5Β1-Integrin Regulates Ovarian Cancer Invasion

Ligand-independent activation of c-Met by ﬁbronectin and a5b1-integrin regulates ovarian cancer invasion and metastasis

AK Mitra1, K Sawada1, P Tiwari1, K Mui1, K Gwin2 and E Lengyel1

1Department of Obstetrics and Gynecology, Section of Gynecologic Oncology, Center for Integrative Science, University of Chicago, Chicago, IL, USA and 2Pathology, University of Chicago, Chicago, IL, USA

The role of the fibronectin receptor, a5b1-integrin, as an route commonly observed in other cancers. The cancer adhesion receptor and in angiogenesis is well established. cells exfoliate from the surface of the ovary by shedding However, its role in cancer cell invasion and metastasis from the primary tumor or through hydroflotation is less clear. We describe a novel mechanism by which caused by circulating peritoneal fluid. They subse- fibronectin regulates ovarian cancer cell signaling and quently attach to the peritoneal surfaces and invade to promotes metastasis. Fibronectin binding to a5b1-integrin form metastatic tumors. This mode of metastasis places led to a direct association of a5-integrin with the receptor unique demands on the ovarian cancer cells and may tyrosine kinase, c-Met, activating it in a hepatocyte require a specific molecular mechanism different from growth factor/scatter factor (HGF/SF) independent manner. the intra- and extravasation involved in hematogenous Subsequently, c-Met associated with Src, and activated metastasis (Lengyel, 2010). Src and focal adhesion kinase (FAK). Inhibition of a5b1- Integrins are among the genes reported to be involved integrin decreased the phosphorylation of c-Met, FAK and in ovarian cancer metastasis. They are heterodimeric Src, both in vitro and in vivo. Independent activation of transmembrane proteins comprised of various combina- c-Met by its native ligand, HGF/SF, or overexpression tions of an a- and b-subunit, with each heterodimer of a constitutively active FAK in HeyA8 cells could possessing one or more ligands (Hynes, 2002). Integrins overcome the effect of a5b1-integrin inhibition on tumor anchor the cancer cells to the extracellular matrix cell invasion, indicating that a5b1-integrin is upstream of (ECM), which provides a contact point for invasion c-Met, Src and FAK. Inhibition of a5b1-integrin on cancer and metastasis. Signals from the ECM are transmitted cells in two xenograft models of ovarian cancer metastasis via integrins through interactions with multiple proteins resulted in a significant decrease of tumor burden, that eventually determine the cellular response to which was independent of the effect of a5b1-integrin on the ECM, including changes in cell shape, motility, angiogenesis. These data suggest that fibronectin pro- proliferation and so on (Giancotti and Tarone, 2003; motes ovarian cancer invasion and metastasis through an Mitra and Schlaepfer, 2006). The fibronectin receptor a5b1-integrin/c-Met/FAK/Src-dependent signaling path- (a5b1-integrin) primarily binds fibronectin to anchor way, transducing signals through c-Met in an HGF/ cells and subsequently activates the non-receptor tyrosine SF-independent manner. kinases (RTKs), focal adhesion kinase (FAK) and Src, Oncogene (2011) 30, 1566–1576; doi:10.1038/onc.2010.532; which play an important role in tumorigenesis by published online 29 November 2010 promoting the proliferation and invasion of cancer and endothelial cells (Mitra and Schlaepfer, 2006; Keywords: ovarian cancer; a5b1-integrin; c-Met; Src; Caswell et al., 2007). FAK; fibronectin Knocking out a5b1-integrin is associated with major vascular defects, such as a marked decrease in capillary branching and incompletely formed vessels in mouse Introduction embryos, leading to embryonic lethality (Yang et al., 1993; Francis et al., 2002). As a5b1-integrin is abun- Ovarian cancer is the fifth leading cause of cancer- dantly present on endothelial cells, research on this related death among women in the United States and integrin has primarily focused on its role in angiogenesis has the highest mortality rate of all gynecologic and elucidated its function in vascular endothelial cells malignancies (Jemal et al., 2009). It involves a unique (Kim et al., 2000; Taverna and Hynes, 2001; Magnussen metastatic process, which rarely takes the hematogenous et al., 2005). Endothelial cells in tumor vessels over- express a5b1-integrin, which helps to anchor them to the Correspondence: Dr E Lengyel, Department of Obstetrics and basement membrane. When antibodies against a5b1- Gynecology, Section of Gynecologic Oncology, Center for Integrative integrin are injected intravenously, they rapidly label Science, University of Chicago, 5841 South Maryland Avenue, MC tumor vessels (Magnussen et al., 2005), highlighting the 2050, Chicago, IL 60637, USA. E-mail: [email protected] importance of a5b1-integrin for tumor vasculature. Received 20 May 2010; revised 16 September 10; accepted 11 October However, the role of a5b1-integrin, when expressed on 2010; published online 29 November 2010 tumor cells, is less well understood. We have previously a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1567 shown that, during ovarian cancer progression, loss of Currently, the most completely characterized function E-cadherin-upregulated a5b1-integrin, which promoted of a5b1-integrin in cancer is its contribution to the adhesion and invasion of ovarian cancer cells. angiogenesis (Wingerter et al., 2005; Avraamides et al., Overexpression of a5b1-integrin indicated a very poor 2008). However, the human a5-integrin antibody used in prognosis in patients with epithelial ovarian cancer Figures 1a and b does not cross-react with murine a5b1- (Sawada et al., 2008). However, the molecular mechan- integrin (Bhaskar et al., 2007), indicating that the effects ism by which a5b1-integrin promotes cancer cell seen on metastasis and tumor growth are the result of a metastasis is currently unknown. Given that fibronectin direct effect of the antibody on the human tumor cells is one of the most abundant proteins in the ECM of rather than the blocking of host a5b1-integrin function. both the peritoneum and omentum (Kenny and Lengyel, To compare the contribution of a5b1-integrin in cancer 2009; Hafter et al., 1984; Franke et al., 2003), an cells with that in host stromal cells, we took advantage understanding of how fibronectin–a5b1-integrin interac- of an anti-murine a5Ab that inhibits angiogenesis by tions regulate metastasis in the ovarian cancer cell might blocking a5b1-integrin on mouse vascular endothelial provide additional insights into the biology of ovarian cells (Bhaskar et al., 2007). The specificity of the human cancer. and murine a5-integrin antibody was confirmed by In the present study, we show that upon binding to testing their effect on adhesion of the human ovarian fibronectin, a5b1-integrin interacts directly with the cancer cell line HeyA8 and the mouse ovarian cancer cell receptor tyrosine kinase c-Met, thereby, activating Src line ID-8 on fibronectin (Supplementary Figure S3). and FAK. This is the first report to demonstrate that the Treatment with the human or the murine a5-integrin fibronectin receptor transduces signals through c-Met, antibody showed a comparable reduction in tumor which is itself a major player in ovarian cancer weight (57 and 48%, respectively) and numbers (73 and metastasis. 64%, respectively) in a SKOV3ip xenograft model (Figure 1c). This reduction was similar to the effect of bevacizumab (Avastin, Genentech, South San Francisco, Results CA, USA), a humanized anti-VEGF antibody. Combin- ing the human and murine a5b1-integrin antibodies did Inhibition of a5b1-integrin inhibits ovarian cancer not significantly augment the antitumor effect of each metastasis individual antibody (Figure 1c). Immunohistochemical The ECM of the normal human omentum, the most staining of the tumors for CD31 confirmed an anti- common site of ovarian cancer metastasis, expresses angiogenic effect of murine a5Ab and bevacizumab, abundant fibronectin and vitronectin (Supplementary whereas, as expected, the human antibody had no effect Figure S1). We and others have previously shown that on blood vessel density (Supplementary Figure S4) or on the fibronectin receptor, a5b1-integrin, is upregulated in VEGF expression (data not shown). ovarian cancer cells during tumor progression (Yokoyama Next, we sought to determine whether a5b1-integrin et al., 2007; Sawada et al., 2008). Therefore, we decided plays a more important role in early or in late metastasis to inhibit a5b1-integrin to determine whether this would (Figure 1d). To study early metastasis, a single treat- result in a reduction of ovarian cancer dissemination. ment of the integrin antibody was given to the nude mice Two orthotopic xenograft models of ovarian cancer simultaneously with the intraperitoneal injection of metastasis were treated with a monoclonal antibody ovarian cancer cells to block the initial attachment and blocking human a5b1-integrin (Hu a5Ab) (Bhaskar invasion of tumor cells to the peritoneum (prevention et al., 2007). SKOV3ip1 or HeyA8 human ovarian study). Late metastasis was studied with the adminis- cancer cells, both of which have high expression levels of tration of repeated treatments beginning 8 days after the a5b1-integrin (Supplementary Figure S2), were injected tumors had established within the peritoneal cavity intraperitoneally into nude mice. Although SKOV3ip1 (intervention study). Interestingly, both prevention and cells form many small metastases, mimicking miliary- intervention studies resulted in similar reductions disseminated ovarian cancer, HeyA8 cells form a few, in tumor weight and the number of metastases, large metastases in the omentum and mesentery that (Figure 1d). Given that in vitro, the human a5-integrin resemble the more solid tumor growth found in a antibody inhibited both ovarian cancer cell adhesion subgroup of patients. Tumors were counted, resected and proliferation (Supplementary Figure S5), our results and the tumor weight measured. Treatment with the suggest that a5b1-integrin is important for both the human a5b1-integrin-blocking antibody reduced tumor initial attachment/invasion and subsequent tumor weight in HeyA8 and SKOV3ip1 xenografts by 64% growth of ovarian cancer cells. Repeating the prevention (Po0.001) and 73% (Po0.001), respectively, compared experiment described above with murine a5Ab, expect- with isotype control treatment (Figures 1a and b). In edly, did not alter metastasis or tumor growth (Figure 1e). the SKOV3ip1 model, there was also a 57% (Po0.005) Based on these results, we performed a survival study decrease in the number of metastases (Figure 1b). after intraperitoneal injection of SKOV3ip1 by admin- These mice also produced smaller tumors than the istering the a5b1-integrin antibody or control immuno- controls, a finding which was paralleled by reduced globulin G (IgG) twice a week. The median survival Ki-67 staining, suggesting a lower proliferation rate was 39 days in the control-treated animals and 56 days upon treatment with the a5-integrin antibody (Figure 1b when the fibronectin receptor was inhibited (log rank: right panel). Po0.0001, Figure 1f).

