The urokinase receptor (uPAR) is upregulated upon tumor cell invasion and correlates with poor lung cancer survival. Although a cis-interaction with integrins has been ascribed to uPAR, whether this interaction alone is critical to urokinase (uPA)- and uPAR-dependent signaling and tumor promotion is unclear. Here we report the functional consequences of point mutations of uPAR (H249A-D262A) that eliminate β1 integrin interactions but maintain uPA binding, vitronectin attachment and association with αV integrins, caveolin and epidermal growth factor receptor. Disruption of uPAR interactions with β1 integrins recapitulated previously reported findings with β1-integrin-derived peptides that attenuated matrix-dependent ERK activation, MMP expression and in vitro migration by human lung adenocarcinoma cell lines. The uPAR mutant cells acquired enhanced capacity to adhere to vitronectin via uPAR–αVβ5-integrin, rather than through the uPAR–α3β1-integrin complex and they were unable to initiate uPA signaling to activate ERK, Akt or Stat1. In an orthotopic lung cancer model, uPAR mutant cells exhibited reduced tumor size compared with cells expressing wild-type uPAR. Taken together, the results indicate that uPAR–β1-integrin interactions are essential to signals induced by integrin matrix ligands or uPA that support lung cancer cell invasion in vitro and progression in vivo.
Lung cancer is the leading cause of cancer-related death in the United States (Jemal et al., 2007). The progression of localized cancer to an invasive form is usually associated with genetic, morphological and phenotypical changes in tumor cells (Sung et al., 2007). The coordinated expression and function of adhesion receptors (particularly integrins) and protease systems that counteract cell-substratum and cell-cell interactions play a critical role in this process. Elevated expression of both the urokinase (uPA) system (which includes uPA, the urokinase receptor uPAR and plasminogen activator inhibitor type 1 or PAI1) and β1 integrins (especially integrin α3β1 and integrin α5β1) correlates with lung cancer progression (He et al., 2001; Wang et al., 2004; Wei et al., 2007; Zhao et al., 2002). The basis for this correlation is largely assumed to be the activation of uPA-dependent plasmin and matrix metalloproteinase (MMP) pathways of extracellular proteolysis (Inuzuka et al., 2000). Suppression of uPAR by antisense expression is reported to markedly suppress tumor metastasis in lung cancer models (Lakka et al., 2001; Rao et al., 2005). However, the mechanisms by which uPAR overexpression lead to lung cancer malignancy in vivo are largely unknown.
The receptor uPAR is a glycosylphosphatidylinositol anchor protein (Ploug et al., 1991) and is known to regulate tumor cell adhesion, migration, invasion, protease secretion and proliferation through interactions with various membrane partners, including several integrins (Carriero et al., 1999; Saldanha et al., 2007; Simon et al., 1996; Wei et al., 1996; Xue et al., 1997). The uPAR binding sites for integrin α3β1 and integrin α5β1 have been congruently identified in Domain III (Chaurasia et al., 2006; Wei et al., 2008; Wei et al., 2007) and a sequence in Domain II might also be important for integrin association (Degryse et al., 2005). It has been reported that uPAR enhances human head and neck carcinoma cell proliferation by interacting with α5β1 integrin, regulating downstream signaling cascades both in vitro and in vivo (Aguirre Ghiso et al., 1999; Chaurasia et al., 2006). In oral squamous carcinoma cells, uPAR–α3β1-integrin interaction potentiates cellular signal transduction pathways that lead to the activation of uPA expression and enhance uPA-dependent invasive behavior (Ghosh et al., 2006). We have previously shown that uPAR directly binds to integrin α3β1 and integrin α5β1, initiating epithelial-to-mesenchymal transition and promoting tumor cell migration, respectively (Wei et al., 2005; Zhang et al., 2003). Our recent findings suggest that uPAR binding to α5β1 integrin is required for maximal responses to fibronectin and tumor cell invasion, and this operates through an enhanced Src-Rac-ERK signaling pathway (Wei et al., 2007). However, to what extent uPAR–β1-integrin association and function is involved in lung cancer invasion and progression remain to be elucidated.
The serine protease uPA binds to its receptor uPAR, affecting cell migration and signaling (Nguyen et al., 1999; Yebra et al., 1999). Although uPA has also been found to bind integrin αVβ3 or integrin αVβ5 with low affinity and promote signaling and directional cell migration (Franco et al., 2006; Tarui et al., 2006), other reports suggest that uPA-induced phosphorylation of extracellular signal-regulated kinase MAP kinase (ERK), and migration and/or invasion might require uPAR–β1-integrin complex (Ahmed et al., 2003; Mazzieri et al., 2006). Other than ERK activation (Aguirre Ghiso et al., 1999; Nguyen et al., 1998), uPA binding to uPAR also activates the Jak-Stat signal transduction pathway, such as Stat1 in vascular cells (Dumler et al., 1998), Stat3 in prostate cancer cells or lung epithelial cells (Pulukuri et al., 2005; Shetty et al., 2006) and Stat5b in CHO cells (Jo et al., 2005), as well as the phosphatidylinositol 3-kinase (PI3K) and Akt–GSK-3β signaling pathways (Galaria et al., 2005). Whether these signaling pathways stimulated by uPA require uPAR–β1-integrin interaction is not clear. The objective of the present study was to examine the role of uPAR–β1-integrin association on different aspects of tumor invasion and to identify the intracellular pathways involved in these events.
