Diverse mechanisms for activation of Wnt signalling in the ovarian tumour microenvironment.

Wnt signalling pathways have been shown to play key roles in both normal development and tumorigenesis. Progression of many human cancers is associated with defined mutations in Wnt pathway components that result in dysregulated β-catenin-mediated gene transcription. Although Wnt pathway mutations are rare in epithelial ovarian cancer (with the exception of the endometrioid histotype), accumulating evidence supports a role for Wnt signalling in ovarian tumorigenesis in the absence of genetic mutations. The present review summarizes evidence in support of activated Wnt signalling in ovarian tumours and discusses alternative mechanisms for Wnt pathway activation in the ovarian tumour microenvironment.

INTRODUCTION

Overview of canonical Wnt signalling

Wnt signalling regulates a diversity of processes fundamental to embryogenesis, including proliferation, differentiation, polarity, adhesion and motility [1-3]. The highly conserved and complex Wnt pathway transduces signals from the extracellular environment through transmembrane receptors and co-receptors to have an impact on cytoskeletal rearrangements and gene expression changes and thereby modulate cell behaviour. In the adult organism, aberrant activation of these same biological processes can induce neoplasia and promote tumour progression [4]. There are two distinct pathways for transduction of Wnt signals: the canonical Wnt/β-catenin pathway (Figure 1) and the non-canonical β-catenin-independent pathway. The latter can be subdivided further into Wnt/PCP (planar cell polarity) and Wnt/Ca2+ signalling pathways. Many excellent reviews have been recently published that detail both the canonical and non–canonical Wnt pathways, thus the present review will provide only a brief summary of the canonical pathway as it relates to EOC (epithelial ovarian carcinoma).

Figure 1

Canonical Wnt signalling and nuclear β-catenin in ovarian tissue

(A) In the absence of Wnt pathway activation, β-catenin is either associated with the cytoplasmic tail of E-cadherin or is targeted to an Axin/APC/GSK3β complex whereupon it is phorphorylated and targeted for degradation (left-hand panel). Activation of Wnt signalling, for example by binding of Wnt to the Fzd receptor, results in association of the co-receptors LRP-5/6 with Fzd and recruitment of Dvl to Fzd at the plasma membrane. Binding of Axin to this complex disrupts the β-catenin degradation complex, enabling accumulation of β-Catenin. β-Catenin can then translocate to the nucleus, bind to Tcf/Lef transcription factors and transcriptional co-activators (not shown) to activate transcription of Wnt target genes. Wnt signalling can be blocked by Dkk that binds Kremen and inhibits the interaction between Fzd and LRP-5/6, or by sequestration of Wnt ligands via WIF or SFRP (right-hand panel). (B) Example of a serous ovarian tumour exhibiting nuclear β-catenin staining. Magnification ×200.

Canonical Wnt signalling is commonly activated by secreted proteins in the Wnt family, which currently consists of 19 members [1-3,5]. Wnts bind to transmembrane Fzd (Frizzled) GPCRs (G-protein-coupled receptors). Interaction of ligated Fzd with co-receptors designated LRP (lipoprotein receptor-related protein)-5 or LRP-6, members of the low-density LRP family, initiates Wnt signalling. In the absence of Wnt signalling, β-catenin functions as a structural component of E-cadherin (epithelial cadherin) junctions and is complexed with the cytoplasmic tail of E-cadherin. In normal epithelial cells, the majority of β-catenin is associated with E-cadherin at cell–cell junctions, and the levels are maintained at low concentrations in the cytoplasm by phosphorylation-dependent degradation of β-catenin. Cytoplasmic β-catenin is targeted to a complex comprised of Axin, APC (adenomatous polyposis coli) and GSK3β (glycogen synthase kinase 3β), resulting in phosphorylation of β-catenin that targets it for degradation through the ubiquitin–proteasome pathway (Figure 1A). Activation of Wnt signalling leads to phosphorylation of LRP-5/6, recruitment of Axin and Dvl (Dishevelled 1) to the plasma membrane, and functional disruption of the β-catenin degradation complex. This in turn enables accumulation of cytoplasmic β-catenin, which can then translocate to the nucleus, bind proteins in the Tcf (T-cell factor)/Lef (lymphoid enhancer factor) family, and activate transcription of Wnt target genes. Wnt signalling can be inhibited by sequestration of Wnt ligands via interaction with SFRP (secreted Fzd-related protein) or WIF (Wnt-inhibitory factor)-1 as well as by blocking co-receptor binding between Kremen and LRP-5/6 via Dkk (Dickkopf).

Accumulation of non-junctional β-catenin, and in particular nuclear translocation, is used as a surrogate marker for activation of the Wnt/β-catenin pathway. This is observed in a number of human cancers, as summarized below, and often results from mutations in Wnt pathway components. With the exception of the endometrioid histotype, Wnt pathway mutations are rare in ovarian cancer [6,7]. However, accumulating evidence suggests a role for Wnt signalling in ovarian tumorigenesis in the absence of genetic mutations, suggesting alternative mechanisms for Wnt pathway activation in EOC (Figure 1B) [8-10].

