Although HGSOC is typically highly sensitive to chemotherapy, a small subgroup (<10%) is refractory to first-line therapy, pointing to a mechanism of inherent resistance. However, even after a clinical remission, most patients suffer from a relapse of the disease (1). While some of these patients are refractory to chemotherapy due to acquired chemoresistance, the majority undergo remission under the same treatment regimen. This regrowth of lesions displaying a similar chemosensitivity as the primary disease points to a mechanism of therapy failure that is fundamentally different form intrinsic or acquired resistance. However, the mechanisms underlying this transient chemoresistance are unknown.

A number of studies have associated chemoresistance with epithelial–mesenchymal transition (EMT), cell cycle arrest, blocked apoptosis, drug efflux, and several signaling pathways, including TGFβ, WNT, and NOTCH, but these observations did not yield a deep understanding of the mechanisms leading to relapse of the disease (17). It has also been a topic of intense research to clarify whether the regrowth of tumors after a complete clinical response is caused by a small population of cancer stem cells that are endowed with stem-like properties (18–20). However, multiple studies showed that ovarian cancer cell subpopulations express stemness markers at highly variable levels in different combinations and with none of these markers being obligatory (21–26). These findings suggest that a common or early ovarian cancer stem cell may not exist or has not been identified yet.

Comprehensive genomic studies by The Cancer Genome Atlas (TCGA) consortium have confirmed the prevalence of the genetic alterations described earlier and identified a number of recurrent, but infrequent changes (2). A more recent study has identified PTEN loss as another common driver event associated with a poor prognosis (27). This study also defined four transcriptional subtypes of ovarian carcinoma (differentiated, proliferative, immunoreactive, and mesenchymal), which were, however, not associated with differences in survival. In contrast, a recent reanalysis of the TCGA data indicated a favorable outcome for the immunoreactive subtype, presumably due to partial antitumor immune response (28). However, new insights into the precise mechanisms of chemoresistance were not gained by these studies. Whole-genome sequencing of DNA from patients with primary refractory, sensitive, and matched acquired resistant disease suggested that (i) inactivation of the tumor suppressor genes RB1, NF1, RAD51B, and PTEN, (ii) reversions of germline BRCA1/2 mutations or loss of BRCA1 promoter methylation, and (iii) overexpression of MDR1 are associated with, and may contribute to, acquired chemotherapy resistance (29). In contrast, amplification of CCNE1 was frequently observed in primary refractory cancer. However, these findings have not been linked to therapy failure or chemoresistance. Finally, a microarray study published in 2005 (30) identified an 11-gene signature characteristic of relapsed HGSOC, but mechanistic links to therapy resistance remain hypothetical.

It has been suggested that treatment failure is caused, in part, by cells escaping chemotherapy-induced cell death by entering a transient state of resistance, e.g., by entering a non-cycling state with low metabolic activity. This is a characteristic feature of detached tumor cells and spheroids floating in the peritoneal fluid or ascites. However, tumor cells surviving chemotherapy can trigger disease recurrence only if they are able to invade the peritoneum or omentum to establish proliferative lesions. Therefore, cancer cell adhesion to, and invasion of, the mesothelial cell layer lining these organs, and thus the mechanisms of transcoelomic dissemination appear to be central aspects of therapy failure. Conditional resistance can also be modulated by interactions with host cells (31–33), whose presence and functional states are spatially and temporally highly heterogeneous.

Several studies investigated the immunomodulating potential of EVs derived from different cell types. Strikingly, EVs released by dendritic cells (DCs) and B cells reveal antitumor and immune stimulatory effects and can provide costimulatory proteins that are able to activate T cells when presented with tumor antigens (94). Based on these observations, DC-EV-based antitumor vaccination studies were developed and tested in clinical trials (95–97).

On the contrary, exosomes secreted from ovarian tumor and bystander cells suppress the immune response (9). EVs collected from ovarian ascites impaired the cytotoxic activity of peripheral lymphocytes and induced apoptosis of lymphocytes and DCs. This correlated with the vesicle-associated expression of FAS ligand (FAS-L) and TRAIL, both known to be involved in apoptosis induction of immune cells. In line with these findings, secreted vesicle-associated FAS-L purified from patient ascites could trigger FAS-mediated apoptosis in a T cell line (98). These results suggest that vesicle-associated FAS-L counterattacks Fas-bearing immune cells including T cells, thus promoting immune evasion to support tumor cell survival. Moreover, there is evidence that ovarian ascites EVs can inhibit T cell activation by arresting the T cell-signaling cascade (99). This was mechanistically attributed to phosphatidylserine (PS), which is present on EVs. However, the functional role of PS remains to be confirmed since this molecule is required (but not sufficient) for the uptake of EVs by recipient cells (100).

Extracellular microvesicles do not only interact with cytotoxic lymphocytes but also regulate the activation of MDSCs via vesicle-associated HSP70, which engages toll-like receptor (TLR) 2 on MDSC (101). There is experimental evidence that these mechanisms might provide a target for successful immunotherapies. Depletion of HSP70-positive EVs and blockade of the HSP70/TLR2 interaction using a specific peptide aptamer diminished the activation and number of MDSCs significantly and inhibited tumor progression in tumor-bearing C57Bl/6 mice. Thus, targeting or interfering with EVs holds promise for novel immunotherapies to treat ovarian cancer. However, the development of innovative therapeutic concepts requires a more detailed understanding of EV biogenesis, molecular composition, and the mechanisms of target cell interaction.

