Domain 2 of the urokinase receptor contains an integrin-interacting epitope with intrinsic signaling activity: generation of a new integrin inhibitor.

We investigated the interaction between the urokinase receptor (uPAR) and the integrin alphavbeta3. Vitronectin (VN) induces cell migration by binding to alphavbeta3, but expression of the uPAR boosts its efficacy. Thus, uPAR may regulate VN-induced cell migration by interacting laterally with alphavbeta3. In contrast, cells expressing a uPAR mutant lacking domain 2 do not migrate in response to VN. This effect is overcome by D2A, a synthetic peptide derived from the sequence of domain 2. In addition, D2A has chemotactic activity that requires alphavbeta3 and activates alphavbeta3-dependent signaling pathways such as the Janus kinase/Stat pathway. Moreover, D2A disrupts uPAR-alphavbeta3 and uPAR-alpha5beta1 co-immunoprecipitation, indicating that it can bind both of these integrins. We also identify the chemotactically active epitope harbored by peptide D2A. Mutating two glutamic acids into two alanines generates peptide D2A-Ala, which lacks chemotactic activity but inhibits VN-, FN-, and collagen-dependent cell migration. In fact, the GEEG peptide has potent chemotactic activity, and the GAAG sequence has inhibitory capacities. In summary, we have identified an integrin-interacting sequence located in domain 2 of uPAR, which is also a new chemotactic epitope that can activate alphavbeta3-dependent signaling pathways and stimulate cell migration. This sequence thus plays a pivotal role in the regulation of uPAR-integrin interactions. Moreover, we describe a novel, very potent inhibitor of integrin-dependent cell migration.

The urokinase receptor (uPAR) 1 regulates pericellular proteolysis and fibrinolysis by localizing urokinase (uPA) and plas-minogen activation on the cell surface. However, uPAR Ϫ/Ϫ mice do not show thrombotic disorders (1,2), and it has now become clear that uPAR function is in the regulation of cell migration (3,4). By binding uPA, uPAR induces migration of both adherent and nonadherent cells in culture. In uPAR Ϫ/Ϫ mice, neutrophil recruitment in response to pulmonary Pseudomonas aeruginosa infection is weaker than in wild type, and leukocyte recruitment to sites of acute inflammation is dramatically reduced (5)(6)(7)(8). Thus, uPAR is involved in physiological and pathological processes requiring cell migration such as angiogenesis, tumor invasion, and inflammation. Furthermore, uPAR promotes cell adhesion directly by binding vitronectin (VN), a serum and extracellular matrix molecule (9,10). Finally, uPAR affects cell adhesion through lateral interactions with integrins (11)(12)(13).
uPAR has no cytoplasmic domain and is bound to the plasma membrane by a glycosylphosphatidylinositol anchor. This receptor consists of three homologous domains, and its N terminus, the domain 1, is the primary binding site for uPA. However, domain 3 is also involved, and the presence of domains 2 and 3 enhances the affinity of domain 1 for uPA (14). uPAR affinity for VN is also enhanced by the binding of uPA (10). The linker region of uPAR, located between domains 1 and 2, contains a strong chemotactic epitope (15)(16)(17). uPA binding to uPAR induces a conformational change in the receptor that exposes this previously masked SRSRY epitope (16), switching uPAR into a ligand for FPRL1, a seven-membrane-spanning domains G protein-coupled receptor, which in turn stimulates cell migration (15,16,18,19). Therefore, uPAR might be considered as a membrane-anchored chemokine (4,20).
Despite the lack of a cytoplasmic domain, uPAR is capable of complex signaling. Hck, c-Src, focal adhesion kinases (FAK), and protein kinases A and C have been shown to regulate uPAR-dependent signaling pathways (21). uPA binding to uPAR activates downstream signaling pathways including mitogen-activated protein kinases (17,21). uPAR also controls small G proteins, like Rac (22), that are involved in the regulation of the cell cytoskeleton (23) and cell morphology.
uPAR interacts with numerous molecules in the plasma membrane, including FPRL1 (15), the epidermal growth factor receptor (24), gp130 (25), and integrins (26), activating or modulating signaling pathways (3,4,13). We have previously shown that uPA-and VN-induced chemotaxis, cytoskeleton reorganization, and cell shape changes required the formation of a uPAR-␣v␤3 signaling complex (19,21). Several studies * The work in the laboratory of F.B. was supported by grants from the Italian Association for Cancer Research and by the European Union Framework program 6 Grants IP 2003-503297 and NoE 2003-502935. The work in the laboratory of D.L. was supported by National Institutes of Health Grant HL 31950. This is TSRI manuscript 16980-CB. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

EXPERIMENTAL PROCEDURES
Materials, Cell Culture, and Adenoviral Infection-Mouse LB6 and NIH 3T3 parental and transfected cells and human embryonic kidney cells (HEK-293 and transfected HEK-293-uPAR) were cultured in DMEM plus 10% FCS. LB6 and NIH 3T3 cells express a low level of murine uPAR (29 -32). The constructs used to transfect the cells with human wild-type or mutated uPAR have been described (29,31). LB6 clone 19 and NIH 3T3-uPAR cells express about 500,000 and 300,000 human receptors/cell, respectively (29). LB6-D1HD3 and LB6-D1D2 clones express about 160,000 and 90,000 uPAR molecules/cell on their surface, respectively (29). Despite the lack of one domain, the dissociation constants of D1HD3 (2 nM) and D1D2 (8 nM) clones for uPA were not very different from wild type (1.5 nM) (29).
