The Endometriotic Tumor Microenvironment in Ovarian Cancer

Women with endometriosis are at increased risk of developing ovarian cancer, specifically ovarian endometrioid, low-grade serous, and clear-cell adenocarcinoma. An important clinical caveat to the association of endometriosis with ovarian cancer is the improved prognosis for women with endometriosis at time of ovarian cancer staging. Whether endometriosis-associated ovarian cancers develop from the molecular transformation of endometriosis or develop because of the endometriotic tumor microenvironment remain unknown. Additionally, how the presence of endometriosis improves prognosis is also undefined, but likely relies on the endometriotic microenvironment. The unique tumor microenvironment of endometriosis is composed of epithelial, stromal, and immune cells, which adapt to survive in hypoxic conditions with high levels of iron, estrogen, and inflammatory cytokines and chemokines. Understanding the unique molecular features of the endometriotic tumor microenvironment may lead to impactful precision therapies and/or modalities for prevention. A challenge to this important study is the rarity of well-characterized clinical samples and the limited model systems. In this review, we will describe the unique molecular features of endometriosis-associated ovarian cancers, the endometriotic tumor microenvironment, and available model systems for endometriosis-associated ovarian cancers. Continued research on these unique ovarian cancers may lead to improved prevention and treatment options.


Introduction
Endometriosis is a debilitating disease that is estimated to affect up to 5 million U.S. women and girls. Endometriosis results in considerable morbidity, including pelvic pain, multiple operations, infertility, and negative effects on psychosocial quality of life [1][2][3][4][5]. Unfortunately, endometriosis is also a significant risk factor for development of ovarian cancer [6]. The presence of endometriosis increases the risk of ovarian endometrioid, low-grade serous, and clear-cell adenocarcinoma by up to 8.9-fold but not high-grade serous adenocarcinoma [7][8][9][10][11][12]. Thus, ovarian endometrioid, low-grade serous, and clear-cell adenocarcinomas are considered endometriosis-associated ovarian cancers. Ovarian cancer is considered a top-five cancer killer in U.S. women, claiming more than 14,000 lives in 2015 [13]. Therefore, 5 million U.S. women and girls with endometriosis are at risk for developing deadly ovarian cancer. Fortunately, ovarian endometrioid and clear-cell adenocarcinoma represent roughly 20% of all ovarian cancers and account for less than 10% of deaths [14][15][16]. Clinically, studies suggest that co-occurrence of endometriosis with ovarian cancer is associated with an improved prognosis [17][18][19][20]. Important factors in this improved prognosis include discovery at early age and early stage disease in women with endometriosis at time of ovarian cancer staging [21][22][23][24], but may in endometriosis-associated ovarian cancer to examine changes in these and other alternative DNA marks. MicroRNA (miRNA) molecules, which are also considered epigenetic changes, are dysregulated in endometriosis (reviewed in [53]). While dysregulated miRNAs in epithelial ovarian cancers have been recently reviewed [54,55], dysregulated miRNA molecules in endometriosis-associated ovarian cancers have not been individually reviewed. Given that miRNA molecules can be secreted from cells, we have included miRNA molecules under endometriotic tumor microenvironment (below).
A challenge to studies on the endometriotic tumor microenvironment is the rarity of clinical samples of ovarian cancer with concurrent endometriosis and the rigor of details provided for patient characterization. Given over 22,000 women will be diagnosed with ovarian cancer in 2016 [13], only 10% will be endometrioid and roughly 10% will be clear-cell [14][15][16]. Additionally, a majority of women with endometriosis-associated ovarian cancers do not have endometriosis at time of staging. Roughly 30% of ovarian endometrioid or clear-cell adenocarcinomas will have concurrent endometriosis, further narrowing the number of tumors to study with concurrent endometriosis [56][57][58][59]. Many studies do not describe the patient population in terms of absence or presence of endometriosis, leaving readers to believe that the women may not have endometriosis, which may not be accurate. Efforts for data harmonization for rare tumors may improve reproducibility. Using well-characterized samples, Banz et al. used transcriptome microarray analysis to evaluate normal ovary, endometriomas, and endometrioid ovarian cancer with and without endometriosis [60]. The results showed a small group of cytokines dysregulated in ovarian cancers with endometriosis, consistent with the inflammatory milieu of endometriosis [60]. Additionally, Zhang et al. showed a unique gene signature in ovarian endometrioid adenocarcinoma with concurrent endometriosis compared to ovarian endometrioid adenocarcinoma without concurrent endometriosis [61]. Highly dysregulated signaling pathways included nuclear factor kappa B (NFkB), transforming growth factor beta (TGFβ), and KRAS signaling [61]. Most likely there are contributions from genetics and epigenetics that may be mediated from the endometriotic tumor microenvironment [62]. However, further studies are needed to examine how endometriosis affects ovarian cancer.

