|Home | About | Journals | Submit | Contact Us | Français|
The goal of ovarian cancer screening is to detect disease when confined to the ovary (stage I) and thereby prolong survival. We believe this is an elusive goal because most ovarian cancer, at its earliest recognizable stage, is probably not confined to the ovary. We propose a new model of ovarian carcinogenesis based on clinical, pathological, and molecular genetic studies that may enable more targeted screening and therapeutic intervention to be developed. The model divides ovarian cancer into two groups designated Type I and Type II. Type I tumors are slow growing, generally confined to the ovary at diagnosis and develop from well established precursor lesions so-called “borderline” tumors. Type I tumors include low-grade micropapillary serous carcinoma, mucinous, endometrioid, and clear cell carcinomas. They are genetically stable and are characterized by mutations in a number of different genes including KRAS, BRAF, PTEN, and beta-catenin. Type II tumors are rapidly growing, highly aggressive neoplasms that lack well defined precursor lesions; most are advanced stage at, or soon after, their inception. These include high-grade serous carcinoma, malignant mixed mesodermal tumors (carcinosarcomas) and undifferentiated carcinomas. The Type II tumors are characterized by mutation of TP53 and a high level of genetic instability. Screening tests that focus on stage I disease may detect low-grade Type I neoplasms while missing the more aggressive Type II tumors which account for most ovarian cancers. A more rational approach to early detection of “ovarian cancer” should focus on low volume rather than low stage of disease.
In the United States, the incidence of ovarian cancer ranks eighth among cancers (excluding skin cancer) in women, but fifth in terms of age-adjusted mortality (http://www.cancer.org/docroot/home/index.asp). The high mortality rate is generally attributed to its occult development resulting in advanced, widespread disease occurring in approximately 75% of women at diagnosis. However, in about 25% of women disease is confined to the ovary (stage I) and 5 year survival is over 90% compared to 30% for women with advanced disease 1. This observation suggests that diagnosing ovarian cancer when confined to the ovary may improve survival. Accordingly, there has been a concerted effort to develop screening methods for the early detection of stage I ovarian cancer.
The serum assay for the tumor marker CA125, alone or in combination with pelvic or transvaginal ultrasound, continues to be evaluated as a potential screening test for ovarian cancer. At present, screening tests are not recommended for use in the general population and are considered to have limited use in the high risk population due to their insufficient sensitivity and their inability to detect early stage disease.
Evaluating potential screening tests for ovarian cancer has been extremely challenging for several reasons: 1) the failure to identify a histologic precursor lesion or a molecular event that precedes malignant transformation, 2) the small number of true early stage, high-grade carcinomas detected makes it often necessary to make inferences using cases that have advanced disease rather than early stage disease (which may behave differently) when compared to controls, 3) the low prevalence in the general population means that extremely large prospective cohorts over a long time are needed to evaluate the ability of the test to detect preclinical disease and 4) the surgical morbidity associated with a positive screening test and the high disease specific mortality means that a test with high specificity and sensitivity is required (most clinicians consider that a minimum positive predictive value [PPV] of 10% is reasonable).
The received view of ovarian carcinogenesis is that carcinoma begins in the ovary and spreads to adjacent organs in the pelvis and the abdominal cavity before metastasizing to distant sites. This is reflected by the FIGO staging system in which cancer confined to the ovary is stage I, when it has spread to the pelvis it is stage II and when it involves abdominal organs it is stage III. Spread beyond the abdominal cavity is stage IV. This concept of ovarian tumor progression, however, is probably not valid. Moreover, attempts to develop methods of early detection for ovarian cancer have been modeled on cervical cancer screening but that analogy is not appropriate. First, it appears that high-grade ovarian carcinoma spreads to extra-ovarian sites early in its development unlike cervical cancer where the transit time from a precursor lesion to invasive carcinoma is about 10 years. Therefore the time frame for detecting ovarian cancer before it has spread beyond the ovary is very brief. Second, unlike cervical cancer which originates in the cervix, many “ovarian” cancers may not begin in the ovary. For example, it is well known that in women with BRCA mutations the risk of developing ovarian type cancer after prophylactic oophorectomy, although reduced, is not entirely eliminated. A recent prospective study of women with BRCA mutations has calculated that the risk of developing peritoneal carcinoma after prophylactic oophorectomy is 4.3% over a median follow-up period of 3 years (1– 20 years) 2. These peritoneal carcinomas are identical histologically to high-grade ovarian serous carcinoma. Also, the majority of sporadic ovarian cancers that present with extensive disease in the peritoneum, omentum, mesentery and other abdominal organs often display only minimal ovarian involvement suggesting that they may be derived from the peritoneum and involve the ovaries secondarily. Another observation that lends support to the view that not all “ovarian carcinoma” begins in the ovary has come from studies of women with BRCA mutations undergoing prophylactic oophorectomy. Careful examination of the ovaries and fallopian tubes of these women has shown that there are small high-grade serous carcinomas in the fimbria of the fallopian tube and not in the ovary suggesting that in these women, “ovarian cancer” begins in the fallopian tube and spreads to the ovaries 3, 4. Thus, it is likely that a substantial number of what is considered “ovarian cancer” is in fact primary peritoneal or fallopian tube carcinoma. Based on FIGO ovarian cancer staging these tumors are advanced stage at inception.
Correlation of the results of recent molecular genetic studies with clinical and histopathologic findings has led us to propose a new model of ovarian carcinogenesis. In this model all ovarian surface epithelial tumors are divided into two groups designated Type I and Type II. Type I tumors tend to present as stage I, low grade neoplasms that develop slowly from well recognized precursors and behave in an indolent fashion. They include low-grade micropapillary serous carcinoma, mucinous, endometrioid and clear cell carcinoma. In contrast, Type II tumors nearly always present as high-stage, high-grade tumors that are extremely aggressive. Included in this group are high-grade serous carcinoma, malignant mixed mesodermal tumors and undifferentiated carcinomas. In addition to the clinical and pathologic differences, there are molecular genetic differences between these two groups as well 5. The low-grade tumors are relatively genetically stable and are characterized by mutations in a number of genes. For example, in low-grade micropapillary serous carcinoma (the prototypic Type I tumor) and its precursor lesions, atypical proliferative serous tumor, so-called “serous borderline tumor (SBT)” the most well characterized molecular alterations are sequence mutations in KRAS, BRAF and ERBB2 oncogenes. Oncogenic mutations in BRAF, KRAS and ERBB2 result in constitutive activation of the mitogen activated protein kinase (MAPK) signal transduction pathway which plays a critical role in the transmission of growth signals into the nucleus 6 and contributes to neoplastic transformation. Previous studies 7, 8 have demonstrated that KRAS mutations at codons 12 and 13 occur in one third of invasive a low-grade micropapillary serous carcinomas and another one third of SBTs. Similarly, BRAF mutations at codon 600 occur in 30% of low-grade serous carcinomas and 28% of SBTs 7 9, 10. Mutations in ERBB2 occur in less than 5% of these tumors. Mutations in KRAS, BRAF and ERBB2 are mutually exclusive. Therefore mutations in these genes are detected in about two thirds of low-grade micropapillary serous carcinomas and SBTs. Mutations of KRAS and BRAF appear to occur very early in the development of low-grade micropapillary serous carcinoma as evidenced by the demonstration that the same KRAS and BRAF mutations detected in SBTs are detected in the cystadenoma epithelium adjacent to SBTs 11. As compared to SBTs, the cystadenoma epithelium lacks cytologic atypia. These findings suggest that mutations of KRAS and BRAF occur in the epithelium of cystadenomas adjacent to SBTs and are very early events in tumorigenesis, preceding the development of SBTs.
The most common molecular genetic alteration in mucinous borderline tumors (MBTs) and mucinous carcinomas (Type I tumors) is point mutation of KRAS 10, 12, 13. In mucinous carcinoma, morphologic transitions from cystadenoma to MBT, to intraepithelial carcinoma and invasive carcinoma have been recognized for some time and an increasing frequency of KRAS mutations at codons 12 and 13 has been described in cystadenomas, MBTs and mucinous carcinomas, respectively 12–16. In addition, mucinous carcinoma and the adjacent mucinous cystadenoma and MBT share the same KRAS mutation 14 indicating that these tumors have a common lineage and support the view that mucinous carcinomas develop in a step-wise fashion from mucinous cystadenomas and MBTs. Besides KRAS, other genetic alterations in ovarian mucinous tumors have not been described.
In endometrioid borderline tumors (EBTs) and endometrioid carcinomas (Type I tumors), mutation of beta-catenin has been reported in approximately one third of cases 17, 18. Mutations of KRAS and BRAF have also been reported in approximately 10% of endometrioid carcinomas 7, 10, 13, 19–21. Mutation of the tumor suppressor, PTEN, occurs in 20% of endometrioid carcinomas, rising to 46% in those tumors with 10q23 loss of heterozygosity 22. Similar molecular genetic alterations including loss of heterozygosity at 10q23 and mutations in PTEN have been reported in endometriosis, atypical endometriosis and ovarian endometrioid carcinoma in the same specimen 22–27. These molecular genetic findings together with the morphological data demonstrating a frequent association of endometriosis with endometrioid adenofibromas and EBTs adjacent to invasive well-differentiated endometrioid carcinoma provide evidence of step-wise tumor progression in the development of endometrioid carcinoma 28. The critical role of the genetic changes in PTEN and KRAS is highlighted by a recent report showing that inactivation of PTEN and an activating mutation of KRAS are sufficient to induce the development of ovarian endometrioid carcinoma in a mouse model 29. More recently, inactivation of the Wnt/beta-catenin and the PI3K/Pten pathways has been reported to be sufficient to induce endometrioid carcinoma in an engineered mouse model 30.
