Elevated CDCP1 predicts poor patient outcome and mediates ovarian clear cell carcinoma by promoting tumor spheroid formation, cell migration and chemoresistance

Hematogenous metastases are rarely present at diagnosis of ovarian clear cell carcinoma (OCC). Instead dissemination of these tumors is characteristically via direct extension of the primary tumor into nearby organs and the spread of exfoliated tumor cells throughout the peritoneum, initially via the peritoneal fluid, and later via ascites that accumulates as a result of disruption of the lymphatic system. The molecular mechanisms orchestrating these processes are uncertain. In particular, the signaling pathways used by malignant cells to survive the stresses of anchorage-free growth in peritoneal fluid and ascites, and to colonize remote sites, are poorly defined. We demonstrate that the transmembrane glycoprotein CUB-domain-containing protein 1 (CDCP1) has important and inhibitable roles in these processes. In vitro assays indicate that CDCP1 mediates formation and survival of OCC spheroids, as well as cell migration and chemoresistance. Disruption of CDCP1 via silencing and antibody-mediated inhibition markedly reduce the ability of TOV21G OCC cells to form intraperitoneal tumors and induce accumulation of ascites in mice. Mechanistically our data suggest that CDCP1 effects are mediated via a novel mechanism of protein kinase B (Akt) activation. Immunohistochemical analysis also suggested that CDCP1 is functionally important in OCC, with its expression elevated in 90% of 198 OCC tumors and increased CDCP1 expression correlating with poor patient disease-free and overall survival. This analysis also showed that CDCP1 is largely restricted to the surface of malignant cells where it is accessible to therapeutic antibodies. Importantly, antibody-mediated blockade of CDCP1 in vivo significantly increased the anti-tumor efficacy of carboplatin, the chemotherapy most commonly used to treat OCC. In summary, our data indicate that CDCP1 is important in the progression of OCC and that targeting pathways mediated by this protein may be useful for the management of OCC, potentially in combination with chemotherapies and agents targeting the Akt pathway.


INTRODUCTION
Epithelial ovarian cancer is the most lethal gynecological malignancy with~140 000 deaths each year worldwide. 1 Ovarian clear cell carcinoma (OCC) is an aggressive form of this malignancy 2 that in Western societies accounts for 5-13% of all ovarian cancers and up to 25% in Asian populations. 3,4 Although stage I patients generally do well, later OCC stages have poorer prognosis than any of the other three major ovarian cancer histotypes: serous, mucinous and endometrioid. These later-stage OCC tumors are characteristically resistant to conventional platinum-based chemotherapy, and o10% of patients with recurrent disease respond to second-line agents. 5,6 Currently, drugs targeting defined signaling pathways are also not available for treatment of OCC patients.
A feature that distinguishes epithelial ovarian cancer from most other solid malignancies is its mode of metastasis. In contrast with the other tumor types, which generally employ the vasculature to spread to distant sites, hematogenous metastases are rarely present at diagnosis of epithelial ovarian cancer. 7,8 Instead epithelial ovarian cancer metastasis occurs via intraperitoneal dissemination, a process characterized by direct extension of the primary tumor into nearby organs and the spread of exfoliated tumor cells throughout the peritoneum, initially via the peritoneal fluid, and later via ascites that accumulates as a result of disruption of the lymphatic system. During these processes survival of malignant cells is promoted via non-adherent growth as spheroids. These multicellular structures also facilitate attachment to, and invasion into the mesothelium of abdominal organs and the omentum. 