The Unique Endometriotic Tumor Microenvironment
While the pathogenesis of endometriosis is still largely poorly understood, the most accepted theory is the implantation theory following retrograde menstruation (reviewed in [63]). Most menstruating women have retrograde menstruation [64], but only 10% have endometriosis [1][2][3], suggesting that unique conditions occur in women with endometriosis. The endometriotic microenvironment contains multiple cell types-endometrial epithelial cells, stromal fibroblasts, endothelial cells, and immune cells-as well as inflammatory mediators, metabolic waste products such as iron from the breakdown of red blood cells, steroid hormones, and small RNA molecules. Thus, it is not surprising that the conditions found in endometriosis are also advantageous to the growth and development of ovarian cancer. However, very little is known about how these stressful conditions directly affect ovarian cancer. In this section, we will describe these important factors within the scope of endometriosis and how these important factors pertain to ovarian cancer. Figure 1 summarizes graphically key players in the endometriotic tumor microenvironment as it pertains to ovarian cancer. between precursor endometriosis lesions and matched clear-cell adenocarcinomas for expression of HIF1A and phosphorylated mechanistic target of rapamycin kinase (P-mTOR) [75]. Importantly, vascular endothelial growth factor (VEGF), leptin (LEP), cysteine rich angiongenic inducer 61 (CYR61), and osteopontin (SPP1) work together in response to hypoxia to establish a local vascular network within the endometriotic lesion [74]. In addition to neoangiogenesis mediated through HIF1A, as endometriotic lesions undergo hypoxia and inflammation from repeated menstrual cycles, the expression of tissue factor increases. Tissue factor is a critical protein for extrinsic coagulation cascade, leading to hypercoagulation. Clinically, women with clear-cell ovarian cancer have more frequent venous thromboembolism [76]. Hypoxia may also lead to cellular proliferation through estrogen receptor, leptin, and prostaglandin modulation [77]. These studies suggest that the hypoxic microenvironment of endometriosis plays a role in not only the potentiation of endometriosis by promoting cell proliferation and nutrient availability through vascularization but may also play roles in outcomes for women with clear-cell ovarian cancer. The increased expression of HIF1A in endometriosis may represent a novel therapeutic target for endometriosis or ovarian cancer [78].

Fibroblasts and Extracellular Matrix Components
Endometriosis is pathologically complex, containing endometrial epithelial and stromal fibroblasts outside the uterine cavity, alongside invading hemosiderin-laden macrophages [79]. The endometriotic extracellular matrix (ECM) plays a significant role in paracrine/autocrine signaling between epithelial and stromal cells [80][81][82][83]. Studies have shown unique functional properties of primary cultures of human endometrial stromal fibroblasts from women with endometriosis compared to cultures from women without endometriosis. Specifically, fibroblast cultures from women with endometriosis have a deficiency in decidualization, the differentiation process by which the uterus prepares for pregnancy [84]. Additionally, these fibroblasts from women with endometriosis have increased ERK signaling, high proliferative potential from progesterone resistance, and acquire an inflammatory phenotype [85][86][87][88][89]. While the importance of stromal-epithelial crosstalk is noted in embryo implantation in the uterus [80], the role of similar crosstalk in endometriosis or epithelial ovarian cancers is still understudied but may represent a key component of the endometriotic tumor microenvironment.
To examine the tumor microenvironment in ovarian cancer, Zhang et al. used computer-aided image analysis and showed that the number of cancer-associated fibroblasts, as indicated by cells positive for smooth muscle antigen, was higher in epithelial ovarian cancers compared to benign adnexal masses. Unfortunately, the specific histology of ovarian cancers and the pathology of the benign adnexal masses were not described in these studies. Large numbers of similarly staining cancer-associated fibroblasts were also found in omental metastatic lesions [90]. Co-culture of cancer-associated fibroblast with ovarian cancer cell lines (SKOV3, CAOV3) led to increased invasion and migration when compared to ovarian cancer cell lines grown in co-culture with normal fibroblasts [90]. One of the main questions regarding cancer-associated fibroblasts is how and why they are becoming activated to benefit tumor cells. Mitra et al. proposed that ovarian cancer cells reprogram fibroblasts into cancer-associated fibroblasts through miRNA expression changes [91]. Specifically, cancer-associated fibroblasts have a significant downregulation of miR-31 and miR-214 and upregulation of miR-155. C-C motif ligand 5 (CCL5), a chemokine known to be highly upregulated in ovarian cancers, is a direct target of miR-214. Similarly, endometriomas have high expression of chemokines and dysregulated miRNA expression [92]. Advancements in the understanding of the role of non-epithelial ovarian cancer cells in ovarian cancer may lead to better treatments which block tumor promotion brought on by tumor adjacent cells.

Immune Cells and Inflammatory Mediators
Dysregulated inflammation plays a key role in endometriosis-associated pathology [93]. For example, Capobianco and Rovere-Querini provide an in-depth review of the role of macrophages in endometriosis, showing a relationship between components of the endometriotic microenvironment such as high iron, hypoxia, and angiogenesis with macrophage recruitment and activation [94]. Additionally, a syngeneic mouse model of endometriosis showed that endometriotic lesions failed to grow without macrophages, and if macrophages were removed after implantation, angiogenesis was halted, blocking the progression of the endometriotic lesion [95]. Further, Canet et al. suggest that retainment of a specific macrophage population in endometriomas, the cell division cycle 42 (CDC42)-positive population, protects endometriomas from malignant transformation [96]. Similarly, platelet factor 4 (PF4) also known as chemokine (C-X-C Motif) ligand 4 (CXCL4) is highly expressed on macrophages in endometriomas, but not on tumor-associated macrophages of clear cell ovarian cancers [97]. Thus, specific details of the macrophage population in endometriosis and ovarian cancer are important and require further study.
Transcriptomic work on endometriomas showed that the inflammatory cytokine transforming growth factor beta 1 (TGFβ1), regulates other inflammatory mediators relevant to endometriosis, including tumor necrosis factor alpha (TNFα) and interleukin-6 (IL6) [92]. These inflammatory mediators are highly elevated in peritoneal fluid from women with endometriosis [98][99][100][101]. The acute and chronic inflammation of endometriosis is a response to the invading tissue, leading to the release of regulated on activation normal T cell expressed and secreted (RANTES), monocyte chemotactic protein-1 (MCP1), and interleukin-8 (IL8), which act as chemoattractants recruiting more macrophages to the area [102]. In terms of the endometriotic tumor microenvironment, the promotion of tumor invasion via macrophages may be dependent on TNFα [103], which is elevated in women with endometriosis [98,99]. Along the same lines, work using an estrogen receptor beta (ERβ)-overexpressing syngeneic mouse model of endometriosis suggests that non-genomic effects of ERβ play a role in the TNFα-mediated dysregulation of endometriosis progression [104]. Encouragingly, treatment of a syngeneic mouse model of endometriosis with a long-acting TNFα-blocking agent decreased endometriotic implant size [105]. However, treatment of women with rectovaginal nodules with infliximab, a TNFα monoclonal antibody, had no improved clinical effect over placebo [106]. Understanding the immune response to misplaced endometrial tissue will be a large factor in understanding the onset and progression of endometriosis and lead to a better understanding of how endometriosis creates a unique and potentially tumor-promoting microenvironment.

