an Amsterdam Criteria family or have any molecular tumor characteristics suggestive of MMR dysfunction (specifically, loss of protein expression or MSI). For our initial calculation of sensitivities and specificities, we assumed that these 46 early-onset patients were not carriers; this was supported by the fact that sequencing the MMR genes of 23 randomly chosen patients who did not have these phenotypic characteristics of a carrier failed to identify a single carrier. Given this, it is unlikely that many (if any) carriers were missed; the probability that the number of carriers in the 46 untested patients is 0, 1, and 2 is .33, .22, and .15, respectively. Therefore, we recalculated the sensitivity and specificity assuming that one carrier had been missed in the 46 patients that had no suggestion of being a carrier (2%), and was incorrectly classified as a noncarrier with an MS stable and IHC normal tumor. We found that the high sensitivity and specificity of IHC and MSI decreased little. For example, the sensitivity became 95% for lack of expression of any MMR protein, 89% for high MSI or low MSI, and 42% for Amsterdam Criteria family history.
Our finding of a high sensitivity for IHC and MSI is supported by data from nine studies reporting on a total of 105 patients with colorectal cancer with known germline mutations in hMLH1 or hMSH2 who were subsequently tested for both MSI and IHC for MLH1 and MSH2.9-17 In these clinic-based series there were only two reported carriers of hMLH1 or hMSH2 mutations with MS stable tumors that expressed both MLH1 and MSH2.13,28 Such outlier observations may be due to failure in laboratory technique,29 but it is theoretically possible that mutations may destroy enzymatic activity while preserving the immunoreactivity and stability of proteins. It is also remotely possible that some cancers in carriers were phenocopies (ie, coincidental cases of colorectal cancer not caused by or associated with the germline MMR mutation and therefore not exhibiting loss of MMR protein expression or MSI). Despite our exhaustive screening, it is possible that carriers could have been missed because we did not screen for large genomic deletions in hMSH6 or hPMS2, and sequencing is not infallible.
The advantages and disadvantages of targeting mutation testing based on IHC, MSI, or family history are summarized in Table 3. On the basis of our data, using IHC alone as a basis for targeting mutation testing would result in all MMR gene carriers being tested for mutations. If we had not conducted IHC testing for all four proteins we would have missed detecting evidence for some carriers (eg, by not conducting IHC testing for MSH6 or PMS2, we would have failed to identify as likely carriers the two patients with the large deletions in hMLH1 and all four patients with hMSH6 mutations—almost one third of all carriers). Similarly, Christensen et al9 determined that IHC testing for PMS2 identified 23% more hMLH1 carriers than testing for MLH1 alone. On the basis of the positive predictive value of 69%, for every two carriers detected, one noncarrier would have been tested. Importantly, IHC testing for all four MMR proteins indicates which MMR genes are mutated in the germline, and this can decrease the cost of genetic testing.
If instead, mutation testing was targeted based on tumor MSI status, our data would suggest that although nearly all MMR gene mutation carriers would have been tested, for every carrier detected, one noncarrier would have been tested. Our extended microsatellite panel for MSI testing, which contains seven mononucleotide repeats, successfully differentiated the hMLH1 and hMSH2 mutation carriers (all high MSI) from the hMSH6 mutation carriers (all low MSI). This would not have been possible using the standard National Cancer Institute panel (data not shown), which tests five microsatellites only and is commonly used to determine who should be tested for germline mutations in all of the MMR genes.5 A panel of 10 markers currently is also being used in the world's largest colorectal cancer family study, the Colon Cancer Family Registry.30
Although the MSI testing performed according to the Revised Bethesda guidelines for HNPCC5 would have identified almost all MMR mutation carriers, one would have had to test for germline mutations in all four MMR genes to do so. Using IHC of MMR protein expression to guide mutation testing would have identified all mutation carriers, with approximately the same sensitivity, but with the advantage that by testing for all four proteins, the observed pattern would have identified correctly the specific genes to be tested. In contrast, family history alone was not nearly as sensitive as IHC or MSI for targeting mutation testing. Half were from families that satisfied the Amsterdam Criteria (including the patient diagnosis); one sixth were from patients with a family history of colorectal cancer not meeting the Amsterdam Criteria; and one third were from patients with no known family history of colorectal cancer.
There are practical difficulties with all three approaches. IHC and MSI (with microdissection) require a pathologist, at least to select sections for staining. MSI testing needs to be conducted in a molecular laboratory, is more expensive, and does not provide gene-specific information. Access to normal tissue, which is required for interpretation, can be restricted in retrospective studies. Obtaining a complete and accurate family history is expensive and time consuming.
These extensive molecular and genetic analyses of a population-based series reveal that the traditional approach to mutation testing, focused initially on the family cancer history of opportunistically identified multiple-case families and nonsystematic referral of patients meeting the Bethesda Criteria, lacks the efficiency for identifying mutation carriers that a systematic assessment of all early-onset patients identified through normal clinical practice could provide.31 We would argue that identifying mutation carriers would be better facilitated by a pathologist or an appropriately trained technician who would, as part of the description of the phenotypic features of the tumor, conduct an IHC examination of tumor for loss of any of the four MMR proteins in early-onset colorectal cancers. Given that the IHC findings do not constitute definitive germline findings, it is arguable that the genetic counseling process, including consent processes, could be delayed until IHC testing suggests a likelihood of the existence of a germline mutation. After genetic counseling and informed consent, germline genetic testing could then be directed toward one or two of the MMR genes, rather than all four.
