The only well-established form of inherited endometrial carcinoma is associated with HNPCC. Over the past several years, the genes responsible for the majority of families that meet the clinical criteria for HNPCC have been identified and cloned. Furthermore, the HNPCC families in which linkage studies have been informative demonstrate linkage to loci subsequently found to harbor one of the known genes. For these reasons, the discussion of the genetic loci and the specific genes involved in inherited endometrial carcinoma are combined.
HNPCC is discussed in great detail in Chap. 32. Therefore, it is presented only briefly here, with an emphasis on endometrial carcinoma. HNPCC is the most common hereditary family cancer syndrome and is transmitted as an autosomal dominant trait. It is clinically defined by these criteria: (a) at least three relatives with colorectal cancer with one a first-degree relative of the other two; (b) the presence of tumors in at least two successive generations; and (c) one family member affected by colorectal cancer before the age of 50.13 Endometrial carcinoma is not a required criterion for the clinical definition of HNPCC. However, the International Collaborative Group on HNPCC has recognized that if these criteria are strictly followed, families with a high incidence of endometrial carcinoma, as well as colorectal carcinoma, would be excluded. Although the actual percentage of all endometrial carcinoma due to HNPCC is not known, it has been shown that the cumulative incidence of endometrial carcinoma in women belonging to HNPCC kindreds is 20 percent by age 70, in contrast to 3 percent in the general population.5
Genetic Loci and Specific Genes
As previously stated, endometrial carcinoma is the most common extracolonic tumor that occurs in HNPCC families. Linkage analysis of several large HNPCC kindreds identified susceptibility loci on the short arms of chromosomes 2 and 3.14,15 Informative linkage studies have determined linkage to either of these two chromosomal arms in the majority of HNPCC families. However, in a small number of families linkage to either of these regions is lacking.
Over the past several years four genes have been identified that cause HNPCC in most of the kindreds meeting the clinical criteria of this inherited family cancer syndrome. The cloning of these genes resulted from the propitious coincidence of several different lines of scientific investigation, including studies aimed at identifying genes that play a role in human tumors, and others aimed at understanding fundamental molecular processes in microbial organisms.
Investigators were analyzing microsatellites (small repetitive sequences) in DNA isolated from tumors arising in HNPCC family members and found alterations in the length of microsatellite DNA sequences when compared to germ-line DNA from the same patients.16 Other investigators reported a similar molecular phenotype in approximately 20 percent of sporadic colorectal tumors.17,18 This molecular phenotype was referred to as microsatellite instability or replication errors. Microsatellite instability and its role, if any, in the neoplastic process was unclear. Shortly after the discovery of microsatellite instability in both sporadic and HNPCC-associated colorectal carcinomas, a study was published demonstrating that mutations in DNA mismatch repair genes led to a 100- to 700-fold increase in the instability of simple dinucleotide repeat sequences in the simple eukaryote S. cerevisiae.19 This observation, along with previous work in both E. coli and S. cerevisiae, provided a crucial connection between microsatellite instability and mutations in DNA mismatch repair genes. This connection led to the ultimate identification of human DNA mismatch repair genes and opened a new avenue of cancer research.
In a very short time, four human homologues of microbial DNA mismatch repair genes were cloned. At a somewhat later date a fifth gene involved in DNA mismatch repair, GTPB, was cloned, but it will not be discussed here.20,21 The four human DNA mismatch repair genes known to cause HNPCC are named hMSH2, hMLH1, hPMS1, and hPMS2, in keeping with their microbial homologues, and are located on chromosomes 2p, 3p, 2q, and 7q, respectively.22-26 The physical maps of hMSH2 and hMLH1 have shown that their locations correlate with chromosomal loci determined to have genetic linkage to HNPCC. The linkage and physical mapping data provided additional information suggesting that DNA mismatch repair genes play a role in the pathogenesis of neoplasms arising in HNPCC kindreds. Subsequent studies documented germ-line mutations in one of these four human mismatch repair genes in affected members of most HNPCC kindreds, with the hMSH2 and hMLH1 genes accounting for the vast majority. At present, there is not a reported difference in the frequency of endometrial carcinoma among the HNPCC kindreds that carry mutations in the different genes.
The contribution of how microsatellite instability and defects in the DNA mismatch repair system contributed to tumorigenesis remained unproved. In microorganisms, the DNA mismatch repair system was known to detect and repair mispaired bases introduced during replication of the cellular genome. Furthermore, microbial organisms lacking a functional DNA mismatch repair system have a marked increase in the rate at which mutations accumulate. In mammalian cells, the DNA mismatch repair system has been much less well characterized, but it is thought to have a similar function to its microbial counterpart. Microsatellite DNA sequences in both humans and microorganisms are prone to undergo alterations in their length (explaining their highly polymorphic nature) during DNA replication. Therefore, it follows that microsatellite DNA sequences might demonstrate numerous alterations in the absence of an intact DNA mismatch repair system. This suggests that microsatellite instability may simply serve as a marker of an increased rate of mutation caused by an underlying defect in the DNA mismatch repair system. This led directly to the notion that lack of a functional mismatch repair system would result in an increased rate of mutations in oncogenes and tumor-suppressor genes, thus predisposing cells to the accumulation of mutations now thought to be a cornerstone of the neoplastic process. In support of this idea, studies have demonstrated an increase in the rate of point mutations in an expressed gene (HPRT) in a mismatch repair-deficient mammalian cell line. The identification of mutations in DNA mismatch repair genes in human tumors created a new class of cancer-causing genes called mutator genes.
The high frequency of endometrial carcinoma in HNPCC families indicates that the genes responsible for HNPCC are involved in the pathogenesis of endometrial carcinoma in this setting. A review of the literature does not provide a straightforward analysis of mutations in women with endometrial carcinoma belonging to HNPCC families. Clearly, further studies of HNPCC-associated endometrial carcinomas are needed.
Sporadic Endometrial Carcinoma
As alluded to earlier, the identification and characterization of the genetic loci and specific genes involved in endometrial tumorigenesis have been hampered by the inadequate recognition of the distinct types of endometrial carcinomas. Much of the problem is related to the classification scheme, described earlier, being initially described in 1983 and only recently gaining widespread acceptance. In addition, uterine serous (Type II) carcinomas are relatively rare, comprising approximately 10 percent of all sporadic endometrial carcinomas. Consequently, and understandably, many of the studies have not clearly stated the type (or types) of endometrial carcinomas that were included. Additionally, many studies that have classified the tumors lack significant numbers to enable the results of the different tumor types to be assessed independently. When possible, this chapter discusses the molecular genetics of sporadic endometrial carcinoma in the context of these two tumor types.
Over the past several years a number of loss of heterozygosity (LOH) studies have attempted to locate regions of the genome that may harbor tumor-suppressor genes that play a role in endometrial tumorigenesis. In combining the results of the major studies, LOH has been detected on these chromosomes: 1, 3, 6, 8p, 9p, 9q, 10q, 11, 13, 14q, 15, 16q, 17p, 18p, 18q, 20, 21, and 22q.28-32 A review of the literature reveals substantial variation in the regions that have been found to undergo LOH in endometrial carcinoma, and there are only several regions from this long list that have shown significant LOH in more than one study. These include loci on chromosomes 3p, 10q, 17p, and 18q. The 3p LOH is striking, as several candidate tumor-suppressor genes and the hMLH1 gene map to this chromosome. The target(s) of 3p LOH have not yet been determined in endometrial carcinoma. Two separate groups of investigators have reported between 35 and 40 percent LOH of a region of 10q, and one group has suggested that there may be two discrete regions of 10q that undergo LOH.29,31 A range of 9 to 35 percent of endometrial carcinomas has been reported to show 17p LOH. A recent study of uterine serous carcinoma detected LOH of 17p, specifically 17p13.1, in 100 percent of informative cases.33 Because most of the LOH studies did not specify the tumor type, it will be of interest in the future to determine the percentage of each type that have 17p LOH. LOH of chromosome 18q has been found in three studies, all of which included tumors from Japanese women, with the highest reported frequency of 33 percent.28,30,32 Other studies have failed to detect 18q LOH, including two studies confined to the analysis of tumors from American women, as well as one exclusively of Japanese women. Although 14q LOH has been identified in only one study, the association of 14q LOH with a poor prognosis led the authors to suggest that 14q LOH may indicate aggressive tumor behavior.30 Interestingly, the authors note that several of the tumors with 14q LOH were uterine serous carcinomas. Further studies are necessary to confirm the possible association of 14q LOH and aggressive behavior of endometrial carcinoma.
As is easily imaginable, the variability among the LOH studies has hindered the identification of novel regions of the genome that may be important in the development of endometrial carcinoma. The reason(s) for the variability between studies are uncertain, but there are many possible explanations. For example, the polymorphic markers used in the various studies are not identical, and if relatively small deletions are responsible for LOH in endometrial carcinoma, the critical regions may only be detected with very specific markers. Furthermore, many of the studies have failed to carefully report the histologic types of the tumors analyzed. If the histologic types, which reflect the distinct categories of endometrial carcinoma, have different underlying molecular genetic alterations, the results of such studies may depend heavily on the types of tumors studied. This point is of further interest, as many studies have included tumors from Japanese and American patients and there is some evidence suggesting differences in the molecular basis of endometrial carcinomas in these two populations.
The discussion of the specific genes is divided into three sections according to the general classification of genes currently recognized as cancer-causing genes: (a) mutator genes, (b) oncogenes, and (c) tumor-suppressor genes.
Due to the association of endometrial carcinoma and HNPCC, presumably sporadic cases of endometrial carcinoma were analyzed for instability of microsatellite DNA sequences. In several studies, microsatellite instability was detected in approximately 20 percent of endometrial tumors.34-36 Given the association of microsatellite instability and mutations in human DNA mismatch repair genes, it seemed likely that mutations in these genes may be involved in the development of sporadic endometrial carcinomas that displayed microsatellite instability. A mutational analysis of four of the known DNA mismatch repair genes (hMSH2, hMLH1, hPMS1, and hPMS2) found that only a small number of sporadic endometrial carcinomas with microsatellite instability had mutations in one of these four genes.37 In addition, mutations of hMSH2 and hMLH1 have been found in two endometrial carcinoma cell lines (HEC59 and AN3CA) that demonstrate microsatellite instability.38 These findings are similar to those seen in cases of microsatellite instability-positive sporadic colorectal cancers. The recent literature has found that the vast majority of microsatellite instability-positive sporadic endometrial carcinomas demonstrate hypermethylation of the hMLH1 promoter. This, in turn, is thought to be related to lack of expression of hMLH1 and the disruption of DNA mismatch repair.38a, 38b
Finally, a recent study found that 34 cases of uterine serous carcinoma failed to demonstrate microsatellite instability.39 The observed difference in the frequency between endometrial and uterine serous carcinoma is statistically significant and provides support for differences in the molecular pathogenesis of the two most common types of endometrial carcinoma.
A number of oncogenes have been studied over the years, yet there are very few that have been found to be altered in a substantial number of endometrial carcinomas. The proto-oncogene recognized as mutated most commonly in endometrial carcinoma is K-ras. It has been shown, in a number of independent studies, to be mutated in 10 to 30 percent of endometrial carcinomas.40-45 K-ras is a member of the ras gene family that consists of three closely related genes (H-ras, K-ras, and N-ras). The H-ras gene was discovered due to its ability to transform an immortalized rodent cell line, and its identification led to the cloning of the two other family members. Each of the ras genes encodes a 21-kDa guanine nucleotide-binding protein (p21) that transduces signals from activated transmembrane receptors to protein kinases that regulate cell growth and differentiation. The oncogenic mutations occur most commonly at codons 12, 13, and 61 and result in a gain-of-function. The mutant ras proteins have a decreased ability to interact with the GTPase-activating protein called ras-GAP, reducing their ability to interact with the GTPase-activating protein ras-GAP, and reducing their ability to hydrolyze guanosine triphosphate (GTP) to guanosine diphosphate (GDP). Hence, the mutant ras protein remains in the GTP-bound or activated state. In endometrial carcinoma, most mutations are found in codon 12. A recent study of American patients, that separated the two types of endometrial carcinoma, found that 11.6 percent of endometrioid carcinomas contained codon 12 mutations, whereas uterine serous carcinomas were all negative for codon 12 mutations.46 The numbers were not statistically significant; however, it suggests that K-ras mutations may be differentially mutated in the different types of endometrial carcinoma. K-ras mutations have also been found in complex atypical hyperplasia (the precursor of endometrioid carcinoma) leading investigators to suggest that K-ras mutations may be a relatively early event in endometrial tumorigenesis.43-45 Investigators have analyzed the association of K-ras mutations with prognosis, but the results have been conflicting.
There are a small number of studies showing alterations in the expression and/or amplification of the HER-2/neu gene in endometrial carcinoma. HER-2/neu is a member of the epidermal growth factor receptor gene family. It encodes a transmembrane tyrosine kinase receptor and is overexpressed in a subset of breast and ovarian cancers. The data on this gene in endometrial carcinoma are limited, but several studies have shown that it is overexpressed in 11 to 59 percent of tumors, and amplified in 14 to 21 percent of tumors.47,48 One study revealed that overexpression and amplification of HER-2/neu were associated with a poor prognosis, and a multivariate analysis indicated that overexpression was an independent prognostic factor.49,50 Independent studies have suggested that overexpression may be more common in uterine serous carcinomas.51
Recently, there have been several studies looking at expression of the bcl-2 gene in endometrial carcinoma and hyperplasia. The bcl-2 gene product prevents cells from undergoing apoptosis and is overexpressed in a number of different types of human tumors. The results of the studies in endometrial carcinoma are contradictory, with some demonstrating increased expression in endometrial carcinomas and others finding it decreased.52,53 However, the results of several studies have found expression in normal proliferative endometrium and an absence of expression in normal secretory endometrium. These results suggest that bcl-2 may play a role in the normal endometrial cycle. Hence, further studies on endometrial carcinoma seem needed to determine if bcl-2 has a role in endometrial tumorigenesis.
As is true in many tumors, the p53 gene has been the most extensively studied gene in endometrial carcinoma. p53 is the prototype tumor-suppressor gene and it is the most frequently mutated gene in human cancers. It encodes a nuclear phosphoprotein with an apparent molecular weight of 53 kDa. For obvious reasons, this gene has been under intensive investigation for many years. Recent studies have begun to elucidate the mechanisms by which p53 controls cell growth (reviewed in reference 54). Briefly, it has been found that p53 expression increases, posttranscriptionally, in response to DNA damage, resulting in a G1/S cell-cycle arrest. It is thought that this arrest gives cells the opportunity to repair the damaged DNA such that mutations are not fixed in the genomic template and, in turn, passed to daughter cells after cell division is complete. It has also been found that elevations in p53 gene expression can lead to apoptosis. Recent data suggest that transcriptional activation of p21WAF1 by p53 is important in the G1/S arrest, but is not essential for apoptosis. Evidently, given its ubiquitous involvement in human tumorigenesis, mutations that inactivate the p53 gene provide a significant growth-promoting affect on many cell types.