Altered Metabolism
Endometriotic cysts contain blood. When blood is metabolized, heme and iron are released into the microenvironment [107]. Because of this, endometriotic cysts contain higher iron levels than other benign ovarian cysts [108]. Consequently, an iron-rich microenvironment can lead to increased proliferation, DNA synthesis, and adhesion, and promote chronic inflammation, allowing for the spread of endometriosis [107]. High iron also leads to excessive oxidative stress, which creates a microenvironment conducive to the induction of mutations and has been linked to cancer development in the liver and lung [107,109]. Shigetomi et al. outlines how endometriotic cells under oxidative stress from excess iron are able to bypass cell cycle checkpoints after DNA damage by overexpressing hepatocyte nuclear factor-1 beta (HNF1B), which activates forkhead box transcription factors and alters miRNA expression promoting cell survival [110]. Due to the excess iron exposure, endometriotic cysts have higher expression of lactose dehydogenase, lipid peroxidase, and 8-hydroxy-2 -deoxyguanosine. High expression of these markers of oxidative stress link endometriosis, high iron, and higher frequencies of gene mutations [108]. These data corroborate the hypothesis that endometriosis produces a high iron microenvironment that may lead to increased DNA damage through oxidative stress, but also promotes cell survival, leading to a highly mutated subpopulation of cells that continue to grow [111].
Alongside high iron levels, endometriotic peritoneal fluid has elevated lactate. Further, endometriotic lesions express high levels of glycolysis genes compared to eutopic endometrium [112]. Increased expression of HNF1α in the endometriotic peritoneum leads to the conversion of glucose to lactate in a process known as the "Warburg Effect," known for its promotion of cell survival in stressful microenvironments [113]. Lipidomics has also been pursued for understanding the metabolomic profile of the endometriotic microenvironment. Lipid profiling studies on endometrial aspirates have shown a reduction of saturated diacylglycerols and triacylglycerols in endometriosis patients compared to healthy controls [114]. In fact, this study generated a panel of 123 metabolites which were differentially expressed in endometriosis women and correctly identified 86% of samples to either the endometriosis or control group [114]. A similar study on endometrial biopsies used five lipid metabolites as biomarkers and were able to predict endometriosis with 75% specificity and 90.5% sensitivity [115]. A true model of the endometriotic tumor microenvironment should include increased iron levels, higher levels of glycolysis-associated proteins, and endometriosis-associated lipidomic profiles.
ovarian cancers [97]. Thus, specific details of the macrophage population in endometriosis and ovarian cancer are important and require further study.
Transcriptomic work on endometriomas showed that the inflammatory cytokine transforming growth factor beta 1 (TGFβ1), regulates other inflammatory mediators relevant to endometriosis, including tumor necrosis factor alpha (TNFα) and interleukin-6 (IL6) [92]. These inflammatory mediators are highly elevated in peritoneal fluid from women with endometriosis [98][99][100][101]. The acute and chronic inflammation of endometriosis is a response to the invading tissue, leading to the release of regulated on activation normal T cell expressed and secreted (RANTES), monocyte chemotactic protein-1 (MCP1), and interleukin-8 (IL8), which act as chemoattractants recruiting more macrophages to the area [102]. In terms of the endometriotic tumor microenvironment, the promotion of tumor invasion via macrophages may be dependent on TNFα [103], which is elevated in women with endometriosis [98,99]. Along the same lines, work using an estrogen receptor beta (ERβ)overexpressing syngeneic mouse model of endometriosis suggests that non-genomic effects of ERβ play a role in the TNFα-mediated dysregulation of endometriosis progression [104]. Encouragingly, treatment of a syngeneic mouse model of endometriosis with a long-acting TNFα-blocking agent decreased endometriotic implant size [105]. However, treatment of women with rectovaginal nodules with infliximab, a TNFα monoclonal antibody, had no improved clinical effect over placebo [106]. Understanding the immune response to misplaced endometrial tissue will be a large factor in understanding the onset and progression of endometriosis and lead to a better understanding of how endometriosis creates a unique and potentially tumor-promoting microenvironment. Figure 1. Composition of the endometriotic tumor microenvironment. Endometriosis represents a pathologically benign disease. Endometriosis may be classified into endometriomas, superficial peritoneal disease, or deep infiltrating endometriosis (invasion > 5 mm). Although deep infiltrating endometriosis is invading, typically into the muscularis layer of the bowel, it is clinically not associated with ovarian cancer. Endometriomas are epithelial lined cysts of the ovary, which can be filled with a brown cyst fluid, and thus the name "chocolate cysts." Endometriomas can be associated with ovarian cancer, with atypical endometriomas having a higher risk of malignant transformation. Atypical endometriomas are characterized by epithelial cells with enlarged hyperchromatic and pleomorphic nuclei, with cellular crowding and high nuclear-to-cytoplasmic ratio. The altered endometriotic tumor microenvironment may lead to malignant transformation or propagation of proliferative potential [107]. RANTES: regulated on activation normal T cell expressed and secreted; MCP1: monocyte chemotactic protein-1; IL: interleukin; TGFβ1: transforming growth factor beta 1; TNFα: tumor necrosis factor alpha; CDC42: cell division cycle 42; CXCL4: chemokine (C-X-C motif) ligand 4. Figure 1. Composition of the endometriotic tumor microenvironment. Endometriosis represents a pathologically benign disease. Endometriosis may be classified into endometriomas, superficial peritoneal disease, or deep infiltrating endometriosis (invasion > 5 mm). Although deep infiltrating endometriosis is invading, typically into the muscularis layer of the bowel, it is clinically not associated with ovarian cancer. Endometriomas are epithelial lined cysts of the ovary, which can be filled with a brown cyst fluid, and thus the name "chocolate cysts." Endometriomas can be associated with ovarian cancer, with atypical endometriomas having a higher risk of malignant transformation. Atypical endometriomas are characterized by epithelial cells with enlarged hyperchromatic and pleomorphic nuclei, with cellular crowding and high nuclear-to-cytoplasmic ratio. The altered endometriotic tumor microenvironment may lead to malignant transformation or propagation of proliferative potential [107]. RANTES: regulated on activation normal T cell expressed and secreted; MCP1: monocyte chemotactic protein-1; IL: interleukin; TGFβ1: transforming growth factor beta 1; TNFα: tumor necrosis factor alpha; CDC42: cell division cycle 42; CXCL4: chemokine (C-X-C motif) ligand 4.