As compared to other types of ovarian epithelial tumors, the main molecular genetic changes associated with ovarian clear cell borderline tumors and clear cell carcinomas (Type I tumors) remain to be identified. Although several molecular genetic changes have been reported in clear cell tumors, most studies have analyzed a limited number of cases and therefore the true prevalence of those changes is not known. For example, mutations in transforming growth factor-beta receptor type II have been reported in one small study of clear cell carcinomas but rarely in other histologic types of ovarian carcinomas 31. Microsatellite instability is present in endometrioid and clear cell carcinoma but is only rarely detected in serous and mucinous tumors 28, 32, 33. This finding supports the view that endometriosis is the common precursor for both endometrioid and clear cell carcinoma 28. We have tentatively included clear cell carcinoma in the Type I group because clear cell carcinomas typically present as stage I tumors and they are frequently associated with precursor lesions such as endometrioisis and clear cell borderline tumors. Also, like the other Type I tumors, clear cell carcinomas do not demonstrate a high level of genetic instability (Shih I-M, et al, unpublished data). On the other hand, clear cell carcinomas are nearly always high-grade and the frequency of KRAS, BRAF and TP53 is low 10. Accordingly, clear cell carcinomas do not really fit into either Type I or Type II groups.
In contrast to low-grade micropapillary serous carcinomas and other Type I tumors, the only mutation that has been consistently detected in Type II tumors are mutations in TP53 which are very rare in Type I tumors. In addition, Type II tumors are characterized by considerable genetic instability which is not observed in Type I tumors. Most studies have shown that approximately 50–80% of advanced stage, presumably high-grade, serous carcinomas have mutant TP53 34–39. When purified tumor samples are analyzed the frequency of TP53 mutations is over 80% in high-grade serous carcinomas 40. It has also been reported that mutant TP53 is present in 37% of stage I and II high-grade serous carcinomas 41. In fact, in BRCA heterozygote women who had undergone prophylactic oophorectomy microscopic, stage I, high-grade serous carcinomas were identified in the ovaries identical to the usual high-grade serous carcinoma that is advanced stage. In addition, over expression of p53 and mutation of TP53 were found in the microscopic carcinomas as well as in the adjacent “dysplastic” epithelium 42. It is plausible that inherited mutations in BRCA genes compromise DNA repair and predispose to genetic instability that may contribute to the “dysplastic changes”. Although sporadic ovarian carcinomas were not analyzed in this study, the clinical and pathologic features of BRCA-linked ovarian carcinomas and their sporadic counterparts are indistinguishable, suggesting that their histogenesis is similar. Thus, although the molecular genetic findings are preliminary, they suggest that conventional high-grade serous carcinoma, in its very earliest stage resembles advanced stage serous carcinoma at a molecular as well as at a morphologic level.
Similar to high-grade serous carcinoma, other Type II tumors, specifically malignant mixed mesodermal tumors (carcinosarcomas) also demonstrate TP53 mutations in almost all cases analyzed 43, 44 45. This has led investigators to suggest that malignant mixed mesodermal tumors are in essence variants of carcinoma so-called “metaplastic carcinoma”.
The proposed model draws attention to the fact that “ovarian cancer” is a heterogeneous group of diseases which not only behave differently but also develop differently. Therefore, different approaches to detection and treatment are required. Type I tumors tend to be low-grade, low-stage and slow growing. Current approaches for their detection based on pelvic examination and tranvaginal ultrasound, are appropriate in most cases. However, Type I tumors constitute only 25% of ovarian cancers so these approaches are inadequate for large scale screening. The vast majority of ovarian cancers are Type II tumors that are high-grade, extremely aggressive and are advanced stage at presentation. Therefore, it is our opinion that the current approach to early detection of ovarian cancer, which focuses on the ovary alone, may selectively identify slow growing “good prognosis” tumors while missing more aggressive tumors. Also, since there is no morphologically defined precursor lesion, early detection that would permit intervention and prevention of invasive disease seldom occurs.
Despite these formidable difficulties, a strategy can be developed to enhance early detection and improve survival based on advances in the surgical treatment of ovarian cancer. It is well recognized that the most important prognostic indicator is not stage but the volume of residual disease following cytoreductive surgery. As surgical techniques have evolved, what constitutes optimal cytoreduction has shifted from <2cm to <1.5cm to <1cm. With each reduction in the amount of residual disease that is considered optimal, survival has improved 46. It is therefore clear that the smaller the tumor volume the more effective chemotherapy will be. Therefore, the current approach to evaluating screening tests should be shifted from detection of stage I tumors to detection of “minimal ovarian carcinoma” irrespective of stage. “Minimal ovarian carcinoma” can be defined as microscopic to 1 cm. As technology advances and the sensitivity of assays is improved, the definition of what constitutes “minimal” can be changed.