7,8 Currently, the molecular mechanisms orchestrating these processes are uncertain; in particular, the signaling pathways used by malignant cells to survive anchoragefree stress and colonize remote sites are poorly defined. Understanding these events at the molecular level is likely to provide new opportunities to therapeutically target ovarian cancer. 2 CUB (complement factor C1r/C1s, embryonic sea urchin protein uEGF, and bone morphogenetic protein-1)-domain-containing protein 1 (CDCP1) is a transmembrane glycoprotein that is widely expressed in epithelial cells as both full-length 135 kDa and proteolytically processed 70 kDa forms. 9 Although its physiological function is unknown, CDCP1 knockout mice have no obvious reproductive, developmental or survival abnormalities, 10,11 suggesting that targeting this protein in disease settings, including cancer, may be well tolerated. In several tumor types that commonly metastasize via vascular routes, including lung, kidney, pancreas and colon cancer, elevated or cell-surface expression of CDCP1 is associated with poor patient outcome. [12][13][14][15][16][17] Consistent with a role in hematogenous metastasis, in animal models of vascular dissemination, survival of cancer cells undergoing extravasation is markedly enhanced by a mechanism involving serine protease cleavage to generate 70 kDa CDCP1. This initiates pro-survival signaling via focal adhesion kinase 1 and phosphoinositide 3-kinase (PI3K) dependent protein kinase B (Akt) activation resulting in suppression of poly (ADP-ribose) polymerase 1 (PARP1) mediated apoptosis. 18,19 Here, we have examined the function and expression of CDCP1 in OCC. Our data indicate that CDCP1 promotes OCC dissemination via a novel non-cleavage-mediated mechanism that involves activation of Akt to promote formation and survival of tumor spheroids, as well as cell migration and chemoresistance. In addition, we demonstrate that silencing CDCP1 and blocking its function using a monoclonal antibody are effective at reducing progression of OCC in a mouse model. Importantly, our immunohistochemical analysis of CDCP1 in normal ovary and fallopian tube and a large cohort of OCC cases showed that its elevated expression correlates with poor patient outcome.

RESULTS
CDCP1 expression is elevated in OCC and correlates with poor outcome CDCP1 protein expression was evaluated by immunohistochemistry using three tissue microarrays (TMA) containing OCC samples from a total of 207 patients and a TMA-containing normal tissue from 25 women with benign gynecological conditions whose samples showed no evidence of disease. From the OCC TMAs, cores for 9 cases were lost during staining leaving 198 evaluable cases, of which 178 (90%) expressed CDCP1 at varying intensity ( Figure 1a). In all 178 OCC-expressing cases, it was striking that CDCP1 expression was almost exclusive to the surface of malignant cells. CDCP1 signal was high in 20 (10%), moderate in 78 (39%) and weak in 80 cases (40%). It was not detected in the stroma of malignant cases (Figure 1a) or in the epithelium or stroma of the 25 normal ovaries examined ( Figure 1b). Survival data were available for 134 OCC cases with 41 of these dying of OCC, 26 dying of other causes and 67 alive at last follow-up. Kaplan-Meier analysis revealed that both disease-free survival and overall survival were much poorer in cases expressing CDCP1 compared with non-expressers (Figures 1c and d).