Rat smooth muscle cells (RSMC) were cultured in DMEM plus 10% FCS. For adenoviral infection, RSMC were grown up to 60 -80% confluence in 100-mm Petri dishes and infected for 3 h with recombinant adenovirus at a multiplicity of infection of 500 plaque-forming units/cell. The recombinant adenoviruses encoding wild-type Tyk2 and a dominantnegative deletion mutant of Tyk2 have been described (33,34). RSMC were kept in serum-free medium, and after 1 day, another infection was performed. Cells were used 2 days after the second infection.
Chemotaxis Assay-A chemotaxis assay was performed as described (19) with modified Boyden chambers. Filters (5-m pore size; Neuro Probe, Gaithersburg, MD) were treated with collagen I (100 g/ml) and fibronectin (10 g/ml; Roche Applied Science). Approximately 40,000 -50,000 cells in serum-free DMEM were added to the upper well, and the chemoattractants to be tested were added to the lower well. When present, antibodies or inhibitors were added to both wells. After overnight migration at 37°C, cells remaining on the upper surface of filters were scraped off, the filters were fixed in methanol, and the cells on the undersurface were stained in a solution of 10% (w/v) crystal violet in 20% (v/v) methanol. The experiments were performed at least twice in triplicate, and the results, expressed as -fold over control, are the mean Ϯ S.D. of the number of cells counted in 10 high power fields per filter. Random cell migration (i.e. migration in the absence of chemoattractant) was given the arbitrary value of 100%.
Immunofluorescence Microscopy-As previously described (19,38), 15,000 -25,000 RSMC cells (20 -40% confluence) were seeded on glass coverslips in a 2-cm 2 well, cultured for 24 h in DMEM plus 10% FCS, washed with phosphate-buffered saline (PBS), and cultured for another 24 h without FCS. After stimulation, RSMC were fixed for 20 min at room temperature with 3% paraformaldehyde, 2% sucrose in PBS, pH 7.5; washed three times with PBS plus 0.2% BSA; permeabilized with 20 mM Hepes, pH 7.4, 300 mM saccharose, 50 mM NaCl, 3 mM MgCl 2 , 0.5% (v/v) Triton X-100 for 3 min at 4°C; and washed three times with PBS plus 0.2% BSA. RSMC were incubated in PBS plus 2% BSA for 15 min at 37°C and then with anti-Stat1 primary antibodies for 30 min at 37°C, washed three times with PBS plus 0.2% BSA, and further incubated with PBS plus 2% BSA for 15 min at 37°C. The cells were then stained with the secondary TRITC-antibody and with FITC-phalloidin for visualization of filamentous actin for 30 min at 37°C. DAPI was used to label the nucleus. Finally, after three washes with PBS plus 0.2% BSA, one with distilled water, the coverslips were mounted with 20% (w/v) Mowiol. Fluorescence photographs were taken with a Zeiss Axiophot microscope.
To quantify the activation of Stat1, at least 10 random pictures were taken, nuclei positively stained for Stat1 were counted, and the data were expressed in percent. Cells were seeded in 96-well culture dishes treated overnight with 1.5% gelatin in PBS, pH 7.4, grown to confluence in DMEM plus 10% FCS, and then serum-starved for 24 h in DMEM. Total binding was determined by incubating the cells on ice for 90 min in binding buffer (DMEM containing 0.1% (w/v) BSA and 10 mM Hepes, pH 7.4) plus 0.5 nM 125 I-VN. Nonspecific binding was measured by competition with a 100-fold excess of unlabeled VN. The cells were then washed three times with ice-cold binding buffer and once with ice-cold PBS and lysed in 1% SDS, 1% Triton X-100, and the radioactivity was counted. Cell number was determined in parallel wells by direct counting, and the results were expressed as cpm/10 5 cells.
Pull-down Assay-suPAR was radioiodinated (as for VN; see above) at a specific activity of 1.5 ϫ 10 5 cpm/g. A mix composed of 2 g of either purified ␣v␤3 or ␣5␤1 integrin and 2 g of 125 I-suPAR was incubated for 4 h at 4°C in the absence (control) or in the presence of 4 g of uPA with or without 5 g of either D2A or D2A-Ala peptide in binding buffer (RPMI plus 0.02% BSA, 10 mM HEPES, pH 7.4; total volume 300 l). Complexes were immmunoprecipitated with 3 g of anti-␣v␤3 (LM609) or anti-␣5␤1 monoclonal antibody (HA5) and 60 l of protein G-agarose beads (Amersham Biosciences). Beads were washed three times in binding buffer and extracted in reducing sample buffer, and the proteins were fractionated by SDS-PAGE. The gels were analyzed by autoradiography and densitometry, and the results are presented as percentage of relative density units normalized to the band obtained with the addition of uPA alone (100%).
