Minireview: Human Ovarian Cancer: Biology, Current Management, and Paths to Personalizing Therapy

Abstract

More than 90% of ovarian cancers have been thought to arise from epithelial cells that cover the ovarian surface or, more frequently, line subserosal cysts. Recent studies suggest that histologically similar cancers can arise from the fimbriae of Fallopian tubes and from deposits of endometriosis. Different histotypes are observed that resemble epithelial cells from the normal Fallopian tube (serous), endometrium (endometrioid), cervical glands (mucinous), and vaginal rests (clear cell) and that share expression of relevant HOX genes which drive normal gynecological differentiation. Two groups of epithelial ovarian cancers have been distinguished: type I low-grade cancers that present in early stage, grow slowly, and resist conventional chemotherapy but may respond to hormonal manipulation; and type II high-grade cancers that are generally diagnosed in advanced stage and grow aggressively but respond to chemotherapy. Type I cancers have wild-type p53 and BRCA1/2, but have frequent mutations of Ras and Raf as well as expression of IGFR and activation of the phosphatidylinositol-3-kinase (PI3K) pathway. Virtually all type II cancers have mutations of p53, and almost half have mutation or dysfunction of BRCA1/2, but other mutations are rare, and oncogenesis appears to be driven by amplification of several growth-regulatory genes that activate the Ras/MAPK and PI3K pathways. Cytoreductive surgery and combination chemotherapy with platinum compounds and taxanes have improved 5-yr survival, but less than 40% of all stages can be cured. Novel therapies are being developed that target high-grade serous cancer cells with PI3Kness or BRCAness as well as the tumor vasculature. Both in silico and animal models are needed that more closely resemble type I and type II cancers to facilitate the identification of novel targets and to predict response to combinations of new agents.

Among the gynecological malignancies, ovarian cancer is the leading cause of mortality in developed countries with 225,500 new cases and 140,200 estimated deaths worldwide (1). Despite the global impact of this disease, the lifelong risk of developing ovarian cancer in the United States is one in 70, and the prevalence one in 2500, even in the postmenopausal population that is at greatest risk. Consequently, ovarian cancer is a disease that is neither common nor rare but that has an overall cure rate of less than 40% across all stages. If we are to improve outcomes for women with ovarian cancer, it will be essential to take into account the clinical, cellular, and molecular biology of the disease to move beyond current management and to personalize care.

Biology of Ovarian Cancer

The normal ovary develops from the gonadal ridge near the mesonephros and contains three major cell types: 1) germ cells that are derived from the endoderm and that migrate to the gonadal ridge where they proliferate and differentiate into oocytes, 2) endocrine and interstitial cells that produce estrogen and progesterone, and 3) epithelial cells that are derived from the Mullerian duct and that cover the ovary and line inclusion cysts immediately beneath the ovarian surface. During normal ovulation, oocytes are released from mature follicles and enter the Fallopian tube where fertilization generally occurs. The fimbriae of the Fallopian tube cover the ruptured follicle and facilitate uptake of oocytes. Both benign and malignant tumors can arise from each of the three ovarian cell types (Fig. 1). Germ cell tumors arise most frequently in the second and third decade and account for 3–5% of ovarian cancers (2). Sex-cord-stromal tumors arise from the ovarian connective tissue, often secrete hormones, and can occur in women of all ages, comprising approximately 7% of all ovarian malignancies. Epithelial ovarian cancers generally develop after age 40 and include approximately 90% of malignant ovarian tumors. In addition to benign and malignant epithelial lesions, borderline tumors of low-malignant potential contain morphologically and molecularly partially transformed epithelial cells that do not invade underlying stroma. Approximately 10% of borderline tumors can recur after resection and prove lethal.

Fig. 1.

Different ovarian tumors originate from different cell subtypes. Prevalence of malignant components in parentheses. [Reproduced from V. W. Chen et al.: Pathology and classification of ovarian tumors. Cancer 97:2631, 2003 (89), with permission. © American Cancer Society.]