Steroid Hormones
Endometriosis is an estrogen-responsive disorder with lesion-level hyperestrogenism. Specifically, endometriotic tissue differs from eutopic endometrial tissue by the high expression of aromatase (CYP19A1) and 17β-hydroxysteroid-dehydrogenase (17β-HSD) type 1 and the absence of 17β-HSD type 2 [107,116]. Aromatase converts androstenedione or testosterone to estrone and estradiol at the level of the endometriotic microenvironment. High levels of estradiol have been linked to IL8 and RANTES production, which facilitate proliferation, inflammation, and feedback to increased expression of aromatase [107,117]. Aromatase activity is also stimulated through prostaglandin E 2 , an inflammatory product of cyclooxygenase 1 and 2 (COX1/2), found in endometriotic lesions in high levels [118]. Inhibitors of prostaglandin E2 receptor show promising effects in a xenograft model of endometriosis [119]. At the endometriotic lesion level, there is significant feed forward production and maintenance of estrogen, associated with pro-tumorigenic qualities. Medical management of endometriosis with oral contraceptives lowers overall steroid hormone levels. This may explain why the protection from combined oral contraceptive therapy on ovarian cancer risk is more robust for women with endometriosis (odds ratio 0.21 (0.08-0.58), p = 0.003) compared to non-endometriosis population (odds ratio 0.47 (0.37-0.61, p < 0.001)) [120]. Thus, the role of steroid hormones on endometriosis-associated ovarian cancers needs further study.

Small RNA Molecules
Small RNA molecules are non-coding RNA molecules that can play an important role in the post-transcriptional regulation of gene expression. Multiple groups of small RNAs have been identified, such as microRNAs (miRNAs), small nucleolar RNA (snoRNAs), small interfering RNAs (siRNAs), and Piwi-interacting RNA (piRNAs) [121]. The most studied type of small RNA molecules in endometriosis-associated ovarian cancers are miRNAs. In general, miRNAs regulate gene expression by mRNA cleavage and translational repression [122,123]. Studies have shown that miRNAs are frequently dysregulated in endometriosis and endometriosis-associated ovarian cancers (reviewed in [53][54][55]). Compilation of dysregulated miRNAs in ovarian endometrioid and clear-cell adenocarcinomas, as well as endometriosis (Supplemental Table S1) shows dysregulated miRNA molecules for each tissue type [53,55,[124][125][126][127][128][129][130][131][132][133]. Figure 2 shows the number of miRNAs dysregulated in ovarian clear-cell and endometrioid adenocarcinomas, and endometriosis tissues. Supplemental Table S1 details the specific miRNA molecules in the each unique and overlapping group. MiR-126 was found downregulated in all three groups. While the function of miR-126 is still unknown, miR-126 was significantly downregulated in endometriosis compared with eutopic endometrium [134]. Additionally, downregulation of miR-126 induced non-ovarian cancer cell proliferation, migration, and invasion, mediated through numerous validated targets, such as PI3K, KRAS, and VEGF. Reduced levels of miR-126 were a significant predictor of poor survival of cancer patients, although women with ovarian cancer were not included in the study [135]. Thus, miR-126 may play a role in endometriosis and ovarian cancer, even though these functional studies did not have ovarian cancer samples with concurrent endometriosis.