Cell Adhesion Assays-The cell adhesion assay was performed using uPAR-overexpressing HEK-293 cells (both clone 14 and clone 19). Vitronectin (1 g/ml)-coated 96-well tissue culture plates were preincubated for 2 h at 37°C, followed by a 1-h incubation at 37°C with 2% BSA in PBS. After two washes with PBS, 3 ϫ 10 5 /ml cells suspended in 100 l of DMEM plus 0.1% BSA were seeded in duplicate on proteincoated 96-well plates and incubated for 1 h at 37°C. After three washes with PBS, attached cells were fixed and stained with Diff-Quick (Dade Diagnostics). The data were quantified by measuring absorbance at a wavelength of 570 nm. The peptides ␣325, scrambled ␣325, D2A, and D2A-Ala (100 M) were added to the cell suspension immediately before seeding cells in the wells.
Statistical Analysis-Statistical analysis was performed with the Prism software using Student's t test for pairwise comparisons of treatments or an analysis of variance model for the evaluation of treatments with increasing doses of a reagent.

RESULTS
uPAR Expression Affects VN-induced Migration-We have examined the chemotactic effect of VN on control mouse cells and on the same cells transfected with human uPAR cDNA. NIH 3T3 and LB6 murine cells express only low levels of murine uPAR, whereas NIH 3T3-uPAR and LB6 clone 19 cells express 300,000 and 500,000 human receptors/cell, respectively. VN has chemotactic activity in both transfected and untransfected cells (Table I); however, an increased migratory response to VN was observed in the cells transfected with uPAR. At 5 g/ml, VN-directed cell migration of LB6 clone 19 cells was significantly enhanced (p values Ͻ0.001) 3.3-fold, with respect to untransfected LB6 cells (Table I). Similar results were observed with NIH 3T3 and human HEK-293 cells, the latter being naturally devoid of uPAR (Table I). Thus, uPAR expression affects the migratory response to VN, in agreement with a direct role of uPAR in VN-induced chemotaxis (21). (9,39), the effect of uPAR on VN-dependent chemotaxis might be due to an increase of VN-binding sites. To investigate this possibility, we tested pools of LB6 and NIH 3T3 cells transfected with either D1HD3-uPAR or D1D2-uPAR, two mutated forms of human uPAR lacking either domain 2 or 3, respectively. Both variants of uPAR are still able to bind human uPA, pro-uPA, or ATF (29). Since VN binding requires an intact uPAR, these mutants are not expected to bind VN (39 -41). As expected, VN bound to LB6 clone 19 cells, but it bound very poorly to cells expressing the mutated forms of uPAR (Fig. 1A).

VN-directed Cell Migration Does Not Depend on VN Binding to uPAR but Requires an Intact uPAR-Since VN binds both integrins and uPAR
We thus tested the relationship between VN-dependent cell migration and VN binding to the cells. In the presence of D1HD3-uPAR, VN failed to stimulate the chemotaxis of NIH 3T3 or LB6 cells even at 10 g/ml (Fig. 1B). This effect was clone-independent, because pools of transfected cells were used rather than clones. The fact that the migratory response to VN was totally absent in D1HD3 cells is surprising, since VN stimulated 2.5-fold chemotaxis even in untransfected parental LB6 cells (Table I). The D1HD3-uPAR mutant decreased the response of the untransfected mouse cells to VN and thus appears to have a dominant negative effect on VN chemotaxis. The results with the D1D2-uPAR mutant were different. Although LB6-D1D2 cells displayed lower sensitivity to low doses of VN compared with LB6 clone 19 cells (Fig. 1B), they still responded to VN with an about 4-fold maximal stimulation at 1 g/ml (Fig. 1B). In addition, cells transfected with D1D2-uPAR still responded to human ATF in a chemotaxis assay, whereas D1HD3 and LB6 did not (Fig. 1C). These data show that the stimulation of chemotaxis by VN did not require its interaction with uPAR. To further confirm this observation, we used VN-(40 -459), a form of VN lacking the uPAR-binding SMB domain and hence unable to bind uPAR (37). VN-(40 -459) stimulated cell migration as well as full-length VN (Fig.  1D). These results suggest that VN induces cell migration by binding to its own integrin receptors. This conclusion is strengthened by our previous observation that anti-␣v␤3 antibodies inhibited the chemotactic effect of VN (21) (see Fig. 4F).