Ovarian tumours inhabit a unique metastatic niche

EOC is a leading cause of death from gynaecologial malignancy. Worldwide, each year approximately 204000 women are diagnosed with EOC, and 125000 die due to complications from the disease. Incidence is highest in the United States and Northern Europe and lowest in Asia and Africa. The majority of women are initially diagnosed after the primary tumour has already metastasized, resulting in a 5-year survival of <30%. The normal ovarian surface epithelium is mesenchymally derived and tissue cohesion is provided by the mesenchymal N-cadherin (neural cadherin). In contrast with most carcinomas that lose epithelial characteristics with tumour progression, EOCs undergo a mesenchymal–epithelial transition, acquire a more differentiated phenotype and gain expression of E-cadherin (reviewed in [11]). During progression, tumours acquire differentiated characteristics reminiscent of specialized Mullerian duct epithelia, resulting in various histotypes of EOC. Thus serous EOC resemble fallopian tubes, endometrioid EOC have characteristics of the endometrium, mucinous EOC is similar to the endocervix and clear-cell EOC tumours resemble vaginal tissue (Figures 2A–2D).

Figure 2

Ovarian tumours and the metastatic niche

Examples of epithelial ovarian carcinoma histotypes after haematoxylin and eosin staining at ×200 magnification: serous (A), endometrioid (B), mucinous (C) and clear cell (D). (E) Model of ovarian cancer metastastatic niche. The model depicts a primary ovarian tumour arising from malignant transformation of ovarian surface epithelium. Single cells and MCAs are shed from the primary tumour into the peritoneal cavity. Accumulation of carcinomatous ascites is commonly observed, particularly in women with advanced disease. Metastasis is the result of multiple intraperitoneal adhesive events, whereupon tumour cells attach to peritoneal mesothelium, disrupt mesothelial cell–cell contacts and migrate into the submesothelial matrix to anchor secondary lesions on the bowel, diaphragm, omentum and other sites.

Clinically, tumours often involve the ovary and omentum, with diffuse multi-focal intraperitoneal metastasis and malignant ascites [12]. Malignant cells are exfoliated as single cells and MCAs (multi-cellular aggregates) from the primary tumour into the peritoneal cavity (Figure 2E), wherein distribution is facilitated by peritoneal fluid. Formation of malignant ascites is believed to result from shed tumour-cell blockage of peritoneal lymphatics and may further facilitate metastatic dissemination. Shed tumour cells interact with mesothelial cells lining the peritoneal cavity, induce mesothelial cell retraction and anchor in the collagen-rich submesothelial matrix to establish secondary lesions [13]. Adhesion of EOC cells to collagen types I and III in the submesothelial matrix is mediated by α2β1 and α3β1 integrins [14-16]. This affinity for interstitial collagens is reflective of the mesenchymal origin of the ovarian surface epithelium. β1 Integrin-mediated adhesion represents an important early event in EOC metastatic dissemination, and adhesion-mediated integrin signalling probably contributes to metastatic success.

As EOC dissemination occurs largely within the peritoneal cavity, a unique microenvironmental niche is established comprised of tumour cells, inflammatory cells and soluble factors secreted by both the tumour and host cells including bioactive lipids, growth factors, ECM (extracellular matrix) proteins and inflammatory mediators. The contribution of the unique ovarian carcinoma microenvironment to the development and progression of metastatic disease is an area of active investigation. Aberrant activation of numerous signalling pathways has been observed in human EOC that result in stimulated proliferation or dysregulated cell death and most have been reviewed elsewhere. For example, activation of receptor tyrosine kinases such as the EGFR [EGF (epidermal growth factor) receptor] family and the VEGFR (vascular EGFR) family is frequently observed in EOC and regulates cell proliferation, survival, motility and metastasis [17,18]. Moreover, activation of VEGFR also alters vascular permeability, thereby contributing to build-up of ascites in EOC patients [19].

Activation of PI3K (phosphoinositide 3-kinase) signalling is observed in the majority of ovarian cancers as a result of gene amplification, activating mutations or activation downstream of receptor tyrosine kinases [20]. Signalling through the IL (interleukin)-6 receptor is also observed in EOC and results in nuclear translocation of STAT3 (signal transducer and activator of transcription 3) and corresponding stimulation of proliferation and angiogenesis [21]. Cytokine-mediated activation of NF-κB (nuclear factor κB) is also commonly detected in ovarian tumours, leading to up-regulation of anti-apoptotic and antioxidant gene products [22]. Lipid signalling via interaction of LPA (lysophosphatidic acid) with LPA GPCRs is a potent means of microenvironmental regulation, as LPA can be produced by tumour cells as well as stromal components, including mesothelial and inflammatory cells [23]. LPA binds to surface GPCRs, leading to activation of diverse signalling pathways depending on the LPA receptor complement and the specific G-proteins expressed by the cell. For example, LPA stimulation of Gα12/13 leads to activation of the small GTPase RhoA, resulting in cytoskeletal changes and cell rounding [24]. In addition, secondary effects of LPA occur as a result of induction of VEGF (vascular endothelial growth factor), TGF (transforming growth factor) α, TGFβ, IL-8, IL-6 and NF-κB expression, thereby contributing to activation of additional signalling pathways [25,26].

WNT SIGNALLING IN EOC

Progression of most cancers has been associated with activation of Wnt signalling acquired through two major routes: mutations in key components of the pathway or through mutation-independent aberrant gene expression. For example, the majority of colon, uterine and bladder tumours are strongly associated with APC, Axin or β-catenin mutations [27-31]. Interestingly, many cancers, including breast, prostate, lung, thyroid and pancreatic cancers (reviewed in [32-37]), do not depend on mutations in APC, Axin or β-catenin for activation of Wnt signalling.

Examples of Wnt signalling activation independent of mutations in key Wnt signalling genes

The presence of nuclear or cytoplasmic β-catenin in human breast cancer specimens is considered to be a strong indicator of activated Wnt signalling in this malignancy [38]. Multiple pathways, such as phosphorylation of β-catenin by EGFR/HER2 (human EGFR) [39], regulation of GSK3β activity by insulin, IGF-1 (insulin-like growth factor-1) [40], ILK (integrin-linked kinase) [41] or PI3K/Akt [42] and loss of PTEN (phosphatase and tensin homologue deleted on chromosome 10) [43] or p53 [44,45], have been associated with Wnt activation in breast cancers. Interestingly, activation of Wnt signalling in prostate cancer also depends on PTEN, PI3K/Akt [42] and IGF-1 [46], and can additionally be regulated by the androgen receptor [47]. In lung carcinoma, activation of Wnt signalling probably occurs through routes involving overexpression of Dvl [48] and repression of Wnt antagonists, such as WIF-1 and Dkk [49,50]. Activation of Wnt signalling in pancreatic adenocarcinoma has been related to overexpression of Wnt-1 and Fzd-2, which promotes stabilization of β-catenin [37].

Histotype-dependent Wnt signalling activation in ovarian carcinoma

Mechanisms for activation of Wnt signalling in ovarian carcinoma exhibit histotype dependence. Thus only endometrioid EOC is strongly associated with activating mutations in β-catenin leading to constitutively active Wnt signalling. The majority of low-grade endometrioid ovarian carcinomas often display nuclear immunoreactivity for β-catenin (70% of cases), and these cases often harbour mutations in the β-catenin gene at codons that encode for residues phosphorylated by GSK3β (54% of cases) [51]. Several studies confirmed the predominance of nuclear β-catenin and frequent β-catenin gene mutations in endometrioid EOC as well as in cell lines derived from this histotype [52-54]. Nuclear β-catenin in low-grade endometrioid EOC also associates with squamous differentiation and correlates with good prognosis and lack of relapse [51,55-57]. Moreover, expression of Wnt target genes including FGF9 (fibroblast growth factor 9) has been described, suggesting that this pathway is active in endometrioid carcinomas [55-57]. Mutations in Axin in cell lines of endometrioid EOC have also been reported [54]; however, a distinct study found no mutations in either APC or Axin in human endometrioid EOC [58]. Furthermore, high-grade endometrioid ovarian carcinomas do not display nuclear β-catenin immunoreactivity and progression is not associated with β-catenin mutations [51]. This evidence supports the existence of two distinct subtypes of endometrioid EOC that may originate from different sources, (reviewed in [59]) based on differences in molecular pathobiology.

In contrast with endometrioid EOC, ovarian carcinomas of serous, clear-cell and mucinous histotypes have only rarely been associated with activating mutations in the key proteins of the Wnt signalling pathway. One report identified cases of clear-cell EOC positive for nuclear β-catenin [60]. Another study identified mucinous EOC positive for mutations in the β-catenin gene in the GSK3β-binding region [61]. Nevertheless, several lines of evidence implicate activation of Wnt/β-catenin signalling in serous EOC in the absence of activating mutations in either APC, Axin or β-catenin. The strongest evidence is the presence of nuclear β-catenin (Figure 1B). A broad range (3–59%) of serous EOCs have been reported to contain nuclear and cytoplasmic β-catenin [60,62,63]. It is noteworthy that a significantly higher percentage of high-grade (23%) serous EOC correlated with the presence of nuclear β-catenin compared with low grade (2.1%) [63], opposite from trends observed for endometrioid EOC [51]. Together, these observations suggest that additional factors initiate Wnt signalling activation in serous EOC progression.

Potential mechanisms of β-catenin stabilization in serous EOC

Nuclear factors

As described above, the results of several studies indicate that ovarian carcinoma cells contain nuclear β-catenin, although the percentage of positive tissues and positive cells within each tested case vary widely. In the transcriptional activation complex, β-catenin partners with Tcf/Lef, Legless and Pygopus to initiate transcription. The presence of cytoplasmic Lef-1 has been reported in serous EOC and in NIH:OVCAR-3 (a serous EOC cell line). These authors also demonstrated co-immunoprecipitation of Lef-1 and β-catenin in poorly differentiated stage III serous EOC and NIH:OVCAR-3, but not in the normal ovary [9]. Comparison of Lef-1 expression in ten specimens of normal ovary and 23 specimens of serous EOC revealed no differences between the two groups [9]. The transcriptional co-activator Pygopus2 was found to be widely expressed in ovarian carcinoma tissues and cell lines of all histotypes, and was demonstrated to be required for tumour cell growth [64]. Furthermore, an additional co-activator designated Legless (BCL9) is widely expressed in both human EOC and normal ovary specimens at the RNA level (Table 1). Thus there are no apparent barriers for formation of the transcriptional activation complex, as the key players are expressed in serous EOC. Nevertheless, Wnt signalling is not constitutively activated in the majority of serous EOC cases or in the majority of individual cells in positive cases. However, it is important to note that currently available approaches do not enable detection of transient activation of Wnt signalling, suggesting that evaluation of Wnt target gene expression may provide potential evidence with which to evaluate signalling pathway activation.

Table 1

RNA expression of IDAX, BCL9 and RPL19 in EOC and normal ovary human specimens

Real-time reverse transcription–PCR was used to detect the levels of IDAX and BCL9 mRNAs in samples from non-cancerous ovarian tissues and ovarian carcinomas (Origene), according to the manufacturer's suggestions. Expression of RPL19 was detected as a positive control. Real-time PCR was carried out with the ABI Prizm (Applied Biosystems) and the Ct (threshold cycle) values were obtained according to the manufacturer's instructions. SYBR Green was used for quantitative PCR as a double-stranded DNA-specific fluorophore. The following primers were used for detection of IDAX, BCL9 and RPL19 mRNAs: IDAX, forward 5′-CAGCCAAGAAGAAGAGGA-3′ and reverse 5′-GGGAACAGGTGTTCTCTCTA3-′; BCL9, forward 5′-ACGACCTCAGAGCAGAGTAT-3′ and reverse 5′-GACAAGACAGTGCTGAAGAG-3′; RPL19, forward 5′-CAATGAAATCGCCAATGCCAACTC-3′ and reverse 5′-TGGACCGTCACAGGCTTGC-3′. *Information obtained from the manufacturer (OriGene) regarding Ovarian Carcinoma Panel I; †information on stage is related to ovarian carcinoma specimens only; ‡positive expression is defined by Ct value below 35, whereas negative expression (neg) is defined as the absence of a detectible signal or Ct value equal or higher than 35. FIGO, International Federation of Gynecology and Obstetrics; AJCC, American Joint Committee on Cancer.

			Ct values‡
Diagnosis*	Tumour grade*	Stage†	BCL9	IDAX	RPL19
Adenocarcinoma of endometrium, papillary serous	FIGO G3: poorly differentiated	Non-cancerous	27.7	29.3	20
Carcinoma of cervix, squamous cell	FIGO G3: poorly differentiated	Non-cancerous	28.7	27.4	20
Abscess of tissue	Not reported	Non-cancerous	30.7	30.4	20
Endometriosis	Not reported	Non-cancerous	33.1	30.9	20
Endometriosis	Not reported	Non-cancerous	29.8	28.9	20
Endometriosis	Not reported	Non-cancerous	33.2	32.5	20
Endometriosis	Not reported	Non-cancerous	32.1	31.4	20
Carcinoma of ovary, endometrioid	FIGO G2: moderately differentiated	I	34.4	33.4	20
Adenocarcinoma of ovary, papillary serous	FIGO G2: moderately differentiated	IA	30.4	33.0	20
Tumour of ovary, papillary serous, borderline	AJCC GB: borderline malignancy	IA	31.7	33.0	20
Tumour of ovary, papillary serous, borderline	AJCC GB: borderline malignancy	IA	33.2	30.6	20
Carcinoma of ovary, endometrioid	FIGO G1: well differentiated	IA	28.9	neg	20
Tumour of ovary, serous, borderline	AJCC GB: borderline malignancy	IA	30.3	30.3	20
Tumour of ovary, borderline	AJCC GB: borderline malignancy	IA	32.2	33.0	20
Tumour of ovary, mucinous, borderline	AJCC GB: borderline malignancy	IA	33.3	31.0	20
Adenocarcinoma of ovary, mucinous	FIGO G3: poorly differentiated	IB	30.7	34.0	20
Adenocarcinoma of ovary, endometrioid	FIGO G3: poorly differentiated	IB	33.6	neg	20
Tumour of ovary, borderline	Not reported	IB	31.8	31.4	20
Tumour of ovary, mucinous, borderline	AJCC GB: borderline malignancy	IC	30.7	28.4	20
Tumour of ovary, serous, borderline	AJCC GB: borderline malignancy	IC	31.0	neg	20
Tumour of ovary, serous, borderline	AJCC GB: borderline malignancy	IC	32.2	33.0	20
Adenocarcinoma of ovary, mucinous	FIGO G2: moderately differentiated	IC	32.1	34.9	20
Adenocarcinoma of ovary, endometrioid, sq. feat.	FIGO G2: moderately differentiated	IC	30.0	31.9	20
Adenocarcinoma of ovary, serous	FIGO G3: poorly differentiated	IIB	31.4	31.7	20
Adenocarcinoma of ovary, endometrioid	FIGO G3: poorly differentiated	IIB	30.9	neg	20
Adenocarcinoma of ovary, endometrioid	FIGO G1: well differentiated	IIC	31.4	31.3	20
Adenocarcinoma of ovary, papillary serous	FIGO G2: moderately differentiated	III	29.2	32.0	20
Adenocarcinoma of ovary, serous	FIGO G3: poorly differentiated	III	28.7	neg	20
Carcinoma of ovary, endometrioid	FIGO G2: moderately differentiated	IIIA	33.9	30.3	20
Tumour of ovary, serous, borderline	AJCC GB: borderline malignancy	IIIA	31.6	31.7	20
Adenocarcinoma of ovary, papillary serous	FIGO G3: poorly differentiated	IIIB	30.5	neg	20
Adenocarcinoma of ovary, serous	FIGO G2: moderately differentiated	IIIB	30.5	neg	20
Adenocarcinoma of ovary, endometrioid	FIGO G3: poorly differentiated	IIIB	31.7	neg	20
Adenocarcinoma of ovary, papillary serous	FIGO G2: moderately differentiated	IIIB	31.5	neg	20
Adenocarcinoma of ovary, papillary serous	FIGO G3: poorly differentiated	IIIB	30.9	neg	20
Tumour of ovary, serous, borderline	Not reported	IIIB	31.0	33.3	20
Adenocarcinoma of ovary, metastatic	Not reported	IIIB	32.0	34.2	20
Adenocarcinoma of ovary, papillary serous	FIGO G1: well differentiated	IIIC	29.1	neg	20
Adenocarcinoma of ovary, papillary serous	FIGO G3: poorly differentiated	IIIC	33.2	31.3	20
Adenocarcinoma of ovary, papillary serous	FIGO G3: poorly differentiated	IIIC	32.7	31.1	20
Adenocarcinoma of ovary, papillary serous	FIGO G3: poorly differentiated	IIIC	32.5	32.2	20
Adenocarcinoma of ovary, papillary serous	FIGO G3: poorly differentiated	IIIC	31.0	34.6	20
Carcinoma of ovary	FIGO G3: poorly differentiated	IIIC	31.9	33.6	20
Adenocarcinoma of ovary, papillary serous	Not reported	IIIC	32.3	32.0	20
Adenocarcinoma of ovary, serous	Not reported	IIIC	32.4	neg	20
Adenocarcinoma of ovary, papillary serous	FIGO G2: moderately differentiated	IV	30.6	27.7	20
Adenocarcinoma of ovary, metastatic	Not reported	IV	32.0	30.9	20
Adenocarcinoma of ovary, papillary serous	FIGO G3: poorly differentiated	IV	32.6	neg	20

Cytoplasmic factors

There are several proteins that participate in the process of phosphorylation and degradation of β-catenin in the cytoplasm, thus preventing translocation into the nucleus. Expression of GSK3β in 23 cases of primary and metastatic serous EOC was approximately 6-fold higher on average than that in ten normal ovarian tissues. Overexpression of FRAT1 (frequently rearranged in advanced T-cell lymphomas 1), which inhibits phosphorylation of β-catenin by GSK3β, strongly correlated with nuclear β-catenin and its accumulation in the cytoplasm in human serous EOC, suggesting a potential alternative mechanism for β-catenin stabilization [60]. Cytoplasmic immunoreactivity for APC was documented in 67 out of 113 (59%) serous EOC and was comparable in primary and metastatic tumours [62], suggesting that components of the β-catenin degradation complex are in place. The same study also reported that the absence of cytoplasmic APC only rarely coincided with nuclear β-catenin [62], suggesting that β-catenin degradation did not depend on the presence of APC. Although no data are available regarding the presence of Axin or Dvl, our findings suggest the presence of a Dvl inhibitor, IDAX (inhibitor of the Dvl and Axin complex), in both normal ovary and serous EOC tissues (Table 1). On the basis of the currently available data it is tempting to speculate that the equilibrium between cytoplasmic GSK3β and FRAT1 levels could functionally regulate Wnt signalling activation in serous EOC. In addition, a possible inhibition of Dvl through IDAX could contribute to the liberation of GSK3β and Axin from the complex with Frizzled, thus allowing for β-catenin degradation.

Membranous factors

Expression of Fzd1 and Fzd5 was tested in 26 normal ovary and 38 EOC (including 22 serous) specimens [65]. Interestingly, a higher number of malignant specimens was positive for both receptors relative to normal ovary (97.1% and 14.3% of malignant EOC were FZD1- and FZD5-positive, whereas only 54.5% and 8.7% of the normal ovarian tissues were positive respectively). Moreover, patients with FZD5-positive tumours had a 6-year probability of survival of 0.2 compared with 0.55 for those with FZD5-negative tumours. In addition, our data suggest that both LRP-6 and Kremen are expressed at the RNA level in the serous ovarian carcinoma cell line DOV13 (M.V. Barbolina and M.S Stack, unpublished work). Furthermore, engagement of collagen-binding β1 integrins, a process that occurs during adhesion to submesothelial interstitial collagens in intraperitoneal EOC metastasis, up-regulates LRP6 mRNA levels in serous OVCA 433 cells [66].

Secreted factors

Expression of an activator of canonical Wnt signalling, Wnt-1, was found in a larger number of malignant EOC (29.4% out of a total 38, of which 22 were serous) compared with normal ovary (9.1% out of 26 total) specimens [65]. However, the same study reported that the increase in Wnt-1 expression in malignant EOC did not lead to stabilization of β-catenin [65]. Expression of Wnt-5a was also highly positive in malignant EOC specimens (80%) compared with those of the normal ovary (27.3%). Of note, both Wnt-1- and Wnt-5a-positive EOC patients had a significantly lower chance of survival compared with their Wnt-1-and Wnt-5a-negative counterparts [65]. Similar to the results summarized above for LRP-6, collagen adhesion by serous EOC cells also induces expression of Wnt-5a [66]. Furthermore, Wnt-5a is prevalent in ascites fluid obtained from women with EOC (Figure 3). Although Wnt-5a is most commonly characterized as an inhibitor of canonical Wnt signalling [67-72], emerging evidence suggests that the pathway (canonical, non-canonical or planar cell polarity) activated downstream of Wnt ligand-binding is receptor context-dependent. Using in vitro models, Wnt-5a has been shown to inhibit canonical Wnt signalling via an alternate Wnt receptor, mRor2 (mammalian receptor tyrosine kinase-like orphan receptor 2); however, Wnt-5a activates canonical Wnt signalling in the presence of Fzd4 and LRP-5. This activation potentiates nuclear β-catenin accumulation and activates Wnt reporter constructs [73]. This is consistent with data showing that Wnt-5a enhances the proliferation, migration and invasion of pancreatic cancer cell lines via β-catenin-dependent signalling pathways. In that study, Wnt5a treatment promoted nuclear β-catenin localization, which was inhibited by siRNA (small interfering RNA) targeted against Wnt-5a [74,75]. Additional studies are needed to clarify the Wnt receptor/co-receptor profiles of EOC cells of various histotypes to clarify the role of Wnt-5a in EOC pathobiology.

Figure 3

Wnt-5a in malignant EOC ascites

Malignant ascites were collected from women with EOC and non-malignant ascites were collected from women with ovarian hyperstimulation syndrome under an institutional review board-approved protocol at Northwestern University School of Medicine. The protein concentration of each specimen was determined and samples containing equal protein (1 mg) were electrophoresed by SDS/PAGE, electroblotted on to Immobilon membranes and analysed by Western blotting using anti-Wnt-5a primary antibody (1:1000 dilution; R&D systems) and peroxidase-conjugated goat anti-(rat IgG) (1:5000 dilution; Santa Cruz Biotechnology). The blots were developed using enhanced chemiluminescence. Molecular mass in kDa is given on the left-hand side.

Expression of Wnt target genes in EOC

Indirect evidence for the existence of active Wnt signalling in serous EOC is expression of Wnt/β-catenin target genes in tumour cells or tissues. Currently over 100 target genes have been identified, the transcription of which is regulated by Wnt signalling (http://www.stanford.edu/~rnusse/pathways/targets.html). It should be noted, however, that many of these genes are also known to be regulated by other pathways, such that target gene expression alone is not sufficient evidence of activated Wnt signalling in serous EOC. Nevertheless, expression of many Wnt target genes, including CCND1 (cyclin D1) [76-79], COX2 (cyclo-oxygenase 2) [66,80], various MMPs (matrix metalloproteinases) [13,14,81-85] and MET [86-87] is observed in EOC, as summarized in Table 2, providing additional evidence in support of Wnt pathway activation.

Table 2

Expression of Wnt target genes in EOC

Overview of published data on Wnt target gene expression in EOC. BIRC5, survivin; CLDN, claudin 1; EDN1, endothelin 1; FST, follistatin; MMP11, stromelysin 3; PLAUR, plasminogen activator, urokinase receptor; PTTG, pituitary tumour-transforming 1; RARG, retinoic acid receptor γ; SOX, sex-determining region box; JAG, jagged; SNAI, SnaiL.

Gene	Reference(s)
CCND1	[78,79,167,168]
COX2	[111,169]
MMP2	[82,83]
MMP9	[83,170]
VEGF	[171]
CD44	[172]
MET	[173]
c-Myc	[174]
PLAUR	[175,176]
MMP7	[177]
CLDN	[178]
BIRC5	[179]
EON1	[122,123]
JAG	[180]
SOX9	[181]
SOX2	[182]
PTTG	[183]
MMP26	[184]
SNAI	[185]
FST	[186]
RARG	[187]
MMP11	[188]
LEF1	[189]

MICROENVIRONMENTAL ACTIVATION OF THE WNT PATHWAY

Intraperitoneal dissemination provides an unique microenvironment for ovarian carcinoma metastases. As outlined in Figure 2, progression of EOC is hallmarked by shedding of single and multi-cellular aggregates of malignant epithelial cells from the primary tumour [11,13]. Accumulation of malignant ascites in the peritoneal cavity is common, particularly in women with late-stage EOC. Thus metastasizing EOC cells exist in a milieu rich in inflammatory cells [88] and growth/signalling factors, including VEGF [89-91], EGF [11,92], TGFα, TGFβ [93,94] and LPA [23,95], providing ample opportunity for cross-talk between signalling networks. Ascites accumulation also modifies peritoneal mechanobiology, altering the force environment of both metastatic tumour cells and peritoneal mesothelium [96,97]. Formation of secondary tumours at peritoneal organs (colon, omentum, uterus and liver) is achieved by β1 integrin-mediated anchoring to the mesothelium and submesothelial matrix [14-16,98], representing a significant transition from a free-floating cell or aggregate to a three-dimensional matrix-anchored structure. Within this unique metastatic niche, current evidence suggests multiple opportunities for regulation of Wnt signalling via molecular, mechanical and adhesion-dependent microenvironmental cues.

LPA and LPA GPCRs

LPA, a bioactive lipid signalling molecule, plays a role in numerous cell processes, including proliferation, migration, adhesion and cell survival [23-26,99,100], by acting at a family of GPCRs known as LPARs (LPA receptors) [101,102]. LPA has wide-ranging influence on cell physiology and pathophysiology, including increases in cytokine and growth factor expression, alteration of surface-protein trafficking and modulation of transcription [103,104]. LPA, which is produced by both normal and malignant cells, is present in high concentration (2–80 μM) in ascites and serum from 98% of ovarian cancer patients, including those with early-stage disease [23,26,104-108]. Furthermore, increasing LPA expression is correlated with poor prognosis, suggesting its potential role as a therapeutic biomarker [109]. Results from our laboratory and others demonstrate LPA regulation of proteases [MMP2, MMP9, MT1-MMP (membrane type 1 MMP) and uPA (urokinase-type plasminogen activator)], inflammatory signalling molecules (COX2), IL-8 and adhesion molecules including E-cadherin and β1 integrin [24,100,104,110-111]. Notably, E-cadherin-based adherens junctions are stabilized by β-catenin [112], and loss of junctional stability may increase the cytoplasmic and/or nuclear pool of β-catenin [66].

LPA induces diverse cellular functions by activating one of five known receptors (LPA1–5), which modulate various signalling pathway proteins, including Rho/ROCK (Rho-associated kinase), IP3 (inositol 1,4,5-triphosphate)/Ca2+, PKC (protein kinase C), PI3K, Ras/MAPK (mitogen-activated protein kinase) and cAMP [101-117]. LPA1 (Gα12/13), LPA3 (Gαq/11) and LPA4 (Gα12/13) receptor subtypes are expressed in the ovary, with LPA4 being the most abundant. LPA2 and LPA3 are aberrantly overexpressed in several ovarian carcinoma cell lines [103,104]. This observation has been confirmed in vivo by detection of overexpressed LPAR2 and LPAR3 mRNA in human ovarian tumour tissues, compared with normal and benign tissues [105-107]. Furthermore overexpression of LPA2 and/or LPA3 potentiates a more proliferative and invasive phenotype in ovarian tumour cells by modulating IL-6, IL-8 and VEGF expression [113]. Interestingly, heterotrimeric G-proteins containing Gα12/13 interact with the cytoplasmic domain of E-cadherin, rescuing breast carcinoma cells from E-cadherinmediated migration suppression, preventing E-cadherin-based cell aggregation and displacing β-catenin from the adherens junction complex [114,115]. In colon cancer cell lines, LPA treatment (1 μM) leads to robust inactivation of GSK3β and nuclear localization of β-catenin [116]. These results correlate with a previous study demonstrating that LPA treatment (0.1–20 μM) or LPA2/LPA3 expression induced a deactivating phosphorylation of GSK3β in HEK (human embryonic kidney)-293 cells [117]. Recently, it has been shown that the Gβγ subunit of the G-protein heterotrimer can also mediate Wnt signalling, inhibiting β-catenin degradation and allowing β-catenin-mediated transcriptional activity [118]. In support of these findings, our unpublished work demonstrates that the LPA–LPAR interaction disrupts junctional localization of E-cadherin and induces nuclear translocation of β-catenin, suggesting a mechanism for Wnt ligand-independent activation of Wnt/β-catenin signalling in EOC (R. J. Burkhalter, Y. Liu and M.S. Stack, unpublished work).

Endothelins

The endothelins (ET-1, ET-2 and ET-3) are a family of peptide signalling molecules that play a diverse role in physiology and pathology. Biological function is achieved via autocrine and paracrine signalling through two GPCRs, ETAR and ETBR (ET type A and B receptors respectively) [119,120]. ET-1 and ETAR are overexpressed in ovarian cancer, driving epithelial-to-mesenchymal transition by stimulating tumour cell proliferation, invasive propensity (migration and adhesion), tumour cell escape from apoptosis and angiogenesis through multiple signalling pathways [120-123]. The ET-1/ETAR axis has been shown to promote the invasive phenotype via inhibition of GSK3β in an ovarian carcinoma cell line [123]. Studies have shown a ETAR/β-arrestin-1/Src signalling complex activates ILK, leading to GSK3β inhibition [124,125]. Subsequently, this activity leads to down-regulation of E-cadherin and increased Snail and β-catenin levels, promoting epithelial-to-mesenchymal transition [126,127]. Interestingly, studies in colon cancer and prostate cancer suggest reciprocal β-catenin regulation of ET-1 transcription through interaction with Tcf-4 [128,129]. These results suggest a potential signalling feed-forward loop for co-ordinate regulation of β-catenin and ET-1 to sustain an aggressive invasive phenotype necessary for tumour progression to metastasis.

Integrin-mediated matrix engagement activates β-catenin signalling

During the ovarian carcinoma metastatic cascade, disseminating cells anchor to the mesothelium and submesothelial matrix of peritoneal organs [11,13]. Although mechanisms regulating mesothelial receptivity have not been fully investigated, mechanical deformation of the mesothelium caused by accumulated ascites fluid may expose submesothelial matrix and/or modulate surface-expressed proteins on mesothelial cells themselves [130]. Additionally, a fibronectin-rich adhesion-promoting stroma is deposited on the otherwise non-adhesive mesothelium [131,132] Several studies have recently demonstrated a link between integrin signalling and the Wnt/β-catenin pathway activation in physiological and pathological cell signalling processes [66,133-138]. Our laboratory has modelled submesothelial anchoring of metastasizing EOC cells using three-dimensional collagen matrices or microsphere-immobilized anti-(β1 integrin) antibodies to mimic matrix-induced integrin aggregation. Integrin clustering in the serous EOC cell lines OVCA 429 and OVCA 433 results in rapid loss of junctional E-cadherin, dissolution of adherens junctions and loss of surface-localized β-catenin [66]. β-Catenin is translocated to the nucleus wherein it activates transcription of Tcf/Lef target genes, including LRP6 and Wnt5a [66]. Similarly, culturing the serous EOC cell line DOV13 in three-dimensional collagen gels leads to down-regulation of two secreted antagonists of Wnt signalling, CTGF (connective tissue growth factor) [139] and DKK1 [16]. Interestingly, down-regulation of DKK1 and CTGF also occurs in DOV13 cells cultured as spheroids relative to two-dimensional tissue culture (Table 3). On the basis of these results, it is interesting to speculate that activation of Wnt signalling in free-floating MCAs may contribute to the acquired chemoresistance of this cell population.

Table 3

Changes in expression of genes related to Wnt signalling in ovarian carcinoma MCAs compared with monolayers

Cells were cultured as monolayers or MCAs. To create MCAs, cells were released from the monolayers using trypsin/EDTA solution, resuspended in minimal essential medium containing 2% fetal bovine serum, plated over solidified 0.5% agarose, and allowed to form spheroids overnight at 37 °C and 5% CO2. Total RNA for cDNA microarray experiments was extracted using TRIzol® (Invitrogen), according to the manufacturer's instructions. All DNA microarray gene expression studies used human oligonucleotide arrays custom printed by a dedicated core facility within the Eppley Institute for Research in Cancer and Allied Diseases, University of Nebraska Medical Center (Omaha, N.E., U.S.A.) as described previously [16]. Microarray slides were scanned with a ScanArray 4000 confocal laser system (PerkinElmer). Analysis of microarray gene expression data, accumulated from three independent experiments, was performed as described [16]. n/a, not applicable. VCAN, versican.