In this report we demonstrate that expression of a uPAR mutant bearing H249A-D262A double mutations (HD mutant), unable to bind either integrin α5β1 or integrin α3β1, changed the mechanism for vitronectin adhesion, impaired fibronectin- or laminin-5-dependent ERK phosphorylation and MMP expression and inhibited uPA-stimulated activation of ERK, Stat1 and Akt–GSK-3β. Either knocking down uPAR by RNAi or expressing HD mutant uPAR in lung cancer cells significantly suppressed Matrigel invasion in vitro and tumor growth in vivo in an orthotopic lung cancer model. The results suggest that both uPA–uPAR–β1-integrin signaling and uPAR–β1-integrin–matrix signaling contribute to lung cancer cell invasion in vitro and progression in vivo.
Generation, expression and characterization of H249A-D262A mutant uPAR
We have previously re-expressed wild-type uPAR (WT), H249A (H) or D262A (D) mutant uPARs into H1299 cells with stable uPAR knockdown to study the importance of uPAR–α5β1-integrin complexes in tumor cell function and signaling (Wei et al., 2007). Since H or D single mutation on uPAR did not completely block H1299 tumor cell invasion in vitro (Fig. 5A) and the effect of single mutations on xenograft tumor growth was not observed in a trial experiment, we asked whether a H249A-D262A (HD) double mutation would have stronger effects. HD mutant uPAR was generated by site-directed mutagenesis (Wei et al., 2007), reconstituted in uPAR-knockdown H1299 cells and examined by FACS analysis (Fig. 1A) and western blot (results not shown). Reconstitution of WT and mutant uPAR does not affect expression levels of cell surface integrins including integrin α3, integrin α5, integrin α6, integrin αVβ3, integrin αVβ5 and integrin β1 (supplementary material Fig. S1). As shown in Fig. 1B, uPA binding to HD H1299 cells was not impaired compared with WT cells. Cells in which uPAR was knocked down (shu) exhibited only very low levels of endogenous uPAR and showed substantially weaker binding of uPA fragment 1-48-Fc.
As uPAR is known to associate with multiple partners important for vitronectin adhesion (Blasi and Carmeliet, 2002), we assessed the physical association between uPAR and caveolin-1, αV integrin or epidermal growth factor receptor (EGFR) by co-immunoprecipitation, similar amounts of which were detected in the uPAR precipitates from WT and HD mutant cells (Fig. 1C). Overall, these data indicate that the HD mutant uPAR bearing cells have a very selective defect in β1 integrin binding but otherwise appear to interact normally with known membrane uPAR partners. We then used this mutant to explore the functional consequences of selective loss of β1 integrin binding on uPAR functions.
Cells expressing WT and mutant uPAR use different integrins to adhere to vitronectin
Although uPAR contains a binding site for direct association with vitronectin, we previously reported that uPAR–β1-integrin interactions are important for cell adhesion to vitronectin (Wei et al., 2001; Wei et al., 1996). Specifically, vitronectin binding to HEK293 cells overexpressing uPAR could be blocked by antibodies to α3 integrin (Wei et al., 2001) or by peptides that block uPAR–α3-integrin association (Degryse et al., 2005). By contrast, a recent report used a mutant uPAR, in which all known integrin-binding residues including H249 and D262 were mutated to alanine, to show that vitronectin adhesion was equivalent between HEK293 and CHO cells expressing WT and mutant uPAR (Madsen et al., 2007). In agreement with these findings, we observed that HEK293 cells stably expressing either WT or HD mutant uPAR were obviously more adhesive on vitronectin than parental cells and exhibited similar morphologies. However, whereas vitronectin adhesion of cells expressing WT uPAR was inhibited by treatment with an α3-integrin blocking antibody, uPAR–α3β1-integrin blocking peptide (α325), or uPAR–β1-integrin blocking peptide (β1P1), as previously reported, vitronectin adhesion of HD cells was not affected by β1 integrin blocking agents. Instead, vitronectin adhesion of HD cells was inhibited by RGD peptides and by an αVβ5 integrin blocking antibody (Fig. 2A,B), indicating that HD cells attach to vitronectin via αVβ5 integrin when uPAR–α3β1-integrin complexes are unable to form. Similar results were obtained in WT and HD H1299 cells.
To explore whether uPAR is involved in αVβ5-integrin-mediated HD cell adhesion to vitronectin, we took advantage of the fact that uPAR–αVβ5-integrin interaction is strongly promoted by uPA binding to uPAR (Carriero et al., 1999). Indeed, vitronectin adhesion of HEK293 HD cells was increased after incubation with pro-uPA because these cells do not express endogenous uPA (Fig. 2B). By contrast, H1299 cells express high level of endogenous uPA and vitronectin adhesion of H1299 HD mutant cells was almost completely blocked by a uPAR monoclonal antibody ATN 617 which prevents uPA binding, suggesting that uPA-uPAR also supports HD cell adhesion to vitronectin. It is worth noting that both HEK293 cells and H1299 cells with uPAR knockdown weakly attached to vitronectin in a αVβ5-integrin-dependent and uPAR-independent manner (Fig. 2B). Together, these results indicate that vitronectin adhesion in HEK293 and H1299 cells operate through distinct mechanisms depending on whether uPAR is expressed and whether uPAR is able to interact with α3β1 integrin or αVβ5 integrin.