However, since anti-uPAR antibodies also inhibited VN-dependent chemotaxis (21), uPAR must be also directly involved. Therefore, uPAR involvement in the migration-promoting effect of VN might be due to a lateral interaction with the integrins (26,42). Since human wild-type uPAR and D1D2-uPAR enhanced the effect of VN, whereas a mutant lacking domain 2 prevented VN-induced cell migration (Fig. 1B), we hypothesized that domain 2 of uPAR might be directly involved in an interaction with an integrin. A region of domain 2, rich in charged amino acid residues and hence likely to be exposed at the surface of uPAR, attracted our attention (peptide D2A). Other sequences within the same domain were not tested.
D2A, a Peptide Derived from the Sequence of Domain 2 of Human uPAR, Has Chemotactic Activity-We have explored the sequence of domain 2 of uPAR and have focused on peptide 130 IQEGEEGRPKDDR 142 (D2A) as its sequence suggested that it might be present on the surface of the protein. Peptide D2A per se was able to stimulate migration of LB6 and LB6-D1HD3 cells with a maximal 2-fold increase at 1 pM for the former and 10 pM for the latter (Fig. 2, A and B). Importantly, D2A restored the response to VN of the LB6-D1HD3 cells, counteracting the dominant negative effect of D1HD3 (Fig. 2B).
We also tested the effect of peptide D2A in a well characterized cell system, RSMC (19,21,41,42), which express uPAR and ␣v␤3 and migrate in response to both uPA and VN challenge. Fig. 2C shows that D2A dose-dependently stimulated the migration of RSMC with a maximum at 1 pM. The chemotactic effects of D2A and VN were not additive, since the combination of optimal doses of peptide D2A (1 pM) and VN (1 g/ml) did not further increase cell migration (Fig. 1D). Similar results were obtained with a combination of VN-(40 -459) and D2A (Fig. 1D).
Identification of Essential Chemotactic Residues of Peptide D2A-We next compared the chemotactic activity of peptides D2A and D2B (see "Experimental Procedures") that share the same amino acids composition but with a reversed sequence. Both peptides were equally active in stimulating migration of RSMC (Fig. 3A), suggesting the presence of a common epitope in both D2A and D2B. Examination of the two sequences revealed a common GEEG sequence, which might explain why the reverse D2A sequence, D2B, was still chemotactic. We substituted the two glutamic acids in D2A for two alanines, giving a GAAG sequence. When this new peptide, D2A-Ala, was compared with D2A and D2B in chemotaxis, it failed to stimulate cell migration (Fig. 3A). These data suggested that the GEEG sequence of both D2A and D2B peptides was responsible for their chemotactic activity. This conclusion was further supported by the use of smaller synthetic GEEG and GAAG peptides. The GEEG peptide dose-dependently stimulated cell migration with a maximum at 1 pM (Fig. 3B). Thus, this GEEG peptide acted in a way similar to the longer D2A peptide (Fig. ). In addition, a scrambled version of peptide D2A has no effect on cell migration (Fig. 3B). However, the mutated GAAG peptide used at the same concentrations did not affect cell migration (Fig. 3B). Therefore, both GAAG and D2A-Ala have no chemotactic activity. D2A Peptide Has ␣v␤3-dependent, Not uPAR-dependent, Chemotactic Activity-We previously showed that uPA and VN synergized in RSMC chemotaxis because they activated different signaling pathways (21). We reported, that unlike uPAinduced cell migration, VN-induced cell migration required down-regulation of PK-A and was not extracellular signalregulated kinase-dependent (21). On the other hand, uPA-promoted cell migration was independent of PK-A and required extracellular signal-regulated kinase activation (21). Therefore, we pharmacologically investigated the pathways involved in D2A chemotaxis. The effects of both VN and D2A were blocked by increasing the intracellular cAMP concentration with forskolin and IBMX (Fig. 4A), a combination that does not inhibit uPA-induced cell migration (21). In contrast, the mitogen-activated protein kinase/extracellular signal-regulated kinase kinase inhibitor, PD98059, which completely blocks pro-uPA-induced cell migration, actin cytoskeleton reorganization, and extracellular signal-regulated kinase translocation into RSMC nucleus (21), had no effect on VN-and D2A-induced cell migration (Fig. 4B). These data suggest that VN and D2A stimulate analogous signaling pathways. In addition, AG-490, a specific inhibitor of the Janus kinases (Jaks), inhibited both VN and D2A chemotaxis in RSMC (Fig. 4C). Once again, the fact that AG-490 acted in a similar way and dose-dependently inhibited cell migration induced by either D2A or GEEG peptide, the shorter form of D2A, reinforces the idea that the GEEG sequence is the chemotactically active epitope of D2A (Fig. 4D). Random cell migration was not affected by the same concentrations of AG-490 (Fig. 4D). These results suggest that both VN and D2A (or the GEEG peptide) can activate the Jak-dependent signaling pathway. We have not yet identified the Jak(s) involved. However, we have excluded the Janus kinase Tyk2, which has been shown to play an important role in regulating uPA-induced SMC migration (25,33,34). We investigated whether Tyk2 would be involved in mediating the migratory signal induced by D2A, GEEG, and VN. RSMC were adenovirally infected to express wild-type Tyk2 (RSMC/Tyk2) or a dominant-negative mutant of Tyk2 (RSMC/K⌬L). When challenged by D2A, GEEG, or VN, RSMC/Tyk2 showed an ϳ2-fold increase in chemotaxis compared with RSMC/Tyk2 migrating toward medium alone (Fig. 4E). However, both RSMC/K⌬L and uninfected RSMC exhibited a similar increase (Fig. 4E). In contrast, RSMC/K⌬L did not migrate in response to PAI-1 (38). Thus, these data show that Tyk2 is not required for D2A-, GEEG-, and VN-induced cell migration. These data also suggest that another Jak kinase(s) is involved in mediating D2A and VN chemotaxis and that in good agreement with a previous study VN (or D2A) and uPA induce cell migration through different signaling pathways (21). Moreover, both VNand D2A-induced cell migration required ␣v␤3 integrin, since they were inhibited by LM 609, a monoclonal antibody against ␣v␤3 (Fig. 4F). Therefore, D2A and VN appear to act through the same signaling pathways.