Oncogene a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1568 HeyA8 Tumors SKOV3ip1 Tumors Ki-67 1.6 * * 2.5 250 2 1.2 200 IgG 150 1.5 0.8 1 100 0.4 0.5 50 Tumor Weight (g) Tumor Weight (g) 0 0 0 IgG Hu α Ab IgG Hu α Ab Number of Metastasis IgG Hu α Ab α 5 5 5 Hu 5Ab

** 80 60 40 20 positive α α % Ki-67 0 IgG Hu 5Ab IgG Hu 5Ab α IgG Hu 5Ab * * * * * * 1.2 250 1 200 0.8 150 0.6 100 0.4 Tumor Weight (g) 0.2 50 Number of Metastasis 0 0 Ab Ab Ab Ab Ab Ab 5 5 5 5 5 5 Ab + Ab + α α α α α α 5 5 α α Avastin Avastin Hu Hu Hu Hu Murine IgG Murine IgG Human IgG Human IgG Murine Murine Murine Murine

1.6 1.2 PREVENTION PREVENTION 1.6 INTERVENTION 250 INTERVENTION 0.8 200 0.4 1.2 150 0 Tumor Weight (g) IgG Murine α Ab 0.8 5 100 200 0.4 50 150 Tumor Weight (g)

0 Number of Metastasis 0 100 IgG Hu α Ab IgG Hu α Ab 5 5 50 Number of Metastasis 0 IgG Murine α Ab SKOV3ip1 Tumors 5 1.0

0.8

0.6 Control IgG 0.4 α Hu 5Ab 0.2 p<0.0001 Probability of Survival

10 2030 40 50 60 70 80 Days after tumor inoculation

6 Figure 1 Blocking a5b1-integrin in ovarian cancer cells inhibits metastasis in xenograft models. (a) HeyA8 cells (1 Â 10 ) were injected intraperitoneally (i.p.) in nude mice which were randomized into 2 groups of 10/group receiving 10 mg/kg human a5b1-integrin 6 antibody (Hu a5Ab) or control IgG. The effect of Hu a5Ab on tumor weight was assessed on day 20. (b) SKOV3ip1 cells (1 Â 10 ) were similarly injected i.p., treatment started on day 8, and tumor weight and the number of metastases assessed on day 35. Representative images of SKOV3ip1 tumors stained for Ki-67 ( Â 400). Images of sections from ﬁve mice were quantiﬁed (**Po0.01). (c) SKOV3ip1 cells (1 Â 106) were injected i.p. in nude mice randomized into six groups of 10 each, and were treated with the indicated antibodies starting on day 8. Mice were euthanized on day 35, and the effect on tumor weight and number of metastases was assessed. (*Po0.001, **Po0.01,). (d) The effect of coadministering SKOV3ip1 cells with a single dose of Hu a5Ab injected i.p. in mice (prevention) was compared with treatment of SKOV3ip1 tumor-bearing mice starting 8 days after injection of cancer cells (intervention). Tumor weight and number of metastases was measured. (e) A single dose of murine a5Ab was coinjected with SKOV3ip1 cells i.p. in nude mice (prevention), and tumor weight and number of metastases was measured on day 35. (f) Kaplan–Meier curve for survival. SKOV3ip1 cells were injected i.p. and 10 mice/group were treated with the human a5Ab or control IgG twice a week starting from day 8 (Po0.0001; log-rank test).