Reduction of uPAR–β1-integrin association and lamellipodia formation in cells expressing H249A-D262A mutant uPAR
The uPAR–α3β1-integrin and uPAR–α5β1-integrin complexes were assessed by co-immunoprecipitation using antibodies to α3 integrin and α5 integrin. Consistent with results we obtained by using HT1080 cells (Wei et al., 2007), the single D and H mutation specifically disrupted uPAR–α3β1-integrin and uPAR–α5β1-integrin association, respectively, whereas the HD mutant was unable to associate with either α3β1 integrin or α5β1 integrin (Fig. 3A). uPAR–β1-integrin association was also assessed by co-immunoprecipitation using antibody to β1 integrins. HD uPAR mutations almost completely inhibited β1 integrin association with uPAR, whereas the H249A or D262A single mutation only had partial inhibition (supplementary material Fig. S2), suggesting that uPAR association with α3β1 integrin and α5β1 integrin constitute the vast majority of uPAR–β1-integrin complexes in these cells. To determine whether HD mutations on uPAR disrupt uPAR–α3β1-integrin complexes on the cell surface, receptor co-clustering experiments were performed (Fig. 3B). Clustering of surface α3 integrin in WT uPAR-expressing cells showed uPAR colocalization by confocal microscopy. Colocalization of uPAR with α3 integrin was significantly reduced in HD-uPAR-expressing cells (Fig. 3C) (P<0.005). Interestingly, clustering of α6 integrin failed to co-cluster cell surface uPAR.
We previously found that uPAR-knockdown H1299 cells showed excessive stress fibers and focal adhesions, whereas control and WT-uPAR-expressing cells, which express high levels of uPAR, showed elevated Rac1 activation (Wei et al., 2007). Indeed, we observed that uPAR knockdown suppressed lamellipodia formation in both H1299 and H1264 cells (not shown). Remarkably, cells expressing HD mutant uPAR plated in serum (Fig. 3D) or on fibronectin (not shown) showed distinctly reduced lamellipodia formation and were less spread compared with WT cells. A similar pattern of cell morphology was observed in WT and HD H1264 cells (not shown). In addition, disruption of uPAR–β1-integrin complexes by β1P1 peptide in H1299 WT cells also inhibited the formation of cell protrusions (Fig. 3E). Together, these results suggest a role for uPAR–β1-integrin interaction in determining cell morphology and cell motility.
Abrogation of matrix-mediated integrin signaling in cancer cells expressing H249A-D262A mutant uPAR
Previous studies showed that either suppression of uPAR or H249A mutation on uPAR in H1299 cells blocks fibronectin-induced, uPAR–α5β1-integrin-dependent ERK activation and MMP9 production (Wei et al., 2007). As shown in Fig. 4A, engaging α3β1 integrin, another major β1 integrin that associates with uPAR, with laminins also induced ERK activation and MMP9 secretion in H1299 cells expressing WT uPAR. Similar results were obtained with H1264 cells (data not shown). Whereas WT cells showed ERK activation and high levels of MMP9 induced by fibronectin or laminin-5, the induction of ERK-P and MMP9 in HD cells, unable to associate with either α3β1 integrin or α5β1 integrin, was absent on all the matrices tested, suggesting that uPAR–β1-integrin complexes are crucial to matrix-induced signaling. In addition, ERK activation and MMP9 induction were not detected in H249A cells on fibronectin or D262A cells on laminin-5, illustrating the requirement of uPAR–α5β1-integrin for cell signaling on fibronectin and uPAR–α3β1-integrin for cell signaling on laminin-5. The pattern of MMP9 induction and ERK phosphorylation on Matrigel is similar to that on laminin-5 (not shown). Collectively, these data demonstrate that uPAR–β1-integrin interactions are required for elevated expression of MMP9 induced by matrices through ERK pathway.
To explore whether other genes are regulated by uPAR–β1-integrin–matrix and whether they contribute to lung cancer cell invasiveness and motility, four sets of mRNA from WT and HD mutant cells cultured on fibronectin for 24 hours were transcriptionally profiled on Agilent microarrays. Using a log-odds ratio of B>0 as a criterion for statistical significance and a differential expression of greater than twofold, 140 genes were found to be upregulated and 24 genes downregulated, in WT cells on fibronectin relative to HD cells (supplementary material Table S1). A comparison of genes upregulated in WT cells but not HD cells against the full list of transcripts represented on the array with the EASE gene ontology algorithm (http://david.niaid.nih.gov/david/ease.htm) suggests that genes involved in the processes of positive regulation of proliferation, negative regulation of apoptosis and angiogenesis were over-represented (EASE score <0.05). mRNA involved in hydrolase activity, including metalloendopeptidase, collagenase and plasminogen-activator activities, were also over-represented (supplementary material Table S2). With a log-odds ratio of B>0 and differential expression of fourfold or greater, 13 genes were revealed to be induced by fibronectin in WT cells but not HD cells (Table 1). The MMP9 transcription level in WT cells was over eight times greater than in HD cells, consistent with the gelatin zymography results (Fig. 4A). Strikingly, MMP1 mRNA was upregulated almost 30-fold in WT cells grown on fibronectin when compared with HD cells. These results were confirmed using RT-PCR (data not shown).