Chemoattractants induce the reorganization of actin cytoskeleton concomitantly with cell motility (45). Thus, as a chemoattractant, D2A should affect cell shape and actin cytoskeleton organization. In addition, since members of the Jak kinase family are involved in mediating the chemotactic signal B, D2A and its shorter form, the GEEG peptide, are equally chemotactic. Both D2A and GEEG peptide stimulate migration of RSMC in a dose-dependent manner. In all cases, in the presence of D2A or GEEG peptide, the results are statistically significant (p Ͻ 0.05) compared with control. GAAG peptide, the smaller version of D2A-Ala, has no chemotactic activity. Migration of RSMC toward medium alone (control) is considered to be 100% migration.

FIG. 4. Effect of signaling inhibitors on RSMC chemotaxis.
A, both D2A and VN chemotactic effects on RSMC are completely inhibited by the addition of a combination of forskolin plus IBMX. The dilution buffer used to solubilize the mix of forskolin and IBMX has no effect. *, statistical significance (p Ͻ 0.0001) compared with control lacking VN. B, the mitogen-activated protein kinase/extracellular signal-regulated kinase kinase inhibitor, PD98059, fails to block D2A-and VN-induced RSMC migration. C, AG-490, a specific inhibitor of the Jak kinases, blocks both D2A-and VN-dependent chemotaxis of RSMC. *, statistical significance (p Ͻ 0.0001) compared with control lacking VN. D, the inhibitor, AG-490, dose-dependently inhibits RSMC chemotaxis induced by either D2A or GEEG peptide, the shorter version of D2A. E, Tyk2 is not involved in the mediation of the migratory signal induced by either D2A, GEEG peptide, or VN. Noninfected RSMC or infected RSMC expressing wild-type Tyk2 (RSMC/Tyk2) or a dominant negative mutant (RSMC/K⌬L) were subjected to chemotaxis assay and migrated toward D2A, GEEG peptide, or VN. Control is represented by noninfected cells migrating in the absence of chemoattractant. In the presence of PAI-1, the migration of RSMC/K⌬L is significantly different (p Ͻ 0.001) from that of RSMC/Tyk2. In the other conditions tested, no significant differences were observed between noninfected RSMC and infected RSMC. F, a monoclonal antibody against ␣v␤3 (LM 609) inhibits both D2A-and VN-induced RSMC migration. **, a difference highly significant compared with control lacking antibody (p Ͻ 0.0001); *, significant difference (p Ͻ 0.05). The 100% value represents the number of cells migrating in the absence of attractant.
induced by D2A and VN (Fig. 4, C-E), we also explored the state of activation of Stat1, a downstream effector of some of the Jaks (see Ref. 46). Subconfluent cultures of serum-starved RSMC were stimulated with 1 pM D2A peptide or 1 g/ml VN for 30 min at 37°C and analyzed for actin cytoskeleton organization and Stat1 distribution. Cells were triple-labeled with FITC-phalloidin, DAPI, and primary anti-Stat1 antibodies, followed by TRITC-secondary antibodies to visualize the actin cytoskeleton, the nucleus, and Stat1, respectively. Unstimulated cells kept at 37°C for 30 min represented the control conditions. Most RSMC under control conditions exhibit numerous stress fibers and a nonpolarized cell shape (Fig. 5). After 30 min of stimulation with D2A, the RSMC acquired an elongated, polarized morphology, semiring actin structures, and membrane ruffling at the leading part of the cell. Actin filaments were also observed flanking the nucleus and in the dragging trail. As expected (21), VN induced similar changes. As far as Stat1, it was mainly cytoplasmic (inactive) in unstimulated RSMC, whereas both VN and D2A induced its translocation to the nucleus (Fig. 5).