The difficulty with incorporating genetic counseling before IHC testing is the burden of consent required of both doctor and patient at a time when the patient's attention is focused on concerns of self-preservation. Furthermore, it could be argued that no matter how statistically correlated to germline information, IHC measures biologic expression and does not provide definitive germline information, which can only come from mutation testing. There are equally or more cogently genetic diagnoses made by pathologists on operative specimens where such presurgical consent for the reporting of a genetic diagnosis is never required, but has equally relevant implications for a germline diagnosis (eg, surgical pathological diagnosis of familial adenomatous polyposis).
The proposed approach will lead to a more efficient identification of individuals at truly high genetic risk of cancer.31 It is important to recognize that the details of how this regimen might be implemented would depend a great deal on local factors. Appropriate cancer surveillance procedures for relatives of identified carriers would also need to be implemented. To achieve this there is a continuing need for research on the efficacy and optimization of surveillance procedures and the identification of environmental, lifestyle, and genetic factors that modify risk in these genetically at-risk individuals. An approach built around IHC testing alone as the first step is likely to be much more cost effective for detecting mutation carriers than one based on family history alone, or one based on MSI testing alone, although the relative cost efficiencies would need to be determined for each local situation depending on resources, skills, infrastructure, and other considerations.31
In conclusion, for early-onset colorectal cancer, IHC testing of tumors for four MMR proteins is a strong predictor of germline MMR mutation status for particular gene(s). It is more efficient, in terms of the extent of genetic testing, than MSI testing and much more efficient than using family history. Tumor-based strategies for triaging colorectal cancer patients for MMR gene mutation testing and subsequent predictive testing may be an efficient way to identify individuals with hereditary colorectal cancer due to an MMR gene mutation (ie, with HNPCC), as well as their relatives who carry the family-specific mutation. Both sets of individuals are at a high colorectal cancer risk, and their lives may be saved or prolonged by intensive surveillance.
Appendix
Suppose the binomial parameter initially has a uniform prior distribution. If we observe n trials with j successes, the posterior distribution is then
Here D ( a, b) is the Dirichlet distribution with parameters a, b. If this is now taken as the prior for , the marginal probability of observing k successes during a subsequent m trials is just
Note that this is a truncated version of the hypergeometric distribution.
As a special case, if there are no successes in the first n trials, the posterior probability of no successes in the next m is (n + 1)/(n + m + 1); the probability of k successes from the subsequent m is
Authors' Disclosures of Potential Conflicts of Interest
Although all authors completed the disclosure declaration, the following author or immediate family members indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.
Acknowledgment
We thank Judith Maskiell and the research interviewers; Gillian Dite and the data management staff; Graham Byrnes for statistical advice; and Sabar Napaki, Deon Venter, and Jane Armes for pathology input. We also thank the men and women who participated in this study.
NOTES
Supported by grants from the National Health and Medical Research Council (Australia) and the Victorian Health Promotion Foundation.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Mitchell RJ, Farrington SM, Dunlop MG, et al: Mismatch repair genes hMLH1 and hMSH2 and colorectal cancer: A HuGE review. Am J Epidemiol 156:885-902, 2002
Aaltonen LA, Peltomaki P, Leach FS, et al: Clues to the pathogenesis of familial colorectal cancer. Science 260:812-816, 1993
Kolodner RD, Tytell JD, Schmeits JL, et al: Germ-line msh6 mutations in colorectal cancer families. Cancer Res 59:5068-5074, 1999
Thompson E, Meldrum CJ, Crooks R, et al: Hereditary non-polyposis colorectal cancer and the role of hPMS2 and hEXO1 mutations. Clin Genet 65:215-225, 2004
Umar A, Boland CR, Terdiman JP, et al: Revised Bethesda Guidelines for hereditary nonpolyposis colorectal cancer (Lynch syndrome) and microsatellite instability. J Natl Cancer Inst 96:261-268, 2004
Vasen HF, Watson P, Mecklin JP, et al: New clinical criteria for hereditary nonpolyposis colorectal cancer (HNPCC, Lynch syndrome) proposed by the International Collaborative group on HNPCC. Gastroenterology 116:1453-1456, 1999