a5b1-Integrin regulates and interacts with c-Met investigate how expression of a5b1-integrin on tumor As the role of a5b1-integrin in angiogenesis is well cells regulates tumor cell function(s). Integrins can described (Kim et al., 2000; Taverna and Hynes, 2001; upregulate cellular signaling pathways directly, but Magnussen et al., 2005), we decided to further there are reports that integrin-mediated adhesion

Oncogene a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1569 activates RTKs (Vuori and Ruoslahti 1994; Wang et al., are part of the same pathway. The absence of any 1996; Nakamura et al., 2008; Ricono et al., 2009). We additive effect was also confirmed at lower concentra- and others have previously reported that c-Met con- tions of c-Met siRNA and the human a5-integrin tributes to the dissemination of ovarian cancer cells antibody (Supplementary Figure S6). Importantly, the (Saga et al., 2001; Wong et al., 2004; Sawada et al., inhibition of invasion by the antibody could be reversed 2007), which is why we investigated the interesting by activating c-Met with its ligand hepatocyte growth possibility that fibronectin-mediated activation of a5b1- factor/scatter factor (HGF/SF) (Figure 3b). This sup- integrin recruits c-Met signaling. To determine whether ports the concept that a5b1-integrin is upstream of the cell adhesion to fibronectin influences c-Met activation, c-Met receptor, but that c-Met can still be activated HeyA8 cells, which express both a5b1-integrin and independently of a5b1-integrin by its ligand HGF/SF. c-Met, were transfected with a small interfering RNA (siRNA) against a5-integrin and plated on fibronectin. After 24 h, c-Met was detected by immunoblotting. a5b1-Integrin activates a c-Met/FAK/Src signaling Silencing a5-integrin resulted in decreased phosphoryla- pathway tion of c-Met in the activation loop of the tyrosine Having shown that a5b1-integrin associates with and kinase domain (Tyr1230,1234,1235), but total c-Met expres- activates c-Met, we next sought to identify downstream sion was unchanged (Figure 2a, left panel). Treatment signaling pathways. Initial experiments focused on the with the a5b1-integrin-blocking antibody also blocked MAPK, STAT-3 and AKT pathways by measuring their c-Met phosphorylation and confirmed this finding key signaling proteins. However, none of these pathways (Figure 2a, middle panel). The c-Met phosphorylation were affected by a5b1-integrin inhibition (Supplemen- was specific to fibronectin adhesion and was not induced tary Figure S7 and data not shown). Given previous by tumor cell binding to collagen or poly-L-lysine reports that integrins use FAK and Src for signaling, we (Figure 2a right panel). Thereafter, we considered the turned our attention to these non-RTKs (Mitra and possibility that a5-integrin and c-Met associate, given Schlaepfer, 2006). Indeed, blocking a5b1-integrin with previous reports (Chan et al., 2006; Nakamura et al., the antibody in HeyA8 cells plated on fibronectin 2008) that integrins and RTKs can interact. Indeed, decreased FAK phosphorylation at Tyr397, which is the 861 c-Met coimmunoprecipitated with a5-integrin in IgG- FAK autophosphorylation site, and at Tyr , which is treated cells, and inhibiting a5-integrin led to a the Src phosphorylation site (Figure 4a). This was significant reduction in the association of the two confirmed when a5-integrin was silenced with a siRNA proteins, which was confirmed by immunoprecipitating (Figure 4a, lower panel) and also confirmed in the both with a5-integrin and c-Met (Figure 2b). A direct SKOV3ip1 ovarian cancer cell line (Supplementary interaction was confirmed by confocal microscopy, Figure S8a). Ectopic expression of a constitutively showing that, in ovarian cancer cells plated on fibronectin, active, membrane-bound form of FAK, CD2-FAK a5-integrin and c-Met colocalize (yellow overlay, (Chan et al., 1994; Huttenlocher et al., 1998), reversed, Figure 2c). The interaction of a5-integrin with c-Met is at least in part, the inhibitory effect of the a5b1-integrin also important in vivo, as mice bearing SKOV3ip1 on invasion, whereas neither the mutated CD2-FAK xenografts treated with the inhibitory human a5b1- (Y397F) plasmid nor the empty vector (CD2) had any integrin antibody have lower c-Met phosphorylation effect (Figure 4b and Supplementary Figure S8b for 1230,1234,1235 (Tyr ) in the tumor than mice treated with IgG SKOV3ip1 cells). Following the observation that a5b1- (Figure 2d). Moreover, blocking a5b1-integrin in normal integrin inhibition decreased FAK phosphorylation at human umbilical vein endothelial cells also inhibited Tyr861, which is a Src phosphorylation site (Mitra and c-Met phosphorylation (Figure 2e), suggesting that Schlaepfer, 2006), we determined whether blocking regulation of c-Met by a5b1-integrin can also occur in a5b1-integrin also inhibits Src phosphorylation and, untransformed cells. Analysis of nine ovarian cancer cell indeed, found decreased Src phosphorylation after lines showed that both proteins are coexpressed, and that antibody treatment (Figure 4c). thosecellsthathadhighexpressionofa5-integrinalsohad Although these experiments suggest that a5b1-integrin high expression of c-Met and vice versa (Supplementary regulates FAK and Src, they do not indicate whether Figure S2). The Pearson correlation coefficient for these two proteins are downstream of a5-integrin or c- coexpression of c-Met and a5-integrin was 0.61. Met. Both these interactions have been previously To understand the functional significance of a5b1- reported (Chen and Chen, 2006; Mitra and Schlaepfer, integrin signaling through c-Met, invasion assays were 2006; Jin et al., 2007). First, we tested whether a5b1- performed. HeyA8 cells were transfected with an siRNA integrin interacted with FAK and/or Src when HeyA8 against c-Met or a5-integrin, and then compared with cells were plated on fibronectin but neither immuno- HeyA8 cells treated with the human a5b1-integrin precipitation nor confocal microscopy showed any inter- antibody. After verifying silencing, we found that the action (data not shown). Next, we tested whether a5-integrin and the c-Met siRNA reduced invasion by 89 FAK and Src are downstream of the fibronectin/ and 91%, respectively (Figure 3a). Blocking the a5b1-integrin/c-Met axis. HeyA8 cells were transfected fibronectin receptor with a5b1-integrin antibody had a with the c-Met siRNA, plated on fibronectin, and the similar inhibitory effect (87%). There was no additional effect of c-Met silencing on FAK and Src phosphoryla- reduction of invasion when both a5b1-integrin and tion was compared with that of treatment with the c-Met inhibition were combined, suggesting that both a5b1-integrin antibody. Silencing of c-Met resulted in