MMP1 protein was similarly upregulated in WT H1299 cells plated on fibronectin or laminin-5, whereas HD cells failed to induce MMP1 on any matrix tested (Fig. 4B). MMP1 induction by WT cells grown on fibronectin was inhibited by the MEK inhibitor PD98059, suggesting that this acts through an ERK-dependent signaling pathway. MMP1 expression of individual H and D mutants on poly-L-lysine, fibronectin or laminin-5 closely paralleled those of MMP9 and ERK-P in H1299 cells (not shown), suggesting that expression of MMP1 and MMP9 is similarly activated by uPAR–β1-integrin complexes upon matrix engagement. Interestingly, pretreatment with the uPAR–α3β1-integrin blocking peptide α325 only inhibited laminin-5-induced secretion of MMP9 and MMP1, whereas the uPAR–β1-integrin blocking peptide β1P1 inhibited both fibronectin- and laminin-5-induced MMPs (Fig. 4C), confirming that matrix-induced signaling and MMP expression require uPAR interactions with specific integrins.
Altogether, these findings indicate that the effects of uPAR on growth, survival and cell motility related pathways in lung cancer cells are mediated by matrix engagement through uPAR–β1 integrin association.
uPAR–β1-integrin–matrix and uPA-uPAR signaling are required for lung cancer cell invasion
Both the uPA system and β1 integrins are well known to be important for lung cancer invasion and metastasis (Liu et al., 1995; Rao et al., 2005; Takenaka et al., 2000). Therefore, the function of uPAR–β1-integrin on the invasive capacity of lung cancer cells was assessed in a Matrigel invasion assay. As shown in Fig. 5A, control and WT cells were able to invade through Matrigel whereas uPAR knockdown and HD cells were not, suggesting that uPAR–β1-integrin complexes are required for lung cancer cell invasion. Interestingly, H and D single mutants only showed partial inhibition, suggesting that both uPAR–α5β1-integrin and uPAR–α3β1-integrin contribute to cell invasion. In WT cells, disruption of the uPAR–β1-integrin interaction by β1P1 significantly inhibited lung cancer cell invasion (Fig. 5B), confirming that uPAR–β1-integrin association is essential to this event. Matrigel invasion was significantly reduced in cancer cells treated with MEK1 inihibitor (PD98059) and broad-spectrum MMP inhibitor (GM6001) (Fig. 5B), suggesting that invasion is dependent on the induction of these pathways through uPAR–β1-integrin complexes. Moreover, lung cancer cell invasion was inhibited by treatment with a uPA antibody neutralizing uPA activity (394) or an uPAR antibody blocking uPA binding to uPAR (ATN617), suggesting that uPA activity and uPA binding to its receptor are important for lung cancer cell invasion. By contrast, AG1478, an inhibitor of EGFR, did not alter cancer cell invasion, suggesting that this type of invasion was not mediated by EGFR signaling. Altogether, these results indicate that both uPA-uPAR and uPAR–β1-integrin association and their functional and/or signaling events are essential for lung cancer cell invasion.
Association of uPAR and β1-integrin is critical to uPA-uPAR signaling. uPA has been suggested to be prognostic marker in non-small cell lung cancer (NSCLC) (Offersen et al., 2007). Both H1299 and H1264 cells express high levels of uPA, because uPA binding to these cells was subtle unless cells were acid-washed before the binding assay to remove the endogenous membrane-bound uPA (Fig. 1B) (Wei et al., 2007). The robust defect of cell invasion resulting from lack of association of uPAR and β1 integrin in HD cells (Fig. 5) raised the question whether uPA signaling in these cells was also impaired. Binding of uPA to uPAR has been shown to stimulate multiple signaling pathways including ERK (Nguyen et al., 1998). We found that ERK phosphorylation in acid-washed H1299 and H1264 cells after pro-uPA treatment peaks at 5 minutes and returns to baseline by 30 minutes (Fig. 6A). To test whether uPA signaling requires β1 integrin association with uPAR, ERK activation in WT or HD H1299 cells 5 minutes after incubation with pro-uPA was assessed and compared. Interestingly, HD mutant H1299 cells failed to increase ERK phosphorylation in response to uPA (Fig. 6B). Similar observations were made in HEK293 cells expressing HD mutant uPAR (supplementary material Fig. S3). uPA binding to uPAR has also been shown to activate Jak-Stat (Dumler et al., 1998; Jo et al., 2005; Shetty et al., 2006) and PI3K–Akt–GSK-3β signaling pathways (Galaria et al., 2005). We then inspected these pathways 30 minutes after uPA treatment and found that uPA-mediated activation of Akt, GSK-3β and Stat1 were all impaired in HD cells (Fig. 6B). These data suggest that uPAR–β1-integrin association is critically involved in multiple uPA-induced signaling pathways. Although Stat3 is expressed in H1299 cells, it was not induced by uPA in either WT or HD cells. Stat2, Stat5 or Stat6 were not detected in these cells by western blotting (not shown). Similarly, treatment with β1P1 peptide, but not the scrambled control, inhibited uPA activation of ERK, Akt, GSK-3β and Stat1 in WT cells (Fig. 6C), further confirming that uPAR–β1-integrin interaction is crucial to uPA-uPAR signaling.
EGFR is overexpressed in NSCLC (Erman et al., 2005) and EGF stimulates the activation of ERK pathway (Adachi et al., 2002). Although EGFR inhibitor AG1478 did not affect cell invasion (Fig. 5B), we nevertheless tested ERK activation stimulated by EGF in the WT and HD mutant H1299 cells. Our data indicate that H, D or HD mutations on uPAR do not affect EGF-induced ERK phosphorylation (Fig. 6D), suggesting that uPAR-integrin interaction is not required for EGF mediated ERK activation.
Both uPAR–β1-integrin–matrix and uPA-uPAR signaling may contribute to lung cancer progression in vivo. Because uPAR–β1-integrin association is crucial to both matrix- and uPA-induced signaling and is involved in lung cancer cell invasion (Figs 4, 5, 6), we further investigated the function of uPAR–β1-integrin in lung cancer progression in vivo using a well-established orthotopic lung cancer model (Chen et al., 2005). H1299 cells were transfected with short-hairpin RNA for uPAR or empty vector and the stable uPAR knockdown cells were reconstituted with wt (WT) or H249A-D262A (HD) mutant uPAR. GFP-tagged control, uPAR knockdown, WT or HD H1299 cells were directly injected into the left lungs of 8-week-old female athymic nude mice (Onn et al., 2003). Five weeks later, the lungs were harvested and the GFP-tagged tumor nodules on serial tissue sections were visualized by fluorescence microscopy (Fig. 7B) and tumor area was measured. Consistent with other models of cancer in which uPAR expression is inhibited, uPAR knockdown H1299 cells developed smaller tumors than control H1299 cells (Fig. 7A). Tumor size was significantly reduced in the HD group when compared with the WT group (P<0.044), suggesting an in vivo role for uPAR–β1-integrin interaction. In addition, tumor incidence was diminished in HD-injected mice (P<0.034; data not shown), which suggests that expression of WT uPAR confers some survival advantage to tumor cells in vivo and is consistent with gene ontology analysis of the array data that indicates over-representation of pro-proliferative and/or anti-apoptotic pathways. Therefore, our data strongly suggest that uPAR–β1-integrin complexes modulate lung tumor progression in vivo.
In this study, we created a uPAR point mutant (HD mutant) that is specifically unable to bind either α5β1 integrin or α3β1 integrin and used this mutant to reveal several important findings: (1) The defects in adhesion, matrix-induced signaling, MMP expression and in vitro migration previously reported with β1-integrin-derived peptides that block uPAR-integrin interactions were recapitulated in the uPAR mutant with blocked β1 integrin interaction. This finding indicates that the previously reported peptides are very likely acting in a specific manner and faithfully reporting the effects of disruption of this physical interaction on uPAR and integrin functions. (2) Cells expressing uPAR can use uPAR to support more than one mechanism for attaching to vitronectin. Although the β1 integrin pathway is dominant in our cell systems, clearly αVβ5 integrin interacts with uPAR to promote vitronectin adhesion, as previously reported. Cells appear to acquire this capacity if the β1-integrin pathway is unavailable. (3) uPA signaling appears to require uPAR interaction with one or more β1 integrins. This conclusion is consistent with several previous reports from other groups using peptides that interrupt uPAR–β1-integrin complexes to reveal impaired uPA-induced ERK phosphorylation in tumor cells (Aguirre Ghiso et al., 1999; Ahmed et al., 2003; Mazzieri et al., 2006). We further extend this finding to other pathways, such as Akt and Stat1, where the requirement for uPAR-integrin interactions has not been reported. (4) In spite of the retained capacity of β1-integrin-interaction-blocked uPAR to interact with uPA, αV integrins, caveolin and EGFR, these mutant uPAR cells have impaired capacity to form lung tumors in vivo. Taken together, our results demonstrate that uPAR–β1-integrin interactions are required for the signaling pathways triggered by both uPAR and integrin-ligand engagement that are believed to contribute to lung cancer cell invasion in vitro and progression in vivo.