These results show that D2A and GEEG peptides stimulate cell migration through signaling pathways that are common to the ␣v␤3-dependent and different from the uPA-uPAR-FPRL1dependent signaling pathways (18,21,47), supporting the idea that the GEEG epitope harbored by peptide D2A may be involved in the uPAR-␣v␤3 interaction.
Peptide D2A Interferes with the Formation of suPAR-Integrin Complexes in Vitro-Since uPAR and ␣v␤3 form complexes on the cell surface and in vitro (48 -50), we investigated the effects of peptide D2A in a cell-free, uPA-dependent, suPAR-integrin co-immunoprecipitation assay (see "Experimental Procedures"). Fig. 6A shows that the addition of uPA is required for 125 I-suPAR to be immunoprecipitated by ␣v␤3 antibodies (lane 1 versus lane 2). D2A (lane 3) reduced the amount of coimmunoprecipitated suPAR by about 50%. D2A also inhibited su-PAR-␣5␤1 co-immunoprecipitation by about 55% (Fig. 6B), indicating that D2A can interact with multiple integrins.
D2A-Ala Inhibits VN-induced Cell Migration-We further tested D2A-Ala by investigating its effects on VN-induced migration. Surprisingly, unlike D2A, D2A-Ala completely abrogated the chemotactic effect of VN on RSMC (Fig. 7A) as well as primary cultures of human smooth muscle cells from the coronary artery (CASMC) and from the aorta (AoSMC) (Fig. 7, B  and C). Thus, D2A-Ala behaves as an inhibitor of VN-induced cell migration. Since the only difference between peptide D2A and D2A-Ala is the presence of a GAAG sequence in the latter, we investigated the inhibitory effect of this GAAG tetrapeptide on VN-induced chemotaxis. GAAG peptide completely inhibited VN-induced RSMC migration (Fig. 7A), showing that the GAAG sequence is responsible for the inhibitory properties of the D2A-Ala peptide.
To determine the IC 50 of D2A-Ala inhibitor, we tested increasing concentrations of D2A-Ala on VN-stimulated RSMC. Fig. 7D shows that D2A-Ala inhibited VN-induced cell migration in a dose-dependent manner. A complete inhibition was  5. Comparison of the effects of D2A, D2A-Ala, and VN on actin cytoskeleton organization, cell morphology, and the distribution of Stat1 in RSMC. Cells were treated with either D2A, D2A-Ala (1 pM), or VN (1 g/ml) for 30 min at 37°C. After fixation, they were permeabilized and triple-stained with FITC-phalloidin, DAPI, and a primary anti-Stat1 antibody. The cells were then stained with a secondary TRITC-anti-Ig antibody. In this way, we have visualized the actin cytoskeleton (green), the nucleus (blue), and Stat1 (red), respectively. Untreated RSMC kept for 30 min at 37°C served as control.
obtained at 1 pM with an IC 50 of ϳ10 -20 fM. Thus, D2A-Ala inhibition is extremely potent.
We also tested the effect of D2A-Ala on uPA-dependent su-PAR-␣v␤3 co-immunoprecipitation in vitro. D2A-Ala also prevented uPAR-␣v␤3 co-immunoprecipitation (Fig. 6A), indicating that it can still interact with ␣v␤3. However, D2A-Ala did not promote the change of morphology and reorganization of actin cytoskeleton (Fig. 5) or Stat1 translocation to the nucleus (Figs. 5 and 8A). Furthermore, D2A-Ala inhibited VN-induced translocation of Stat1 (Figs. 5 and 8A). These data indicate that D2A-Ala interacts with ␣v␤3 in an antagonistic way, preventing the transduction of a chemotactic signal. The exact molecular mechanism is presently unknown but will be the focus of future experiments.
D2A-Ala Is a General Inhibitor of Integrin-dependent Cell Migration-Since ␣v␤3 mediates the chemotactic effect of VN, and since uPAR interacts with several integrins, D2A-Ala might inhibit migration induced by other matrix proteins such as FN and LN. Indeed, FN and LN bind to and act through different integrins, including ␣5␤1. D2A-Ala inhibited both FN-and LN-induced migration (Fig. 8B). Peptide GAAG, the shorter form of D2A-Ala also inhibited FN-induced cell migration (Fig. 8C). In conclusion, D2A-Ala (or GAAG peptide) is a powerful and general inhibitor of matrix-induced, integrin-dependent, cell migration.