The ultimate goal of early detection, given the lack of morphologically recognizable precursor lesions, is the identification of biomarkers that precede the development of precursor lesions. It has been shown that mutations of TP53 are currently the most common molecular genetic change in Type II tumors 7, 8, 40, 47, 48. Moreover, mutation of TP53 occurs very early in the genesis of Type II neoplasms. In fact, mutant TP53 is observed in intraepithelial neoplasia in the fallopian tube fimbria of BRCA patients 49. Importantly, TP53 mutations are inherited during cancer evolution and contribute to the transformed state. As a result, the initiating genetic changes are retained in both the primary and recurrent tumors and it is likely that the tumor DNA containing mutant TP53 or polypeptides released from these tumors can be detected in body fluids. It has been shown that small amounts of mutant alleles in cell-free body fluids can be quantified with unprecedented sensitivity by new technologies such as BEAMing 50–52. Expression products of these mutant marker genes might be detected by highly sensitive mass spectrometry technologies or even by specific capture ELISA.
The proposed model for ovarian carcinogenesis also has important implications for targeted treatment. With identification and development of a panel of sensitive and specific molecular genetic changes for Type II tumors, treatment could be administered using drugs that target the pathways affected by the mutations, for example mutant TP53. Therapeutic options would be offered based on the presence of these biomarkers alone. A precedent for this approach currently exists for women who are identified as having BRCA mutations, many of whom choose to undergo prophylactic hysterectomy and bilateral salpingo oophorectomy.
Type I tumors do not present the same challenges as Type II tumors because they generally are localized and indolent and usually do not spread until late in their evolution. Since they are slow growing, therapeutic agents that are effective against Type II tumors are not as effective against Type I tumors and therefore new approaches to their treatment would be desirable. For example, Type I carcinomas harbor several mutations in protein kinases and therefore the pathways that they control could be amenable to inhibitor treatment or targeted by immunotherapy. Protein kinases are the largest superfamily of conserved genes in the mammalian genome, and represent the largest family of genes implicated in human cancer. Deregulation of protein kinase activity as a result of somatic mutations in these genes can lead to malignant transformation and therefore these genes could provide potential targets for therapeutic intervention. Therapeutic molecules and proteins aimed directly at inhibiting protein kinase activity have been developed. These include STI571 (Gleevec), an ATP-binding competitive inhibitor that is a potent inhibitor of the BCR-Abl and c-KIT tyrosine kinases. Currently, Gleevec is a standard treatment for patients with chronic myelogenous leukemia and gastrointestinal stromal tumors 53.
In many Type I carcinomas, there is constitutive activation of the MAPK signaling pathway due to mutations in either KRAS or BRAF genes, the upstream regulators of MAPK. It is therefore conceivable that BRAF inhibitors and other MEK inhibitors could prolong disease-free interval and improve overall survival in patients with these types of advanced-stage Type I tumors. The mutated biomarker sequences might also be specifically targeted by immunotherapy since the mutated sequence is non-self and its expression is restricted to the tumor cells.
A new model for the pathogenesis of ovarian cancer which divides surface epithelial tumors into two groups, designated Type I and Type II, is proposed. Type I tumors are slow growing, generally confined to the ovary at diagnosis and develop from well established precursor lesions that are termed “borderline” tumors. Type I tumors included low-grade micropapillary serous carcinoma, mucinous, endometrioid, and clear cell carcinomas. They are genetically stable and characterized by mutations in a number of different genes including KRAS, BRAF, PTEN, and beta-catenin. In contrast, Type II tumors, which constitute most ovarian carcinomas, are rapidly growing, highly aggressive neoplasms that lack well defined precursor lesions. These tumors include high-grade serous carcinoma, malignant mixed mesodermal tumors (carcinosarcomas) and undifferentiated carcinomas. This group is characterized by mutation of TP53 and a high level of genetic instability. The model accounts for why the current approach to screening aimed at detecting stage I ovarian carcinoma has not been effective because most ovarian carcinomas belong to the Type II group which are advanced stage at, or shortly after, their inception. Therefore a more rational approach to the early detection of “ovarian cancer” may be low volume not low stage of disease.
In summary, the proposed model by assigning different ovarian tumors into two categories based on clinical, morphologic and molecular genetic characteristics facilitates understanding the pathogenesis of ovarian cancer. The model does not replace the histopathologic classification but by drawing attention to the molecular genetic events that play a role in tumor progression can shed light on new approaches to early detection and treatment.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.