Histological subtypes of epithelial ovarian cancer

Traditionally, ovarian cancers have been thought to develop from flattened nondescript ovarian surface epithelial cells into cancers that resemble epithelium of the Fallopian tube (serous), endometrium (endometrioid), mucin-secreting endocervical glands (mucinous) and glycogen-filled vaginal rests (clear cell) (Fig 2). Ovarian cancer histotypes have been linked to expression of the HOXA9, HOXA10, and HOXA11 genes that regulate normal gynecological differentiation (3). In contrast to many other cancers, malignant transformation triggers the program of normal differentiation. Tumor histotype (4) and tumor grade or degree of differentiation (5) affect the stage at diagnosis, rate of growth, prognosis, and responsiveness to chemotherapy.

Fig. 2.

Origin and histological subtypes associated with type I and type II molecular classification. [Reproduced from S. Vaughan et al.: Rethinking ovarian cancer: recommendations for improving outcomes. Nat Rev Cancer 11:719, 2011 (14), with permission. © Nature Publishing Group.]

Pattern of spread

Similar to cancers that arise from other sites, epithelial ovarian cancer can spread through lymphatic and blood vessels to nodes and parenchyma of distant organs, including the liver and, because patients are surviving longer with recurrent disease, lung and brain (Fig 3). A distinctive feature of ovarian cancer is the ability to spread through the abdominal cavity, forming nodules on the surface of the parietal and visceral peritoneum including the omentum. Blockage of diaphragmatic lymphatics prevents outflow of proteinaceous fluid from the peritoneal cavity, causing the accumulation of ascites fluid in advanced disease.

Fig. 3.

Pattern of spread of epithelial ovarian cancers. Ovarian cancer cells can spread through lymphatics to nodes at the level of the renal hilum, through blood vessels to the liver, lung, and brain or over the peritoneal surface. [Reproduced from R. C. Knapp et al.: Natural history and detection of ovarian cancer. Gynecology and Obstetrics (edited by J. W. Sciarra), Harper, Row, Philadelphia, p 1 (90), with permission.]

Origin of ovarian cancer

Major risk factors for ovarian cancer include advancing age, number of ovulatory cycles, and a positive family history of ovarian, breast, uterine, or colon cancer related to mutations of BRCA1, BRCA2, mismatch repair genes, or TP53 in the germ line. Risk is halved by the use of oral contraceptives for as long as 5 yr before menopause, possibly related to reduced ovulation and treatment of transforming cells with progestational agents. If understood in greater depth, oral contraceptives could provide a strategy for prevention. Approximately 15% of ovarian cancers are familial and 85% sporadic. Traditionally, ovarian cancers have been thought to arise from ovarian surface epithelial cells or, more frequently, from similar cells that line cysts immediately beneath the ovarian surface. A morphological and genetic continuum can be demonstrated between normal epithelium, dysplasia and invasive high-grade carcinoma with cortical inclusion cysts of the ovary in both BRCA mutation carriers and noncarriers (6). In recent years, it has become apparent that a fraction of ovarian cancers or primary peritoneal carcinomas can also arise from endometriosis, epithelial rests in the normal peritoneum, or the fimbriae of Fallopian tubes. Serous tubal epithelial carcinomas and tubal carcinomas with TP53 mutations have been found in up to 80% of prophylactic salpingo-oophorectomy specimens from carriers of mutant BRCA1 or BRCA2 genes (7–10) and may account for many of the 20% of sporadic cancers thought to be of primary peritoneal origin. TP53 mutations have also been found within small cysts in the ovaries, consistent with an early event in carcinogenesis that may permit transforming cells to survive telomeric crisis, leading to multiple amplicons containing genes that promote proliferation, invasion, and metastasis.

Type I and type II ovarian cancers

At a clinical, cellular, and molecular level, ovarian cancers fall into two major groups based on histological grade, molecular phenotype, and genotype (Table 1 and Fig. 2) (11–13). Type I cancers are low grade of serous, mucinous, endometrioid, or clear-cell histotype. They are often diagnosed in an early stage (I or II), grow slowly, and resist conventional chemotherapy but may respond to hormonal treatment. The more prevalent type II cancers are high grade of serous, endometrioid, or undifferentiated histotype. These cancers present at late stage (III–IV), grow aggressively, and respond to conventional chemotherapy but less often to hormonal manipulation. The distinction between type I and type II cancers provides an initial step in understanding the heterogeneity of ovarian cancers and applying this new knowledge to personalized care (13, 14).

Table 1.