The Migration-promoting Effect of Peptide D2A Does Require the Presence of uPAR on the Cell Surface-LB6 and NIH 3T3 cells express low levels of murine uPAR. Therefore, to test whether the effects of peptides D2A and D2A-Ala required the expression of uPAR on the cell surface, we used HEK-293 cells that do not express uPAR (18). D2A elicited a strong chemotactic effect on HEK-293-uPAR cells but failed to stimulate the migration of HEK-293 cells (Table II), indicating that uPAR expression is a prerequisite for the D2A chemotactic effect. D2A-Ala did not induce migration in either HEK-293 or HEK-293-uPAR cells (Table II). VN induced chemotaxis in HEK-293 cells, and this effect was increased in the HEK-293-uPAR cells (Table II). The addition of D2A did not modify the response to VN in either of the cell types. However, D2A-Ala blocked VNdependent chemotaxis also in the cells that do not express uPAR (Table II). Thus, whereas D2A needs uPAR expression to stimulate migration, the inhibitor D2A-Ala can inhibit VN- FIG. 7. Peptide D2A-Ala has no chemotactic activity but is an inhibitor of VN-induced cell migration. A, D2A-Ala or its shorter version, the GAAG peptide, blocks VN-induced RSMC chemotaxis. B and C, VN-induced migration of human SMC from the coronary artery (B) and from the aorta (C) is totally inhibited in the presence of peptide D2A-Ala. A chemotaxis assay was performed as described under "Experimental Procedures." *, statistical significance (p Ͻ 0.0001). D, D2A-Ala inhibits VN-induced migration of RSMC in a dose-dependent manner. A chemotaxis assay was performed in the absence (control) or in the presence of VN (1 g/ml) with or without increasing doses of D2A-Ala. *, statistical significance (p Ͻ 0.001) compared with control lacking VN. Migration of SMC toward medium alone (control) is considered to be 100% migration.
induced chemotaxis even in the absence of uPAR.
In conclusion, these data show that the agonistic effect of D2A and the antagonistic effect of D2A-Ala use two slightly different mechanisms, which require the presence of uPAR on the cell surface for the former but not for the latter. Therefore, the sequence of peptide D2A contains both "interaction" and "signaling" information, and the two can be dissociated.
D2A/D2A-Ala and ␣325 Have Different Effects on Cell Adhesion onto VN-D2A/D2A-Ala both inhibit uPAR-integrin association (see Fig. 6). This is the same effect exerted by the ␣325 integrin peptide (27). It is possible, therefore, that the D2A peptide represents the binding site for ␣325. To test this possibility, we have performed cell adhesion assays onto VN using two different clones of HEK-293 uPAR cells and verified the effect of these peptides. As shown in Table III, ␣325 totally blocked cell adhesion on VN in both clones. However, neither D2A nor D2A-Ala inhibited adhesion nor had any effect on the inhibitory effect of ␣325. We conclude, therefore, that although all of these peptides inhibit uPAR-integrin co-immunoprecipitation, the D2A sequence does not represent the binding site for ␣325 on uPAR and that the two peptides act through a different mechanism.

DISCUSSION
Although the glycosylphosphatidylinositol membrane-bound uPAR was first thought to be a key regulator of plasminogen activation, it is now recognized that it is also (perhaps mostly) a signaling and adhesion receptor. The fact that uPAR does not have a cytoplasmic domain suggests that its effects must be mediated through interactions with other receptors on the cell surface. A wide diversity of transmembrane receptors have been reported to interact with uPAR, including for instance LDL-receptor-related protein and other internalization receptors, the epidermal growth factor receptor and the G proteincoupled receptor FPRL1 (for reviews, see Refs. 3, 11, 13, and 51). In the case of FPRL1, uPA binding to uPAR induces a conformational change that exposes the chemotactic sequence located in the linker region between domains 1 and 2 of uPAR. This conformational change turns uPAR into a ligand for FPRL1, which mediates at last the chemotactic signal of uPA (15,16,18,19).
Integrins are another important family of receptors that can interact with uPAR (8,26,35,50,52,53), in particular with ␣m␤2 (Mac-1), ␣3␤1, ␣5␤1, and ␣v␤3 (26 -28, 35). Integrins are RSMC were treated with either D2A or D2A-Ala (1 pM) in the absence or in the presence of VN (1 g/ml) for 30 min at 37°C. After fixation, the cells were permeabilized and triple-stained with FITC-phalloidin, DAPI, and a primary anti-Stat1 antibody followed by a secondary TRITC-anti-Ig antibody to visualize the actin cytoskeleton, the nucleus, and Stat1, respectively. Untreated RSMC kept for 30 min at 37°C served as control. At the end of the incubation period, random pictures were taken, and nuclei positively stained for Stat1 were counted and expressed as a percentage of total nuclei. *, statistical significance well known for their role in the regulation of cell adhesion and migration, but unlike uPAR, they possess a cytoplasmic domain connected to downstream signaling molecules. Furthermore, integrins are capable of bidirectional signaling, conveying outside-in and inside-out signals. Perhaps for this reason, integrins interact with numerous membrane proteins such as integrin-associated protein, tetraspanins, and uPAR (for a review, see Ref. 54). The role of uPAR in these interactions has been examined, and it has been suggested that uPAR behaves as a modulator of integrin function (26,27,42,49,55). The role of uPAR in cell migration is not confined to mediating the uPA signal. In fact, it is also involved in other signals, like that of formylmethionylleucylphenylalanine and VN (21,40,56). In this study, we have investigated the role of uPAR in VN-mediated cell migration and have explored the consequences of uPAR-integrin interactions on cell signaling, cytoskeleton organization, cell morphology, and cell migration. We confirm here that uPAR is an important regulator of integrin function. Although it is not absolutely required (see the effect of VN on HEK-293 cells, which do not produce uPAR), uPAR overexpression considerably enhances VN-induced cell migration. In untransfected mouse cells (LB6 and NIH 3T3 cells), VN also stimulates migration (Table I); hence we speculate that mouse uPAR can interact with mouse integrins, although we have not directly addressed this point. Moreover, we also assume that human uPAR can interact with mouse integrins. It is known that the uPA-uPAR interaction is species-specific (57,58), but there is no evidence that the human uPAR cannot interact with mouse integrins. Although it is known that integrin engagement induces proteases including uPA (59), the effect of endogenous uPA on the uPAR-dependent increase of VN chemotactic activity in transfected cells can probably be ignored because of the species specificity of the uPA-uPAR interaction (57,58).