Small RNA Molecules
Small RNA molecules are non-coding RNA molecules that can play an important role in the post-transcriptional regulation of gene expression. Multiple groups of small RNAs have been identified, such as microRNAs (miRNAs), small nucleolar RNA (snoRNAs), small interfering RNAs (siRNAs), and Piwi-interacting RNA (piRNAs) [121]. The most studied type of small RNA molecules in endometriosis-associated ovarian cancers are miRNAs. In general, miRNAs regulate gene expression by mRNA cleavage and translational repression [122,123]. Studies have shown that miRNAs are frequently dysregulated in endometriosis and endometriosis-associated ovarian cancers (reviewed in [53][54][55]). Compilation of dysregulated miRNAs in ovarian endometrioid and clear-cell adenocarcinomas, as well as endometriosis (Supplemental Table S1) shows dysregulated miRNA molecules for each tissue type [53,55,[124][125][126][127][128][129][130][131][132][133]. Figure 2 shows the number of miRNAs dysregulated in ovarian clear-cell and endometrioid adenocarcinomas, and endometriosis tissues. Supplemental Table S1 details the specific miRNA molecules in the each unique and overlapping group. MiR-126 was found downregulated in all three groups. While the function of miR-126 is still unknown, miR-126 was significantly downregulated in endometriosis compared with eutopic endometrium [134]. Additionally, downregulation of miR-126 induced non-ovarian cancer cell proliferation, migration, and invasion, mediated through numerous validated targets, such as PI3K, KRAS, and VEGF. Reduced levels of miR-126 were a significant predictor of poor survival of cancer patients, although women with ovarian cancer were not included in the study [135]. Thus, miR-126 may play a role in endometriosis and ovarian cancer, even though these functional studies did not have ovarian cancer samples with concurrent endometriosis.  Table S1 details the miRNAs in each group above.

Model Systems for Studying Rare Ovarian Cancers
Multiple model systems have been employed to study endometriosis and endometriosisassociated ovarian cancers (reviewed in [139,140]). This review will highlight the tumor microenvironment of the genetically engineered mouse models of endometriosis-associated ovarian cancers. We have chosen to focus on spontaneous models instead of transplant models (reviewed in [140]). Because there has yet to be a comprehensive mouse model that replicates ovarian cancer with endometriosis, this review will also focus on the role of immortalized cell lines, xenograft models, co-culture systems, and three-dimensional (3D) models.

Candidate Genes in Genetically Engineered Mouse Models
High-grade serous ovarian cancer is a genomically complex disease [141] and although neither endometrioid nor clear-cell ovarian cancer have been as extensively profiled, they are likely complex as well. For the study of genetically engineered mouse models, fortunately, both endometrioid and clear cell ovarian cancer have high frequency mutations in only a handful of genes: ARID1A, PIK3CA, CTNNB1, PTEN, and KRAS [33,[40][41][42]44,45]. Use of traditional Cre recombinase technology with candidate-gene floxed alleles has had mixed results in terms of single gene knockout developing endometriosis-associated ovarian cancers. Table 1 lists the promoters driving Cre recombinase, and Table 2 details the brief rationale behind the use of specific genes in these mouse models. Table 3 lists these genes with combinations of tissue-specific promoters driving Cre recombinase. Despite the promising allele targets and the tissue-specific promoters driving Cre recombinase, there are no genetic mouse models of endometriosis and concurrent ovarian cancer. Investigators have created genetically engineered mouse models, which developed ovarian low-grade serous, clear-cell, or endometrioid adenocarcinoma (Table 3). However, none of these models have concurrent endometriosis. This suggests that different genetic combinations are required to model concurrent endometriosis and ovarian cancer. The discussion below highlights the role of the microenvironment of each model, and how this microenvironment may be playing a role in ovarian cancer development. Even though the presented models do not completely represent the endometriotic tumor microenvironment, they are still useful for understanding development of endometrioid or clear-cell ovarian cancer.

Endometriosis
The only genetically engineered mouse model to spontaneously develop endometriosis with a single gene change is a highly innovative mouse model developed by Dinulescu et al. [162]. Using an oncogenic KRAS knock-in allele mouse (Kras G12D ), peritoneal endometriosis developed after injection of adenovirus-driven Cre (Ad Cre ) through the uterotubal junction to infect the ovarian bursa. This true peritoneal endometriosis model contained glandular epithelium and stromal components validated by molecular immunohistochemistry to cytokeratin 7, 8, and 20, estrogen receptor, progesterone receptor, smooth muscle actin, and CD10 [162]. Conversely, when Ad Cre was injected through the infundibulum to the ovarian bursa, the model develops ovarian endometriosis-like lesions without the stromal component [162]. A transplantation experiment hints that the peritoneal endometriosis is uterine or tubal in origin while the ovarian endometriosis-like lesions are ovarian surface epithelium derived [162]. While long-term follow up showed no development of ovarian cancer, future studies into the molecular lineage using secretory or ciliary markers may allow better definition of cell of origin [26,62]. A similar mouse model adds human mucin 1 (MUC1) to oncogenic Kras G12D with Ad Cre intrabursal injection [168]. This mouse model similarly exhibits endometriosis-like lesions of the ovary. Importantly, these mice developed an immune response to MUC1 with high numbers of CD4+ Foxp3+ regulatory T cells in para-aortic lymph nodes compared to uninjected mice without lesions [168]. Models which recapitulate the immune response are needed to study the endometriotic tumor microenvironment.
Because mice do not normally menstruate, modeling retrograde menstruation requires significant manipulation. In homologous mouse models of endometriosis, endometrium from an estrogenprimed donor mouse is injected into a syngeneic estrogen-treated recipient mouse. However, homologous mouse models such as these grow poorly without exogenous estrogen [162]. A variation is the menstrual mouse model. In this model, the donor mouse undergoes significant hormonal manipulation followed by a stimulation of the uterus leading to decidualization. Hormone withdrawal leads to degeneration of the endometrium with leukocyte invasion, similar to menstruation in women [169][170][171]. Donor sloughed endometrium is then placed into recipient syngeneic mouse. Using this approach, Cheng et al. placed oncogenic Kras G12V endometrial tissue into the subcuticular ventral abdomen of syngeneic mice without exogenous hormonal stimulation or matrix [172]. These lesions contained glandular epithelium, stroma, immune cells, extracellular matrix, and blood vessels with both estrogen receptor alpha and beta expression [172]. Similarly, Greaves et al. used a similar approach with endometrial tissue from a menstrual model of wild type mice. Using hormonally stimulated receptor mice, injection of tissue intraperitoneal with this non-genetically modified endometrial tissue leads to peritoneal endometriosis [173]. Again, these tissues were histologically and molecularly similar to human endometriosis [173]. Hormonal levels (i.e., endogenous versus exogenous high levels), tissue placement (i.e., subcuticular versus intraperitoneal), and genetic changes important to endometriosis-associated ovarian cancers (i.e., oncogenic KRAS, loss of function ARID1A) must be considered when using these menstrual endometriosis models. Additionally, genetically engineered mouse models that are unable to undergo decidualization such as Pgr Cre ;Arid1a f/f mice [160] do not allow such studies.