Most studies on ovarian cancer ascites EVs are related to their impact on immune evasion, but there is also evidence that EVs regulate invasion and metastasis. First, ovarian cancer EVs carry molecules that directly regulate tumor cell migration in an autocrine or paracrine fashion, including soluble L1 (102), CD24, EpCAM (103), and soluble ALCAM (104). Second, membrane vesicles from malignant ascites were also found to contain the activated proteases: matrix metalloproteinase (MMP)-2, MMP-9, uPA (105), and ADAM17/TACE (104). These vesicle-associated proteases promote extracellular matrix (ECM) degradation directly or through the release of soluble factors from the EVs, which may allow tumor cell invasion into the stroma to facilitate metastasis. Third, EVs may carry miRNAs, which interfere with invasion and migration either positively or negatively. One prominent example is the transfer of miR-6126, which targets integrin β1, a key regulator of cancer cell metastasis. The transfer of miR-6126 mimic to endothelial cells (ECs) reduced invasion and migration of ovarian cancer cells in vitro and tumor growth in vivo (106).

The loading of EVs with specific miRNAs and the transfer of these miRNAs to recipient cells may also affect drug resistance. The vesicle-mediated transfer of miR21 from omental cancer-associated adipocytes and cancer-associated fibroblasts (CAFs) to ovarian cancer cells was recently identified as a pathway causing resistance to taxane-based chemotherapy. In the recipient cancer cells, miR-21 directly targets APAF1 mRNA, which is indispensable for the initiation of apoptosis in response to taxanes (32, 107).

Taken together, the vesicle-mediated cross talk between cells of the ovarian cancer ascites emerges as a pivotal component that establishes a protumorigenic environment promoting—but probably not limited to—immune suppression, invasion, and drug resistance. It remains a challenge to gain a better understanding of the mechanisms regulating cargo sorting into vesicles and their biosynthesis to establish links between vesicle-associated signatures, effects on signaling pathways, and altered functions in recipient cells. This knowledge is a prerequisite to be able to develop new therapeutic options for ovarian cancer by interfering with vesicle biogenesis, loading with specific cargo or vesicle uptake.

Tumor-associated macrophages represent the major subpopulation of myeloid cells in ovarian cancer ascites. A hallmark of macrophages is their plasticity in response to their microenvironment (108). Classical activation results in immune stimulatory, pro-inflammatory cells, while alternatively activated macrophages comprise a wide spectrum of subtypes with functions in tissue repair, angiogenesis, and immune regulation. TAM activation is skewed by factors of the TME to adopt a spectrum of phenotypes that represent mixed forms of alternatively activated and pro-inflammatory macrophages (108), which has also been clearly demonstrated for TAMs in ovarian cancer ascites (12). TAMs do not possess tumoricidal activity, but are rather thought to promote immune suppression and various aspects of cancer growth and progression, including tumor cell invasion, angiogenesis, and metastasis (108) (Figure 3). Consistent with these tumor-promoting functions of TAMs, expression of the alternative activation marker CD163 in TAMs from malignancy-associated ascites showed a strong correlation with early relapse of serous ovarian carcinoma after first-line therapy (12). Among the soluble factors contributing to TAM polarization, tumor progression, and a poor clinical outcome, IL-6, IL-10, TGFβ, and AA play a prominent role (12, 40, 48–50, 109).

As shown in mouse models, normal macrophages can have two developmentally different origins. While infiltrating macrophages are derived from blood monocytes produced by the bone marrow, resident tissue macrophages, including peritoneal macrophages, are of fetal (yolk sac) origin (110–116). Although TAMs can be derived from recruited blood monocytes (117–119), tissue-resident macrophages can also be a substantial source for the generation of TAMs (112, 120–125). A recent study (109) has revealed a surprising similarity between TAMs from ovarian cancer ascites and resident peritoneal macrophages with respect to their global transcriptional profile, the expression of differentiation markers, and their activation state. For example, both resident peritoneal macrophages and TAMs are characterized by high expression of the alternative activation markers CD163 and CD206, and both TAMs and peritoneal macrophages express genes with essential functions in phagocytosis and antigen presentation at similarly high levels. The only clearly discernible difference between both cell types was an upregulation of genes linked to ECM remodeling in TAMs (109). On the other hand, blood monocytes are most likely attracted by chemokines of the tumor miroenvironment and therefore consequently contribute to the generation of TAMs. However, it remains an unresolved issue at present to which extent infiltrating monocytes and resident peritoneal macrophages represent the origins of ovarian cancer ascites-associated macrophages.

A therapeutically beneficial effect of “re-educated” TAMs has been reported for patients with pancreas adenocarcinoma and a related mouse model (126), suggesting that TAMs may also represent a potential target in ovarian cancer, as both ascites and tumor tissue typically show a high density of TAMs. TAMs from ovarian cancer ascites can indeed be shifted under experimental conditions to trigger a partial cytotoxic activity under experimental conditions (127, 128).¹ However, the observations are partly contradictory, and activity toward autologous tumor cells has not been analyzed. It is, therefore, unclear, whether, and if so under which conditions, ovarian cancer TAMs could be (re)differentiated to macrophages with tumoricidal properties or the ability to trigger a cytotoxic response by other immune cells under clinically relevant conditions.

Interleukin-12 is a particularly interesting cytokine in the context of ovarian cancer due to its immune stimulatory antitumor effects and its inverse associations with disease progression (129–131). A hallmark of TAMs in ovarian cancer ascites is their defect to release IL-12 in response to inflammatory stimuli, which results from a transcriptional block of the IL12B gene encoding the p40 subunit (109, 127, 132). Another cytokine with beneficial immune stimulatory and antitumor effects in ovarian cancer patients is interferon (IFN)γ (133–136) consistent with the observation that IFNγ can prevent the skewing of monocyte differentiation by ovarian cancer ascites from immune stimulatory IL-10^lowIL-12^high macrophages to TAM-like IL-10^highIL-12^low cells and to abrogate the suppressive effect of ovarian cancer ascites on the inducibility by TLRs ligands of IL-12 secretion (127) (see text footnote 1). As IL-12 is an essential determinant of a cytotoxic immune response, understanding the regulatory network impinging on the IL12B gene may be a key to the development of efficacious TAM-targeted therapies.

Besides macrophages, other myeloid cells and NK cells represent the two main subsets of innate immune cells displaying efficient yet mechanistically distinct tumor suppressive effects. Consequently, both cell types are targeted by tumor immune escape mechanisms. One prominent mechanism is the differentiation of myeloid cells to MDSCs. These cells comprise a heterogeneous population of immature myeloid cells, including immature precursors of macrophages, granulocytes, and DCs. MDSCs are capable of inhibiting both adaptive and innate immunity via suppression of lymphocytes and NK cell activity. Like TAMs and Tregs, MDSC significantly impact on patient survival by enhancing cancer progression and metastasis (137).

Recruitment of monocytes and their differentiation into MDSC or TAMs strongly depends on components within the TME. In ovarian cancer, the tumor-associated inflammatory mediator PGE₂ was shown to attract MDSC into ascites. This was dependent on the induced expression of functional CXCR4 and its ligand CXCL12, facilitating migration of MDSC into the ascites (58). Direct or indirect blocking of PGE₂ production could effectively inhibit MDSC responsiveness, providing a rationale to target PGE₂ signaling therapeutically (58). Interestingly, tumor-infiltrating MDSC are the predominant producers of IL-10 and also depend on IL-10 to develop their immunosuppressive function in vivo further, confirming the fundamental role of IL-10 for the development and maintenance of a permissive TME (138). The molecular mechanisms that regulate the differentiation and function of MDSCs in cancer have been partly identified (137); however, the mechanisms that recruit MDSCs to the tumor tissue and ovarian cancer ascites remain poorly understood.

Another important key component of the innate immune system for recognizing and eliminating cancer cells is NK cells. The most important cytotoxicity receptors that mediate NK cell-dependent immunosurveillance are the CD16 receptor, the NKG2D receptor, and the natural cytotoxicity receptors, including NKp30. Defects in NK cell function such as impaired cytotoxicity/cytokine secretion, aberrant receptor and ligand expression, reduced NK cell number, or NK cell anergy are reported in malignant diseases and involved in immune escape (139). In ovarian cancer ascites, two main pathways inhibiting NKG2D (140) and NKp30 (141, 142) activity were described. First, ovarian cancer cells release the pro-inflammatory cytokine macrophage migration inhibitory factor (MIF), which stimulates tumor cell proliferation, migration, and metastasis and promotes tumor angiogenesis. MIF targets also NKG2D expression by transcriptionally downregulating NKG2D in NK cells to diminish cytotoxicity toward tumor cells (140). Second, NK cell activity in ascites is inhibited due to high levels of soluble B7-H6, one of the ligands for the NKp30 receptor. High expression of soluble B7-H6 was associated with a diminished NKp30 expression on tumor-associated NK cells and impaired NK cell activity (142). In line with this observation, lower B7-H6 expression of patients correlated with a better OS and reduced metastasis and cancer progression (141). Thus, the impaired NK cell functions in the ovarian cancer TME may contribute to immune escape, and the underlying mechanism should be further investigated.

Interleukin-18-primed NK cells release the chemokines CCL3 and CCL4 upon exposure to ovarian cancer cells to attract immature immature DCs. The NK-DC cross talk results in upregulation of CXCR3 and CCR5 ligands (CXCL9, CXCL10, and CCL5) on DCs, which in turn can recruit CD8⁺ effector T cells to the tumor environment (143). Thus, defective NK cell activation does not only affect NK-dependent killing of target cells but may also impact on the cellular interactions mandatory for the development of an antitumor immune response. A better understanding of the cross talk among innate immune cells in ovarian cancer ascites and the consequences for tumor maintenance and metastasis is one prerequisite for the development of elaborated immunotherapies.

The detrimental effect of CD4⁺ T cells was caused by an immunosuppressive subpopulation of T cells expressing the transcription factor FOXP3, termed Tregs, but not by FOXP3-negative CD4⁺ T cells (149, 152). Tregs are responsible for the maintenance of immune homeostasis by controlling autoimmunity, allergy, and inflammation, as well as responses to tumors. There is a high diversity within Tregs, dependent on their origin (thymus versus periphery), type of immune response they control, and their localization (lymphoid versus non-lymphoid tissue). Tumor-infiltrating Tregs belong to the non-lymphoid subpopulation, which differs from Tregs present in the circulation or lymphoid organs by the production of the immunesuppressive cytokines IL-10 and TGFβ as well as the expression of chemokine receptors (e.g., CCR4) and other membrane-associated immune modulators [e.g., cytotoxic T lymphocyte antigen 4 (CTLA4), GITR] (153, 154). Several reports described the presence of Tregs within ovarian cancer tumor-associated lymphocytes (152, 155, 156), and their accumulation correlated inversely with patient survival (157, 158). Tumor-associated Tregs suppressed the proliferation of, as well as IFNγ and IL-2 production by, CD8⁺ T cells specific for the tumor antigen HER2 and thereby counteracted the protective effect of tumor-specific effector T cells (158). The ovarian cancer environment caused migration of CTLA4⁺ FOXP3⁺ GITR⁺ Tregs via the chemokine CCL22, secreted by tumor cells and TAMs, and its receptor CCR4, expressed by infiltrating Tregs (155, 158) (Figure 4). Accordingly, the application of an anti-CCR4 antibody in an experimental model of ovarian carcinoma blocked Treg migration and enhanced antitumor response including increased IFNγ secretion by CD8⁺ T cells (159). In addition, “non-canonical” CD8⁺FOXP3⁺ cells (160) seem to be involved in ovarian cancer progression. These cells are dependent on TGFβ1 and p38 MAPK signaling (161), but their precise function remains unclear.

Besides negative effects of Tregs, there are further factors within the ovarian cancer environment that dampen CTL-mediated antitumor responses (Figure 4). These include inhibitory pathways mediated by CTLA4, PD-1, and lymphocyte activation gene 3 (LAG3) protein (162–165). On activation via the T cell receptor (TCR) and costimulatory molecule CD28, which provides an essential signal for growth and survival, T cells upregulate CTLA4. CTLA4 is a negative regulator of T cell activity by competing with CD28 for the B7-1 and B7-2 ligands on antigen-presenting cells (APCs), resulting in cell cycle arrest and attenuation of effector function (166). Both, PD-1 and LAG3 are strongly upregulated by CD8⁺ T cells, and their differentiation is skewed toward a dysfunctional pathway, termed functional exhaustion (167). This state was originally described for CD8⁺ T cells arising during chronic infections upon uncontrolled antigen exposure and is characterized by gradual impairment of proliferation and production of effector cytokines IFNγ and TNFα (167). However, the exhausted CD8⁺ T cells maintain some effector function, allowing them to restrict ongoing viral replication while preventing tissue damage (168). This principle might also apply to the ovarian cancer environment in which overwhelming antigen exposure and an inflammatory milieu may cause CD8⁺ T cell exhaustion (163, 164). Nevertheless, the increased numbers of tumor-infiltrating CD8⁺ T cells correlate with a better survival prognosis (144, 149), suggesting that TILs maintain their functionality at least partially as suggested by increased levels of IFNγ in tumor tissue from patients with a favorable clinical outcome (144).

Exhaustion can be reversed by blocking signaling ligands expressed on APCs, tumor cells, or tumor-infiltrating immune cells (167, 169). Therefore, the main therapeutic focus is currently placed on the development of antibodies blocking interactions of coinhibitory receptor–ligand pairs (e.g., CTLA4:B7, PD-1:PD-L1). These antibodies are termed immune checkpoint inhibitors and have demonstrated considerable benefit in clinical trials for several cancers including ovarian carcinoma (165, 169, 170). In a mouse model of ovarian carcinoma, neutralization of CTLA4 signals caused regression of growing tumors, however, only in an early tumor stage. This was accompanied by activation of CD8⁺ T cells for IFNγ production in vitro (171).

Anti-CTLA4 antibodies (ipilimumab) have been administered to ovarian carcinoma patients after vaccination with irradiated, autologous tumor cells engineered to secrete GM-CSF. This treatment resulted in antitumor effects only in a minority of patients, which is in sharp contrast to melanoma (162). The histologic analysis revealed a linear relationship between tumor necrosis and an increased ratio of intratumoral CD8⁺ T cells to FOXP3⁺ Tregs (162). This finding suggests that anti-CTLA4 antibodies could influence not only effector CD8⁺ T cells but also Tregs, which constitutively express CTLA4. Indeed, anti-CTLA4 monoclonal antibody therapy is associated with depletion of intratumoral FOXP3⁺CD4⁺ Tregs and probably with elimination of immune checkpoint blockade of effector CTLs (172). However, in animal studies, the antitumor activity of anti-CTLA4 monoclonal antibodies was attributed to the depletion of Tregs rather than to the direct activation of effector CTLs (173–175). Taken together, these results suggest that the interference with the CTLA4 pathway, leading to depletion of Tregs and probably to activation of effector CD8⁺ T cells, is involved in the control of ovarian carcinoma to some extent. This points to the existence of additional mechanisms suppressing CD8⁺ T cell responses, such as the upregulation of PD-1 and LAG3 expression on TIL (163).

Besides CTLA4, PD-1- and LAG3-expressing CD8⁺ T cells specific for tumor-derived antigen NY-EASO-1 were also detectable within the ovarian carcinoma infiltrates. These molecules were upregulated on CD8⁺ T cells by IL-6 and IL-10, present in ovarian carcinoma ascites, as well as by tumor-infiltrating DCs. The strongest functional impairment in the capability to produce IFNγ and TNFα was detectable within the population of CD8⁺ T cells coexpressing both LAG3 and PD-1, an effect reversible by the simultaneous blockade of these receptors (163). The synergistic suppression of CD8⁺ T cell activity by LAG3 and PD-1 was further evaluated in a mouse model of ovarian carcinoma and revealed that physical interactions between these molecules contribute to this cooperative effect (164). The suppressive effect of ovarian cancer ascites also involves dampening of the TCR-induced activation of transcription factors NFκB and NFAT, which is crucial for T cell activation, presumably resulting from an inhibition of signal transduction upstream of PLCγ (176). Consistent with the observations that both Tregs and inhibitory molecules contribute to the immune suppression in ovarian carcinoma, high levels of miR-424(322), which targets the PD-1:PD-L1 and CTLA4:B7 pathways, displayed a synergistic antitumor effect with chemotherapy by activating CD8⁺ T cells and reducing Treg infiltration in the mouse model (177).

The blockade of PD-1 by anti-PD-1 antibody (nivolumab) in patients with platinum-resistant ovarian cancer showed encouraging results (165). In addition to suppressing the activity of infiltrating CD8⁺ T cells, the migration of these cells into the TME is prevented by epigenetic suppression of Th1-associated cytokines CXCL9 and CXCL10 (178). Thus, epigenetic regulators of Th1-chemokine production caused increased migration of CD8⁺ T cells into the tumor environment associated with tumor reduction and improved the antitumor efficacy of immune checkpoint blockade in response to anti-PD-L1 antibodies in a mouse model (178). These data indicate that regulation of both the influx and function of CD8⁺ T cells in ovarian carcinoma need to be taken into account for the development of more efficacious therapies. Taken together, the suppression of the antitumor activity of CD8⁺ T cells in the ovarian cancer environment seems to be evoked by multiple mechanisms, which include (i) the inhibition of CD8⁺ T cell migration, (ii) the induction of inhibitory molecules on CD8⁺ T cells, and (iii) modulated CD8⁺ T cell activity by increased presence of Tregs.

There is accumulating evidence in experimental models for an influence of the TME on the glycolytic metabolism in CD8⁺ T cells, which supports proliferation and effector function (179). The competition for nutrients between tumor cells and T cells has a crucial impact on cancer progression. Thus, decreased availability of glucose in the TME suppressed IFNγ production and antitumor function of CD8⁺ T cells along with reduced glycolysis (180). Likewise, restriction of glycolysis led to insufficient production of phosphoenolpyruvate and defects in Ca²⁺-NFAT signaling, thereby suppressed antitumor CD8⁺ T cell function (181). Moreover, elevated extracellular potassium concentrations in the tumor environment dampened CD8⁺ T cell antitumor function via impairment of signaling pathways linked to metabolism (182). Therefore, elucidating the influence of the ovarian cancer environment on CD8⁺ T cell metabolism might extend our understanding on immunosuppression and give new insights for therapy strategies.

The metastatic spread of ovarian cancer cells depends on the invasion into the mesothelium, a mostly squamous epithelium that covers all organs within the peritoneal cavity. The mesothelium is formed by a single layer of mesothelial cells with an underlying basement membrane composed of collagen, fibronectin, and laminin. An early event of metastasis involves the attachment of ovarian cancer cells present in the ascites to abdominal organs. It has been shown that cancer cells attach less firmly to mesothelial cells than to the basement membrane (183) and that metastatic sites are typically devoid of mesothelial cells (183, 184). It is therefore widely believed that mesothelial cells act as a barrier for cancer cells and represent the first line of defense against ovarian cancer and against other abdominally metastasizing cancers. It is well conceivable that ovarian cancer cells attach to ECM at the site of pre-existing lesions of the mesothelial cell layer, which might occur spontaneously at low frequency or might be fostered by the inflammatory environment of the ascites. However, there is no experimental evidence for this scenario so far.

Alternatively, ovarian cancer cells can attach directly to mesothelial cells, mainly through β1 integrin (185–187) and CD44 (186, 188, 189). This attachment to mesothelial cells is favored by a genetic program of ovarian cancer cells with upregulation of mesenchymal genes, including TWIST1 and ZEB1 (190), as well as downregulation of epithelial genes like CDH1 encoding E-Cadherin (191). The downregulation of E-Cadherin induces upregulation of α5β1 integrin, which mediates selective binding to fibronectin (192), providing a mechanistic basis for the enhanced ability to attach to mesothelial cells (191).

During the metastatic process, ovarian cancer cells and mesothelial cells mutually influence their behavior in multiple ways. Upon contact with mesothelial cells, ovarian cancer cells upregulate expression of the MMP-2 to cleave fibronectin and vitronectin, with the resulting fragments of fibronectin and vitronectin providing enhanced attachment to mesothelial cells through the integrins α5β1 and αvβ3 (193). The binding of α5β1 integrin on cancer cells to fibronectin on mesothelial cells promotes the activation of myosin in the ovarian cancer cells (194). As a result, the ovarian cancer cells exert myosin-dependent mechanical forces on the mesothelial cell layer to promote the disruption of mesothelial intercellular junctions and retraction of mesothelial cells leading to gaps within the mesothelial cell layer that can be utilized by ovarian cancer cells to invade into the submesothelial matrix (184, 194). This process, by which ovarian cancer cells impact on the integrity of the mesothelial cell layer, has been referred to as “mesothelial clearance” (194). As an alternative to invading through gaps in the mesothelial monolayer by inducing mesothelial retraction, ovarian cancer cells might gain access to the submesothelial tissue by actively killing mesothelial cells. Such a mechanism involving the expression of FAS-L on cancer cells and FAS on mesothelial cells has been shown for colon cancer cells in vitro (195).

In disagreement with the concept that mesothelial cells shield the underlying ECM from invasion by ovarian cancer cells, prometastatic functions of mesothelial cells have also been described. Ovarian cancer cells secrete TGFβ1 to activate a TGFβ1/RAC1/SMAD-mediated signaling pathway in mesothelial cells that results in secretion of fibronectin by mesothelial cells and increased ovarian cancer cell adhesion, invasion, and proliferation (196, 197).

After passing through the mesothelial cell layer, ovarian cancer cells adhere to the submesothelial basement membrane, locally disrupt the basement membrane, and invade into the underlying connective tissue (184) (Figure 5). This involves the activation of MET in ovarian cancer cells, which is triggered by binding of α5β1 integrin on cancer cells to fibronectin (198). Several lines of evidence strongly suggest that TAMs contribute to the adhesion and invasion of ovarian cancer cells by promoting the remodeling of ECM (199). A comparison of the transcriptomes of ovarian cancer TAMs to that of peritoneal macrophages from non-cancer patients revealed ECM remodeling genes as the main difference between these cell populations (109). The clinical significance of this observation is supported by a strong correlation of poor survival with the expression of these genes in the PRECOG data set (109, 200). The molecular basis of the cross talk of ovarian cancer cells with cells of their microenvironment (Figure 6), in particular with immune cells such as TAMs and T cells in the ascites, and its impact on ovarian cancer cell adhesion and invasion are just beginning to be understood. In the following, we will summarize data on metastasis-promoting factors in the ascites and the molecular signaling pathways induced by them.

A large body of evidence indicates that LPA, which is frequently present in the ascites at high levels, plays a major role in ovarian cancer metastasis (see The Lipid Signaling Network of Ovarian Cancer Ascites). LPA acts via activation of LPA receptors on ovarian cancer cells, the expression of which is upregulated under hypoxic conditions that are frequently encountered in the TME (77). The effects of LPA on adhesion, migration, and invasion of ovarian cancer cells are mediated by multiple intracellular signaling pathways downstream of LPA receptors as described above. Mechanistically, LPA increases the expression of β1 integrin on ovarian cancer cells with enhanced adhesion to collagen as well as of MMPs and promotes the degradation of and invasion into the ECM (201). Along the same lines, LPA also induces the downregulation of tissue inhibitor of metalloproteinases expression (89). Moreover, LPA has been shown to suppress senescence of ovarian cancer cells through LPAR₂ in the presence of wild-type p53 (202) and to induce the expression of VEGF (203).

Vascular endothelial growth factor plays a central role in neoangiogenesis, progression, and metastasis of numerous types of cancer. In ovarian cancer, VEGF is found in both the primary tumor and the ascites at high levels, and VEGF expression correlates with poor survival (204–206). Accordingly, bevacizumab, a monoclonal antibody targeting VEGF, has been approved for the treatment of advanced and relapsing ovarian cancer and prolongs progression-free survival (207). Ovarian cancer cells and TAMs constitute major sources of VEGF (52, 208) (Figure 1), but other host cells might also contribute to VEGF production (209–211). The expression of VEGF is induced by hypoxia (212, 213) and multiple other signaling molecules (214). VEGF supports ovarian cancer metastasis by acting on both ovarian cancer cells in an autocrine or paracrine manner and on other cell types including peritoneal ECs and immune cells (see below). The invasion-promoting effect of VEGF might be mediated, at least in part, by an upregulation of MMPS in ovarian cancer cells (215, 216). In addition to direct effects on cancer cells, VEGF promotes a permissive environment for ovarian cancer cell metastasis by acting on peritoneal ECs to promote angiogenesis and vascular permeability, leading to the formation of ascites (47). At sites of metastatic lesions, VEGF-induced angiogenesis is likely to support the growth of larger metastatic nodules. Moreover, VEGF might further contribute to ovarian cancer metastasis by affecting immune cell functions as there is evidence suggesting that VEGF suppresses T cell activation and proliferation (211) and that VEGF levels inversely correlate with CD3+CD56+ NK-like T cell numbers (217).

Multiple aspects of ovarian cancer biology, including cancer cell adhesion and invasion, are under the control of a highly complex communication network between different cell types within the ovarian cancer TME. For some cytokines, ovarian cancer cells or immune cells have been identified as sites of production (Figure 1), while for other cytokines, the cell types releasing them remain unknown. TAMs represent a central player in the ovarian cancer cytokine network to support ovarian cancer cell adhesion and invasion (8). They secrete multiple metastasis-promoting cytokines including IL-6, IL-10, CCL18, CCL22, TNFα, and TGFβ (218). The number of TAMs is positively regulated by CCL2/MCP-1, which is released from ovarian cancer cells into the ascites (219). IL-6 released by TAMs promotes the adhesion, invasion, and proliferation of ovarian cancer cells (51, 220), which might be mediated in part by an increase of MMP expression by ovarian cancer cells (220). TNFα produced by TAMs also promotes invasion of ovarian cancer cells, the exact mechanism of which awaits further clarification (221, 222). Through the release of CCL22, TAMs attract Tregs to ovarian cancer cell clusters, which in turn suppress cytotoxic T cells (158).

CXCL12/SDF-1 released by an unknown cellular source into the ascites acts on cancer cells via CXCR4 (223). The expression level of CXCR4 is elevated by binding of ovarian cancer cells to ECM via β1 integrin (223). The activation of CXCR4 results in the upregulation of the prometastatic cytokines IL-6, CXCL12, and TNFα in ovarian cancer cells (8) and increases metastasis (224). IL-6 secretion by ovarian cancer cells is also increased by platinum-based chemotherapy and promotes the polarization of TAMs toward a protumorigenic and chemoresistant phenotype (225). Another mediator present in the ascites is the lipid sphingosine-1-phosphate (SIP), which stimulates invasion of ovarian cancer cells, probably through several different signaling pathways downstream of SIP receptors (226–229).

In addition to ovarian cancer and immune cells, adipocytes of the omentum are actively involved in cytokine signaling. They secrete IL-6, IL-8/CXCL8, CCL2, and adiponectin, which act on cancer cells via their respective receptors to support ovarian cancer cell metastasis (230). Adipocytes thereby establish a specific microenvironment with strong growth-promoting effects on disseminated ovarian cancer cells. Examples in this context are (i) the promotion of tumor cell homing and invasion into the omentum by adipocyte-derived IL-8 (230) in cooperation with CAFs, (ii) the direct transfer of fatty acids from adipocytes to adjacent tumor cells via fatty acid-binding protein 4 for energy production (230), and (iii) the activation of the ERBB3 receptor on metastasizing cancer cells by the mitogenic ligand neuregulin 1 (5).

Besides tumor cells, large populations of Tregs, TAMs, and possibly other host cells provide high levels of immunosuppressive mediators, thereby strongly limiting the fraction of cells competent to produce the crucial pro-inflammatory cytokines, such as IFNγ and IL-12. The influx of potentially cytotoxic T cells, monocytes, and NK cells cannot counteract this efficient tumor protection scheme: the dominant TME at advanced stages of the disease converts their phenotypes toward tumor promotion and immune suppression despite the presence of pro-inflammatory molecules, including cytokines, lipids, and nitric oxide as well as cellular debris and discharged intracellular molecules acting as ligands for TLRs (231, 232). This cocktail exerts strong suppressive functions on all types of immune cells, but at the same time promotes the protumorigenic properties of cancer cells, enabled by cell type-specific signal transduction mechanisms. It has been known for a long time that tumors resemble wounds that do not heal (233, 234). The final stage of wound healing, termed resolution phase, provides effective termination of inflammation. Therefore, it is tantalizing to speculate that the TME of advanced tumors lock immune cell types in a state resembling a never-ending resolution phase, while deregulated pro-inflammatory, protumorigenic signaling persists in the tumor cells. Consistent with this idea is the observation that ovarian cancer TAMs differ from normal peritoneal macrophages mainly by the upregulation of ECM remodeling genes (109), which is also the hallmark of resolution phase (235). Moreover, type I, III, and V collagens are major components of the TME, similar to a healing wound (233). These collagens are also among the matrix proteins upregulated in ovarian cancer TAMs, and their expression shows a striking association with poor survival (109).

A vast number of studies attribute the activation states of both innate and adaptive immune cells to differential activation of JAK/STAT and NFκB signaling. The JAK/STAT network is primarily regulated by cytokines that have either pro- or anti-inflammatory properties (236) and convey intercellular communication. NFκB proteins are effectors of receptors that detect either host-elicited or pathogen-derived signals and trigger mostly pro-inflammatory responses (237). In cancer cells, the very same networks exert mostly protumorigenic functions due to differential and constitutive activation of effector molecules and regulation of target genes. Thus, in cancer cells, STAT and NFκB have been implicated in the stimulation of proliferation, suppression of apoptosis, angiogenesis, promotion of invasion, and other protumorigenic processes (238). Due to the simultaneous presence of a plethora of both pro- and anti-inflammatory mediators, the ovarian cancer TME represents an excellent model to study how differential signal integration mediates cell type-specific biological effects (12, 14–16, 39, 52, 239). The following paragraphs describe the molecular mechanisms that, in a simplified view, enable unhindered tumor cell growth, driven by constitutive NFκB and STAT3 signaling, and simultaneously preclude immune cell activation by preventing pro-inflammatory NFκB via STAT3-driven mechanisms (Figure 7).

Regulation of IL-12 expression is a paradigm for the cross talk of NFκB and STAT signaling (Figure 8). In humans, macrophages exert their tumoricidal activity largely by activating NK cells via the secretion of IL-12 and IL-18 (240). Likewise, macrophages and other APCs use IL-12 to promote the differentiation of naïve T cells to cytotoxic Th1 cells. Occupation of the IL-12 receptor on T and NK cells induces STAT4 phosphorylation, leading to expression of IFNγ (241), a crucial activator of macrophages and class II major histocompatibility complex gene expression and thus an essential player in tumor surveillance (242, 243). IL-12 exerts its pivotal role in antitumor immune surveillance as a heterodimer of p35 and p40, encoded by the IL12A and IL12B genes, respectively. Expression of IL12B is under stringent transcriptional control and represents the rate-limiting step for IL-12 production (244).

Transcription of IL12B in macrophages requires the cooperative action of two transcription factors activated by different signaling pathways, i.e., the NFκB subunit REL and STAT1. While NFκB activation is triggered by pro-inflammatory cytokines and TLR ligands, STAT1-containing dimers are induced by interferons (245). STAT1 acts to orchestrate remodeling of promoter-proximal nucleosomes as a prerequisite for activation of Il12b transcription by NFκB proteins (246). Thus, the sequential action of STAT1 and REL in macrophages is crucial for the establishment of a cytotoxic immune response that depends on IL-12 production by macrophages. This provides for a highly efficient control mechanism that physiologically safeguards against detrimental inflammatory reactions in the absence of either activated STAT1 or REL. Tumors exploit this stringent regulatory network through signaling mediators that prevent the simultaneous activation of both pathways. IL-12 production is efficiently suppressed by numerous molecules found in ovarian cancer ascites, including IL-10, PS, fatty acid derivatives, and G-protein receptor ligands (247–259). The underlying mechanisms include both global and gene-specific suppression of the NFκB response in macrophages.

Apart from its role in APCs, REL is required for the expression of IL2 in T and NK cells and thereby necessary for their proliferation, survival, and activation (Figure 8). This includes the production of IFNγ, thus establishing a positive feedback loop between both cell types. Importantly, transcription of both IL12B and IL2 is fully dependent on REL; its function cannot be substituted by other NFκB proteins (260–262). However, dependent on the microenvironment, activation of REL in T cells can also dampen the immune response, e.g., in the presence of TGFβ, as is the case in ovarian cancer ascites. Thus, a number of mouse studies showed that the TGFβ-activated SMAD3 transcription factor cooperates with REL to induce expression of FOXP3 and that IL-2 contributes to FOXP3 induction by activating STAT5 (263, 264). As a consequence, FOXP3 drives the differentiation of naïve CD4-positive T cells to Treg, thereby preventing IFNγ production. These mechanisms are physiologically important to enable states of immune privilege or tolerance, which ensure that initiation of inflammation is efficiently prevented in reproductive organs, brain, and eyes. This regulatory network orchestrated by immunosuppressive macrophages and Tregs (265) is the blueprint for the immunesuppressive mechanisms invoked in the ovarian cancer TME.

In contrast to immune cells, NFκB is constitutively active in >50% of ovarian carcinomas and is correlated with poor survival (266–268). In fact, inflammatory signaling involving NFκB and STATs has been implicated in the genesis of many tumor entities. Activation of NFκB in cancer cells is most frequently caused by mediators of TME, since mutations of NFκB factors are confined to lymphoid malignancies (269). In ovarian cancer ascites, these mediators include IL-6, TNFα, as well as TLR and EGF receptor ligands (8, 12, 15, 52, 270). Proteins encoded by NFκB target genes include anti-apoptotic, growth-promoting, proinvasive, proangiogenic and immunesuppressive proteins, exemplified by BCLX-L, IL-6, IL-8, VEGF, and PD-1L, which explains its multifaceted role in ovarian cancer progression and therapy resistance (268, 271). Platinum-based chemotherapy also induces NFκB in ovarian carcinoma (272). Strikingly, RELA was found among the upregulated proteins in carboplatin-resistant tumors, as well as STAT5B (253). STAT5 and STAT3 are frequently hyperactive in human cancers, driving apoptotic blockade, invasion, stemness, and pro-inflammatory signaling, at least in part through persistent NFκB activation (273–275).

NFκB signaling impacts both protumorigenic and antitumorigenic mechanisms in different cell types; while enabling unhindered growth of tumor cells, it is also crucial for innate and adaptive immune responses as exemplified by its requirement for induction of IL12B. Accordingly, cautionary advice has been worded regarding broad pharmacological intervention (252, 256). Non-selective targeting, for example, inhibition of proteasome activity or bromodomain-mediated acetyl-lysine binding, has been tested in clinical studies, but it remains unclear whether these compounds have long-lasting benefits. Drugs that target factor(s) with protumorigenic roles in both tumor and stromal cells may be more promising targets. STAT3 was proposed as a common factor mediating negative effects (274, 275). However, caveats with the development of STAT3 antagonists are its high homology with STAT1 and its roles in barrier function and host defense, and considerable efforts dedicated to the development of specific and bioavailable STAT antagonists have met with little success (258). To circumvent the multifaceted roles of signaling networks, cell type-specific signaling will have to be identified and targeted to avoid the pitfalls of previous drug development strategies. For example, counteracting IL-10 (and other, currently unknown) TME-mediated immunosuppressive signaling events and thereby allowing for STAT1 activation, c-REL translocation, and IL-12 production might be a promising approach.