Gene	Up-regulation (fold)	Down-regulation (fold)	Cell line
DKK1	n/a	3.3	DOV13, ES2
CTGF	n/a	3.4	DOV13
VCAN	2.2	n/a	DOV13
CCND1	2.3	n/a	ES2

The mechanical microenvironment may modulate Wnt signalling through β-catenin

More than two-thirds of all cases of EOC simultaneously present with malignant ascites. Ascites development in progressive or recurrent disease is a poor prognostic indicator and is correlated with a lack of response to treatment [140-142]. In comparison with the peritoneal cavity of disease-free women that contains approximately 20 ml of peritoneal fluid, ascitic volumes average 4.9 litres in EOC patients, with a range of 0.8 to 15 litres [143-145]. Although biological components of ascites are increasingly shown to influence progression of ovarian cancer via diverse signalling pathways [74,146,147], potential biomechanical signals activated by increasing intraperitoneal fluid pressure have not been investigated. Tumour cells sense alterations in the force environment via mechanosensing cell surface-expressed proteins, including integrins, and subsequent ‘inside-out’ integrin signalling induces cytoskeletal modifications that effect cell behaviour [148-151]. Force modulation in the context of three-dimensional matrix rigidity is under investigation in several other tumour types [140,149,152-154]; however, the potential effects of the complex mix of strain, compression and shear forces conferred by ascites fluid are unknown. Our preliminary results show that static strain results in loss of surface-associated E-cadherin in serous EOC cells (J. Symowicz and M.S. Stack, unpublished work). In colon cancer, laminar shear stress increased DKK1 expression and decreased activated β-catenin [155]. Conversely, transient compression in APC-deficient cell lines facilitated phosphorylation of β-catenin, leading to destabilization of adherens junctions, nuclear translocation of β-catenin and transcriptional activation of the β-catenin target genes Myc and Twist1 [156]. These results highlight the need to consider multiple distinct mechanical stimuli when modelling the force microenvironment of tumour tissues.

FUTURE PERSPECTIVES

As summarized above and in Figure 4, the unique microenvironmental niche inhabited by ovarian tumours provides a number of diverse mechanisms for transcriptional activation of Wnt/β-catenin target gene expression, even in histotypes devoid of activating mutations in Wnt pathway components. As canonical Wnt signalling is known to regulate proliferation, differentiation and ‘stemness’ [157], clearly additional studies are needed to provide a more detailed understanding of the role of Wnt target gene expression in ovarian tumour progression and metastasis. In particular, the contribution of microenvironmentally regulated Wnt/β-catenin signalling to maintenance of stemness properties in the self-renewing population of ovarian-cancer-initiating cells [158] remains to be explored. As many extracellular proteinases are β-catenin target genes, analysis of the involvement of this pathway in proteinase-mediated intraperitoneal anchoring of metastatic EOC cells may also prove informative.

Figure 4

Potential mechanisms for activation of Wnt/β-catenin signalling in the ovarian tumour microenvironment

In the absence of activating mutations in key components of the Wnt signalling pathway, microenvironmental factors may contribute to transcriptional activation of Wnt/β-catenin target genes. For example, interaction of LPA with LPAR or ET with ET receptors (ETR) have been shown to block the activity of GSK3β, thereby resulting in increased nuclear translocation of β-catenin and enhanced β-catenin-mediated transcriptional activity (left-hand panel). Lateral aggregation of collagen-binding β1 integrins leads to dissolution of adherens junctions and enhanced nuclear translocation of β-catenin. In addition, integrin aggregation enhances the expression of LRP-6 and Wnt-5a while down-regulating the Wnt pathway inhibitors DKK and CTGF. Mechanical compression, exerted on MCAs of metastasizing EOC cells in the form of increased intraperitoneal fluid pressure due to accumulated ascitic fluid, may also destabilize adherens junctions and promote nuclear translocation of β-catenin (right-hand panel).

A number of compounds have been identified that target various components of the Wnt signalling pathway [1,159-161]. For example, non-steroidal anti-inflammatory drugs do not block Wnt signalling themselves, but can induce degradation of Tcf or inhibit Wnt targets such as COX2 [162]. Various small-molecule inhibitors of β-catenin–Tcf interaction have been identified that disrupt β-catenin–Tcf binding, decrease reporter gene transcription, inhibit cell proliferation in vitro and block tumour growth in murine models [163-165]. An alternative approach has been to utilize compounds that stabilize Axin, thereby stimulating β-catenin degradation [166]. As summarized above, with the exception of endometrioid ovarian tumours, progression of EOC has not been commonly associated with activated Wnt signalling. Consequently, there is no available literature reporting the efficacy of Wnt signalling inhibitors in serous EOC, and a search of the clinical trials database (www.clinicaltrials.gov) for trials involving EOC patients and Wnt signalling inhibitors returned no hits. Nevertheless, in light of the accumulating evidence in support of activated Wnt signalling in EOC and increasing evidence for ligand-independent activation of the pathway through microenvironment-regulated cross-talk, preclinical studies evaluating the potential efficacy of small- molecule inhibitors of this signalling pathway are warranted. Following successful preclinical trials, future human clinical trials should be designed to identify the subset of patients most likely to benefit from this biologically targeted therapy (for example, showing nuclear β-catenin staining) and to use validated biomarkers of response.