Oncogene a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1570 Fibronectin + + - - - - α 5-integrin siRNA + - - Collagen I - - + + - -

Ab Poly-L-Lys - - - - + + Scrambled RNA - + - 5

α α Hu 5Ab + - + - + - α -integrin IgG 5 Hu IgG - + - + - +

p c-Met 145 kDa p c-Met 145 kD

c-Met 170 kDa c-Met 170 kD 145 kDa 145 kD Actin Actin 42 kDa 42 kD

α IP 5-integrin IP c-Met Fibronectin + + - - Fibronectin + + - - Poly-L-Lys - - + + Poly-L-Lys - - + + α α Hu 5Ab + - + - Hu 5Ab + - + - IgG - + - + IgG - + - + α WB c-Met 145 kDa WB 5-integrin 150 kDa α WB 5-integrin 150 kDa WB c-Met 145 kDa

α 5 Integrin c-Met Merged

p c-Met α Hu 5Ab IgG In Vivo HUVEC α α Hu 5Ab + - Hu 5Ab + - IgG - + IgG - + 145 kDa p c-Met p c-Met 145 kDa

170 kDa c-Met 145 kDa 160 * c-Met 145 kDa 120 80 Actin 42 kDa Control 40 Actin 42 kDa % of IgG 0 α Hu 5Ab IgG

Figure 2 a5b1-Integrin signals through the receptor tyrosine kinase c-Met. (a) Left: HeyA8 cells were transfected with a5-integrin siRNA 1230,1234,1235 or treated with human a5Ab and plated on fibronectin, followed by western blots for a5-Integrin, p-c-Met Y and c-Met. Right: HeyA8 cells were plated on fibronectin, collagen or poly-L-Lysine, and were treated with human a5Ab, followed by immunoblotting for 1230,1234,1235 phospho-c-Met Y , c-Met and actin. (b) HeyA8 cells were plated on fibronectin or poly-L-Lysine, and treated with human a5Ab for 24 h. Cell lysates were immunoprecipitated (IP) with an antibody against a5-integrin or c-Met, respectively, followed by immunoblotting (WB) for a5-integrin or c-Met. (c) Confocal microscopy. HeyA8 cells were plated on fibronectin-coated cover slips, detected with c-Met and a5-integrin antibodies followed by fluorescein or Alexa Fluor 594-labeled secondary antibodies, respectively. (d) SKOV3ip1 xenografts were treated with Hu a5Ab or IgG as described in Figure 1a. Left: Western blots for p-c-Met Y1230,1234,1235 and c-Met using tumor lysates. Right: Image of tumor sections stained for p-c-Met Y1200,1234,1235 ( Â 200 and Â 400). Images of sections from five mice were quantified using Image J (National Institutes of Health, Bethesda, MD, USA) with color deconvolution (*Po0.001). (e) Human umbilical vein endothelial 1230,1234,1235 cells were plated on fibronectin and treated with human a5Ab, followed by western blots for p-c-Met Y and c-Met.

a decrease in FAK (Tyr397) and Src phosphorylation histochemical staining showed a reduction of FAK 416 397 (Tyr ) (Figure 5a). The inhibitory effect of the a5b1- phosphorylation (Tyr ) in tumors treated with the integrin antibody on c-Met, FAK and Src phosphorylation a5b1-integrin inhibitory antibody, but not in the control could be overcome by stimulating cells with HGF/SF IgG-treated tumors (Figure 5c, right). To determine (Figure 5b). Addition of HGF/SF in the presence of the whether c-Met directly associated with and activated a5b1-integrin inhibitor resulted in a dose-dependent FAK and Src, lysates from a5b1-integrin antibody or rephosphorylation of c-Met, FAK and Src. The IgG-treated HeyA8 cells were immunoprecipitated with in vitro results were conﬁrmed in the a5b1-integrin anti c-Met antibody. We observed that Src coimmuno- antibody-treated SKOV3ip1 tumors. Western blotting precipitated with c-Met (Figure 5d) but not FAK (data showed a decrease in FAK (Tyr397, Tyr861) and Src not shown). Taken together, these data indicate that 416 (Tyr ) phosphorylation in tumor tissue from integrin ﬁbronectin activates an a5b1-integrin/c-Met/Src/FAK- antibody-treated tumors (Figure 5c, left). Immuno- dependent signaling pathway.

Oncogene a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1571 α 5-integrin siRNA + - - c-Met siRNA + - - Scrambled RNA - + - Scrambled RNA - + - 170 kDa α -integrin 5 150 kDa c-Met 145 kDa

Actin 42 kDaActin 42 kDa

800

600

400

200

0 Number of Invaded Cells Ab Ab IgG 5 5 α α Mock Hu c-Met siRNA -Integrin siRNA 5 c-Met siRNA+IgG α Scrambled siRNA c-Met siRNA+Hu

1000 800 600 400 200 0 Number of Invaded Cells Ab IgG 5 α HGF/SF Hu Untreated Ab+HGF/SF Ab+HGF/SF Ab+HGF/SF IgG+HGF/SF 5 5 5 α α α Hu C-Met siRNA+Hu Scrambled siRNA+Hu

Figure 3 Inhibiting a5b1-integrin, silencing c-Met or both blocks invasion, which can be overcome by activation of c-Met by HGF. (a) HeyA8 cells were transfected with the indicated siRNAs and seeded on matrigel-coated Boyden chambers (BD Falcon), treated with human a5Ab and allowed to invade for 24 h. Silencing was conﬁrmed by immunoblotting. (b) HeyA8 cells or HeyA8 cells transfected with c-Met siRNA were placed in matrigel-coated Boyden chambers, treated with human a5Ab with or without HGF/SF or untreated control as indicated, and allowed to invade for 24 h. For (a) and (b), the average number of invaded cells/ﬁeld was plotted against the treatment given. Results are from three independent experiments.

