| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Commentaries |
From the Department of Pathology, Weill Medical College of Cornell University, New York, New York
The prevailing paradigm of cancer as a genetic disorder, arising from the acquisition and subsequent selection of mutations in cancer-causing genes, has propelled much of cancer research over the past two decades and has expanded our understanding of cancer at a phenomenal pace. Cancer-causing genes have been identified at an ever-increasing rate and fall into three major categories: oncogenes, tumor suppressor genes, and the more recently described mutator genes. The development of a malignant neoplasm usually involves mutations in one or more genes of each category, resulting in abnormal function of the encoded protein through changes in the level of gene expression (eg, amplification) or in the characteristic of the protein (eg, missense mutation or translocation). These fundamental changes disturb normal cellular functions (eg, cell cycle control, apoptosis, DNA mismatch repair) that, in sum, drive the neoplastic process. In turn, the perturbation of normal cellular functions has ramifications for the expression status of countless genes, ultimately resulting in the complexity and heterogeneity inherent to malignant neoplasms. The genetic model of cancer would suggest that only those changes in expression caused by a genetic event are critical to the development of the neoplastic process; all others are simply a consequence of the transformed state. However, recent studies of DNA promoter hypermethylation have called this belief into question. Data are accumulating that support a role for this epigenetic alteration in the inactivation of genes that play a causative role in the development of human tumors.
DNA methylation is an epigenetic modification essential to mammalian development.1 This modification occurs as a postreplicative addition of methyl groups to cytosine residues.2 Although this change is reversible by demethylation and, as such, is not a fixed, heritable alteration, patterns of methylation in normal cells are relatively stable. However, aberrations in DNA methylation including global hypomethylation and localized regions of hypermethylation have been consistently detected in human tumors and transformed cell lines.3-5 Although the significance of the changes remains controversial, numerous studies support a role for methylation in tumorigenesis. Much of the recent excitement about methylation in cancer has centered around the detection of hypermethylated CpG islands in the promoters of several tumor suppressor genes (eg, p16, von Hippel-Lindau, E-cadherin) and the hMLH1 mutator gene in tumors that lack intragenic mutations of the respective genes.6-9 The majority of CpG islands in normal cells are unmethylated and the acquisition of methylation at these sites in tumors correlates with loss of gene expression, suggesting that promoter hypermethylation is an alternative mechanism for inactivation of genes that contribute to tumor development.10 Despite these profound associations, skepticism remains in regard to an active role for methylation in tumorigenesis. This is due in part to our incomplete understanding of the factors that control methylation and the overall consequences of changes in methylation patterns on chromatin structure and gene expression. In this issue of The American Journal of Pathology, Esteller and colleagues provide support for an important role of promoter hypermethylation in endometrial tumorigenesis.11 As with most previous studies, theirs does not present direct experimental evidence, but through a series of elegant studies on uterine endometrioid carcinoma (UEC) and its hyperplastic precursor lesions, the authors reveal hMLH1 promoter hypermethylation as an early event in the development of endometrial carcinoma.
A subset of sporadic uterine endometrioid carcinomas, the most common type of endometrial cancer, demonstrates a molecular phenotype commonly referred to as microsatellite instability (MSI).12-14 MSI is detected as alterations in the size of microsatellite DNA sequences in DNA derived from tumors compared to DNA from matched normal tissue.15 This phenotype was first detected in a subset of sporadic colorectal cancer and subsequently shown to be consistently present in tumors from patients with hereditary nonpolyposis colorectal cancer (HNPCC).16-18 The sequence of events that followed these discoveries was an extraordinary example of the synergy between basic and biomedical science. In brief, soon after MSI was identified in primary human tumors a group working on DNA mismatch repair in Saccharomyces cerevisiae reported instability of a dinucleotide repeat (GT) in yeast strains harboring mutations in DNA mismatch repair genes.19 This provided a connection between phenotype and genotype in a simple eukaryote with possible implications for human tumors. This collision of events expedited the cloning of the first human DNA mismatch repair gene, hMSH2, using degenerate primers based on sequences of microbial DNA mismatch repair genes. Furthermore, it was shown that affected HNPCC family members carried germline mutations in hMSH2 and that the wild-type copy was absent in their colorectal tumors.20,21 Soon thereafter, three additional human mismatch repair genes were cloned (hMLH1, hPMS1, and hPMS2) and many HNPCC families were identified with germline mutations in hMLH1 and only rarely in hPMS1 and hPMS2.22,23 Approximately 90% of all HNPCC kindred carry mutations in either hMLH1 or hMSH2. These studies not only defined the cause of an inherited family cancer syndrome, they also provided experimental evidence for what is referred to as the mutator phenotype of cancer and opened a new avenue of cancer research. For an excellent discussion of this topic the reader is referred to a Commentary in the previous volume of the Journal.24
Because endometrial carcinoma is the most common non-colorectal cancer in women of HNPCC families, sporadic endometrial carcinomas were tested for MSI and found in a subset of tumors. Notably, the mutational analyses of the four known mismatch repair genes that were shortly to follow were not forthcoming.25-28 As the manuscript by Esteller et al notes, of the 61 MSI-positive sporadic UECs analyzed to date only 4 (<7%) contain mutations (3 in hMSH2 and 1 in hMLH1) in one of the four known mismatch repair genes. Studies of sporadic MSI-positive colorectal cancer have produced similar findings.29 These data led investigators to suggest that additional unidentified DNA mismatch repair genes might be responsible for MSI in the majority of sporadic tumors. Recently, however, studies on colorectal, gastric, and endometrial carcinoma have shown a statistically significant correlation between MSI and hypermethylation of the hMLH1 promoter.30-32 For endometrial carcinoma, 71 to 92% of MSI-positive carcinomas have hypermethylation of the hMLH1 promoter.32-34 Conversely, such methylation is rare in MSI-negative tumors. In addition, several studies using an in vitro approach have produced further support for epigenetic silencing of hMLH1 in the disruption of mismatch repair and subsequent development of microsatellite instability. These investigators have shown that there are MSI-positive colorectal and endometrial carcinoma cell lines that demonstrate hMLH1 hypermethylation and lack hMLH1 expression. When these lines are treated with the demethylating agent 5-azacytidine, hMLH1 becomes expressed, the MSI phenotype disappears, and functional mismatch repair, as measured by strand-specific mismatch repair, is restored.35,36 Furthermore, when the drug is removed, hMLH1 is silenced and the MSI phenotype is re-established.35 These studies provide strong support for the idea that hMLH1 promoter hypermethylation is associated with inactivation of a gene with a defined role in the development of human tumors.
However, several aspects of the investigations raise some doubt about a causative role for hMLH hypermethylation in tumor development, including the presence of hMLH1 promoter hypermethylation in a small number of tumors (sporadic and HNPCC related) with mutations in either hMLH1 or hMSH2; MSI-negative cases with hMLH1 promoter hypermethylation; and increased tendency of MSI-positive cell lines to methylate endogenous and exogenous DNA.37,38 There are several plausible explanations for these findings that do not exclude the importance of hMLH1 promoter hypermethylation in tumorigenesis.
In the case of associated mutations in hMLH1, it has been suggested that promoter hypermethylation may be associated with the silencing of the wild-type allele of hMLH1.10 As with tumor suppressor genes, mutator genes require the inactivation of both alleles to demonstrate the full phenotype: in this circumstance, to abolish mismatch repair activity and promote MSI. The mechanism thought to be most common for inactivation of nonmutated alleles of recessive cancer genes is loss of heterozygosity (LOH) of the region of the chromosome that harbors the gene in question.39 In at least one colorectal cancer with both mutation and hypermethylation of hMLH1, loss of heterozygosity of chromosome 3p (the location of hMLH1) was lacking, consistent with methylation as an alternative to the inactivation of the wild-type allele.36 Although this argument is convincing, it does not explain the rare occurrence of hMLH1 promoter methylation in conjunction with hMSH2 mutation. Investigators have proposed that in this setting hMLH1 hypermethylation may be monoallelic or heterogeneously present within the tumor cell population.2 The current methods used for determination of methylation status are polymerase chain reaction-based and are, therefore, likely to detect low levels of methylation. It has been reported that one such method is capable of detecting 1 methylated allele in 1000 unmethylated alleles.40 Given the heterogeneity of malignant neoplasms it is highly likely that in some tumors there is occasional methylation of the CpG island of hMLH1 without complete loss of hMLH1 activity and therefore of MSI. This explanation may also explain the occasional MSI-negative tumor that reveals hMLH1 promoter hypermethylation.
Finally, the propensity of MSI-positive cell lines to methylate DNA fuels the argument that hypermethylation of hMLH1 may simply reflect an underlying global methylation abnormality. This chicken-and-egg argument is difficult to combat directly, but data exist that investigators use to support hMLH1 hypermethylation as the cause and not the consequence of MSI. For example, a paper by Herman et al reports that 22% of HNPCC-related tumors demonstrate p16 hypermethylation.36 This is a much lower incidence than seen in MSI-positive sporadic tumors, where it is 60%, and is comparable to that of MSI-negative sporadic tumors.37 This finding implies that a mismatch repair deficiency per se does not necessarily promote hypermethylation of hMLH1. Unfortunately, it does not directly address the question of the temporal relationship between hMLH1 promoter hypermethylation and loss of hMLH1 expression.
Through the determination of the methylation status of the hMLH1 promoter in endometrial carcinoma and its hyperplastic precursors, Esteller et al have provided keen insight into this relationship. The authors assayed 27 endometrial carcinomas and 116 endometrial hyperplasias of different severity for hMLH1 promoter hypermethylation using methylation-specific polymerase chain reaction, an assay developed and used extensively by this group. In addition, in all of the carcinomas and a subset40 of the hyperplasias the MSI status was determined. They confirm the association of hypermethylation and MSI in endometrial cancer by showing that 11 of 12 (91%) MSI-positive carcinomas show methylation, whereas all of 15 MSI-negative tumors lack methylation. They go on to show that LOH of chromosome 3p21.3 is absent in all 11 tumors, suggesting that there is biallelic inactivation by hMLH1 methylation.
Notably, the authors detect methylation in 8 of 116 hyperplasias. Although methylation is absent in 62 simple hyperplasias, it is present in 1 of 33 complex hyperplasias without atypia and in 7 of 21 complex atypical hyperplasias (CAH). CAH is morphologically similar to, but distinguished by lack of stromal invasion from, well differentiated endometrioid carcinoma and is thought to be the direct precursor of the invasive disease. As such, CAH and endometrioid carcinoma are often found concurrently in uteri removed for either diagnosis. Perhaps the most interesting finding of this study is the relationship of hypermethylation, MSI, and the association of hyperplasia and carcinoma.
Of the 8 hypermethylated hyperplasias, 4 displayed MSI and all of these were associated with a concurrent hypermethylated, MSI-positive carcinoma. The 4 remaining cases lacked MSI and only 2 were associated with a carcinoma. Notably, one of the carcinomas was MSI- and methylation-positive and one was MSI-negative and lacked methylation. In the latter case, the authors found an additional focus of CAH that was methylation-negative, suggesting that the invasive component arose from the methylation-negative hyperplasia. Several key points are highlighted by these results. First, the fact that all of the MSI-positive hyperplasias have associated carcinoma is consistent with previous reports. This finding supports a hypothesis first proposed in the adenoma-carcinoma sequence of MSI-positive colorectal carcinoma stating that the temporal window for the development of carcinoma may be relatively short in the presence of MSI. Second, the presence of methylation in the absence of MSI in two hyperplasias with associated carcinomas suggests that methylation of hMLH1 can occur before the development of MSI. In one case, the associated carcinoma is MSI- and methylation-positive, suggesting that it arose from the hyperplasia with a methylated hMLH1 gene in conjunction with the development of MSI. In the other case, the associated carcinoma is both MSI- and methylation-negative, and the authors propose that it arose from the hyperplastic component that lacked methylation. This might be a plausible alternative if the hyperplasia had methylation of only one hMLH1 allele or the methylation were heterogeneously present. Finally, two of the MSI-negative hyperplasias with methylation lacked associated carcinoma, perhaps reflecting monoallelic/heterogeneous methylation.
In sum, these findings demonstrate that hMLH1 methylation can precede the development of endometrial carcinoma. However, to determine definitively the consequence of such methylation, additional studies are clearly warranted. To resolve the issue of the relationship of the methylation-positive hyperplasias to the associated carcinomas, the analysis of additional molecular markers altered early in endometrial tumorigenesis (eg, PTEN and K-ras mutations, microsatellites) may be useful. Alternatively, it may be that the definitive answers about the role of promoter hypermethylation in cancer will come unexpectedly from experiments not yet formulated. For today, however, all of the data seem to suggest that promoter hypermethylation is an epigenetic alteration associated with the inactivation of a subset of genes that are central to tumor development.
Footnotes
Address reprint requests to Lora Hedrick Ellenson, M.D., Department of Pathology, Weill Medical College of Cornell University, 1300 York Avenue, New York, NY 10021. E-mail: lhellens{at}mail.med.cornell.edu
Accepted for publication September 20, 1999.
References
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
