What's New in Hereditary Colorectal Cancer?

Although autosomal-dominant syndromes account for a relatively small proportion of colorectal cancers (CRCs), their clinical and biological importance is considerable. Hereditary forms of CRC often present at an early age and therefore contribute disproportionately to the loss of years of life. Nevertheless, screening, early diagnosis, and surgical intervention have been shown to reduce mortality in the case of both familial adenomatous polyposis (FAP) and hereditary nonpolyposis colorectal cancer (HNPCC).1,2 To manage these conditions, it is axiomatic that one must first achieve a reliable diagnosis. This edition of Archives of Pathology & Laboratory Medicine includes 4 articles that examine diagnostic and phenotypic features of FAP and HNPCC. Taken together, these articles reinforce the central importance of the pathologist in the diagnosis of these conditions, as well as in furthering our understanding of the clinical manifestations and underlying mechanisms. Indeed, it is in complex areas such as hereditary neoplasia that the pathologist plays an essential role in linking clinical medicine with basic science. Between them, the articles in this edition provide a comprehensive overview of the main types of hereditary CRC. This editorial attempts to place the articles within the context of some recent insights and developments.

Classic FAP is diagnosed readily by the demonstration of many hundreds if not thousands of colorectal adenomas. Nevertheless, certain mutations in the adenomatous polyposis coli (APC) tumor suppressor gene may result in an attenuated form of the disease. The number of adenomas may fall below the diagnostic threshold of 100, and the adenomas tend to be flat and are found more frequently in the proximal colon. The onset of malignancy is later than in the classic form of FAP.3 Attenuated FAP (AFAP) is the subject of 1 of the case reports by Ionescu et al.4 This study emphasizes the difficulties that may be encountered in reaching a diagnosis of AFAP.

In FAP, it is accepted that most, if not all, adenomas are initiated when the wild-type APC allele (second copy) is inactivated somatically by either mutation or loss. It has been widely assumed that adenomas occurring sporadically are also initiated by APC mutation. However, several studies have failed to demonstrate a high frequency of APC mutation in microadenomas and small sporadic tubular adenomas of the colorectum5,6 or have instead implicated genes such as KRAS7 and β-catenin.8 Therefore, the consistent demonstration of loss of heterozygosity near the APC locus on chromosome 5q in multiple adenomas from the same subject would be indicative of a germline mutation in APC. However, in the study by Ionescu et al,4 it was not possible to show consistent loss of polymorphic markers flanking the APC gene. Undaunted by this negative finding, Ionescu et al4 were able to show the same pattern of loss of an intragenic single nucleotide polymorphism in exon 15 of the APC gene in 6 of 6 adenomas. This would suggest that relatively small deletions were accounting for the loss of the second APC allele. This finding encouraged Ionescu et al to screen for a germline mutation in APC, and they were rewarded by the discovery of a truncating mutation of codon 161 in the 5′ region of APC. Regions 5′ to codon 158 have been linked to AFAP. One other patient has been described with a Q161X mutation, and this patient also had features of AFAP.9 Therefore, the AFAP phenotype is likely to be associated with mutations extending beyond codon 158. The patient described in the study by Ionescu et al4 had more than 100 adenomas, which would place her within the category of classic FAP. However, the figure of 100 adenomas as the cutoff value between AFAP and classic FAP should not be taken too rigidly. By the age of 40, most patients with FAP have many hundreds if not thousands of colorectal adenomas. Age and modifying genetic influences will impact on the phenotype associated with a particular APC mutation.

Although one should not expect too much from polyp numbers as a guide to specific genetic diagnosis, it is nevertheless useful to obtain an approximation of absolute numbers of polyps in a colectomy specimen. Even a rough count carries far more information than a term such as multiple. However, counting polyps is surprisingly difficult. In the case of FAP, one can assume that the vast majority of polyps are adenomas, count the numbers within a small area, and extrapolate to achieve an approximation for the entire colon. Even then, pinhead-sized polyps may go uncounted. In the case of tens or hundreds of polyps, one cannot be certain that all the polyps are adenomas. There is a natural temptation to sample the larger polyps for histologic diagnosis. However, it is worth sampling a subset of the smaller and less conspicuous polyps as well, since these could turn out to be hyperplastic polyps. The presence of numerous hyperplastic polyps would make a diagnosis of AFAP less likely, since APC mutation is not associated with hyperplastic polyps. By contrast, the condition hyperplastic polyposis is associated with the presence of traditional adenomas, serrated adenomas, and mixed polyps, as well as hyperplastic polyps, and is probably underdiagnosed.10 

Another type of polyposis that must be distinguished from both FAP and AFAP is MYH polyposis.11 The phenotype of this autosomal-recessive condition can mimic both FAP and AFAP. The causative gene was discovered when it was found that somatic APC mutations in adenomas obtained from polyposis patients with no detectable germline mutation in APC were more frequently guanine to thymine (G to T) transversions than would be expected by chance.11 ,MYH was one of a few candidate base-excision repair genes that could explain such a narrow spectrum of mutations and turned out to be the causative agent. Most polyps in MYH polyposis are stated to be adenomas. However, it is conceivable that hyperplastic polyps would also be numerous within the background mucosa, since inactivation of MYH is also associated with activating G to T mutations in KRAS, a gene that is closely associated with hyperplastic polyps.12 

The list of extracolonic neoplasms in FAP has grown steadily over the years. In the West, most malignancies in the upper gastrointestinal tract are focused on the duodenum and periampullary region, where adenomas are initiated by inactivation of APC (as in the colon). The lack of a markedly increased risk for pancreatic cancer in FAP patients is explained by the lack of a key role for APC in ductal pancreatic neoplasms. Gupta and Mazzara13 document the finding of grade 3 pancreatic intraepithelial neoplasia (PanIN) in a pancreaticoduodenectomy specimen resected from a patient with FAP and upper gastrointestinal neoplasia. The demonstration of inactivation of the second APC allele in DNA extracted from the PanIN would be evidence that the finding was not a mere coincidence. This was not undertaken, however, and additional studies are required to confirm the link between FAP and pancreatic neoplasia.

While adenomas do occur in HNPCC and slightly more frequently than in the general population, the lack of a distinct premonitory stage delayed the recognition of HNPCC as a distinct clinicopathologic entity and continues to make the diagnosis of the condition difficult as compared with FAP. Henry Lynch made a major contribution to our understanding of HNPCC by assembling information on families and characterizing the clinical and pathologic features of the syndrome.14 By the 1980s, cancer family registries or cancer genetics clinics in many different countries were systematically enrolling families that fulfilled the characteristics described by Lynch. The Amsterdam Criteria (AC) were introduced in 1991 to achieve a standardized approach to the investigation of HNPCC families across different registries.15 The original AC can be remembered very simply as the 3-2-1 rule, that is, 3 first-degree relatives across 2 generations with 1 being younger than 50 years. Although it was hoped that the AC would be relatively specific for the syndrome described by Lynch, they were not intended to serve as gold-standard diagnostic criteria, but merely as an approach to achieving some uniformity across different studies.15 

With the discovery of the DNA mismatch repair genes and their causative role in the etiology of HNPCC/Lynch syndrome, it became possible to refine the diagnostic criteria. While for technical reasons it may not be possible to identify a germline mutation in every family, it became clear that HNPCC/Lynch syndrome families were characterized by a very definite set of clinical, pathologic, and molecular features that were all explained by loss of DNA mismatch repair proficiency. A most important marker in this regard was the demonstration of numerous mutations in repetitive sequences of DNA known as microsatellites.16 It was possible to show DNA microsatellite instability (MSI) in the colorectal adenomas and carcinomas of affected family members, as well as in extracolonic neoplasms.17 When antibodies to the DNA mismatch repair proteins became commercially available, it was possible to show that loss of nuclear expression of DNA mismatch repair proteins was tightly correlated with MSI status.18 The demonstration that loss was restricted to a particular DNA mismatch repair protein, such as MLH1 or MSH2, meant that the screening for a germline mutation could be focused on the gene in question. In this issue of Archives, Burgart19 has provided a very practical guide to the testing of DNA mismatch repair status in colorectal carcinoma.

It has been known for many years that CRCs may cluster in families. Although the AC may appear quite stringent, it is possible to identify families that fulfill the criteria but in which the CRCs are DNA mismatch repair proficient. In these families, the CRCs are less likely to be multiple, right-sided, poorly differentiated, mucinous, or DNA diploid,20 while adenomas are more numerous but show less advanced histology.21 Importantly, the risk of both CRC and extracolonic cancers typical of HNPCC is lower in members of AC-positive families in which there is no evidence of a DNA mismatch repair defect.22 Burgart19 points out that 40% of AC-positive families (HNPCC families) have no evidence of a heritable defect of DNA mismatch repair, and that only the 60% with such evidence should be regarded as having the Lynch syndrome. This still may lead to confusion as many regard HNPCC and Lynch syndrome as synonymous. Despite the fact that AC-positive families are clearly heterogeneous, clinicians have continued to consider the AC (or closely related criteria) as diagnostic for HNPCC and, therefore, Lynch syndrome. This unfortunate state of affairs has come about because the AC are very simple to apply. However, this uncritical use of a very limited set of clinical features has resulted in families being given unwarranted diagnostic labels and has generated considerable confusion in the research community. If it has been shown that tumors in a family with clustering of CRC are not explained by DNA mismatch repair deficiency, then that family should not be labeled as having either Lynch syndrome or HNPCC. The designation familial colorectal cancer type X has been suggested as an interim term for this group.22 