Discussion sites of ovarian cancer metastasis. Our data provide evidence for a novel mechanism by which fibronectin The ECM protein, fibronectin, is a potent mediator of regulates ovarian cancer cell signaling. Binding of a5b1- cell adhesion, migration and growth. It is also one of the integrin to fibronectin recruits the RTK, c-Met, which most abundant ECMs and basement membrane proteins then activates downstream signaling events. The pre- in human omentum and peritoneum ((Franke et al., sence of this mechanism is supported by experiments 2003) and Supplementary Figure S1), the most common showing that inhibition of a5b1-integrin using a siRNA

Oncogene a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1572

Ab CD2FAK + - - - 5

α CD2FAK Y397F - + - - IgG Hu CD2 - - + - p FAK Y397 125 kDa CD2FAK 160 kDa 125 kDa FAK 125 kDa FAK

p FAK Y861 125 kDa p CD2FAK Y397 160 kDa p FAK Y397 125 kDa FAK 125 kDa

α Actin 5-integrin siRNA + - - 42 kDa Scrambled RNA - + -

α -integrin IgG 5 150 kDa α 800 Hu 5Ab p FAK Y397 125 kDa 600 400 FAK 125 kDa 200 0 Ab 5 α Number of Invaded cells CD2 Hu IgG

p Src Y416 60 kDa CD2FAK

Src 60 kDa CD2FAK Y397F

Figure 4 Adhesion of ovarian cancer cells to fibronectin activates FAK and Src through a5b1-integrin. (a) Top: HeyA8 cells were transfected with a5-integrin siRNA or treated with human a5Ab, and plated on fibronectin for 24 h. Western blots were performed for 397 861 p-FAK-Y , p-FAK-Y , FAK and a5-integrin. (b) HeyA8 cells were transfected with CD2FAK (constitutively active FAK), CD2FAK-Y397F (mutated/inactive FAK) or CD2 (vector). Top: Expression and phosphorylation (Y397) of CD2FAK (160kDa) was verified by western blotting. Bottom: Transfected HeyA8 cells were added to matrigel-coated Boyden chamber (BD Falcon) with human a5Ab or IgG, and allowed to invade for 24 h. The number of invaded cells was counted. (c) HeyA8 cells were plated on 416 fibronectin and treated with human a5Ab or IgG for 24 h. Phosphorylated (p)-Src Y and Src were detected by immunoblotting.

or a blocking antibody inhibits fibronectin-mediated flexibility to use the ECM as well as paracrine HGF/SF c-Met phosphorylation both in vitro and in vivo. to gain access to c-Met-directed signaling pathways. Immunoprecipitation and visualization of c-Met and However, although c-Met is the signal mediator after a5-integrin using confocal microscopy imply that the a5b1-integrin activation, it can also initiate integrin- adhesion receptor and the RTK associate to form a mediated signaling. Activation of c-Met leads to molecular signaling unit. Inhibition of a5b1-integrin also association with the a6b4-integrin and tyrosine phos- led to a decrease in phosphorylation of c-Met in human phorylation of the b4-integrin cytoplasmic domain umbilical vein endothelial cells, indicating that this (Trusolino et al., 2001). In this case, the integrin does process is not specific for cancer cells alone. not act as the adhesion receptor, but rather as a The first evidence that adhesion can regulate c-Met mediator of c-Met signaling. came from an observation in the Bishop Laboratory To determine whether inhibition of the fibronectin (Wang et al., 1996) at UCSF, showing that adherence of receptor on cancer cells plays a role in either early or late cancer cells to fibronectin and type IV collagen induces tumor metastasis, ovarian cancer cells received either a phosphorylation of c-Met. This observation was con- single pretreatment with the a5b1-integrin antibody sistent with an inverse experiment using melanoma before intraperitoneal injection (prevention) or were cancer cells, which showed that the constitutive activa- treated repeatedly after solid tumors were established tion of c-Met is lost upon cellular detachment (Rusciano within the peritoneal cavity (intervention). The results et al., 1996). The data presented in our study indicate indicated that there are at least two distinct mechanisms that cell adhesion regulates c-Met through direct by which the a5b1-integrin expression on tumor cells activation by a5b1-integrin. When a5b1-integrin- promotes ovarian cancer metastasis. It contributes to mediated c-Met activation was blocked, c-Met could the initial adhesion/invasion on the fibronectin-rich still be stimulated by its ligand, HGF/SF, suggesting peritoneum and omentum, and it later helps the cancer that in ovarian cancer cells, c-Met can be activated both cells to metastasize and proliferate as the tumor in a ligand-dependent and -independent manner. The progresses. In contrast, host a5b1-integrin contributes ligand-dependent activation triggers the canonical path- to tumor angiogenesis essential for later stages of ways involving Erk and AKT (Birchmeier et al., 2003), metastasis (Kim et al., 2000; Magnussen et al., 2005). which we and others have shown to be inhibited by The integrin-mediated activation of Src and FAK is small molecule inhibitors of c-Met (Birchmeier et al., not a consequence of direct binding of these non-RTKs 2003; Zillhardt et al., 2010). Fibronectin-mediated to the integrin, but occurs via c-Met. Expression of an activation resulted in signaling via an alternate pathway activated form of FAK was able to overcome the involving Src and FAK. The cell, therefore, has the inhibitory function of the a5-integrin-blocking antibody

Oncogene a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1573 Fibronectin ++++ α Hu 5Ab +--- Fibronectin ++++ α c-Met siRNA -+-- Hu 5Ab +++- Scrambled siRNA --+- HGF/SF (ng/ml) -4080-

p c-Met 145 kDa p c-Met 145 kDa 170 kDa c-Met 145 kDa 170 kDa c-Met 145 kDa Actin 42 kDa Actin 42 kDa

p FAK Y397 125 kDa p FAK Y397 125 kDa

p FAK Y861 125 kDa FAK 125 kDa

FAK 125 kDa p Src Y416 60 kDa

p Src Y416 60 kDa Src 60 kDa

Src 60 kDa Ab

In Vivo 5 α p FAK Hu IgG α Hu 5Ab IgG p FAK Y397 125 kDa * p FAK Y861 125 kDa 120 80 40 FAK 125 kDa 0 α Hu 5Ab IgG p Src Y416 60 kDa % of IgG Control