Clear Cell Ovarian Cancer
Poorly differentiated clear-cell ovarian carcinoma develops at 7.5 weeks post-injection in Ad Cre ;Arid1a f/f ;Pik3ca* H1047R female mice with 77% penetrance and with 57% of injected mice having peritoneal metastasis [157]. Similar deletion of ARID1A alone or with knock-in of Pik3ca mutations showed ovarian surface epithelium hyperplasia but no endometriosis [157,158]. Although clear cell features are present two weeks post-injection, endometriotic-like lesions are not described [157]. Microarray analysis, comparing primary ovarian tumors to contralateral un-injected ovary, found almost 600 genes dysregulated with significant enrichment in immune system function [157]. Consistent with an endometriotic tumor microenvironment, IL6 signaling was found to be increased in the primary tumors, peritoneal metastases, body fluids, and ascites [157]. IL6 signaling and tumor cell growth was blocked with IL6 neutralizing antibodies. While IL6 expression was also implicated in normal ovarian surface epithelium hyperplasia with ARID1A deletion or Pik3ca mutation alone, the combination further enhanced IL6 production [157]. Cross-species, global gene expression profiling showed similar dysregulated genes in this mouse model compared to ovarian clear-cell adenocarcinoma from women [174]. Together these data suggest that the deletion of ARID1A and mutation in Pik3ca* H1057R results in increased IL6 expression leading to the ovarian surface epithelial hyperplasia and eventually clear cell ovarian cancer. These tumor cells perpetuate IL6 production, creating a positive feedback loop of increased IL6 and increased cell (normal and cancerous) proliferation [157,174]. This interaction highlights how the tumor and its microenvironment can interact with one another to generate a more tumor-promoting environment.

Endometrioid
ARID1A, PIK3CA, CTNNB1, PTEN, and KRAS [33,[40][41][42]44,45] are commonly mutated in both endometrioid and clear cell ovarian cancers from women. However, manipulation of these genes in mice typically results in endometrioid but not clear cell ovarian cancer. On injection of adenovirus-driven Cre (Ad Cre ) into the ovarian bursa through the infundibulum of Pten f/f ;Kras G12D female mice generated female mice with 100% penetrance of highly aggressive and metastatic endometrioid ovarian cancer at 12 weeks. Interestingly, this mouse model has ovarian endometriosis-like lesions with either addition of oncogenic Kras G12D or deletion of Pten alone, but only results in endometrioid ovarian cancer when both Pten and Kras G12D are simultaneously mutated [162].
A mouse model targeting both Pten and Apc resulted in endometrioid ovarian cancer with high penetrance and metastatic disease [161]. Unfortunately, this conditional knockout (Ad Cre ) did not result in endometriosis, which may be due to the early (6-week post-injection) tumor development [161]. Another model of endometrioid carcinoma in mice utilized a double conditional knockout of Pten and Arid1a and intrabursal Ad Cre injection to show a progression of ovarian surface epithelium hyperplasia, endometrioid carcinoma, and finally poorly differentiated carcinoma [158]. The well-differentiated endometrioid carcinoma was confined to the ovaries, suggesting the place of origin, while the undifferentiated tumors had metastasized into the peritoneal cavity [158]. Guan et al. hypothesizes that ARID1A plays a role in both tumor initiation and progression but requires the collaborative second hit of Pten to produce tumors [158]. Although the hyperplasia was not linked to endometriosis in these mice, it does speak to an environment of uncontrolled cellular proliferation giving rise to endometrioid ovarian cancer when left untreated.
High nuclear β-catenin levels have uniquely been found in endometrioid ovarian cancer from women, where this nuclear accumulation leads to activation of the WNT pathway [149]. Gain-of-function deletion of exon 3 of Ctnnb1 leads to stable β-catenin expression in mice [175].
Amhr2 Cre Ctnnb1 f/f female mice have aggressive endometrioid ovarian cancers with 100% penetrance by 6 months. Addition of Pten deletion to this model allows for tumors that are even more aggressive by 6 weeks [165]. Similar to deletion of exon 3 of Ctnnb1, deletion of Apc leads to stable β-catenin and WNT signaling activation [149]. Only with deletion of Pten did mice develop ovarian tumors [161]. To model the progression of type I tumors to the more aggressive type II tumors, Wu et al. (2013) added Pik3ca E545K/+ to Apc f/f Pten f/f mice with Ad Cre and showed peritoneal and lung metastasis [167].
While these models used Amhr2 Cre or Ad Cre to focus genetic changes in the ovarian surface epithelium, other studies have created conditional genetic changes in the oviduct. When Apc and Pten were concurrently deleted in the fallopian tube using Ovgp1 Cre , endometrioid tumors of the ovaries developed in 10 of 15 mice, with 50% of those resulting in metastasis to the lungs or omentum [146]. Deletion of Pten in the fallopian tube by Pax8 Cre also resulted in endometrioid tumors. Specifically, 75% of female mice developed primary tumors in the fallopian tube by 7 months, and 75% of tumor-burdened mice had metastasis to the ovaries [147]. Deletion of Apc with Pgr Cre female mice revealed tumors in both the oviduct and ovaries. Specifically, 25 of 40 female mice developed endometrioid oviductal tumors, one of 43 developed granulosa cell tumors, and 12 of 43 developed endometrioid ovarian tumors. While these female mice had simple ovarian cysts, the authors did not specifically denominate them as endometriosis [166]. Taken together, these mouse models suggest that the oviduct and/or the ovary may be involved in endometrioid cancer development in the mouse.