uPAR has been reported to bind vitronectin directly (Wei et al., 1994) and yet interaction of uPAR with α3β1 integrin has been found to regulate cell adhesion and signaling on vitronectin (Degryse et al., 2005; Wei et al., 2001). Because α3β1 integrin has not been reported to bind vitronectin directly, it is likely that the complex binds vitronectin through uPAR. A recent report suggests that integrin association is not initially required for uPAR-dependent vitronectin adhesion because HEK293 or CHO cells expressing a uPAR mutant with all five known uPAR binding sites (including H249 and D262) substituted with alanine did not affect cell attachment to vitronectin (Madsen et al., 2007). Consistent with this, we found that both HEK293 and H1299 cells expressing HD mutant uPAR show comparable adhesion to vitronectin (2 μg/ml) compared with WT cells (Fig. 2A,B). Notably, vitronectin adhesion of HD cells was resistant to interference by an α3β1 integrin blocking antibody or integrin-blocking peptides α325 and β1P1, implying, as argued by Madsen and co-workers, that this type of vitronectin adhesion does not require uPAR association with α3β1 integrin (Madsen et al., 2007). However, we found that vitronectin adhesion of HD cells is sensitive to uPA binding to uPAR, RGD peptide and αVβ5 integrin blocking antibody (Fig. 2B), suggesting that the adhesion is probably mediated by uPA–uPAR–αVβ5-integrin via a vitronectin binding site on αVβ5 integrin. By contrast, vitronectin adhesion of WT uPAR-bearing cells was significantly inhibited by these α3-blocking reagents, as previously reported (Degryse et al., 2005; Wei et al., 2001), but not by αVβ5 integrin antagonists. The capacity of uPAR to engage β1 integrins probably empowers additional signaling that promotes motility based on the following observations: (1) HD lung cancer cells show a less invasive phenotype in culture; (2) uPA-uPAR-mediated signaling in both HEK293 cells (supplementary material Fig. S3) and lung cancer cells (Fig. 6B,C) was abrogated by HD mutant uPAR expression. These findings parallel previous observations regarding the influence of uPAR on cellular responses to fibronectin adhesion. Specifically, the presence or absence of uPAR on tumor cells did not alter overall fibronectin adhesion but uPAR–α5β1-integrin interactions did alter the mechanism of fibronectin attachment and the signaling that ensues (Wei et al., 2005; Wei et al., 2007). Together our findings suggest that a switch of cellular adhesive mechanism may be a common phenomenon in tumor cells with upregulated uPAR and β1 integrin expression. Accordingly, it seems that direct vitronectin binding to uPAR and uPAR-dependent signaling could occur by distinct mechanisms depending on which integrin binding partner is engaged with uPAR. As uPAR and uPAR-binding integrins are extensively and differentially expressed in many malignant tumors, it is likely that matrix association of these tumor cells is largely mediated by specific uPAR-integrin complexes.
To identify additional signaling pathways regulated by fibronectin engagement that is dependent on uPAR–β1-integrin, we compared gene expression profiles of WT and HD-expressing cells on fibronectin. The results confirmed our previous observation that MMP9 induction requires uPAR–α5β1-integrin association. In addition, MMP1 was found to be the highest differentially expressed gene. This finding was confirmed by RT-PCR (not shown) and western blotting (Fig. 4B). Both MMP9 and MMP1 have been shown to be highly expressed in NSCLC and have been frequently implicated in tumor invasiveness (Leinonen et al., 2006; Lin et al., 2004). Furthermore, gene ontology analyses of transcription profiles using the EASE algorithm suggest that activation by fibronectin attachment of multiple genes involved in ECM degradation including several MMPs is dependent on uPAR–α5β1-integrin interaction (supplementary material Table S2). In addition, pro-proliferative, anti-apoptotic and angiogenesis-related genes were also upregulated in an uPAR–α5β1-integrin-dependent manner. These results suggest that uPAR association with β1 integrins is intimately involved with tumor cell survival and invasion.
Plasminogen activators/plasmin and matrix metalloproteinases are two major families of enzymes thought to participate in the pericellular proteolysis associated with the invasive program of tumor cells (Dano et al., 1999; Murphy and Gavrilovic, 1999). In the lung cancer cell lines studied here, MMP induction and uPA signalling are both dependent on uPAR–β1-integrin interactions (Figs 4 and 6), and were confirmed to be involved in invasion (Fig. 5B). Disruption of both uPAR–α3β1-integrin and uPAR–α5β1-integrin complexes with peptide β1P1 significantly hampered Matrigel invasion of WT cancer cells. Furthermore, cells expressing individual H or D mutant uPAR only partially inhibited invasion (Fig. 5A), suggesting that the drastic inhibitory effect of HD mutations resulted from blocking uPAR association with multiple β1 integrins. Together, these observations indicate that Matrigel invasion of H1299 cells requires both uPAR–α3β1-integrin and uPAR–α5β1-integrin associations.
Suppression of uPAR expression is reported to block tumor progression in a number of xenograft tumor models (Lakka et al., 2001; Pulukuri et al., 2005), and we confirmed this observation with uPAR-knockdown H1299 cells in the orthotopic lung cancer model used here (Fig. 7A). One limitation to this result is that there might be off-target shRNA effects in vivo, although the uPAR-targeting sequence we used has been validated by other groups (Vial et al., 2003; Wu et al., 2007) in vitro. Another limitation to this study is that occasional large tumors arising in mice injected with uPAR-knockdown H1299 cells expressed a substantial amount of uPAR (not shown), implying escape from knockdown. Some of the in vivo tumors from HD H1299 cells could have also derived from cells expressing endogenous uPAR. This is probably because cells injected in vivo are no longer under selection pressure for the retention of uPAR knockdown. Unfortunately, we are not able to confirm this because endogenous and re-expressed uPAR are indistinguishable in these cells. Studies reported here also demonstrate that the HD mutations on uPAR specifically target its β1 integrin binding capacity, and not other aspects of uPAR function. For example, uPAR–αV-integrin and uPAR-EGFR complexes have been implicated in chemotaxis and tumor cell motility (Czekay et al., 2003; Gargiulo et al., 2005). We found that HD mutations did not affect uPAR association with αV integrins or EGFR (Fig. 1C) and our studies reveal enhanced adhesion to vitronectin via uPAR–αV-integrin interactions (Fig. 2) and normal EGF-induced signaling (Fig. 6D) in HD cells. uPAR association with αV integrins and EGFR could have some compensatory role in tumor growth of the HD cells in vivo. Thus, within the context of the limitations of our in vivo model and the discrete defect we created in uPAR function, a small but statistically significant difference in tumor incidence and size between the WT and HD groups should be considered noteworthy. Consistent with matrix-specific induction of MMPs (Fig. 4), we also found that neither H nor D single mutation is sufficient to abolish in vitro invasion (Fig. 5A) or to inhibit tumor growth (Tang et al., observation), suggesting that multiple uPAR-integrin complexes are involved in tumor invasion in vivo. Taken together, the results reported here demonstrate the importance of uPAR–β1-integrin interactions in vivo and underscore additional aspects of uPAR function that can promote tumor progression. Further dissection of these functions in vivo could better direct therapeutic attempts to suppress tumor metastasis by targeting uPAR.