Src 60 kDa

IP c-Met IP Src Ab Ab 5 5 α α Hu IgG Hu IgG WB c-Met WB Src 60 kDa 145 kDa

WB Src 60 kDa WB c-Met 145 kDa

Figure 5 a5b1-Integrin signals through c-Met to activate FAK and Src. (a) HeyA8 cells were transfected with c-Met siRNA or with scrambled siRNA, plated on fibronectin and treated with human a5Ab. Cells were lysed, resolved on SDS–polyacrylamide gel electrophoresis, transferred to nitrocellulose membrane and probed with the indicated antibodies. (b) HeyA8 cells were plated on fibronectin and treated with human a5Ab as indicated for 24 h. The cells were then treated with HGF/SF for 30 min and immunoblotting performed with the specified antibodies. Results in (a) and (b) are representative of three independent experiments. (c) SKOV3ip1 xenografts were treated with human a5Ab or with IgG as described in Figure 1 (b) Left: Tumor lysates were immunoblotted with the indicated antibodies. Right: Representative image of tumor sections stained for p-FAK-Y397 ( Â 200). Images of sections from five mice were quantified using Image J (*Po0.001). (d) Immunoprecipitation. HeyA8 cells were plated on fibronectin and treated with human a5Ab or control IgG for 24 h, followed by immunoprecipitation with an antibody against c-Met or Src, respectively, followed by immunoblotting (WB) Src and c-Met. All results are representative of at least three independent experiments. on invasion, as was stimulation of cells by HGF/SF. activation of c-Met is instead the regulation of c-Met by These results were confirmed in vivo, as treatment with a5b1-integrin.Taking into consideration that c-Met can the human a5b1-integrin antibody blocked both FAK be activated in a ligand-dependent as well as in a ligand- and Src activation in the tumor. We were never able to independent manner, current strategies to inhibit c-Met show a direct interaction between a5-integrin and FAK could be reconsidered: inhibition of the c-Met cyto- or Src (data not shown), despite reports that FAK can plasmic domains (for example, the kinase domain) directly interact with integrins (Mitra and Schlaepfer, could be more effective than targeting the HGF ligand- 2006). However, activation of c-Met by a5b1-integrin binding site. resulted in the direct binding of Src to c-Met (but not In summary, we define a signaling pathway in ovarian FAK). This suggests that upon binding to fibronectin, cancer that starts with the fibronectin-mediated activa- a5b1-integrin activates c-Met, which then associates with tion of a5b1-integrin and leads to direct contact with Src and subsequently activates FAK. Consistent with c-Met, followed by Src and FAK activation, linking our results, Chen and colleagues reported that when c- mitogenic RTK signaling with adhesion (Figure 6). This Met is constitutively activated in aggressive lung cancer may be one of the mechanisms utilized by cancer cells cells, FAK interacts with c-Met independent of stimula- to constitutively activate c-Met in a fibronectin-rich tion with HGF/SF. This was not observed in less microenvironment. In particular, this mechanism could invasive cells (Chen and Chen, 2006). It is possible be of special importance in ovarian cancer because the that most of what is seen as the ligand-independent first step of ovarian cancer metastasis begins with cancer

Oncogene a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1574 Mouse orthotopic xenograft model of ovarian cancer Fibronectin The xenograft model was described previously (Sawada et al., 2007, 2008). Brieﬂy, HeyA8 or SKOV3ip1 cells (1 Â 106) were injected intraperitoneally into 6-week old, female, athymic nude mice. On days 3 (HeyA8) and 8 (SKOV3ip1) after c-Met - -Integrin injection, by which time the cells had metastasized and visible 5 1 metastatic tumors had developed, mice were randomized into Cell Membrane treatment groups. All antibodies were injected intraperitoneally at a dose of 10 mg/kg, twice a week. Treatment continued p p until day 20 for HeyA8 and day 35 for SKOV3ip1 models.

Immunohistochemistry p Immunohistochemical studies were performed by using 5 mm Invasion Src thick formalin-fixed deparaffinized sections as previously and described (Sawada et al., 2008). Xenograft tumor tissue was Proliferation probed with anti-CD31 antibody (1:100), anti-Ki-67 (1:300), p anti-p-c-Met Y1230, 1234, 1235 (1:50) and anti-p-FAK Y397 (1:200). Specimens of normal human omentum that were obtained FAK from patients undergoing surgery for benign conditions were stained for fibronectin (1:600), collagen IV (1:50, laminin