Evaluation of p53 in endometrial carcinoma has largely been by immunohistochemistry, and overexpression of the protein has been reported in anywhere from 11 to 45 percent of endometrial carcinomas.55-57 Evaluation of the data is troublesome due to a lack of description of the staining patterns (intensity and percent of cells staining) and the types of tumors analyzed. The staining pattern may be of utmost importance, as it is thought that detection of p53 by immunohistochemistry reflects the presence of mutations in the gene. Many studies have shown that there is considerable variability in staining and that only intense, diffuse staining may accurately predict the presence of mutations. One large study demonstrated that positive staining was more common in high-grade (41.7 percent) than in low-grade (12 percent) tumors, and another study revealed it more frequently in high-stage (41 percent) than in low-stage (9 percent) tumors.56,58 Furthermore, when the tumor types have been separated, a higher frequency of staining is noted in uterine serous carcinomas (66 to 86 percent) as compared to the endometrioid type (Fig. 52-2).33,59 Several studies have shown that overexpression of p53 by immunohistochemistry is an independent prognostic variable, predicting a poor prognosis.57,60
Immunohistochemistry of p53 in uterine serous carcinoma and its precursor endometrial intraepithelial carcinoma. Endometrial intraepithelial carcinoma (EIC) arises abruptly from atrophic endometrium (A) and shows intense positive staining (B). A typical uterine serous carcinoma (C) also shows intense, diffuse staining for p53 protein (D). (Reprinted with permission from Tashiro H, et al.33)
Analyses have also shown a wide range (9.5 to 23 percent) in the frequency of p53 mutations in endometrial carcinoma.59,61 Again, these differences may be due to the types, grades, and stages of tumors analyzed. Many of the mutational studies have consistently shown that mutations are more common in high-grade tumors, and a recent study analyzing only uterine serous carcinomas detected mutations in 90 percent of tumors.33 The strong association of p53 mutations and uterine serous carcinoma may offer an explanation for the prognostic significance of p53 overexpression and its association with a poor outcome.
Many of the p53 studies have focused on the clinical utility of the results. Recent studies suggest that they may also provide meaningful information about the molecular pathogenesis of endometrial carcinoma. As mentioned earlier, a putative precursor of uterine serous carcinoma has been described and p53 immunohistochemical studies revealed positivity in a very high percentage of endometrial intraepithelial carcinoma (Fig. 52-2).62 This finding is in contrast to the very infrequent staining of atypical hyperplasia, the precursor of endometrioid carcinoma. Mutational analyses have shown that mutations in exons 5 to 8 of the p53 gene are present in a majority of endometrial intraepithelial carcinomas (78 percent), suggesting, along with the high frequency of p53 mutations in uterine serous carcinoma, that p53 mutations occur early in the pathogenesis of this tumor type.33 It is reasonable to speculate that early mutation of the p53 gene may be an important determinant of the aggressive biological behavior of uterine serous carcinoma, resulting in the poor outcome of patients with this tumor type.
Several recent studies have shown that mutations in PTEN, a tumor-suppressor gene located on chromosome 10q23.3, are common in endometrial carcinoma.63,64,65 Approximately 40 to 50 percent of endometrioid carcinomas contain PTEN mutations, making it the most frequently mutated gene yet identified in this tumor type. The small number of uterine serous carcinomas analyzed for PTEN mutations are negative; however, before mutations in this tumor type are excluded more cases of serous carcinoma should be analyzed. Interestingly, PTEN mutations are more frequent in microsatellite instability-positive tumors than in those that lack instability.63,64 Although the biological basis of this association is not yet understood it may represent an important finding with regards to the molecular pathogenesis of endometrioid carcinoma. Furthermore, PTEN mutations have been identified in approximately 20 percent of complex hyperplasias, with and without atypia, suggesting that PTEN mutations occur early in the pathogenesis of at least some endometrial carcinomas.66,67 The predicted amino acid sequence of PTEN reveals significant homology to both tensin, a protein located in focal cell adhesions, and tyrosine phosphatases.68 Biochemical studies have since shown that PTEN encodes a dual-specificity phosphatase, and substrates of PTEN are currently under investigation.69 In addition, several studies have implicated a role for PTEN in signal transduction pathways.70,71 In sum, the high frequency of PTEN mutations in endometrial carcinoma and their presence in hyperplastic lesions imply that inactivation of PTEN plays a significant role in its development. Clearly, the role of PTEN in endometrial tumorigenesis will be actively pursued in the near future.
Finally, a recent study found mutations in the β-catenin gene in 13 percent of endometrial carcinomas, and an accumulation of β-catenin protein in 38 percent of endometrial carcinomas.53a This finding is of considerable interest as it suggests a role for the Wnt signaling pathway, a pathway commonly involved in colorectal tumorigenesis, in the development of endometrial carcinoma. Additional studies are needed to determine the significance of this pathway in endometrial tumorigenesis.