Low-Grade Serous Ovarian Cancer
Addition of oncogenic Kras (Kras G12D ) with either Amhr2 Cre or Cyp19 Cre resulted in ovaries with abnormal follicles, which were non-tumorigenic but also non-mitotic and non-apoptotic [145]. Deletion of Pten using Amhr2 Cre did result in increased proliferation and increased cell survival of ovarian surface epithelium [163]. However, the loss of the tumor suppressor Pten alone is not tumorigenic in somatic cells of the ovary. When Pten is deleted in the context of oncogenic Kras with Amhr2 Cre , there is development of low-grade serous papillary cystadenocarcinoma [163]. Although no endometriosis was noted, these mice were shown to have ovarian surface epithelium hyperplasia and abnormal follicle-derived ovarian lesions. Mullany et al. continued work on the Kras G12D ;Pten f/f ;Amhr2 Cre mice and showed that ovarian surface epithelium cells, removed from mutant mice prior to tumor formation, developed into tumors when grown in soft agar [143]. This key result suggests that Kras and Pte// play a significant role in the development of tumors in the ovarian surface epithelium, and the genetic mutations are the primary driver, since tumor formation occurred even outside of the ovarian microenvironment [143].

Immortalized Cell Lines
Immortalized human ovarian cancer cell lines have been widely used for studying molecular mechanisms of ovarian cancer. Ovarian cancer cell lines are used to study cancer biology, connecting genetic and epigenetic alterations to cancer development, progression, and drug response. Importantly, ovarian cancer cell lines have been developed from different histological and molecular subtypes of ovarian cancer. Unfortunately, molecular characterization has revealed that common ovarian cancer cell lines (i.e., SKOV3, HEYA8) do not molecularly represent the histology of tumor of origin. The number of cell lines derived from either endometrioid or clear cell ovarian cancers is more limited than high-grade serous cell lines. However, molecular profiling, including attention to gene mutations common in these endometriosis-associated ovarian cancers (i.e., ARID1A, PIK3CA, CTNNB1, PTEN, and KRAS) and mutations common in high-grade serous (i.e., TP53), have allowed better molecular and biological distinction [176][177][178][179][180][181][182]. Table 4 shows the common endometrioid and clear-cell ovarian cancer cell lines, including lines that were not derived from endometriosis-associated ovarian cancers, but which may molecularly represent non-high grade serous cell lines. Even fewer cell endometriotic cell lines exist, with 12Z cells being the only widely shared epithelial-like endometriosis immortalized cell line [183]. For rigor and reproducibility, additional well-characterized endometriotic cell lines and possibly ovarian cancer cell lines derived from women with endometriosis need to be created.

Xenograft Models
Implantation of immortalized human cell lines typically requires immunocompromised mice. A Japanese group created telomerase transformed endometriosis epithelial cell lines and confirmed cellular growth, steroid hormone response, and lack of malignant transformation in nude mice [185]. Further, these cells have been used in xenograft models to study treatment effects of small molecular inhibitors in endometriosis [104,200]. However, limited distribution outside Japan has restricted the use of these cells for studies of endometriosis-associated ovarian cancers. A similarly developed endometriotic epithelial cell line (EEC16) does not grow in SCID mice [184].
In terms of the endometriotic tumor microenvironment, Komiyama et al. placed normal endometrium of women without endometriosis into SCID mice. RMG-1 cells, a clear-cell ovarian cancer cell line, were grown in mice then transplanted into mice with or without endometrial implants. Although the tumors weighed less when grown with endometrium, proliferation was significantly higher in mice with transplanted endometrium. Additionally, these tumors expressed high levels of TGFβ and IL6. Addition of normal human endometrium changed the xenograft model to a more endometriotic microenvironment [201].

Three Dimensional (3D) and Co-Culture Models
Immortalized cell lines in monolayer two-dimensional (2D) culture fail to recapitulate the complexity of tumor tissue. Tumors are three-dimensional (3D) structures, surrounded by other cell types and a unique extracellular matrix (ECM) that is biologically optimized for growth of each cell type [202]. To recapitulate this for in vitro model systems, immortalized cell lines can be grown in Matrigel, ultra-low-adhesive plates, or a hanging drop. Using these methods, many immortalized cell lines will form 3D spheroids. Three-dimensional spheroid models can be highly instructive towards the understanding of current drug resistance and new therapeutics because they better mimic the way 3D tumors or de novo spheroids interact with the surrounding microenvironment. Specifically, the architecture of spheroids results in non-heterogeneity of nutrient and drug penetration, which can cause differential responses to varying layers of the spheroid. For example, Lee et al. compared 31 ovarian cancer cell lines in both 2D monolayer and 3D spheroids to primary tumors. Three-dimensional spheroids showed slower rates of proliferation and decreased drug sensitivity than the same cells grown in 2D [202]. Additionally, these 3D spheroids mimicked histological characteristics of primary tumors. Although the authors did not perform genome-wide transcriptomic analysis, candidate biomarkers such as mucin 16, cell surface associated (CA125), Wilms Tumor 1 (WT1), estrogen receptor, Paired box gene 8 (PAX8), and β-catenin were examined by IHC on a tissue microarray composed of 2D and 3D samples. The expression of these biomarkers correlated well with expression in primary tumors [202]. These data suggest that 3D spheroid models alter the microenvironment in a potentially more biological way compared to other in vitro systems. Additionally, Lal-Nag et al. used high-throughput screening to test multiple oncological drugs against the HEYA8 cell line. The cells responded differently to various drugs if they were grown in monolayer, in the process of forming spheroids, or already in pre-formed spheroids. This work establishes that the dimensionality of ovarian cancer cells plays a role in how they respond to their environment [203]. Similarly, Chowwanadisai et al. created cisplatin-resistant ovarian cancer spheroids by treating cells with sub-threshold doses of cisplatin, which resulted in a mesenchymal-enriched gene expression signature [204]. While molecular changes within spheroids may play a role in chemotherapy resistance, size of spheroids, similar to remaining disease after debulking surgery, plays a role in response. Tanenbaum et al. [205] showed that small spheroids treated with either short-term high-dose or prolonged low-dose cisplatin underwent significant shrinkage. Importantly, large spheroids preferentially responded to short-term high doses of cisplatin [205]. The investigators did not explore if the remaining cells became chemotherapy-resistant [205]. Although immortalized cell lines from endometrioid or clear cell ovarian cancers have not been extensively tested in 3D culture, we anticipate that they would behave similarly.
In addition to single cell types within 3D spheroids, co-culture systems can be useful. For example, endometrial epithelial cells are inhibited at a rate of 65-80% when grown in co-culture with endometrial stromal cells, highlighting the need for complex co-culture models [206]. Additionally, co-culture models of epithelial and stromal endometriosis cells show that stromal cells are responsible for metabolism of iron. The authors hypothesize that storage of iron by stromal cells is protective against malignant transformation of epithelial cells. Specifically, a lack of stromal cells and an abundance of epithelial cells, which cannot metabolize iron, leads to oxidative damage and oncogenic change [207]. This hypothesis fits with data from Anglesio et al. showing tumorigenic mutations in KRAS in epithelial cells of endometriosis but not stromal cells [27]. Similarly, co-culture of macrophages with endometriotic epithelial or endometriotic stromal cells leads to an increase in invasion that is more robust in epithelial than stromal cells [208]. Three-dimensional organoids made from endometrium and decidua have been developed simultaneously by two independent laboratories and represent promising models for in vitro study [209,210]. Development of additional endometriotic tumor microenvironment models are needed to study ovarian cancer cells within spheroids, 3D organoids, or co-culture systems.

Future of Precision Therapy for/or Prevention of Ovarian Cancer
Endometriosis is a known risk factor for ovarian cancer [41]. However, early treatment of endometriosis represents a known prevention strategy for ovarian cancer. For example, a woman on oral contraceptive therapy has a more robust protection against ovarian cancer if she has endometriosis than if she does not [120]. While treatment of endometriosis with contraceptives is effective, women desiring fertility do not enjoy the side of effects of contraception, and when medical management is stopped, 73% of women have return of symptoms. Additionally, surgical treatment of endometriosis with removal of one or both ovaries results in significant decrease in ovarian cancer risk. However, 55% of women undergoing local resection of endometriosis will have at least one more surgery over the course of seven years [211,212]. Morbidity associated with multiple operations makes selection of timing for endometriosis surgery important in pre-menopausal women. New treatments for endometriosis are needed. Importantly, discovery of new treatments for endometriosis should be a priority for ovarian cancer funding agencies as these therapies may lead to prevention of ovarian cancer.
In terms of therapy highlighting the importance of the molecular signaling between cells within tumors, Mok et al. used a systems biology approach to study individual cell types. Machine learning with large databases of drugs and molecular effects highlighted an FDA-approved drug for potential targeted treatment. While this study used high-grade serous ovarian tumors, it brings forward the importance of non-epithelial ovarian cancer cells in cancer treatment [213]. Importantly, the study focused on TGFβ signaling pathways [213]. Endometriosis also has dysregulated TGFβ signaling pathways [92]. Similar treatment of endometriosis may prevent ovarian cancer.

Conclusions
The Gynecologic Cancers Steering Committee of the National Institutes of Health (NIH) proposed strategic priorities for ovarian cancer. These research priorities focus on discovery of biomarkers, identification of cancer subsets to drive treatment recommendations, immunotherapy, combination therapies, and manipulation of the host-tumor microenvironment. While these priorities are not specific for a particular histotype, they are highly applicable to both the more common high-grade serous and less common endometriosis-associated ovarian cancers and warrant further study in endometriosis-associated ovarian cancer models.

Conflicts of Interest:
The authors declare no conflict of interest.