Figure 6 Fibronectin-/a5b1-integrin-mediated regulation of c-Met. (1:50)) or vitronectin antibody (1:500). Proposed role of ligand-independent activation of c-Met involving a fibronectin/a5b1-integrin/c-Met/Src/FAK-dependent signaling pathway. Western blot analysis Cells or tumor tissues were lysed, proteins were separated by 4–20% gradient SDS–polyacrylamide gel electrophoresis and transferred to nitrocellulose, and probed with the respective cell adhesion. Indeed, it could provide an explanation primary and secondary antibodies (Sawada et al., 2008). for how, in ovarian cancer, the early adhesion of cancer cells to the fibronectin-rich basement membranes Immunoprecipitation instantly activates key mitogenic-signaling pathways in Cell lysates were precleared with protein A/G agarose beads the cell, allowing a single ovarian cancer cell or a small and incubated with anti-a5-integrin, c-Met or Src antibodies, cluster of cells to survive, invade and subsequently thrive followed by protein A/G agarose beads (Radjabi et al., 2008). in the initially hostile environment on the surface of the The beads were then spun down and washed, boiled for 5 min abdominal cavity. in SDS–polyacrylamide gel electrophoresis-gel loading buffer, chilled and spun down. Equal amounts of the supernatant were then separated by 4–20% SDS–polyacrylamide gel electrophoresis, transferred to nitrocellulose membrane, Materials and methods probed for coimmunoprecipitation of c-Met, a5-integrin or Src antibodies and later stripped and reprobed for the Reagents and cell lines immunoprecipitated protein. Anti-FAK and paxillin antibodies, growth-factor-reduced matrigel and human fibronectin were from BD Biosciences (San Jose, CA, USA); phospho-FAK Y397, phospho-FAK Y861 Immunofluorescence imaging and phospho-c-Met Y1230,1234,1235 were from Biosource (Carlsbad, HeyA8 cells were plated in serum-free medium on fibronectin- coated cover slips (5 ug/ml) and allowed to attach overnight. CA, USA). Antibodies against c-Met, a5-integrin, CD31 and protein A/G agarose beads were from Santa Cruz Biotechno- Thereafter, the cells were fixed with 4% paraformaldehyde in logy Inc. (Santa Cruz, CA, USA). Anti-Src and phospho- phosphate buffered saline, permeabilized with 0.1% Triton X- SrcY416 antibodies were purchased from Upstate (Billerica, 100 and probed with rabbit anti-a5-integrin and mouse anti-c- 1 MA, USA). Antibodies against AKT, phospho-AKT S473, Met antibodies at 4 C overnight (1:200). The samples were phospho-ERK1/2, ERK1/2, phospho-JNK, JNK, phospho- washed and incubated with Alexa Fluor 594-labeled goat anti- p38 MAPK and p38 MAPK were from Cell Signaling rabbit antibody (1:1000) and fluorescein-labeled goat anti- (Danvers, MA, USA). The fibronectin antibody was from mouse antibody (1:1000), mounted with Prolong Gold Sigma-Aldrich (St Louis, MO, USA); Ki-67 antibody was (Invitrogen) and analyzed using a Leica SP2 laser scanning from LabVision (Fremont, CA, USA); vimentin, collagen IV confocal microscope (Leica, Wetzlar, Germany). and laminin antibodies were from Dako Cytomation (Carpinteria, CA, USA). The 5-chloromethylfluorescein diacetate was from Transient transfection Invitrogen (Carlsbad, CA, USA). The human ovarian cancer HeyA8 cells were transfected with 30 nM siRNA against a5- cell lines SKOV3ip1 (first described by Janet Price (Yu et al., integrin (Ambion, Austin, TX, USA) or against c-Met using 1992; Brader et al., 1997) and HeyA8 were from Gordon B. Lipofectamine 2000 transfection reagent (Invitrogen). The Mills (MD Anderson Cancer Center, Houston, TX, USA). plasmids for the constitutively activated FAK (CD2FAK), The function-blocking antibodies against human a5b1-integrin CD2FAK with a mutation at the autophosphorylation site (M200, referred to as Hu a5Ab), mouse a5b1-integrin (339.1, (CD2FAK Y397F) and the control vector (CD2) not expres- referred to as Murine a5Ab) and Avastin (Genentech) were sing FAK were a kind gift from Dr Alan Horwitz, University provided by Facet Biotech Corp (Redwood City, CA, USA). of Virginia School of Medicine. HeyA8 cells or SKOV3ip1 Non-specific human or mouse IgG was from Chrome Pure, cells were seeded in 6-well plates and transfected with 1 mg Jackson ImmunoResearch, West Grove, PA, USA. DNA using Fugene6 transfection reagent (Roche, Basel,

Oncogene a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1575 Switzerland). The cells were subsequently used for invasion Statistical analysis assays. Data analysis was done by unpaired, two-tailed Student’s t-test assuming equal variance of the test and the control populations.

Matrigel invasion assay Cellular invasion was assayed in vitro as described previously Conflict of interest using matrigel coated Boyden chambers (BD Falcon, Franklin Lakes, NJ, USA) (Sawada et al., 2008). Wherever applicable, Facet Biotech Corp. provided funding and antibodies for the the human a5b1-integrin antibody or control IgG (25 mg/ml) animal experiments (Figure 1). was added to the top chamber. Cells were fixed with 4% paraformaldehyde after 24 h, stained with giemsa and en- umerated. Experiments were performed in triplicate and 5 fields ( Â 200 magnification) per filter counted. Acknowledgements

We thank Dr Alan Horwitz, Department of Cell Biology, Adhesion assay University of Virginia School of Medicine for providing us Cellular adhesion was assayed in vitro as described previously with the FAK constructs. We thank Gail Isenberg for critically (Sawada et al., 2008). Briefly, ovarian cancer cells were reviewing the manuscript. We appreciate the inputs of fluorescently labeled with 10 mM 5-chloromethylfluorescein Dr Vinay Bhaskar and Dr Vanitha Ramakrishnan during diacetate and treated with the human a5b1-integrin antibody the initial part of the project. Funding: Ernst Lengyel holds a or control IgG (25 mg/ml), and plated in a 96-well plate Clinical Scientist Award in Translational Research from the precoated with fibronectin (5 mg/ml). Adherent cells were Burroughs Wellcome Fund and is supported by grants from quantified by measuring fluorescent intensity (excitation, the Ovarian Cancer Research Fund (Liz Tilberis Scholars 490 nm; emission, 520 nm). Program), and the NCI (RO1 CA111882).