CDCP1 mediates migration and non-adherent growth as spheroids of OCC cells in vitro As elevated CDCP1 expression predicts poor OCC patient outcome, we examined its function in this malignancy. Western blot analysis to identify cell lines suitable for in vitro and in vivo assays showed that CDCP1 is expressed in each of the 11 OCC lines examined, with lowest levels in KK cells, and proteolytically processed 70 kDa CDCP1 in OVTOKO, RMG-I and JHOC-7 cells with this cleavage product either not present or apparent at low levels in the remaining cell lines (Figure 2a). Based on this analysis, TOV21G, KOC7C, OVTOKO and RMG-I cells were selected for reduction of CDCP1 expression, and KK cells for overexpression.   As pro-survival signaling involving proteolytic conversion of CDCP1 to 70 kDa requires Akt signaling, 18,20 we assessed whether activation of this kinase is altered in the three cell lines modulated for CDCP1 expression. As shown in Figure 2b, western blot analysis showed that modulation of CDCP1 was sufficient to robustly impact levels of pAkt in these OCC cell lines, and this was independent of cleavage of CDCP1. The level of Akt activation was directly proportional to the level of expression of CDCP1. Similar data have previously been reported for poorly metastatic melanoma A375 cells. 21 For this cell line, CDCP1 overexpression induced Akt activation via a mechanism that was dependent on phosphorylation of tyrosine 734 (Y734) of full-length 135 kDa CDCP1. 21 To examine whether altered Akt activation in our OCC cells is also dependent on CDCP1-Y734, we generated a second dox-inducible KK line able to express phosphorylation-defective CDCP1-Y734F (designated KK-Y734). As shown in Figure 2c, dox treatment-induced expression of both CDCP1 and CDCP1-Y734F within 6 h and pAkt-S473 levels were consistently elevated in both KK-WT and KK-Y734F cells by 24 h. This novel result indicates that pY734 is not required for CDCP1-induced activation of Akt in KK cells. Consistently, western blot analysis showed that CDCP1 endogenously expressed by TOV21G and KOC7C cells and induced in KK-WT cells is not phosphorylated at Y734 (Figure 2d). The generated cell lines were used to examine the role of CDCP1 in processes important in the progression of OCC: proliferation, migration and spheroid formation and growth. As shown in Figure 2e, proliferation of each cell line over 96 h was unchanged by altered CDCP1 expression. In contrast, stable silencing of CDCP1 reduced TOV21G and KOC7C cell migration in a Transwell system (Corning, Crown Scientific, Minto, NSW, Australia) over 48 h by 65 and 73%, respectively ( Figure 2f). Consistently, elevated CDCP1 expression increased KK-WT cell migration 2.9-fold in this in vitro system ( Figure 2f). The effect of modulated CDCP1 expression on OCC spheroids was assessed after growing cells in low-adhesion plates for 72 h. Silencing of CDCP1 reduced the number of spheroids formed by TOV21G and KOC7C cells by 57 and 55%, respectively, while elevated expression resulted in~2.2-fold more KK-WT-cell spheroids ( Figure 2g). Of note, the function of CDCP1 in migration and spheroid formation is independent of pY734, as induction of CDCP1-Y734F in KK-Y734F cells triggered similar increases in these processes as KK-WT cells (Figures 2f and g), correlating with the levels of pAkt seen in Figure 2c. In addition, the function of CDCP1 in migration and spheroid formation appears to be independent of proteolytic cleavage of CDCP1. Silencing of CDCP1 in OVTOKO and RMG-I cells, which express significant levels of cleaved CDCP1, reduced migration and spheroid formation without altering prolifertion (Supplementary Figure S1b) simmilar to TOV21G and KOC7C cells, which almost exclusively express fulllength CDCP1 ( Figure 2).
We also examined whether manipulation of CDCP1 expression altered the epithelial and mesenchymal characteristics of OCC cells, as transitions between these states are thought to be important in ovarian cancer. 22 However, western blot analysis indicated no change in expression of the epithelial marker EpCAM (epithelial cell adhesion molecule) or the mesenchymal marker vimentin (Supplementary Figure S2a). Consistently, modulating CDCP1 levels did not alter cell morphology ( Supplementary  Figures S2b and c). CDCP1 promotes OCC tumor growth and accumulation of ascites in a mouse model We employed two approaches to examine the role of CDCP1 in OCC in vivo. First, we examined the impact of modulating CDCP1 expression on the ability of TOV21G and KK cells to form intraperitoneal (i.p.) tumors in immunocompromised NOD.Cg-Prkdc scid Il2rg tm1Wjl /SzJ (NSG) mice. Immediately before killing the mice on day 24, tumor burden assessed by bioluminescence imaging (BLI) was markedly lower in mice carrying TOV21G cells silenced for CDCP1 (TOV21G-shCDCP1) compared with scramble control (TOV21G-Scr) cells (Figures 3a and b). At the time of killing the mice there was a 2.5-fold increase in the number of tumor nodules 48 mm 3 in mice injected with TOV21G-Scr compared with TOV21G-shCDCP1 cells, while there was no difference in the size of nodules (Figure 3c). Consistent with reduced tumor burden, ascites volume was 450% lower in the TOV21G-shCDCP1 mouse group (Figure 3d). Western blot analysis showed that levels of CDCP1 in cultured cells prior to injection were maintained in TOV21G-shCDCP1 and TOV21G-Scr xenograft tumors (Figure 3e). It was not possible to examine Akt activation in these tumors by western blot analysis as pAkt-S473 was masked by the albumin signal, which is of similar molecular weight. Accordingly, Akt activation was examined immunohistochemically, which showed that levels were markedly higher in TOV21G-Scr tumors compared with TOV21G-shCDCP1 tumors ( Figure 3f) consistent with data from cultured cells (Figure 2b). Data from KK cells also indicated that CDCP1 promotes OCC in vivo. BLI showed that 6 weeks after i.p. injection, bioluminesence was fourfold to fivefold higher in mice carrying KK-WT cell xenografts and fed a dox-containing diet to induce CDCP1 expression, than control mice (Figures 3g and h).
In the second approach, we compared i.p. growth in mice populations of TOV21G cells expressing CDCP1 at either low (CDCP1 low ) or high (CDCP1 high ) levels. These populations were selected by fluorescence-activated cell sorting based on cell-surface expression using the anti-CDCP1 monoclonal antibody 41-2. 19 After this selection process the populations were expanded under adherent cell culture conditions for 8 weeks to generate sufficient numbers of cells for mouse assays. Both adherent CDCP1 low -and CDCP1 high -cell populations had the same morphology as parental TOV21G cells (Figure 4a). After in vitro expansion, to re-establish conditions mimicking growth in ascites the cell populations were replated onto low-attachment plates for 7 days and then examined. Western blot and flow cytometry analysis assessing the molecular phenotype of these populations indicated about fivefold higher levels of total CDCP1 expressed by CDCP1 high TOV21G cells (Figure 4b) and~10-fold more CDCP1 on the surface of these cells compared with CDCP1 low cells (Figure 4c). Western blot data also confirmed increased activation of Akt in CDCP1 high in comparison with CDCP1 low and parental TOV21G cells, while there was no change in Src phosphorylation or in the mesenchymal marker vimentin (Figure 4b). From samples obtained from mice killed on day 24, it was apparent that CDCP1 high cells were much more tumorigenic in vivo than CDCP1 low cells forming many more nodules on adipose tissue and the peritoneal membrane, and occasionally on the ovary, liver and spleen. Tumor nodules on the mesenteric   facilitates the seeding of cancer cells on peritoneal organs and structures. 7, 8 We noted above that modulation of CDCP1 has a marked impact on the ability of OCC cells to form spheres under non-adherent conditions (Figure 2g and Supplementary Figure  S1d). To explore whether the role of CDCP1 in sphere formation promotes OCC, we selected populations of OCC cells that retain the ability to grow as spheres in long-term cell culture conditions by growing TOV21G and KOC7C cells in ultra-low-attachment plates in complete media. Between 7 and 14 days after plating spheroid-forming cells (SFCs) and non-SFCs (NSFCs) were separated by differential centrifugation. To expand these subpopulations for downstream assays, SFCs and NSFCs were plated into adherent culture vessels and grown for 8 weeks. The cells were then replated for 7 days in ultra-low-attachment plates, which demonstrated that SFC and NSFC phenotypes were maintained (Figure 5a). Consistent with CDCP1 having a role in sphere formation, western blot analysis showed that CDCP1 levels were markedly higher in SFCs compared with NSFCs (Figure 5b, non-adhered). In addition, analysis of Akt, Src and ERK (extracellular-signal-regulated kinase) pathways that are important in CDCP1-mediated signaling, 18,20,[23][24][25] demonstrated that each is activated in SFCs compared with NSFCs ( Figure 5b, non-adhered). Interestingly, when both SFCs and NSFCs were replated onto cell culture dishes for adherent growth, levels of CDCP1 and activation of Akt, Src and ERK returned to basal levels (Figure 5b, adhered).
To examine the importance of CDCP1-mediated sphere formation in OCC, we first assessed whether TOV21G SFCs are more tumorigenic than NSFCs, by injecting non-enzymatically dissociated single-cell suspensions of each cell population i.p. into mice. At the time of killing the mice both TOV21G SFCs and TOV21G NSFCs had formed tumors throughout the peritoneal cavity and also caused the accumulation of ascites, with TOV21G SFCs forming 3.3-fold more tumor nodules 48 mm 3 (Figure 5c) and inducing 4.2-fold more ascites than TOV21G NSFCs (Figure 5d). To directly examine the importance of CDCP1 in tumorigenicity of OCC spheroids in mice, we examined the effect of the function-blocking anti-CDCP1 antibody 10D7 18-20 on tumorigenicity of TOV21G SFCs . At the time of killing the mice 24 days after injection, 10D7-treated TOV21G SFCs had formed many fewer i.p. tumor nodules than isotype-matched IgG-treated cells (Figure 5e). Specifically, in comparison with control IgG, the anti-CDCP1 antibody reduced the number of TOV21G SFC tumor nodules 48 mm 3 by 54% (Figure 5f) and the volume of ascites by 57% (Figure 5g).

Targeting CDCP1 increases OCC chemosensitivity
Carboplatin is the most active ovarian cancer chemotherapy and is generally used in combination with paclitaxel. 26 Resistance to these agents is a major OCC treatment problem. 2 As our data indicate that CDCP1 mediates pro-survival signaling in OCC, we were interested in whether it also has a role in resistance of this malignancy to chemotherapy. This was examined by measuring the response in vitro to increasing carboplatin (0-100 μM) and paclitaxel (0-30 nM) 27-29 of TOV21G-Scr versus TOV21G-shCDCP1, TOV21G low versus TOV21G high , and KK versus KK-WT and KK-Y734F cells. Cells were cultured under adherent conditions and viability was measured at 72 h. TOV21G-Scr cells were about twice as resistant to carboplatin and paclitaxel as TOV21G-shCDCP1 cells (Figures 6a and b) and, consistently, that TOV21G high cells were twice as resistant to both carboplatin and paclitaxel as TOV21G low cells (Figures 6c and d). Also consistent with these data, doxtreated KK-WT and KK-Y734F cells were both twice as resistant to these agents as control dox-treated cells (Figures 6e and f).
To examine mechanistically how CDCP1 mediates OCC resistance to carboplatin and paclitaxel, we performed western blot analyses on lysates from three OCC cell lines that had been modulated for CDCP1 expression and treated with these chemotherapeutic agents. We examined CDCP1 expression as well as activation of pathways previously shown to be regulated by CDCP1 including pAkt-S473/Akt, p-Src-Y416/Src and pERK/ ERK. 18,23,25,30 Of note, the response to carboplatin and paclitaxel of TOV21G and KOC7C cells, which endogenously express CDCP1 at medium to high levels, was to increase CDCP1 levels within 3 h of treatment and this was accompanied by a transient increase in pAkt levels (scramble lanes, Supplementary Figures S3a and b). For both of these lines the presence of the CDCP1 silencing construct shCDCP1 was sufficient to overcome this induction of CDCP1 and pAkt caused by chemotherapy (shCDCP1 lanes, Supplementary  Figures S3a and b).
Analogous data were obtained from KK cells that express CDCP1 at low levels. As can be seen in Supplementary Figures S3c and d, both carboplatin and paclitaxel induced CDCP1 expression in parental KK cells and KK-WT cells by 3-6 h after the start of chemotherapy, and this was accompanied by increased pAkt levels. In KK-WT treated with dox to induce CDCP1 expression, we did not observe a further increase in CDCP1 in response to chemotherapy, although a transient increase in pAkt levels was seen at 3h (Supplementary Figures S3c and d). Interestingly, doxtreated KK-Y734F cells showed the same responses, further supporting data from Figure 2c showing that CDCP1-mediated Akt activation in OCC is unlike the CDCP1-Y734-dependent activation reported in other cancer settings. 18,23,25,30 Activation of Src and ERK was not seen consistently in these OCC cell lines in response to chemotherapy, although we did note increased pERK levels over the time course of carboplatin treatment of doxtreated KK-WT and -Y734F cells (Supplementary Figure S3).
These data indicate that CDCP1 has a role in chemotherapy resistance and that CDCP1-mediated activation of Akt in OCC in response to chemotherapeutic agents occurs via a novel mechanism independent of Src and ERK activation. To directly address whether blockade of CDCP1 is a useful strategy to increase OCC susceptibility to standard chemotherapy, we compared the impact against mouse i.p. xenografts of 10D7 and carboplatin as monotherapies, and these as combined agents. Twice weekly 10D7 treatments started on day 3, and weekly carboplatin treatments started on day 7. At the time of sacrificing the mice, combined 10D7 and carboplatin was much more efficacious at reducing tumor burden than either monotherapy. Specifically, weekly BLI showed that combining 10D7 and carboplatin significantly reduced tumor growth within the experimental period compared with the controls treated with IgG, 10D7 only or carboplatin plus IgG (Figure 6g). At the time of killing on day 24, mice treated with combined 10D7 and carboplatin showed 7.9-, 3.1-and 3.7-fold fewer tumor nodules 48 mm 3 than IgG, 10D7 and carboplatin plus IgG group, respectively (Figure 6h). Almost no ascites was present in mice that received combined 10D7/carboplatin (Figure 6i).

DISCUSSION
Our data indicate for the first time that the transmembrane glycoprotein CDCP1 is an important and inhibitable mediator of OCC. Significantly, its elevated expression indicates poorer suvival of OCC patients, and our finding that CDCP1 is expressed by the vast majority of OCC tumors but not by normal ovaries is encouraging that this protein could be selectively targeted in OCC without impacting on normal cells and structures present within the tumor environment. CDCP1 has previously been shown to protect cancer cells that metastasize via the vasculature, [18][19][20] and our data are the first to show that this protein is critical in processes essential for the peritoneal metastasis characteristic of OCC. In particular, for this distinct type of metastasis it is a key mediator for formation and growth of the tumor spheroids that promote the anchorage-independent survival, and seeding of cancer cells on peritoneal organs and structures that is essential for intraperitoneal dissemination of OCC. In addition, it promotes migration and chemoresistance of OCC cells. An important finding is that the anti-CDCP1 antibody 10D7 increases OCC responsiveness to chemotherapy in vitro and in vivo suggesting that combining anti-CDCP1 therapies with agents currently used in patients could be useful against OCC. The importance of elevated CDCP1 expression in OCC was indicated by Kaplan-Meier analysis, which showed that patients expressing this protein have much poorer disease-free survival and overall survival. These patient data were supported by in vitro assays and experiments in mice. Data from four OCC cell lines silenced for CDCP1 expression and another overexpressing this protein demonstrated that CDCP1 mediates migration and spheroid formation in vitro, processes that are important for OCC in vivo. 7,8 Consistently, silencing of CDCP1 in TOV21G cells resulted in markedly lower tumor burden and accumulation of ascites in mice. In addition, TOV21G cells selected for high CDCP1 expression were able to induce 2.5-fold greater tumor burden and 5.8-fold more ascites in mice than TOV21G cells expressing this protein at~10-fold lower levels. In addition, induced expression of CDCP1 markedly increased the tumorigenicity of KK cells in vivo.
These findings indicate important roles for CDCP1 as a protumorigenic factor in OCC.
Our observation that CDCP1 is not expressed by normal ovarian epithelium was somewhat surprising as it has previously been shown to be widely expressed by other normal epithelia including epithelium of the colon, 24,31 breast, lung, 31 skin, 32,33 pancreatic ducts, lung and prostate. 34 The widespread expression outside the ovary may indicate that efforts to target CDCP1 in OCC would require its local rather than systemic delivery. Also of relevance to being able to target CDCP1 in OCC patients, our immunohistochemical analysis showed that CDCP1 in malignant cells it is largely restricted to the surface where it would be accessible to function-blocking anti-CDCP1 antibodies, which we 18,19 and others 35 have shown are effective at inducing cell death in vivo. Interestingly, we have recently shown that epidermal growth factor receptor signaling promotes recycling of CDCP1 to the surface of serous ovarian cancer cell lines; 36 however, the mechanism that induces predominant location of CDCP1 on the surface of OCC cells in patients remains to be defined.
Based on data from two different experimental approaches, we show that altered CDCP1 expression is sufficient to modulate Akt activation in OCC cell lines. Our first approach involved modulation of CDCP1 expression via lenti-viral-mediated silencing and overexpression. We observed that silencing of CDCP1 in TOV21G, KOC7C, OVTOKO and RMG-I cells markedly reduced pAkt-S473 levels and, consistently, its overexpression in KK cells resulted in robust activation of this pathway, which was independent of p-CDCP1-Y734 (Figure 2c). These CDCP1-mediated changes in Akt activation in vitro were accompanied by corresponding changes in migration and spheroid formation (Figures 2f and g), which are cellular processes essential for OCC, and chemo-resistance ( Figure 6), which is a characteristic feature of many advanced OCC tumors. The impact on Akt signaling in vitro of altered CDCP1 expression was also apparent in vivo. Intraperitoneal tumors recovered from mice carrying TOV21G-shCDCP1-cell tumors showed much lower levels of pAkt-S473 than TOV21G-Scr-cell tumors (Figure 3e).
In our second approach, we selected SFC and NSFC populations of TOV21G and KOC7C cells that were able to maintain long-term spheroid-forming capacity. We noted that in both cell lines, CDCP1 expression was markedly induced in SFCs in comparison with NSFCs. Consistent with data from silencing and overexpression assays, the increased CDCP1 expression seen in SFCs was accompanied by activation of Akt. Interestingly, Src and ERK activation was also seen in SFCs compared with NSFCs. Consistent with these data, it has been reported that several mediators of CDCP1 signaling, including integrin β1, 37 Akt 38 and ERK, 39 also mediate formation of ovarian cancer cell spheroids. We confirmed the importance of elevated CDCP1 by showing that a functionblocking anti-CDCP1 antibody that we have previously used to inhibit vascular metastasis 18 caused a marked reduction in the ability of SFCs to form i.p. tumors and accumulation of ascites in vivo.
Currently, we have limited information on the mechansim by which altered expression of CDCP1 impacts activation of Akt in OCC. Previous reports have shown that CDCP1-mediated hematogenous metastasis in mouse and chicken embryo models is dependent on activation of Akt. 18,20 In these in vivo models, proteolytic conversion of 135 kDa CDCP1 to 70 kDa p-CDCP1-Y734 initiated docking of this cleaved and phosphorylated form of CDCP1 to β1 integrin, and the focal adhesion kinase 1 and PI3K-dependent activation of Akt is critical for cell survival during extravasation. [18][19][20] Our data indicate a novel mechanism of CDCP1-mediated activation of Akt in OCC that is distinct to the serine protease-mediated conversion of 135 to 70 kDa p-CDCP1-Y734 that promotes hematogenous metastasis via pAkt. Rather than proteolysis of CDCP1 being the initiator of Akt activation, increased CDCP1 expression was sufficient to induce pAkt-S473, and this was independent of both Src and ERK signaling. These data suggest that the increased activation of Src and ERK seen by us in SFCs (Figure 5b) occurred in parrallel with CDCP1 activation of Akt rather than being involved in the relay of signals between CDCP1 and Akt. In addition, two of our findings demonstrate that CDCP1 activation of Akt signaling in OCC is not dependent on Y734 of CDCP1. First, in KK cells the phosphorylation-defective mutant CDCP1-Y734F was able to induce phosphorylation of Akt-S473 as robustly as wildtype CDCP1 (Figure 2c), even when these cells were under chemotherapeutic challenge (Supplementary Figure S3). In addition, the increased migration, spheroid formation and chemoresistance of KK cells mediated by elevated expression of CDCP1 was  (1 × 10 6 ). At the time of killing the mice 24 days after injection, TOV21G SFCs were much more tumorigenic than TOV21G NSFCs showing more tumor nodules 48 mm 3 in size (c) and larger ascites volume (d). (e-g) Female NSG mice (n = 5 per group) were injected i.p. with TOV21G SFCs (1 × 10 6 ) and anti-CDCP1 monoclonal antibody 10D7 or isotype-matched control IgG (100 μg). Antibody treatment was repeated twice a week until the mice were killed. At the time of killing the mice 24 days after injection, TOV21G SFCs co-injected with 10D7 had formed many fewer i.p. tumor nodules. Representative images of the mesenteric membrane from the mice in each group are shown. *Indicates tumor nodule. TOV21G SFCs co-injected with 10D7 formed fewer tumor nodules 48 mm 3 (f) and smaller ascites volume (g). unaffected by loss of Y734 of CDCP1 (Figures 2f and g, 6e and f). It was also interesting that western blot analysis showed consistently that within 3-6 h the OCC chemotherapies, carboplatin and paclitaxel, induce expression of CDCP1 in OCC cell lines and that this is accompanied by phosphorylation of Akt without altering Src and ERK activation (Supplementary Figure S3). Understanding the mechanisms by which CDCP1 expression is increased and how it signals to promote OCC may provide additional opportunities to treat this cancer.
In conclusion, our findings implicate CDCP1 as a protumorigenic factor in OCC by promoting migration, spheroid growth and chemoresistance. It is not expressed in normal ovary but is elevated in the vast majority of OCC patients and its elevated expression correlates with poor patient outcome. Our data indicate that blockade of CDCP1 reduces tumor burden and accumulation of ascites, particularly when used in combination with carboplatin, suggesting that targeting OCC with anti-CDCP1 agents and chemotherapy or drugs blocking other signaling pathways may be of benefit to patients.

MATERIALS AND METHODS
Antibodies and reagents, in vitro assays and western blot and statistical analysis Antibodies, reagents, ovarian samples, immunohistochemistry, cell culture, luciferase labeling of cells, modulation of CDCP1 expression, in vitro assays, Figure 6. Targeting CDCP1 increases OCC chemosensitivity. Cell viability in response to 72 h of treatment of TOV21G-Scr and TOV21G-shCDCP1 (a, b), TOV21G-CDCP1 low and TOV21G-CDCP1 high (c, d) and dox-treated KK, KK-WT and KK-Y734 cells (e, f), over the indicated concentration ranges of carboplatin (a, c, e) and paclitaxel (b, d, f). *P o0.05, **P o0.01 and ***Po0.001. NA, not applicable. (g-i) Female NSG mice injected with luciferase-labeled TOV21G cells (1 × 10 6 per mouse) were treated with isotype IgG, carboplatin and IgG, 10D7, or carboplatin and 10D7. Doses were IgG and 10D7 at 25 mg/kg/dose starting 3 days after cell injections then continuing twice weekly; carboplatin 15 mg/kg/dose starting 7 days after cell injections then continuing weekly. Tumor growth was monitored weekly by BLI and total photon flux was plotted against time (g). At harvest tumors 48 mm 3 were counted (h) and ascites volume was measured (i).
cell and tissue lysis, mouse xenografts and western blot and statistical analyses are described in Supplementary Materials and Methods.