Histology, precursors, and distinctive molecular features (13)

Molecular type	Histology	Precursor	Molecular features
I	Low-grade serous carcinoma	Borderline-carcinoma sequence	KRAS and BRAF mutations
I	Low-grade endometrioid carcinoma	Endometriosis	Mutations CTNNB1, PTEN, and microsatellite instability
I	Mucinous carcinoma	Cystadenoma-borderline sequence	KRAS mutations
I	Clear-cell carcinoma	Endometriosis possibly	PTEN mutations and LOH, PIK3CA mutations
II	High-grade serous carcinoma	De novo from inclusion cysts	p53 mutations, BRCA1/2 mutations and BRCA1 methylation
II	High-grade endometrioid carcinoma	Epithelial inclusion cysts	p53 mutations, BRCA1/2 mutations and BRCA1 methylation
II	Undifferentiated, carcinosarcoma

LOH, Loss of heterozygosity.

Molecular alterations in type I ovarian cancers

Among the type I ovarian cancers, low-grade serous carcinomas appear to grow from serous borderline tumors in 60% of cases (15). Often, these cancers exhibit papillary architecture. Low-grade serous cancers tend to have a normal karyotype and wild-type TP53 and BRCA1/2 but frequent mutations in the B-RAF (2–35%) and KRAS genes (19–54%) (16). The IGF receptor is also expressed by the majority of low-grade serous cancers. Like other type I cancers, low-grade serous tumors are resistant but not refractory to standard chemotherapy (17, 18).

Other type I tumors are uncommon and include low-grade mucinous and clear-cell histotypes that respond to conventional platinum-based chemotherapy in only 26% (19) and 15% (20) of cases, respectively. KRAS is frequently mutated in mucinous cancers (21) and in associated borderline tumors (22). Clear-cell and low-grade endometrioid carcinomas share a similar gene expression pattern, consistent with a common origin, including two genes associated with chemoresistance, ANXA4 and UGT1A1 (23). Inactivating mutations of ARID1A, a chromatin-remodeling gene, have been found in 49% of ovarian clear-cell carcinomas and 30% of endometrioid ovarian cancers (24, 25) PPP2R1A, the regulatory subunit of a serine-threonine phosphatase required for chromosome segregation, is also mutated in 7% of clear-cell ovarian cancers (24). Low-grade endometrioid cancers exhibit frequent inactivating mutations and epigenetic silencing of PTEN and activating mutations of PIK3CA that up-regulate phosphatidylinositol-3-kinase (PI3K) signaling.

Molecular alterations in type II ovarian cancers

Although low-grade type I cancers appear to be driven by activating mutations on a background of a relatively normal karyotype, high-grade type II cancers are driven by copy number abnormalities and marked genomic instability. The Cancer Genome Atlas Project, which analyzed more than 300 high-grade serous cancers, detected amplification of more than 30 growth-stimulatory genes (26). Amplification and overexpression of genes in the PI3K family occur in more than 40% of type II cancers, conferring PI3Kness, or activation of the PI3K pathway. When ovarian cancers occur in carriers of BRCA1 or BRCA2 germline mutations, they are generally type II high-grade tumors. Somatic mutations of BRCA1 and BRCA2 can occur, BRCA1 can be silenced, and upstream mutations can down-regulate BRCA function, producing BRCAness, or homologous DNA repair deficiency in more than 40% of type II ovarian cancers. Mutations of p53 were found in 96% of type II high-grade serous cancers. Judged from their sequence, most of these mutations are inactivating. Other mutations were uncommon, with NF1, RB1, and CDK12 mutated in 2–4%. Less than 1% of type II cancers had mutations of BRAF, PI3KCA, KRAS, or NRAS found in type I tumors. These genes may, however, be rare but important drivers of high-grade serous cancers. Despite the low prevalence of Rb mutations, dysfunction of the Rb pathway has been found in 67% of high-grade serous cancers (26).

Current Clinical Management of Ovarian Cancer

Over the last three decades, 5-yr survival for ovarian cancer patients has increased from 37 to 45%, related to more consistent use of cytoreductive surgery and combination chemotherapy with platinum compounds and taxanes (27). The majority of patients are diagnosed in advanced stage with multiple tumor nodules studding the parietal and visceral peritoneum in the pelvis, omentum, and diaphragm.

