This is a review on the transition from our empirical approach to treat ovarian cancer to a specific treatment based on molecular signature. We have reviewed not only the evidence-based medicine focused on the origin and tumor morphology of ovarian cancer but also the molecular signature era based on molecular phenotyping of the tumor and its microenvironment, which influences the direct targeted therapy. Evidence-based medicine has shown that the targeted therapy studies are mainly biomarker driven, more focused, and hence treat only those patients who have the underlying molecular abnormality. This molecular abnormality is the target of the drug, leading to higher rates of response. These findings will carry important implications for screening, detection, and treatment of ovarian cancer in the future.
Keywords : Ovarian cancer Molecular profiling Targeted therapy
With the turn of the century, we are witnessing the tran- sition from our empiric approach to cancer treatment to the new paradigm of personalized medicine, based on the molecular signature of individual cancers (Table 1). Evi- dence-based medicine has focused on the organ of origin and tumor morphology, whereas in the molecular signature age, molecular phenotyping of the tumor and its microen- vironment directed the therapeutic decision [1]. Until recently, our treatments resulted from large randomized trials comparing new therapies to the ‘‘gold standard.’’ These costly studies were slow in accrual and because of the mixture of various tumor types originating in the same organ, e.g., the ovary, the results were limited. Targeted therapy studies are biomarker driven, more focused, and would treat only those patients who have the underlying molecular abnormality that is the target of the drug, yielding higher rates of response.
Ovarian cancer was believed to originate from the invagi- nation, metaplasia, and malignant transformation of the surface epithelium of the ovary. This is a unicellular layer of mesothelium similar to the peritoneum that was thought to undergo metaplasia in the inclusion cysts following ovulation. The cancerous cells would then expand, reach the surface, and extend to the peritoneal surfaces. This view had important implications on the efforts for screen- ing and early detection, search of new chemotherapy reg- imens, and routes of delivery of chemotherapy. These time- honored concepts led to large randomized trials in which all patients with ovarian cancer were placed on trial irre- spective of histology, mixing endometrioid, clear-cell, mucinous, and serous cancers alike. Since the introduction of platinum compounds in the early 1980s, combination chemotherapy regimens improved the length of survival of patients with ovarian cancer, but did not impact cure rates that have remained below 30 % over the last 40 years.
Recent evidence suggests that ovarian cancers can be segregated into two major types similar to endometrial cancer (Table 2). Type II ovarian cancers form a majority of cancers and include high-grade papillary serous tumors and high-grade endometrioid cancers, including the carci- nosarcomas. Overall, these aggressive tumors represent *75 % of ovarian cancers and have a poor outcome. Type I ovarian cancers include the low-grade cancers that have indolent courses and usually present at a low stage. Frequently, areas of borderline tumors are visible in the vicinity of these cancers, and there appears to be a con- tinuum of histologically progressive lesions from dysplasia to low-grade neoplasia.
Low-grade serous tumors are characterized by K-ras(30 %), B-raf (30 %), or erb-b2 (5 %) mutations (Fig.1)[2]. These are mutually exclusive. Therefore, mutations inany of the genes are detected in about two-thirds of mi-cropapillary serous carcinomas and atypical proliferativeserous tumors. In contrast, these genes are not mutated inhigh-grade serous cancers [2,3]. Mutations of K-ras and B-raf seem to occur very early in the development of low-grade micropapillary serous carcinomas, as evidenced bythe demonstration that the same K-ras and B-raf mutations detected in borderline tumors are detected in the cystade-noma epithelium adjacent to these borderline tumors [4].On the other hand, no increase in BRCA mutations was detected in patients with borderline tumors [5].
Low-grade endometrioid tumors have a similar molec-ular profile as their endometrial counterparts (Comparewith Fig.2). In addition, endometrioid carcinomas of the ovary are associated with HNPCC [6] and coexist withtheir endometrial counterparts relatively frequently (up to20 % of cases of endometrioid carcinoma of the ovary are associated with synchronous atypical hyperplasia orendometrioid adenocarcinoma of the endometrium [7]).The favorable outcome of such cases suggests that these are independent primaries and also suggests a role of hormonal environment.
Clear-cell carcinomas seem to form a subgroup of non- hormonally dependent cancers with a low mitotic index [8], which may be related to their reduced response to plati- numbased chemotherapy [9]. They resemble renal clear- cell cancers, and microarray analysis has recently been shown to predict their response to sorafenib [10].