Rodriguez-Bigas MA, Boland CR, Hamilton SR, et al: A National Cancer Institute Workshop on Hereditary Nonpolyposis Colorectal Cancer Syndrome: Meeting highlights and Bethesda guidelines. J Natl Cancer Inst 89:1758-1762, 1997
Kievit W, de Bruin JH, Adang EM, et al: Current clinical selection strategies for identification of hereditary non-polyposis colorectal cancer families are inadequate: A meta-analysis. Clin Genet 65:308-316, 2004
Christensen M, Katballe N, Wikman F, et al: Antibody-based screening for hereditary nonpolyposis colorectal carcinoma compared with microsatellite analysis and sequencing. Cancer 95:2422-2430, 2002
Debniak T, Kurzawski G, Gorski B, et al: Value of pedigree/clinical data, immunohistochemistry and microsatellite instability analyses in reducing the cost of determining hMLH1 and hMSH2 gene mutations in patients with colorectal cancer. Eur J Cancer 36:49-54, 2000
Hoedema R, Monroe T, Bos C, et al: Genetic testing for hereditary nonpolyposis colorectal cancer. Am Surg 69:387-392, 2003
Marcus VA, Madlensky L, Gryfe R, et al: Immunohistochemistry for hMLH1 and hMSH2: a practical test for DNA mismatch repair-deficient tumors. Am J Surg Pathol 23:1248-1255, 1999
Salahshor S, Koelble K, Rubio C, et al: Microsatellite Instability and hMLH1 and hMSH2 expression analysis in familial and sporadic colorectal cancer. Lab Invest 81:535-541, 2001
Scartozzi M, Bianchi F, Rosati S, et al: Mutations of hMLH1 and hMSH2 in patients with suspected hereditary nonpolyposis colorectal cancer: Correlation with microsatellite instability and abnormalities of mismatch repair protein expression. J Clin Oncol 20:1203-1208, 2002
Thibodeau SN, French AJ, Roche PC, et al: Altered expression of hMSH2 and hMLH1 in tumors with microsatellite instability and genetic alterations in mismatch repair genes. Cancer Res 56:4836-4840, 1996
Ward R, Meldrum C, Williams R, et al: Impact of microsatellite testing and mismatch repair protein expression on the clinical interpretation of genetic testing in hereditary non-polyposis colorectal cancer. J Cancer Res Clin Oncol 128:403-411, 2002
Wahlberg SS, Schmeits J, Thomas G, et al: Evaluation of microsatellite instability and immunohistochemistry for the prediction of germ-line MSH2 and MLH1 mutations in hereditary nonpolyposis colon cancer families. Cancer Res 62:3485-3492, 2002
Jenkins MA, Baglietto L, Dite GS, et al: After hMSH2 and hMLH1: What next? Analysis of three-generational, population-based, early-onset colorectal cancer families. Int J Cancer 102:166-171, 2002
Young J, Simms LA, Biden KG, et al: Features of colorectal cancers with high-level microsatellite instability occurring in familial and sporadic settings: Parallel pathways of tumorigenesis. Am J Pathol 159:2107-2116, 2001
Armes JE, Trute L, White D, et al: Distinct molecular pathogeneses of early-onset breast cancers in BRCA1 and BRCA2 mutation carriers: A population-based study. Cancer Res 59:2011-2017, 1999
Southey MC, Young MA, Whitty J, et al: Molecular pathologic analysis enhances the diagnosis and management of Muir-Torre syndrome and gives insight into its underlying molecular pathogenesis. Am J Surg Pathol 25:936-941, 2001
Schouten JP, McElgunn CJ, Waaijer R, et al: Relative quantification of 40 nucleic acid sequences by multiplex ligation-dependent probe amplification. Nucleic Acids Res 30:e57, 2002
Knottnerus JA, van Weel C, Muris JW: Evaluation of diagnostic procedures. BMJ 324:477-480, 2002
Liu B, Parsons RE, Hamilton SR, et al: HMSH2 mutations in hereditary nonpolyposis colorectal cancer kindreds. Cancer Res 54:4590-4594, 1994
Farrington SM, Lin-Goerke J, Ling J, et al: Systematic analysis of hMSH2 and hMLH1 in young colon cancer patients and controls. Am J Hum Genet 63:749-759, 1998
Viel A, Genuardi M, Capozzi E, et al: Characterization of MSH2 and MLH1 mutations in Italian families with hereditary nonpolyposis colorectal cancer. Genes Chromosomes Cancer 18:8-18, 1997
Wijnen J, Khan PM, Vasen H, et al: Majority of hMLH1 mutations responsible for hereditary nonpolyposis colorectal cancer cluster at the exonic region 15-16. Am J Hum Genet 58:300-307, 1996
Liu T, Tannergard P, Hackman P, et al: Missense mutations in hMLH1 associated with colorectal cancer. Hum Genet 105:437-441, 1999
Muller W, Burgart LJ, Krause-Paulus R, et al: The reliability of immunohistochemistry as a prescreening method for the diagnosis of hereditary nonpolyposis colorectal cancer (HNPCC): Results of an international collaborative study. Fam Cancer 1:87-92, 2001
Lindor NM, Burgart LJ, Leontovich O, et al: Immunohistochemistry versus microsatellite instability testing in phenotyping colorectal tumors. J Clin Oncol 20:1043-1048, 2002
Hopper JL. Application of genetics to the prevention of colorectal cancer, in Senn HJ, Morant R (eds): Recent Results in Cancer Research, Vol 166. Berlin, Germany, Springer, 2005, pp 17-33(Melissa C. Southey, Mark )