Materials and Methods
Cell culture and reagents
HEK293 cells and human lung cancer cell lines NCI-H1299 and NCI-H1264 were obtained from the American Type Culture Collection (Manassas, VA). The cell lines were cultured in DMEM or RPMI 1640 supplemented with 10% FBS and penicillin-streptomycin. Cells expressing control vector or shRNA for uPAR were supplemented with zeocin (50 μg/ml). All cells expressing wild-type or mutant uPARs were also supplemented with hygromycin B (100 μg/ml).
Fibronectin, EGF and 0.01% poly-L-lysine solution were purchased from Sigma-Aldrich (St. Louis, MO). 804G supernatant rich in laminin-5 was a gift from Jonathan C. Jones (Northwestern University Medical School, Chicago, IL). uPA 1-48-Fc fusion protein and pro-uPA were provided by Steven Rosenberg (Chiron, Emeryville, CA). Peptides RGD, RAD, α325, scα325, β1P1 and scβ1P1 were synthesized at Anaspec (San Jose, CA) and purified by HPLC. MMP inhibitor GM6001, MEK inhibitor PD98059 and EGFR inhibitor AG1478 were purchased from Calbiochem (San Diego, CA). Pre-cast 10% gelatin gels and Alexa Fluor secondary antibodies were purchased from Invitrogen (Carlsbad, CA). The site-directed mutagenesis kit was purchased from Stratagene (La Jolla, CA). uPAR mAb (clone 3B10) for FACS, uPA activity neutralizing mAb (394) and uPAR pAb (399R) were purchased from American Diagnostica (Stamford, CT). uPAR mAb for blotting (R2) was a kind gift from Michael Ploug (Finsen Lab, Copenhagen, Denmark). uPAR mAb ATN615 and ATN617 were kindly provided by Andrew Mazar (Attenuon, LLC, San Diego, CA). Anti-EGFR pAb, MMP1 pAb (AB19140), integrin α3 mAb (P1B5), integrin α5 mAb (P1D6), α5 pAb integrin αVβ3 mAb (LM609), integrin αVβ5 mAb (P1F6), integrin α6 mAb (GoH3), integrin β1 mAb (JB1A), integrin pAb αV and rabbit anti-β1 pAb were purchased from Chemicon (Temecula, CA). Anti-phospho- and total antibodies to ERK, Akt and GSK-3β were purchased from Cell Signaling (Danvers, MA). Anti-human Fc-HRP and anti-human Fc-APC were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA). Caveolin pAb, Stats Activation Sampler Kit, Matrigel and Transwells were purchased from BD Biosciences (San Jose, CA). α3 pAb was from Santa Cruz Biotechnology (Santa Cruz, CA).
Expression of H249A-D262A mutant uPAR in uPAR-knockdown cells
The H249A-D262A mutant uPAR (HD) was generated as described (Wei et al., 2007).
Cells with uPAR knockdown or with WT or mutant uPAR expression were stained with primary antibody to uPAR (3B10) or various integrins and secondary APC-conjugated anti-mouse IgG (Sigma-Aldrich) and analyzed on a flow cytometer (FACSCaliber®; BD Biosciences).
uPAR ligand-binding assay
All the procedures were done at 4°C. Cells were acid-washed (50 mM glycine-HCl, 100 mM NaCl, pH 3.0) for 3 minutes, neutralized (0.5 M HEPES, 0.1 M NaCl, pH 7.5) for 10 minutes and incubated with uPA 1-48-Fc fusion protein (0-50 nM) followed by goat anti-Fc-APC in DMEM-0.1% BSA for 1 hour. After washing, the bound 1-48-Fc was detected by FACS analysis.
The cell adhesion assay was performed as described previously (Wei et al., 2001). Cells were seeded onto vitronectin (2 μg/ml)-coated plates with or without pro-uPA, uPAR mAbs, integrin-blocking antibodies or peptides. Attached cells were fixed, Giemsa stained and quantified by measuring OD550nm.