Mucinous tumors are a relatively rare subtype of ovarian cancer. K-ras mutations are frequently found in these tumors [11]. HER2 amplification with over expression of the protein on the surface of the tumor cells is present in 15–20 % of mucinous ovarian cancers [12]. No data are yet available on the response rate of this small group of tumors to Trastuzumab (Herceptin) therapy.
High-grade cancers of the ovary on the other hand are associated with p53 mutations and cyclinassociated abnormalities (Fig. 3). Cyclins, as their name indicates, contribute to the temporal coordination of each mitotic event [13]. Abnormalities result in chromosome instability and thus may contribute to tumorigenesis (ref).
Based on the above-indicated molecular changes, it becomes apparent that each type of ovarian cancer can be targeted differently and that the response to presently available chemotherapies will vary in function of the underlying mutations, explaining the spectrum of responses seen in large randomized trials. Personalized therapies resulting from specific mutation analysis of the particular tumor and the evaluation of the surrounding tissues in a specific patient are expected to yield higher response rates than presently observed and ultimately result in potential cures.
Based on whole genome expression profiling using thousands of probe sets, the authors have reproducibly been able to document that borderline ovarian tumors were indistinguishable from low-grade ovarian cancers, but were completely distinct from high-grade cancers [14–16]. In addition, none of these had any common clusters compared to ovarian surface epithelium, indicating that ovarian tumors do not arise from the surface epithelium. This has important implications for screening strategies. Low-grade cancers, that for the most part are early stage, represent different molecular entities compared to advanced-stage high-grade cancers. Finding these early low-grade cancers by presently available screening tools or early detection means will have no impact on the survival of the high- grade cancers which represent the cancers that carry the poor prognosis. Only screening for molecular markers specific to high-grade cancers might influence the detection of these cancers and impact their outcome.
An important shift in our understanding of the origin of ovarian cancers came when unexpected incidental fallopian tube cancers were detected at the time when prophylactic salpingo- oophorectomies were performed on BRCA mutation carriers [17]. Some authors started to speculate that the epithelial serous cancer cells found on ovaries and throughout the peritoneal cavity might arise from shedding of cells from the epi- thelium of the fallopian tube, rather than from the metaplasia and malignant transformation of the single layer of mesothelial cells surrounding the ovary [18]. Further investigating this hypothesis, areas of strong p53 immunostaining were identified in normal secretory cells from the fimbriated end of the fallopian tube of BRCA-positive patients [17] as well as the foci of tubal intra-epithelial carcinoma [19]. These areas, referred to as TIC, expressed identical mutations as the associated quote ovarian cancers [20]. If confirmed, risk-reducing surgery might focus on the fallopian tube rather than the ovary. A recent effort initiated in the third quarter of 2010, led by the cancer agency in British Columbia (Canada), promoted salpingectomies rather than tubal ligations performed for birth control as well as at the time of hysterectomy for benign causes (www.ovcare.ca). The authors of this initiative hope to decrease the development of ovarian cancers in these women in the future. In addition, Marquez et al. [21] had shown in 2005 that clear-cell cancers and endometrioid cancers of the ovary resembled normal endometrial tissue, serous cancers resem- bled normal fallopian tube cells, and mucinous cancers resembled normal colon epithelium. Moreover, none of the cancer profiles resembled ovarian surface epithelium. It is presently believed that serous tumors originate in the fimbriated end of the fallopian tube [22]. Other authors [23] further hypothesize that if the K-ras/B-raf pathways become altered in the cells from the fimbria, these cells evolve via a morphologic continuum including low malignant potential and low-grade cancers. On the other hand, if p53 mutations occur, these cells produce high-grade aggressive-behaving cancers. In addition, they speculate that clear-cell cancers and endometrioid cancers migrate from the endometrium. This is supported by the protection obtained from tubal ligation only on clear-cell (OR, 0.32; 95 % CI, 0.006–2.50) and endometrioid cancers (OR, 0.20; 95 % CI, 0.046–1.46) [24]. Mucinous cancers and Brenner tumors would originate from Walthard cell nests [23]. These findings (Fig. 4) carry important implications for screening, detection, and treatment of ovarian cancer, which will necessitate a complete reappraisal and adjustment of our present practices. Gene expression profiling of individual cancers and their micro environment, in the context of host factors, represents the cornerstones of personalized medicine and will predict the response to chemotherapy and prognosis for the specific patient being evaluated [25].