Surgery

Ovarian cancer is one of the few malignancies where surgeons will undertake cytoreductive operations, even if all macroscopic tumor cannot be removed. Reducing tumor burden to where no macroscopic tumor is left before chemotherapy is considered optimal cyroreduction (9). Surgery can be performed after neoadjuvant chemotherapy (10) when optimal cytoreduction is not considered feasible at initial diagnosis. Survival increases with the expertise of the surgeon (28), and optimal cytoreductive surgery is an independent prognostic factor (29). The purpose is to achieve both correct FIGO (International Federation of Gynecology and Obstetrics) staging (30) and therapeutic thorough cytoreduction.

In several retrospective series, cytoreductive surgery for recurrent disease has been associated with improved survival when all macroscopic cancer can be removed (31, 32). Two ongoing prospective trials in Europe and the United States are evaluating criteria and outcomes for secondary cytoreduction.

Primary chemotherapy

Six cycles of carboplatin and paclitaxel chemotherapy are considered standard adjuvant treatment for newly diagnosed ovarian cancer after cytoreductive surgery. Carboplatin is an alkylating agent that binds covalently to DNA, creating adducts that form intrachain and interchain cross-links. Paclitaxel binds noncovalently to microtubules and increases their stability, interfering with mitotic spindle formation. Both agents induce apoptosis. Chemotherapy has generally been administered iv, but three randomized phase III trials have shown a 20–25% relative risk reduction in mortality after intraperitoneal therapy for patients who have been optimally cytoreduced (33–35). Chemotherapy is generally administered every 3 wk, but weekly dose-dense administration of paclitaxel has produced improved survival in one trial from Japan (36), and a confirmatory trial has not yet been completed.

Empirical addition of three other active drugs including pegylated liposomal doxorubicin, topotecan, and gemcitabine to standard therapy failed to improve upon the progression-free or overall survival observed with paclitaxel and carboplatin alone (37). Two recent trials have added a vascular endothelial growth factor (VEGF)-binding antiangiogenic antibody, bevacizumab, to standard treatment during and for up to 15 months after chemotherapy. Improved progression-free but not overall survival was reported (91, 92).

Chemotherapy for recurrent ovarian cancer

More than 70% of patients with advanced ovarian cancer will experience disease recurrence and become candidates for second-line chemotherapy, within 12 and 18 months. Retreatment with carboplatin and paclitaxel is associated with a 20–50% response when platinum-sensitive disease recurs more than 6 months after primary chemotherapy (Table 2). Although recurrent disease is not curable, combinations of drugs can prolong survival. In platinum-sensitive disease, a combination of carboplatin with paclitaxel (38, 39), gemcitabine (40), or liposomal doxorubicin (41) is superior to single-agent carboplatin (42). Disease that recurs in less than 6 months is considered platinum resistant. In this setting, several drugs produce response rates ranging from 10–30% and increase progression-free survival such as liposomal doxorubicin (43), weekly paclitaxel (44), and topotecan (45). Other drugs have demonstrated activity in phase II clinical studies, including gemcitabine (46), bevacizumab (47, 48), docetaxel (49), and etoposide (50).

Table 2.

Recurrent populations according to interval from last platinum

Term	Definition
Refractory	Progression while receiving last line of platinum-based therapy or within 4 wk of last platinum dose
Platinum resistant	Progression-free interval of less than 6 months
Partially platinum sensitive	Progression-free interval since last platinum of 6–12 months
Platinum sensitive	Progression-free interval since last platinum of 6–12 months

Biomarkers

CA125 (MUC16) is a high-molecular-mass (1 MDa) glycosylated transmembrane mucin that is expressed by 80% of ovarian cancers (51) and is important for adhesion, motility, and invasion of ovarian cancer (52). CA125 is shed from ovarian cancers and circulates in serum where it has provided the first generally useful biomarker for monitoring the response of ovarian cancer to chemotherapy (53). Persistent elevation of CA125 after chemotherapy indicates residual disease with more than 90% accuracy. CA125 has been used routinely to detect recurrence after chemotherapy. A recent trial found that early treatment of recurrent ovarian cancer with chemotherapy based on doubling of CA125 did not prolong survival when compared with treatment 5 months later at the time of clinical or symptomatic relapse (54). Limitations of the trial included inadequate stratification for important prognostic variables, use of suboptimal thresholds for CA125, delays in treatment in one quarter of participants, and suboptimal chemotherapy in two thirds of patients (55). Consequently, only one quarter of patients were treated promptly with combinations of drugs that could improve survival. Given the limitations of chemotherapy for recurrent disease, however, it remains uncertain whether monitoring for recurrence with CA125 improves overall survival, although it does identify patients for secondary cytoreductive surgery and provides time for treatment with multiple conventional and novel drugs.

Additional applications of CA125 include its use in combination with age, ultrasound (56), or other biomarkers (57–59) to identify patients with pelvic masses who would benefit from referral to a specially trained gynecological oncologist for cytoreductive surgery. Although individual values of CA125 are not sufficiently specific for detecting early-stage disease, trials are currently underway to test the value of a rising CA125 to trigger ultrasound that would prompt surgery (60, 61). Preliminary data suggest that this strategy is sufficiently specific that only three exploratory laparotomies will be required to detect an ovarian cancer and that an increased fraction of early-stage disease can be detected.

Enhancing drug sensitivity and overcoming drug resistance

During primary chemotherapy, approximately 70% of ovarian cancers will respond to platinum alone or in combination with paclitaxel (62). In two of three major trials, the addition of paclitaxel to cisplatin or carboplatin increased disease-free and overall survival compared with platinum-based therapy alone (63, 64). Despite an additive increase in overall survival with combination chemotherapy, only 42% of previously untreated patients will respond to paclitaxel, and there is no synergy between platinum compounds and the taxanes. Consequently, more than half of patients receive the toxicity, but not the benefit of taxanes, and there is room for significant improvement. Taxanes induce apoptosis in cancer cells after increasing microtubule stability and delaying or preventing passage through the cell cycle. Recent studies suggest that knockdown of several kinases can enhance paclitaxel sensitivity of ovarian cancer cells (65). In some cases, this relates to enhancing microtubule stability, and in others, it depends upon modifying apoptotic mechanisms or centrosome function (66). Using paclitaxel in combination with specific RNA interference or specific low-molecular-weight kinase inhibitors could lead to a greater fraction of ovarian cancer patients responding to primary therapy.

After treatment with carboplatin and paclitaxel, specifically resistant cancer cells emerge. Knowledge regarding the biology of taxane and platinum resistance is beginning at the preclinical level (67–69), but strategies for reversing drug resistance have not been validated clinically.

Targeted Drugs and Antibodies to Personalize Therapy

Individual targeted agents

Improvement in outcomes might result from therapy that targets the abnormal proteins in each patient's cancer. In the Cancer Genome Atlas Research Program analysis, different fractions of high-grade type II ovarian cancers had amplification of some 22 oncogenes for which specific inhibitory drugs were already available (26). To date, however, individual targeted agents have had only a modest impact on recurrent ovarian cancer in unselected patients. With the exception of bevacizumab, eight targeted drugs, gefitinib, imatinib, sorafenib, temsirolimus, mifepristone, enzastaurine, lapatinib, and vorinostat, have produced objective response rates of less than 10% and have stabilized disease for 6 months in less than 25% of cases in phase II trials. It is clear that we must use multiple agents and seek synthetic lethality if we are to produce deep and long-lasting remissions of recurrent disease and ultimately to improve primary therapy.

BRCAness

One of the best examples of synthetic lethality to reach the clinic to date is provided by the activity of poly-ADP-ribose polymerase (PARP) inhibitors in ovarian cancers that display BRCAness, i.e. a deficiency of BRCA1/2 function (70, 71) is associated with a better overall prognosis (72) and response to platinum compounds (73). Although 10–15% of ovarian cancers have germline BRCA1/2 mutations (74–76), up to 47% of type II high-grade serous ovarian cancers have genetic or epigenetic inactivation of BRCA1/BRCA2 (77). Somatic BRCA1/2 mutations were present in 19% of unselected ovarian cancer and 23% of high-grade serous cancers (78). Another recent report described BRCA1/2 mutations or BRCA1 silencing in 33% of high-grade serous cancers (26).

BRCA1 and BRCA2 mediate homologous recombination, which is one mechanism of DNA repair. Cancers with BRCAness are deficient in homologous repair and cannot repair DNA double strand breaks induced by platinum compounds (79). Inhibition of a second DNA repair pathway, base excision repair, by PARP inhibitors causes synthetic lethality in cancers with BRCAness. PARP inhibitors have produced response rates of more than 40% in ovarian cancers with BRCA1/2 mutations (80).

PI3Kness

Activation of PI3K signaling or PI3Kness can be produced by activating mutations of PIK3CA, inactivating mutations of PTEN, or amplification of PIK3CA, PIK3CB, PIK3R4, AKT1, AKT2, or AKT3 (81). Common copy number gains of PIK3CA, PIK3CB, and PIK3R4 in type II high-grade ovarian cancer were associated with decreased survival (82). Currently, mammalian target of rapamycin inhibitors, PI3K and AKT inhibitors are being investigated. In human ovarian cancer cell lines with PTEN deficiency, sensitivity to PARP1 inhibitors and cisplatin, but not to PARP1 inhibitors and paclitaxel, was higher than in the wild type (83). Thus, combinations of drugs that block both PI3K/AKT and PARP should be evaluated in patients with BRCAness and PI3Kness, whereas drugs that block both PI3K/AKT and MAPK kinase should be pursued in patients with abnormalities of the PI3K or Ras/MAPK pathway. The presence of activated pathways is likely to be necessary but not sufficient. Better predictive models will be required in cell cultures, animals, and in silico, to identify relevant targets and to choose optimal combinations for individual ovarian cancer patients.

Antiangiogenesis

Ovarian cancer metastases cannot grow to greater than 1 mm without blood vessel formation. Endothelial cells associated with tumor vessels depend upon proangiogenic factors for survival and can proliferate more rapidly than vessels serving normal tissues, providing targets for antiangiogenic therapy (84). Ovarian cancers produce multiple proangiogenic factors including VEGF, IL-8, and basic fibroblast growth factor. Inhibitors of proangiogenic proteins such as VEGF (bevacizumab and aflibercept), angiopoietins (AMG386), PDGF (imatinib and pazopanib), or their receptors VEGF receptor (pazopanib, sorafenib, sunitinib, and BIBF1120) are being tested in the clinic. Bevacizumab, as a single agent or in combination with daily low-dose cyclophosphamide, can produce an objective response rate of 20% in recurrent ovarian cancer and stabilize disease for 6 months in 40%. Given the extraordinary expense of bevacizumab, identifying biomarkers with high negative predictive value is an important unmet need. Putative biomarkers described include circulating endothelial cell precursors, CA125, DII4, VEGF-C, and neuropilin-1 (85). A decrease in perfusion with magnetic resonance imaging has also been evaluated, but predictive tests are not yet sufficiently precise to use routinely (86, 87). Relevant animal models will be crucial to develop multi-agent antivascular therapy and to facilitate identification of relevant predictive biomarkers.

Implications for the Development of Animal Models

Current knowledge regarding the biology and clinical management of ovarian cancer suggests that targeted agents must be used in combination to select the right drugs for the right patient at the right time. Given more than 400 anticancer drugs and antibodies in the current pharmaceutical pipeline, not all combinations can be tested in the clinic. Development of animal models that mimic the biology of human ovarian cancer will be critical for identifying new targets and useful combinations on the path to personalized therapy. The accompanying minireview considers the currently available models for ovarian cancer (88). In judging the relevance of these models, it will be important to consider the ability of primary cancers to metastasize from the ovaries not only through lymphatic and blood vessels but also to the surface of the peritoneal cavity, producing ascites. Animal models should mimic one of the two major types of ovarian cancer: low-grade type I disease driven by mutations of RAS, Raf, or members of the PI3K pathway on a background of genomic stability, wild-type TP53 and BRCA1/2, and hormonal signaling; or high-grade type II disease driven by amplification and genomic instability with mutant TP53 and dysfunctional Rb with or without BRCAness or PI3Kness. Models that produce mucinous, endometrioid, or clear-cell histotypes would also be of interest. Because mice and chickens can express CA125, animal models might also be used to identify complementary biomarkers. Whether or not animal models mimic human disease precisely, these ovarian cancers should respond to platinum compounds, taxanes, and inhibitors of Ras/MAPK, proangiogenic factors, PARP, and the PI3K pathway. Whatever the genotype and phenotype of models, validating their predictive power with agents known to be active in the clinic would be important if they are to contribute to translational research.