References

Avraamides CJ, Garmy-Susini B, Varner JA. (2008). Integrins in Huttenlocher A, Lakonishok M, Kinder M, Wu S, Truong T, angiogenesis and lymphangiogenesis. Nat Rev Cancer 8: 604–617. Knudsen K et al. (1998). Integrin and cadherin synergy regulates Bhaskar V, Zhang D, Fox M, Seto P, Wong M, Wales PE et al. (2007). contact inhibition of migration and motile activity. 141: 515–526. A function blocking anti-mouse integrin a5b1 antibody inhibits Hynes RO. (2002). Integrins: bidirectional, allosteric signaling angiogenesis and impedes tumor growth in vivo. J Transl Med 5: machines. Cell 110: 673–687. 1–11. Jemal A, Siegel R, Ward E, Hao Y, Xu J, Thun MJ et al. (2009). Birchmeier C, Birchmeier W, Gherardi E, Vande Woude GF. Cancer statistics, 2009. CA Cancer J Clin 59: 225–249. (2003). Met, metastasis, motility and more. Nat Rev Mol Cell Biol Jin E, Choi Y, Park E, Bang O, Kang S. (2007). MMP-2 functions as a 4: 915–925. negative regulator of chondrogenic cell condensation via Brader KR, Wolf JK, Hung M-C, Yu D, Crispens MA, van Golen KL down-regulation of the FAK -integrin b1 interaction. Devel Biol et al. (1997). Adenovirus E1A expression enhances the sensitivity of 308: 474–484. an ovarian cancer cell line to multiple cytotoxic agents through an Kenny HA, Lengyel E. (2009). MMP-2 functions as an early response apoptotic mechanism. Clin Cancer Res 3: 2017–2024. protein in ovarian cancer metastasis. Cell Cycle 8: 683–688. Caswell PT, Spence HJ, Parsons M, White DP, Clark K, Cheng KW Kim S, Bell K, Mousa SA, Varner JA. (2000). Regulation of et al. (2007). Rab25 associates with a5b1 integrin to promote angiogenesis in vivo by ligation of integrin a5b1with the central invasive migration in 3D microenvironments. Dev Cell 13: 496–510. cell-binding domain of fibronectin. Am J Pathol 156: 1345–1362. Chan P-C, Chen SY, Chen C-H, Chen HC. (2006). Crosstalk between Lengyel E. (2010). Ovarian cancer development and metastasis. Am J hepatocyte growth factor and integrin signaling pathways. J Biomed Pathol 177: 1053–1064. Sci 13: 215–223. Magnussen A, Kasman I, Norberg S, Baluk P, Murray R, McDonald Chan P-Y, Kanner SB, Whitney G, Aruffo A. (1994). A transmem- D. (2005). Rapid access of antibodies to a5b1 integrin overexpressed brane-anchored chimeric focal adhesion kinase is constitutively on the luminal surface of tumor blood vessels. Cancer Res 65: activated and phosphorylated at tyrosine residues identical to 2712–2721. pp125fak*. J Biol Chem 269: 20567–20574. Mitra SK, Schlaepfer D. (2006). Integrin-regulated FAK-Src signaling Chen SY, Chen H-C. (2006). Direct interaction of focal adhesion in normal and cancer cells. Curr Opin Cell Biol 18: 516–523. kinase (FAK) with met is required for FAK to promote hepatocyte Nakamura Y, Matsubara D, Goto A, Ota S, Sachiko O, Ishikawa S growth factor-induced cell invasion. Mol Cell Biol 26: 5155–5167. et al. (2008). Constitutive activation of c-Met is correlated with c- Francis SE, Goh KL, Hodivala-Dilke KM, Bader BL, Stark M, Met overexpression and dependent on cell-matrix adhesion in lung Davidson D et al. (2002). Central roles of a5b1 integrin and adenocarcinoma cell lines. Cancer Sci 99: 14–22. fibronectin in vascular development in mouse embryos and Radjabi AR, Sawada K, Jagadeeswaran S, Eichbichler A, Kenny HA, embryoid bodies. Arterioscler Thromb Vasc Biol 22: 927–933. Montag A et al. (2008). Thrombin induces tumor invasion through Franke FE, Von Georgi R, Zygmunt M, Mu¨nstedt K. (2003). the induction and association of matrix metalloproteinase-9 and Association between fibronectin expression and prognosis in b1-integrin on the cell surface. J Biol Chem 283: 1463–1472. ovarian carcinoma. Anticancer Res 23: 4261–4268. Ricono JM, Huang M, Barnes LA, Lau SK, Weis S, Schlaepfer D et al. Giancotti FG, Tarone G. (2003). Positional control of cell fate through (2009). Specific cross-talk between epidermal growth factor receptor joint integrin/receptor protein kinase signaling. Annu Rev Cell Dev and integrin avb5 promotes carcinoma cell invasion and metastasis. Biol 19: 173–206. Cancer Res 69: 1383–1391. Hafter R, Klaubert W, Gollwitzer R, Graeff H. (1984). Crosslinked Rusciano D, Lorenzoni P, Burger MM. (1996). Constitutive activation fibrin derivatives and fibronectin in ascitic fluid from patients with of c-Met in liver metastatic B16 melanoma cells depends on both ovarian cancer compared to ascitic fluid in liver cirrhosis. Thromb substrate adhesion and cell density and is regulated by a cytosolic Haemost 35: 53–64. tyrosine phosphatase activity. J Biol Chem 271: 20763–20769.

Oncogene a5b1-integrin and c-Met in ovarian cancer AK Mitra et al 1576 Saga Y, Mizukami H, Suzuki H, Urabe M, Kume A, Nakamura T Wingerter P, Kazman I, Norberg S, Magnussen A, Zanivan S, et al. (2001). Expression of HGF/NK4 in ovarian cancer cells Rissone A et al. (2005). Uniform overexpression and rapid suppresses intraperitoneal dissemination and extends host survival. accessibility of a5b1 integrin on blood vessels in tumors. Am J Gene Ther 8: 1450–1455. Pathol 167: 193–211. Sawada K, Mitra AK, Radjabi AR, Bhaskar V, Kistner E, Tretiakova MS Wong AST, Roskelley CD, Pelech SL, Miller D, Leung PCK, et al. (2008). Loss of E-cadherin promotes ovarian cancer metastasis via Auersperg N. (2004). Progressive changes in Met-dependent alpha 5-integrin, which is a therapeutic target. Cancer Res 68: 2329–2339. signaling in a human ovarian surface epithelial model of malignant Sawada K, Radjabi AR, Shinomiya N, Kistner E, Kenny HA, Salgia transformation. Exp Cell Res 299: 248–256. R et al. (2007). C-Met overexpression is a prognostic factor in Yang JT, Rayburn H, Hynes RO. (1993). Embryonic meso- ovarian cancer and an effective target for inhibition of peritoneal dermal defects in a5 integrin-deﬁcient mice. Development 119: dissemination and invasion. Cancer Res 67: 1670–1680. 1093–1105. Taverna D, Hynes RO. (2001). Reduced blood vessel formation and Yokoyama Y, Sedgewick G, Ramakrishnan S. (2007). Endostatin tumor growth in a5-integrin-negative teratocarcinomas and embry- binding to ovarian cancer cells inhibits peritoneal attachment and oid bodies. Cancer Res 61: 5255–5261. dissemination. Cancer Res 67: 10813–10822. Trusolino L, Bertotti A, Comoglio PM. (2001). A signaling adapter Yu D, Wolf JK, Scanlon M, Price JE, Hung M-C. (1992). Enhanced c- function for a6b4 integrin in the control of HGF-dependent invasive erb B-2/neu expression in human ovarian cancer cells correlates with growth. Cell 107: 643–654. more severe malignancy that can be suppressed by E1A. Cancer Res Vuori K, Ruoslahti E. (1994). Association of insulin receptor 53: 891–898. substrate-1 with integrins. Science 266: 1576–1578. Zillhardt M, Christensen J, Lengyel E. (2010). An orally available Wang R, Kobayashi R, Bishop M. (1996). Cellular adherence elicits small molecule inhibitor of c-Met, PF-2341066, reduces tumor ligand-independent activation of the Met cell -surface receptor. Proc burden in a pre-clinical model of ovarian cancer metastasis. Natl Acad Sci USA 93: 8425–8430. Neoplasia 12: 1–10.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene