Critical role of integrin α5β1 in urokinase (uPA) /urokinase receptor (uPAR, CD87) signaling

Urokinase-type plasminogen activator (uPA) induces cell adhesion and chemotactic movement. uPA signaling requires its binding to uPA receptor (uPAR/CD87), but how glycosylphosphatidylinositol-anchored uPAR mediates signaling is unclear. uPAR is a ligand for several integrins (e.g. α5β1) and supports cell-cell interaction by binding to integrins on apposing cells (in trans). We studied whether binding of uPAR to α5β1 in cis is involved in adhesion and migration of Chinese hamster ovary cells in response to immobilized uPA. This process was temperature-sensitive and required mitogen-activated protein kinase activation. Anti-uPAR antibody or depletion of uPAR blocked, whereas overexpression of uPAR enhanced, cell adhesion to uPA. Adhesion to uPA was also blocked by deletion of the growth factor domain (GFD) of uPA and by anti-GFD antibody, whereas neither the isolated uPA kringle nor serine protease domain supported adhesion directly. Interestingly, anti-α5 antibody, RGD peptide, and function-blocking mutations in α5β1 blocked adhesion to uPA. uPA-induced cell migration also required GFD, uPAR, and α5β1, but α5β1 alone did not support uPA-induced adhesion and migration. Thus, binding of uPA causes uPAR to act as a ligand for α5β1 to induce cell adhesion, intracellular signaling, and cell migration. We demonstrated that uPA induced RGD-dependent binding of uPAR to α5β1 in solution. These results suggest that uPA-induced adhesion and migration of Chinese hamster ovary cells occurs as a consequence of (a) uPA binding to uPAR through GFD, (b) the subsequent binding of a uPA·uPAR complex to α5β1 via uPAR, and (c) signal transduction through α5β1.

Urokinase-type plasminogen activator (uPA) 1 is a highly restricted serine protease that converts the zymogen plasmino-gen to the active plasmin. Plasmin, in turn, mediates pericellular proteolysis of extracellular matrix proteins in the path of cellular invasion (1,2). uPA has also been shown to induce adhesion and chemotactic movement of myeloid cells (3,4), to induce cell migration in human epithelial cells (5) and bovine endothelial cells (6), and to promote cell growth (7)(8)(9). These signaling functions of uPA do not require its proteolytic activity.
uPA is composed of three independently folded domain structures, growth factor domain (GFD) (residue 1-43), kringle domain (residue 50 -131), and serine protease domain (residue 159 -411). Enzymatic digestion of uPA by plasmin generates an amino-terminal fragment (ATF) that consists of the GFD and kringle domains and the low molecular weight fragment (LMW-uPA), possessing serine protease activity. uPA binds with high affinity through GFD (10) to a cell-surface receptor (uPAR/CD87) that has been identified in many cell types (1). uPAR is a glycosylphosphatidylinositol-anchored 35-55-kDa glycoprotein. It is generally accepted that uPA-mediated signaling requires prior binding to uPAR. However, the mechanism by which uPAR mediates signaling events is still to be fully elucidated. A major problem in understanding how uPA signals derives from the fact that uPAR has no transmembrane structure, leading to the proposal that hypothetical transmembrane adapters may be involved in this process (11).
Among the candidate transmembrane adapters are the integrins, a family of cell adhesion receptor heterodimers that interact with many extracellular matrix and cell-surface ligands (12). At least 18 ␣ and 8 ␤ subunits have been identified. Integrin-ligand interaction is involved in many biological and pathological situations, including cell anchorage and migration, cell-cell interaction during immune response, development, wound healing, vascular remodeling, and cancer metastasis and invasion (13). Integrins transduce signals from outside cells through their interaction with specific ligands. uPAR has been shown to associate with integrins by co-immunocoprecipitation, immunocolocalization, and resonance energy transfer approaches (14 -16). However, it has not been established whether the association of uPAR with integrins is responsible for uPA-mediated signaling. We have recently reported that recombinant soluble uPAR is a ligand for several ␤ 1 and ␤ 3 integrins (17), and we postulated that uPAR can transduce signals through the integrin signaling pathway upon binding to integrin in trans. However, it is still unclear whether uPAR binds to integrins as a ligand when both are present on the same membrane (in cis). It has been proposed that integrins "laterally associate" with uPAR (for review, see Ref. 2) and play a role in uPA⅐uPAR-initiated signaling events. However, the role of the integrin itself in the uPA⅐uPAR signaling is unclear, since in the current models other integrin FIG. 1. Cell adhesion to uPA in a uPAR-dependent manner. a and b, expression of uPAR on mock-transfected CHO (a) and uPAR-CHO (b) cells. Cells were stained with rabbit polyclonal anti-uPAR or control rabbit IgG followed by fluorescein isothiocyanate-labeled goat anti-rabbit IgG. Stained cells were analyzed by flow cytometry. c, adhesion of uPAR-and mock-transfected CHO cells to immobilized uPA. Wells in 96-well Immulon-2 microtiter plates were coated with 100 l of phosphate-buffered saline containing uPA at a concentration of 50 -1000 nM. Cells (10 5 cells/well) were added to the wells and incubated at 37°C for 1 h. Bound cells were quantified by measuring endogenous phosphatase activity. ligands (e.g. fibronectin) appear to be essential for uPA⅐uPAR signaling (8).
In this study we designed experiments to identify the role of integrin ␣ 5 ␤ 1 in uPA⅐uPAR signaling using recombinant uPA fragments and Chinese hamster ovary (CHO) cells overexpressing uPAR or mutant ␣ 5 ␤ 1 . We found that cells adhered to immobilized uPA in a signaling-dependent manner. Anti-uPAR antibody, depletion of uPAR, and deletion of the GFD of uPA effectively blocked cell adhesion to uPA, suggesting that binding of uPA to uPAR through GFD is critical for cell adhesion to uPA. Interestingly, anti-␣ 5 antibody, RGD peptide, and function-blocking ␣ 5 ␤ 1 mutations blocked cell adhesion to uPA, suggesting that ␣ 5 ␤ 1 is critical to this process as well. uPAinduced migration of CHO cells also required GFD of uPA, uPAR, and ␣ 5 ␤ 1 . We demonstrated that uPA induced RGD-dependent binding of uPAR to ␣ 5 ␤ 1 in solution. These results suggest that uPA-induced signaling in CHO cells involves a process in which (a) uPA binds to uPAR, (b) the uPA⅐uPAR complex binds to ␣ 5 ␤ 1 as a ligand in cis, and c) signal transduction is initiated through ␣ 5 ␤ 1 .
CHO cells were obtained from the American Type Culture Collection (Manassas, VA). CHO cells expressing the three domain forms of human uPAR (designated uPAR-CHO) have been described (17). The ␣ 5 -deficient CHO cells (B2 variant) expressing human ␣ 5 (wild type or mutant) have been described (21).
Adhesion Assays-Adhesion assays were performed as previously described (24). Briefly, wells in 96-well Immulon-2 microtiter plates (Dynatech Laboratories, Chantilly, VA) were coated with 100 l of phosphate-buffered saline (10 mM phosphate buffer, 0.15 M NaCl, pH 7.4) containing substrates at a concentration of 50 -1000 nM and were incubated 1 h at 37°C. Remaining protein binding sites were blocked by incubating with 0.2% BSA (Calbiochem) for 1 h at room temperature. Cells (10 5 cells/well) in 100 l of Hepes-Tyrode buffer (10 mM HEPES, 150 mM NaCl, 12 mM NaHCO 3 , 0.4 mM NaH 2 PO 4 , 2.5 mM KCl, 0.1% glucose, 0.02% BSA) supplemented with 2 mM MgCl 2 were added to the wells and incubated at 37°C for 1 h unless stated otherwise. After non-bound cells were removed by rinsing the wells with the same buffer, bound cells were quantified by measuring endogenous phosphatase activity (25). Antibodies were used at a 250-fold dilution for ascites (KH72 and 135-13C) and at 10 g/ml for purified antibodies or IgG. Data are shown as means Ϯ S.D. of three independent experiments. We confirmed that equivalent amounts of the fragments and mutants of uPA were coated on the plate by enzyme-linked immunosorbent assay (data not shown).
Mitogen-activated Protein Kinases (MAPK) Activation Assay-uPAR-CHO cells were plated into 6-ell tissue culture plates at 2 ϫ 10 6 cells/ml in Dulbecco's modified Eagle's medium supplemented with 0.5% fetal calf serum, 1ϫ penicillin-streptomycin-glutamine solution and incubated for 2 days at 37°C in a 95% air, 5% CO 2 humidified atmosphere. The cell culture media was removed, and the cells were washed once with prewarmed serum-free Dulbecco's modified Eagle's medium. The cells were incubated for 3 h with serum-free Dulbecco's modified Eagle's medium with or without the MEK inhibitor PD98059 (500 M) at 37°C in a 95% air, 5% CO 2 humidified atmosphere. The cells were stimulated with different concentrations of soluble scuPA for 5 min at 37°C. The reaction was terminated by removing the stimulation media and washing the cells with 1 ml of ice-cold phosphate-buffered saline containing 1 mM Na 3 PO 4 followed by incubation with 100 l of ice-cold radioimmune precipitation assay buffer for 20 min on ice. The whole cell lysate was collected, and the nuclear material was pelleted by centrifugation at 14,000 ϫ g for 10 min. The supernatant from each treatment was retained and stored at Ϫ20°C until required. Whole cell lysates (40 g of protein) were fractionated using 4 -20% SDS-polyacrylamide gel electrophoresis and electrophoretically transferred to nitrocellulose for Western blotting. The membranes were blocked for 1 h at room temperature with blocking buffer (10 mM Tris-HCl, 150 mM NaCl, pH 7.5, supplemented with 0.1% Tween 20 and 5% BLOTTO; Biorad, Hercules, CA). To determine the phosphorylation changes in MAPK the membranes were washed 3 times for 5 min with wash buffer (10 mM Tris-HCl, 150 mM NaCl, pH 7.5, supplemented with 0.1% Tween 20) and incubated overnight at 4°C with a 1:1000 dilution of rabbit anti-phospho-p44/42 MAPK (Thr-202/Tyr-204) antibody (Cell Signaling Technology, Beverly, MA) in blocking buffer. The blots were washed 3 times for 5 min with wash buffer and probed with a 1:2000 dilution of anti-rabbit horseradish peroxide-conjugated secondary antibody (Cell Signaling Technology) in blocking buffer at room temperature for 1 h. The blots were washed 3 times for 5 min and developed using the Immun-Star horseradish peroxide chemiluminescence substrate kit (Bio-Rad). The blots were stripped by incubating with stripping buffer (0.1 M glycine, pH 2.6, and 2% SDS) for 30 min at 50°C. The blots were washed 3 times with 10 mM Tris-HCl, 150 mM NaCl, pH 7.5, re-blocked for 1 h at room temperature with blocking buffer, washed 3 times for 5 min, incubated with a 1:1000 dilution of rabbit anti-p44/42 MAPK (Cell Signaling Technology) overnight at 4°C in blocking buffer, and processed as above to determine the total MAPK levels of each lane.
Migration Assays-Cell migration was analyzed using tissue culturetreated 24-well Transwell plates (Costar, Cambridge, MA) with polyd, effect of anti-uPAR antibody on adhesion to uPA. The coating uPA concentration was 500 nM. Cells were incubated with immobilized uPA in the presence of control (cont) rabbit IgG or rabbit anti-uPAR polyclonal antibody. The data suggest that cell adhesion to uPA required uPAR. e, depletion of uPAR by the phosphatidylinositol-specific phospholipase C (PIPLC) treatment. uPAR-CHO cells were treated with phosphatidylinositol-specific phospholipase C (1 unit/10 7 cells in 1 ml of medium) for 1 h at 37°C and were stained with anti-human uPAR mAb 3B10 or control mouse IgG followed by fluorescein isothiocyanate-labeled anti-mouse IgG. Cells were analyzed by flow cytometry. The results suggest that more than 90% of uPAR was removed by the phosphatidylinositol-specific phospholipase C treatment. f, effect of the phosphatidylinositol-specific phospholipase C treatment on cell adhesion to uPA. The coating concentration of uPA was 200 nM. phosphatidylinositol-specific phospholipase C-treated uPAR-CHO cells were incubated with immobilized uPA. The results suggest that the phosphatidylinositol-specific phospholipase C treatment markedly reduced cell adhesion to uPA. uPAR/Integrin ␣ 5 ␤ 1 Interaction and uPA Signaling carbonate membranes of pore size 8 m. The lower side of the filter was coated with various concentrations (20 -200 nM) of substrates. Coated filters were placed into a serum-free migration buffer (Dulbecco's modified Eagle's medium supplemented with 10 mM Hepes, 0.5% bovine serum albumin, and 1ϫ penicillin-streptomycin), and cells (100 l) suspended in the same buffer (8 ϫ 10 5 cells/ml) were added to the upper chamber. The cells were incubated at 37°C in 5% CO 2 for 20 h. Cells in the upper chamber were removed by wiping, and those that migrated to the lower surface of the filters were fixed and stained with 0.5% crystal violet in 20% ethanol and counted. The result in each well is the mean cell number of 4 randomly selected high magnification microscopic fields from triplicate experiments. In some experiments, anti-integrin antibodies (10 g/ml) were incubated with cells for 15 min before to the assay.
Co-precipitation of uPAR and Integrin ␣ 5 ␤ 1 1-Soluble uPAR was radioiodinated with Na-125 I using Iodo beads (specific activity 14,500 cpm/ng). Purified human ␣ 5 ␤ 1 integrin was obtained from Chemicon International. Purified ␣ 5 ␤ 1 (6 g/ml), mAb HA5 (9 g/ml), 125 I-labeled soluble uPAR (6 g/ml), and uPA (12 g/ml) were incubated with protein G-agarose beads either in the presence or absence of RGD peptide (150 g/ml) in serum-free RPMI 1640 medium supplemented with 10 mM HEPES, pH 7.4, 0.02% bovine serum albumin at 4°C for 4 h. As a control, experiments were performed in the absence of uPA. Beads were washed 3 times in RPMI supplemented with 10 mM HEPES, 0.02% bovine serum albumin. Bound materials were extracted into reducing SDS-PAGE sample buffer and analyzed by SDS-PAGE and autoradiography.
Other Methods-Flow cytometric analysis and stress-fiber staining were performed as described before (17,26).

RESULTS
uPAR-dependent Cell Adhesion to uPA-It has been reported that CHO cells express endogenous hamster uPAR (17). We detected low level endogenous hamster uPAR on the mocktransfected CHO cells with polyclonal anti-human uPAR antibodies (Fig 1a). Consistent with this finding, mock-transfected CHO cells adhered to immobilized uPA in a dose-dependent manner (Fig. 1c). We transfected CHO cells with cDNA encoding human uPAR and cloned stable cell lines expressing high levels of receptor (designated uPAR-CHO). uPAR-CHO cells express uPAR at a much higher level (Fig. 1b) and showed greater adhesion to uPA than mock-CHO cells (Fig. 1c). Anti-uPAR polyclonal rabbit IgG completely blocked adhesion of uPAR-and mock-transfected CHO cells to uPA (Fig. 1d). To further test whether uPAR is involved in this process, we treated uPAR-CHO cells with phosphatidylinositol-specific phospholipase C to remove the glycosylphosphatidylinositolanchored uPAR. The phosphatidylinositol-specific phospholipase C treatment removed more than 90% of uPAR on the cell surface, as determined by flow cytometry with anti-uPAR ( uPAR/Integrin ␣ 5 ␤ 1 Interaction and uPA Signaling 1e), and markedly reduced the adhesion of uPAR-CHO cells to uPA (Fig. 1f). These results indicate that adhesion of these CHO cells to immobilized uPA is uPAR-dependent and that the level of adhesion to uPA correlates with the amount of uPAR.
We also found that adhesion of uPAR-CHO cells to immobilized uPA occurred at 37°C, but not at 4°C, suggesting that this process is temperature-dependent and requires signal transduction (Fig. 2a). It has been reported that uPA induces activation of MAPK (27). An inhibitor to MEK-1 (PD98059) that prevents the activation of MAPK (extracellular signalregulated kinases1/2) (Fig. 2b) consistently blocked adhesion of uPAR-CHO cells to uPA in a dose-dependent manner (Fig. 2c). These results suggest that this process requires extracellular signal-regulated kinase 1/2 activation (most likely induced by immobilized uPA). We found that immobilized uPA did not induce spreading or stress-fiber formation in uPAR-CHO cells, in contrast to fibronectin used as a positive control (Fig. 2d).
Role of Integrins in uPAR-mediated Adhesion to uPA-It has been proposed that integrins may be critically involved in the uPA⅐uPAR signaling (2). To examine this hypothesis we first determined whether integrins contribute to cell adhesion to uPA. To do so, we tested the effects of anti-integrin mAbs on adhesion of uPAR-CHO cells to uPA. CHO cells have endogenous hamster integrins ␣ 5 ␤ 1 , ␣ v ␤ 1 , and ␣ v ␤ 5 (data not shown). We found that RGD peptide (100 M) blocked adhesion of uPAR-CHO cells to uPA, but control RGE peptide did not (Fig.  3a). Consistent with the findings that RGD-dependent integrin(s) is involved in this process, anti-␣ 5 mAb (KH72) completely blocked adhesion of uPAR-CHO to uPA, whereas anti-␣ v ␤ 5 mAb (P1F6) or control ascites did not ( Fig. 3a and data not  shown). These results suggest that adhesion of uPAR-CHO cells to immobilized uPA is ␣ 5 ␤ 1 -dependent. Then we tested whether the level of ␣ 5 ␤ 1 expression affects cell adhesion to uPA using the B2 variant of CHO cells, which expresses ϳ2% of ␣ 5 ␤ 1 compared with parental CHO cells (28). The B2 cells adhered to uPA at a level lower than CHO cells (Fig 3b). This adhesion was completely blocked by anti-uPAR polyclonal antibodies and anti-integrin ␣ 5 (data not shown). These results suggest that cell adhesion to uPA is dependent on the level of ␣ 5 ␤ 1 but that a small amount of ␣ 5 ␤ 1 on the B2 cells still supports the adhesion to uPA to some extent. Another possibility is that integrin ␣ v ␤ 1 may also be involved in this process, although we were not able to test this hypothesis since function-blocking anti-hamster ␣ v mAb is not currently available. To test the specific contribution of ␣ 5 ␤ 1 to uPA⅐uPAR-dependent cell adhesion, we used immobilized anti-uPAR mAb as a uPAR ligand. We found that uPAR-CHO cells adhered to anti-uPAR and that this adhesion was not inhibited by antiintegrin ␣ 5 mAb (Fig. 3c), suggesting that ␣ 5 ␤ 1 was not required for this process. Taken together, adhesion of CHO cells to uPA requires uPAR and ␣ 5 ␤ 1 , and ␣ 5 ␤ 1 is specifically involved in this process only when uPA is used as a ligand.
FIG. 3. Requirement of integrin ␣ 5 ␤ 1 for adhesion of CHO cells to uPA. a, effect of anti-integrin antibodies on cell adhesion to uPA. The coating concentration of uPA was 500 nM. Cells were incubated with immobilized uPA in the presence of control IgG, P1F6 (anti-integrin ␣ v ␤ 5 ), KH/72 (anti-integrin ␣ 5 ), RGD, or RGE peptide. P1F6 and KH/72 cross-react with endogenous hamster ␣ v ␤ 5 and ␣ 5 ␤ 1 , respectively. The results suggest that integrin ␣ 5 ␤ 1 (RGD-dependent) is critical for cell adhesion to uPA. cont., control. b, effect of ␣ 5 expression levels on cell adhesion to uPA. The B2 variant of CHO cells (28) express low level ␣ 5 ␤ 1 (about 2% of wild type). B2 cells were tested for their ability to adhere to immobilized uPA as a function of the coating concentration of uPA. The results suggest that B2 cells adhered to uPA at a significantly lower level than CHO cells. c, effect of anti-␣ 5 mAb on adhesion to anti-uPAR mAb. Anti-uPAR mAb (anti-D2D3, 20 g/ml) was used instead of uPA as substrate. uPAR-CHO cells were incubated with immobilized anti-uPAR mAb in the presence of KH/72 or control mouse IgG. The results suggest that integrin ␣ 5 ␤ 1 is not critical for cell adhesion to anti-uPAR mAb.

uPAR/Integrin ␣ 5 ␤ 1 Interaction and uPA Signaling
The Domains of uPA Required for Cell Adhesion to uPA-To identify which uPA domains are involved in uPAR/␣ 5 ␤ 1 -dependent cell adhesion to uPA, we used several uPA fragments including the ATF, the kringle domain, the LMW-uPA, ⌬GFD-uPA, which lacks GFD, and ⌬Kringle-uPA, which lacks the kringle domain (Fig 4a). We found that ATF and ⌬Kringle-uPA supported cell adhesion at levels comparable with that of wildtype uPA. Kringle, LMW-uPA, or ⌬GFD-uPA did not support the adhesion at all (Fig 4b). These results suggest that GFD is primarily involved in cell adhesion to uPA, but other domains are not.
As a second approach, we tested whether mAbs to different domains of uPA block cell adhesion to uPA. We found that anti-GFD (AD3471) and anti-kringle mAb (Ab963) blocked adhesion of uPAR-CHO cells to uPA, whereas anti-LMW (UNG-5) did not (Fig 4c). Ab963 has been observed to inhibit the binding of uPA to whole cells despite the fact that its epitope has been mapped to the kringle domain. These results are consistent with the results with uPA fragments (Fig. 4b) with the exception of the effect of the anti-kringle antibody. GFD is required for uPA to bind to uPAR (10). Taken together these studies suggest that uPAR/␣ 5 ␤ 1 -mediated cell adhesion to uPA is also critically dependent on the interaction with GFD.
Mutations in Integrin ␣ 5 Affect Cell Adhesion to uPA-We have recently reported that uPAR is a ligand for several integrins (17). Specifically, soluble uPAR supports integrin-mediated cell adhesion, and glycosylphosphatidylinositol-linked uPAR binds to integrins in apposing cells in trans and supports cell-cell interaction. We have reported that mutations in the ligand binding region of integrin ␣ 4 subunit blocked adhesion of ␣ 4 ␤ 1 -transfected CHO cells to immobilized soluble uPAR (17), suggesting that uPAR binds to ␣ 4 ␤ 1 as a ligand. These critical residues are located within the ligand binding site in integrins based on the crystal structure of integrin ␣ v ␤ 3 (29,30). We suspected that cell adhesion to uPA in the present study involves interaction between uPAR and ␣ 5 ␤ 1 on the same cell surface (in cis). We have reported that similar mutations (the Tyr-186 to Ala (Y186A), F187A, and W188A) in the ligand binding site of ␣ 5 blocked fibronectin binding to ␣ 5 ␤ 1 (21). We therefore studied whether uPAR binds to ␣ 5 ␤ 1 as a ligand during adhesion to uPA using these ␣ 5 mutants.
We then tested the effect of these ␣ 5 mutations on cell adhesion to uPA. In this experiment we used ATF instead of wildtype uPA to exclude the possible contribution of the serine FIG. 4. uPA domains that are involved in adhesion to uPA. a, uPA fragments used in this study. b, adhesion to uPA fragments. The coating concentration of uPA fragments is 500 nM. Adhesion of uPAR-CHO cells to immobilized uPA fragments was determined. The data suggest that only ATF supported significant adhesion. c, effect of anti-uPA mAbs on adhesion to uPA. The coating concentration of uPA is 500 nM. uPAR-CHO cells were incubated with uPA in the presence of mAbs to uPA. The data suggest that anti-kringle and anti-GFD effectively blocked the adhesion. uPAR/Integrin ␣ 5 ␤ 1 Interaction and uPA Signaling protease domain because we found that both B2 and ␣ 5 -B2 cells bind weakly to LMW-uPA when ␣ 5 ␤ 1 is activated with Mn 2ϩ . 2 We found that ␣ 5 -B2 cells adhered to ATF more than parental B2 cells, but that ␣ 5 /Y186A-B2 and ␣ 5 /W188A-B2 cells adhered at the level comparable with or lower than that of parental B2 cells (Fig 5b). The adhesion of parent B2 cells to ATF may be due to endogenous low levels of ␣ 5 ␤ 1 on B2 cells or due to endogenous ␣ v ␤ 1 , as noted above. These results suggest that Tyr-186 and Trp-188 of ␣ 5 are critical for cell adhesion to both soluble uPAR and ATF.
Taken together these results suggest that uPAR-mediated cell adhesion to uPA requires the intact ligand binding function of ␣ 5 . This is consistent with the observation that RGD peptide blocked adhesion of uPAR-CHO cells to uPA in the present study (Fig. 3a) and that the anti-␣ 5 mAb we used in the present study (KH/72) binds to the ligand binding site of ␣ 5 . 3 We propose that uPAR binds to ␣ 5 ␤ 1 in cis as a ligand upon cell adhesion to uPA.

uPA-induced Cell Migration Also Depends on Both uPAR
and Integrin ␣ 5 ␤ 1 -It has been reported that physiological concentrations of uPA stimulated a chemotactic response in human monocytic THP-1 through binding to uPAR (11). We tested whether uPAR and ␣ 5 ␤ 1 are involved in uPA-induced cell migration. We found that wild-type uPA was chemotactic for uPAR-CHO cells, whereas ⌬GFD-uPA or kringle was not (Fig 6a), indicating that GFD is required for this process. We found that anti-human uPAR polyclonal antibodies, anti-uPAR mAb (anti-D2D3), and anti-␣ 5 mAb (KH72) effectively blocked uPA-induced migration (Fig 6b). Neither control rabbit IgG nor anti-integrin ␣ 6 mAb blocked uPA-induced migration. Taken together these results suggest that GFD, uPAR, and integrin ␣ 5 ␤ 1 are critically involved in uPA-induced cell migration.
uPA-induced Binding of uPAR to Integrin ␣ 5 ␤ 1 -The results in the present study predict that binding of uPA to uPAR 2 T. Tarui and Y. Takada, unpublished data. 3 T. Tarui and Y. Takada, unpublished observation.  (21). The effect of the ␣ 5 mutations on adhesion to soluble uPAR (a) and on adhesion to ATF of uPA (b) was determined. The concentration of soluble uPAR (D2D3) and ATF are 5 g/ml and 500 nM, respectively. Divalent cations used were 2 mM Mg 2ϩ for soluble uPAR (a) and 0.1 mM Mn 2ϩ for ATF (b). Note that B2 cells have some background binding to ATF as expected from Fig. 2b. The results suggest that these mutations affect cell adhesion to soluble uPAR and ATF.
FIG. 6. uPA-induced migration of uPAR-CHO cells. a, migration of uPAR-CHO cells toward different uPA fragments was determined using modified Boyden chamber assays. The concentration of uPA fragments used for coating membranes in migration assays is 500 nM. The results suggest that GFD is critical for uPA-induced migration. b, effect of antibodies on uPA-induced migration was determined. Antibodies used are KH/72 (anti-␣ 5 ), 135-13C (anti-␣ 6 ), anti-uPAR (rabbit polyclonal), and anti-D2D3 (anti-uPAR mAb). The results suggest that integrin ␣ 5 ␤ 1 and uPAR are critical for uPA-induced migration.
through GFD induces uPAR binding to ␣ 5 ␤ 1 . We tested whether uPA directly induces binding of uPAR and ␣ 5 ␤ 1 in a cell-free system. We incubated labeled soluble uPAR with isolated ␣ 5 ␤ 1 in the presence and absence of uPA. We isolated ␣ 5 ␤ 1 -uPAR complex with anti-␣ 5 mAb HA5 (non-function blocking). We found that uPAR co-precipitated with ␣ 5 ␤ 1 in the presence of uPA but not in the absence of uPA (Fig. 7). Consistent with the results in the present study, RGD peptide reduced the co-precipitation of uPAR and ␣ 5 ␤ 1 . These results suggest that uPA markedly increases binding of uPAR to ␣ 5 ␤ 1 in solution and probably on the cell surface. DISCUSSION In the present study we demonstrate that immobilized uPA induces cell adhesion and migration in an uPAR-dependent manner. Interestingly, signaling from immobilized uPA is similar to that from soluble uPA. Cell adhesion to uPA required signaling because this process was sensitive to temperature and to a MEK-1 inhibitor. We have shown that uPA-induced cell adhesion and migration required uPA binding to uPAR through GFD of uPA, since 1) antibodies against uPAR and GFD effectively blocked this interaction and 2) depletion of uPAR reduced, and overexpression of uPAR enhanced, cell adhesion to uPA. Also, we have shown that the catalytic domain of uPA was not critical for cell adhesion to uPA.
A major finding of the present study is that cell adhesion and migration to uPA required the ligand binding function of ␣ 5 ␤ 1 . We have previously reported that uPAR binds (as a ligand) to integrins in trans and that trans interaction between uPAR and integrins supported cell adhesion and cell-cell interaction (17). In the present study we have shown that function-blocking anti-␣ 5 mAb blocked uPA-induced cell adhesion and migration and that RGD peptide and the ␣ 5 mutations that block ligand binding effectively blocked cell adhesion to uPA. The crystal structure of integrin ␣ v ␤ 3 has only a single RGD binding site between ␣ v and ␤ 3 subunits (30) but did not show the position or existence of allosteric RGD binding sites. The integrin mutations we used in the present study (Y186A, F187A, and W188A in ␣ 5 ) are very close to the RGD peptide in the ␣ v ␤ 3 ⅐RGD complex (30) (the distances between their backbones are within 10 Å), suggesting that these mutations directly block ligand binding to ␣ 5 ␤ 1 . Thus, it is highly likely that RGD peptide and the integrin mutations directly blocked uPAR binding to ␣ 5 ␤ 1 and thereby blocked cell adhesion to uPA. We do not, however, preclude the possibility that RGD peptide binds to allosteric RGD binding sites and affects uPAR binding to ␣ 5 ␤ 1 .
Because anti-uPAR antibodies effectively blocked adhesion of uPAR-CHO cells to uPA, ␣ 5 ␤ 1 alone cannot directly support adhesion to uPA. Consistent with this observation, no ␣ 5 ␤ 1 ligand other than uPAR was required for uPA⅐uPAR signaling in the present study. It is highly likely that uPAR is a primary ligand for ␣ 5 ␤ 1 upon uPA-induced cell adhesion and migration in the present study and that uPAR binds to ␣ 5 ␤ 1 as a ligand even when uPAR and ␣ 5 ␤ 1 interact in cis. Thus uPA-induced cell adhesion and migration involves the following sequence. 1) uPA binds to uPAR, 2) uPAR then binds to ␣ 5 ␤ 1 in cis, and 3) signal transduction is mediated through ␣ 5 ␤ 1 (Fig. 8). CHO and B2 cells express ␣ v ␤ 1 and ␣ v ␤ 5 , and we have shown that anti-␣ v ␤ 5 (P1B6) does not block adhesion to uPA. However, we do not rule out the possibility that ␣ v ␤ 1 is involved in uPAR binding in cis. We are not able to test whether ␣ v ␤ 1 is involved because anti-␣ v antibody that cross-reacts with hamster ␣ v is not available.
It has previously been reported that the avidity of ␣ 5 ␤ 1 in uPAR-rich human carcinoma Hep3 cells for fibronectin was higher than that of uPAR-poor dormant Hep3 cells (8). The levels of MAPK activation by fibronectin were much higher in uPAR-rich cells than uPAR-poor cells. It is unclear, however, how fibronectin, uPAR, and ␣ 5 ␤ 1 are involved because both Complexes were immunoprecipitated with protein G-agarose beads and mAb against ␣ 5 ␤ 1 . Immunoprecipitated proteins were fractionated by SDS-PAGE, and gels were exposed to film. Autoradiographs were analyzed by densitometry and are presented as percentage of relative density units normalized to uPA alone.

FIG. 8. A model of uPAR-integrin cis interaction in uPA signaling.
Glycosylphosphatidylinositol-linked uPAR binds to GFD of soluble uPA (anti-GFD and anti-uPAR may block this process). The uPA⅐uPAR complex then binds to integrins as a ligand in cis (anti-integrins, RGD peptide, and anti-uPAR may block this process). The autocrine-type signal transduction that is induced by the occupied integrins would mediate cell adhesion and migration. This signal transduction may activate MAPKs and induce activation of unoccupied integrins by inside-out signaling. uPAR would be able to bind to integrins in apposing cells in trans and induce paracrine-type signals as previously proposed (17). It is unclear whether cell adhesion to immobilized uPA as shown in the present study occurs in vivo, but this system would be useful to analyze the uPA-uPAR-integrin-mediated signaling. The uPA-uPAR interaction may be primarily responsible for adhesion of cells to immobilized uPA, and integrins reinforce this interaction by binding to uPAR in cis. Stress-fiber formation may not be involved in this process. uPAR/Integrin ␣ 5 ␤ 1 Interaction and uPA Signaling fibronectin and uPAR are ligands for ␣ 5 ␤ 1 . We suspect that 1) the initial uPAR binding to a limited number of ␣ 5 ␤ 1 in cis induces the increased signaling through ␣ 5 ␤ 1 and activated MAPK, 2) ␣ 5 ␤ 1 is activated by inside-out signaling, and 3) fibronectin binds to activated unoccupied ␣ 5 ␤ 1 , leading to more outside-in signaling in uPAR-rich cells. In uPAR-poor cells, in contrast, the initial uPAR-␣ 5 ␤ 1 binding may not be enough to induce ␣ 5 ␤ 1 signaling. This would be a possible reason that reduction in the level of uPAR induced a protracted state of dormancy in tumor cells. We were not able to detect stress-fiber formation upon adhesion of uPAR-CHO or mock-transfected CHO cells to immobilized uPA in the absence of fibronectin (Fig. 2d). It is unclear whether signal transduction through ␣ 5 ␤ 1 on adhesion to uPA stabilized the interaction between uPAR and immobilized uPA and/or induced re-organization of cytoskeletal proteins in the present study.
Several recent papers suggest that uPAR binds to unique "non-ligand binding sites" in repeat 4 of the ␣ subunits in several integrins (31,32). Several integrin peptides from the predicted 2-3 loop in repeat 4 (e.g. PRHRHMGAVFLLSQEAG, residues 240 -257 of ␣ 3 ) have been reported to block uPARintegrin interaction. Several laboratories including ours have previously identified many amino acid residues within repeats 2-4 of the integrin ␣ subunits that are critical for binding to integrin ligands (21,(33)(34)(35)(36)(37)(38). A recent crystal structure of integrin ␣ v ␤ 3 (29) verified these mapping results; the critical amino acid residues that have been identified by mutagenesis are located within the ligand binding site of integrins and exposed to the surface. We suspect that the PRHRHMGAVFLLSQEAG peptide of ␣ 3 may be located within the ligand binding site of ␣ 3 because mutations of the predicted 2-3 loop in repeat 4 of ␣ IIb effectively blocked fibrinogen binding to ␣ IIb ␤ 3 (33). We have previously shown that mutations in the predicted 2-3 loop of repeat 3 of ␣ 4 (which is clearly within the ligand binding site of ␣ 4 ) effectively blocked uPAR-␣ 4 ␤ 1 interaction (17). In the present study we demonstrated that the corresponding mutations in ␣ 5 also blocked uPAR-␣ 5 ␤ 1 interaction. It is, thus, highly likely that the uPAR binding site may actually overlap the binding sites for other integrin ligands. If uPAR and other integrin ligands bind to the overlapping region within repeat 2-4 in non-I-domain integrins, it is possible that the integrinderived peptides described above might block the binding of other ligands to integrins as well as uPAR.
We recently reported that the angiostatin, a plasminogen fragment that contains the amino-terminal three or four kringles, binds to several integrins (including ␣ v ␤ 3 and ␣ 9 ␤ 1 but not ␣ 5 ␤ 1 ) (39). We found that ␣ 5 ␤ 1 did not bind to the kringle domain of uPA in the present study. It is, thus, not likely that uPA may cross-link uPAR and ␣ 5 ␤ 1 through GFD and kringle. We found that deletion of the kringle domain (⌬kringle-uPA) did not block the uPAR/␣ 5 ␤ 1 -mediated cell adhesion to uPA in the present study. This indicates that uPA does not bind to ␣ 5 ␤ 1 through the kringle. The finding that an anti-kringle antibody effectively blocked cell adhesion to uPA may be due to steric hindrance of GFD function by the antibody or by causing the kringle to block access of GFD for uPAR. We have observed that that Ab963 and other mAbs whose epitopes map to the kringle can block the binding of uPA to uPAR on cells. 4 However, these studies do not preclude the possibility that uPA may cross-link uPAR and other integrins through GFD and kringle, forming a uPA⅐uPAR⅐integrin signaling complex on the cell surface.
It has been reported that uPA increases association of uPAR with integrins using co-immunoprecipitation (8,32,40,41) although to date the mechanism underlying this phenomenon has remained unclear. Consistent with these reports, we demonstrated that uPA induces RGD-dependent binding of soluble uPAR to isolated ␣ 5 ␤ 1 in solution in the present study (Fig. 7). These results predict that uPA binding to uPAR induces uPAR binding to ␣ 5 ␤ 1 in cis on the cell surface. What is the mechanism of uPA-induced uPAR binding to integrins? It has been reported that soluble uPAR forms dimers and oligomers (42).
Interestingly, the addition of an equimolar concentration of uPA leads to the dissociation of these dimers and oligomers. It is tempting to speculate that the ability of uPAR to binding to integrins is related to dissociation of uPAR, i.e. that GFD binding to uPAR may induce dissociation of uPAR dimers and oligomers, exposing the integrin binding sites within uPAR. Additional studies to examine the role of uPA and uPAR in ␣ 5 ␤ 1 activation are currently under way.