Immunoprecipitation and western blotting
Cells were lysed in Triton lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl and 1% Triton X-100) supplemented with protease inhibitors and 1 mM PMSF. Clarified lysates were immunoprecipitated with antibody to integrin α3 (P1B5), integrin α5 (P1D6) or integrin β1 (JB1A) and the immunoprecipitates were blotted for uPAR (R2), integrin α3 (pAb), integrin α5 (pAb), or integrin β1 (pAb). Clarified lysates were also immunoprecipitated with uPAR mAb (R2) and the immunoprecipitates were blotted for uPAR (399R), caveolin (pAb), αV integrins (pAb) or EGFR (pAb).
Receptor co-clustering and confocal microscopy
Cells plated on fibronectin (10 μg/ml) were incubated with mouse mAb to integrin α3 (P1B5) or rat mAb to integrin α6 (GoH3) on ice for 30 minutes. The integrins were clustered with fluorophore-conjugated secondary antibody (Alexa Fluor 350 for α3 integrin or Alex Fluor 568 for integrin α6) at 37°C for 30 minutes or left nonclustered on ice. Cells were then fixed in 4% paraformaldehyde, blocked with 10% goat serum, stained with primary anti-uPAR antibody (Rhodamine-conjugated ATN615 or ATN615 with Alexa Fluor 350 as appropriate). Confocal microscopy was performed on a laser-scanning microscope (LSM 510; Carl Zeiss MicroImaging). Images were captured with a ×63 oil-immersion objective, analyzed with Zeiss LSM Image Browser software (Carl Zeiss MicroImaging) and prepared by Canvas X (ACD Systems). Statistical significance was determined using the Student's t-test (P<0.05).
Gelatin zymography and western blotting of MMPs
MMP9 was detected by gelatin zymography as described (Wei et al., 2007). In brief, serum-starved cells were seeded on fibronectin (5 μg/ml), poly-L-lysine (50 μg/ml), Matrigel, or laminin-5 for 24 hours (MMP9) or 48 hours (MMP1). In some cases, cells were treated with peptides. The conditioned medium was analyzed for MMP9 on gelatin gel and MMP1 by western blot. The volume of the conditioned medium loaded to the gel was normalized to the total protein from the cells left in the same well.
ERK activation was accessed as described (Wei et al., 2007). In brief, cells were seeded on fibronectin, laminin-5, Matrigel or poly-L-lysine for 20 minutes and lysed. Lysates were immunoblotted for ERK-P and total ERK. In some cases, serum-starved cells were treated with EGF (20 ng/ml) before lysis.
Micorarray data analysis
Four replicate total RNA samples were prepared from H1299 cells expressing WT and HD uPAR plated on fibronectin-coated surfaces for 24 hours at 37°C, amplified by using the Agilent low RNA input fluorescent linear amplification kits (Agilent) with direct incorporation of Cy3- and Cy5-labeled CTP. Samples were hybridized to Agilent Whole Human Genome 4×44K chips and analyzed with Feature Extraction v9.5 software (Agilent).
uPA signaling assay
Serum-starved cells were acid washed and neutralized. Pro-uPA (10 nM) was added and incubated at 37°C for 5 minutes or 30 minutes to access ERK activation or activation of Akt, GSK-3β and Stat1, respectively. After incubation, cells were lysed and the lysates were blotted for ERK-P and total ERK, Akt, GSK-3β and Stat1. In some experiments, cells were pre-incubated with β1P1 peptides before uPA stimulation.
Cells (105) were seeded into Biocoat Inserts containing Matrigel (BD Biosciences) pre-coated with fibronectin on the bottom and cultured overnight in serum-free condition with FBS added to the lower chamber. Cells invaded through the Matrigel and attached to the underside of the membrane were fixed with methanol, stained with Giemsa, extracted in 10% acetic acid and absorbance was measured at 595 nm. In some experiments, the invasion was accessed in the presence of uPA blocking antibody, uPAR mAb ATN617, MMP inhibitor GM6001, MEK inhibitor PD98059, EGFR inhibitor AG1478 or peptide β1P1.
Equal numbers (2×105) of GFP-tagged control, uPAR knockdown, WT or HD H1299 cells were resuspended in 20 μl PBS containing 10 μg Matrigel and orthotopically injected into the left lungs of 8-week-old female athymic nude mice (Jackson Laboratory) (Onn et al., 2003). The mice were kept in pathogen-free environments and checked daily for 4-5 weeks. Lungs were then harvested and embedded in OCT compound and rapidly frozen in liquid nitrogen. The GFP-tagged tumor nodules on multiple tissue sections were visualized by fluorescence microscopy using Spot Camera (Diagnostic Instruments) and the largest cross-sectional tumor area was measured using SimplePCI software (Compix). Statistical significance was determined by Mann-Whitney U-test (P<0.05) for tumor area and Fisher's exact test (P<0.05) for tumor incidence.
We thank J. C. Jones for 804G supernatant, M Ploug for R2 antibody, A. Mazar for ATN615 and ATN617 antibodies, and S. Rosenberg for N-terminal fragment of uPA 1-48-Fc fusion protein. We also acknowledge J. Alexander for his cloning assistance. We thank David Erle for helping us with the microarray analysis. This work was supported by National Institutes of Health grants HL44712 (HAC) and CA125564 (HAC).
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/121/22/3747/DC1
- Accepted August 14, 2008.
- © The Company of Biologists Limited 2008