The discovery of the underlying mechanism for HNPCC/Lynch syndrome did not result in an immediate revision of the diagnostic criteria. This is because DNA MSI is not limited to HNPCC, but occurs in 10% to 15% of sporadic MSI-positive CRCs.23,24 In the latter, the underlying mechanism is methylation of the promoter region of MLH1, leading to silencing of the gene.25 Sporadic MSI-positive CRCs have various clinical, pathologic, and molecular features in common with HNPCC, but differ in a number of important respects. Apart from being more age-related and more common in females, sporadic MSI-positive CRCs are characterized by frequent mutation of BRAF and widespread DNA methylation.26–28 By contrast, somatic mutations of APC, β-catenin, and KRAS are more frequent in HNPCC.26 At the morphologic level, sporadic MSI-positive CRCs are more likely than CRCs in HNPCC/Lynch syndrome to be mucinous, to have a serrated architecture, and to be associated with serrated polyps, while they are less likely to show tumor budding (de-differentiation at the invasive margin).26 Recently, sporadic MSI-positive CRCs and HNPCC have been distinguished by gene expression array technology.29 

Despite the differences between sporadic MSI-positive CRC and HNPCC, the distinction may not always be straightforward. A particular reason for this difficulty lies with the discovery of a new mechanism for inactivating a copy of the MLH1 gene in the germline, namely, epimutation or hemiallelic methylation of MLH1.30–32 It is likely (although this has still not been proven) that germline hemiallelic methylation and germline mutation of MLH1 will predispose to early-onset CRC through identical mechanisms and genetic alterations. There is one crucial difference, however. There is no evidence that epimutations are inherited, and therefore the resulting early-onset CRCs will be sporadic. The fact that some familial clustering of CRC has been associated with epimutation is likely to be due to ascertainment bias. That is, epimutations have been sought within subjects registered in cancer family clinics and not through population-based screening of early-onset cases. Colorectal carcinomas with age-related acquired DNA methylation appears to be less frequent in Southeast Asia (eg, Japan and Korea).33 The overrepresentation of sporadic high-level MSI (MSI-H) CRCs due to germline hemiallelic methylation of MLH1 in the East could account for the molecular differences among sporadic MSI-H CRCs between East and West.33 

The Bethesda guidelines (BGs) differ from the AC in a number of subtle but important respects. While the AC were introduced in order to stratify high-risk families registered in cancer genetic clinics, the BGs were aimed at the population at large and were linked very specifically with the testing of tumors for DNA MSI.34 Clearly, testing for MSI is both expensive and time-consuming and is not universally available in diagnostic laboratories. The BGs were designed to identify the subset of CRCs that should be tested for MSI status. A particular aspect of the design of the BGs was to exclude MSI-positive CRCs that were likely to be sporadic. In other words, those CRCs that met the BG standards and were subsequently shown to be MSI-positive would be likely to be from patients with HNPCC/Lynch syndrome. This is because both age at diagnosis of CRC and positive family history were factored into the BGs. While the BGs are less stringent than the AC, by adding MSI testing they should become more specific, as well as more sensitive. The AC lack sensitivity because they are slanted toward the identification of large, multigenerational families in which penetrance of particular mutations is likely to be high. Hereditary nonpolyposis colorectal cancers may sometimes present as sporadic CRC for such reasons as nonpaternity, adoption, denial or ignorance of family history, nonpenetrance, small or geographically dispersed families, and rare new mutations.

It was necessary to revisit the BGs for a number of reasons.35 These included problems of specificity with the National Cancer Institute's panel of DNA microsatellite markers,36 the increasing availability of immunohistochemistry, the inadequacy of the original histologic features linked with MSI status, and the need to identify older-onset cases of HNPCC presenting in the community setting. With respect to the panel of microsatellite markers, it was shown that some of the nonmononucleotide markers, such as MYCL1 and D2S123, were mutated at surprisingly high frequency in CRCs that were DNA mismatch repair proficient.37 This meant that low-level MSI CRCs that happened to have mutation of 2 non-mononucleotide markers could be labeled inappropriately as MSI-H. While it has been contended that immunohistochemistry is not as sensitive as MSI testing (see Burgart19), it is likely that the staining for MSH6 and PMS2 proteins, as well as MLH1 and MSH2, will identify the vast majority of CRCs from subjects with HNPCC.38–40 In this regard, the importance of PMS2 has been strongly reinforced recently.41–43 Burgart19 points out that pathogenic missense mutations may result in the expression of a functionally deficient protein. In such cases, there may be loss of PMS2 expression even though mutant MLH1 protein is expressed.41 Loss of PMS2 expression may therefore serve as a marker for germline mutation of MLH1. Loss of PMS2 expression may also indicate a germline mutation in the PMS2 gene itself.42,43 Despite the possibility of the occasional false-negative immunohistochemical test, one could substantially reduce costs by limiting MSI testing to those cases in which immunostaining is either inconclusive or yields an apparently normal expression pattern at odds with a high index of suspicion of HNPCC/Lynch syndrome (see below).

In the revised BGs the key histologic features of MSI-H CRCs are specified as follows: tumor-infiltrating lymphocytes, a Crohn-like lymphoid reaction, mucinous or signet ring cell histology, and a medullary-type growth pattern.35 Since these features can be recognized relatively easily by pathologists, they provide an important opportunity for diagnosing later-onset cases of HNPCC, that is, presenting up to the age of 60 years. This component of the revised BGs means that the pathologist has an important role to play in the diagnosis of patients with HNPCC/ Lynch syndrome who do not present in multigenerational families. In this issue of the Archives, Gologan et al44 have shown that at least 1 of the preceding histologic features was present in 80% of MSI-H CRCs that occurred within a population-based series of early-onset CRCs. Although the sensitivity and negative predictive value of histologic features were high, the specificity and positive predictive value were low.

The significance of the findings of Gologan et al44 is 2-fold. First, the revised BGs can identify more CRCs with MSI-H and therefore more patients with Lynch syndrome.39 Second, the histologic features can be used to exclude cases from further testing for DNA mismatch repair status. For example, in CRC presenting in a patient younger than 50 years, one could restrict testing for DNA mismatch repair status to those cases with the appropriate histology. This approach may seem cavalier, yet by adding the absence of “dirty necrosis” to tumor-infiltrating lymphocytes and mucinous histology, morphology was shown to be 100% sensitive for MSI-H status in a recent study.45 If one chose to undertake immunohistochemistry in all CRCs presenting in patients younger than 50 years, then one could at least use the histology features to limit the use of MSI testing to the small subset of CRCs with histologic features consistent with MSI-H status, despite normal expression of all 4 DNA mismatch repair proteins. To put this more simply, a sporadic CRC presenting in a patient younger than 50 years is highly unlikely to be from a patient with HNPCC if both the immunohistochemical findings are normal and the histologic features of MSI-H status are lacking.

Based on the preceding premises, the pathologist is able to identify most cases of HNPCC presenting in the community setting. Loss of expression of MLH1 can be caused by mechanisms other than germline mutation and should not be equated with a diagnosis of HNPCC/Lynch syndrome. While most instances of loss of MSH2 expression will be explained by a germline mutation of MSH2, somatic loss may account for a few cases. The pathologist is testing phenotype rather than genotype, and reports should therefore raise the possibility of a diagnosis of HNPCC rather than issue a certain diagnosis. The suggestion that pathologists should not undertake immunohistochemical staining for DNA mismatch repair genes without the patient's prior consent is misguided. The pathologist is a professional who is trained to work up a specimen fully and to provide the clinician with the best available diagnostic evidence. By contrast, counseling patients on the basis of limited clinical information is undesirable. In failing to undertake a full diagnostic workup, the pathologist might be considered to be in breach of duty of care. The fact that the pathologist can contribute to the diagnosis of a serious genetic disorder by using a strategy that is both simple and cost-effective represents an important medical advance.

A number of HNPCC-like families, including some that met the AC, were recently distinguished from HNPCC/Lynch syndrome on the basis of various clinical, pathologic, and molecular features. These features included variable MSI status in the CRCs of affected family members; frequent mutation of BRAF and DNA methylation in both CRCs and polyps; a background of serrated polyps, including hyperplastic polyposis in up to 10% of cases; frequent glandular serration in CRCs; and a high female-male ratio among affected subjects.46 These families will constitute at least some of the familial colorectal cancer type X cases noted above. There is evidence of an increased risk of cancer in the relatives of subjects with CRCs showing DNA methylation.47 Another study did not confirm this finding, but introduced a major bias by excluding families that met a clinical definition of HNPCC.48 As mentioned, not all families meeting a clinical definition of HNPCC have a germline mutation in a DNA mismatch repair gene. Therefore, a genetic predisposition to DNA methylation could underlie the serrated pathway syndrome.46 

In summary, the recent literature, including the 4 articles published in this edition of Archives of Pathology & Laboratory Medicine, not only expands our knowledge of hereditary CRC, but also highlights the fundamental role of the pathologist in achieving the correct diagnosis.