The impact of human uPAR on mouse integrin ␣v␤3 is further demonstrated by the influence of two mutants of uPAR, D1D2-uPAR and D1HD3-uPAR. Whereas the first only slightly alters the response to VN, the second prevents it altogether (Table I and Fig. 1B). The effects of these mutants rule out the possibility that uPAR mediates VN chemotaxis by a direct binding mechanism, since neither of the two mutants can bind VN (Fig. 1) (39). Furthermore, VN- (40 -459), the truncated form of VN that lacks the somatomedin B domain harboring the binding site for uPAR but still conserving the RGD site (52), promoted cell migration as well as full-length VN (Fig. 1D). Moreover, VN can stimulate migration of HEK-293 cells that are devoid of uPAR. Therefore, the antibody inhibition data of Fig. 3D confirm that VN stimulates cell migration by binding to its own integrin receptors, particularly ␣v␤3 as previously re-ported (21,40,60), exploiting the lateral interactions between integrins and uPAR (10).
This hypothesis is plainly supported by the identification of peptide D2A, which is located in domain 2 ( 130 IQEGEEGRPK-DDR 142 ) of uPAR. This peptide on the one hand abolished the inhibitory effect of the expression of D1HD3-uPAR and, on the other hand, showed direct signaling properties identical to VN. Peptide D2A blocks the formation of uPAR-␣v␤3 and uPAR-␣5␤1 complexes, indicating that it can interact directly with at least two integrins. D2A binding also appears to be functionally relevant, since it stimulated migration of cells expressing uPAR and, even more importantly, overcame the inhibitory effect of D1HD3-uPAR expressing LB6 cells (Fig. 2). Like VN-, D2A-induced cell migration was completely blocked by LM609, a monoclonal antibody against ␣v␤3. Thus, D2A not only binds to the integrin but also generates a signal through it. The investigation of the downstream signaling pathways activated by peptide D2A fully agrees with this idea. Using previously identified inhibitors that discriminate between uPAR-and VNdependent signaling (21), we found that D2A stimulated migration via VN-dependent, uPAR-independent signaling pathways. Indeed, on the one hand, increasing intracellular cAMP using forskolin and IBMX, which has no effect on uPA-directed cell migration (21), totally inhibited both D2A-and VN-induced chemotaxis. On the other hand, PD98059, the mitogen-activated protein kinase/extracellular signal-regulated kinase kinase inhibitor that prevents mitogen-activated protein kinase activation and blocks uPA-induced cell migration (21,61), failed to inhibit either D2A or VN-dependent chemotaxis (Fig.  3). Finally, both D2A and VN activated the Jak/Stat signaling pathway as observed for numerous other chemoattractants such as chemokines (46). In fact, both D2A-and VN-promoted chemotaxis were blocked by AG-490, an inhibitor of the Janus family of kinases, suggesting that D2A can activate at least one Jak. Moreover, both D2A and VN induced Stat1 relocalization to the nucleus of RSMC (Fig. 5). Stats are the downstream effectors of Jaks, and once activated, these latent cytoplasmic transcription factors translocate into the nucleus (46). Importantly, D2A (or GEEG peptide) and VN did not induce cell migration through activation of the Tyk2 member of Jak kinases, which was reported to be the major Jak mediating uPA-induced SMC migration (33,34). In fact, the dominant negative Tyk2 mutant did not affect VN-, D2A-, or GEEGinduced cell migration (Fig. 4E). The Jak family member responsible for the Stat-1 activation remains to be identified, as well as the mechanism involved, since Stat-1 nuclear localization depends on a balance of Stat-1 import and export from the nucleus (62). In addition, D2A promotes the appearance of the elongated morphology typical of motile cells (sometimes also called a hand mirror shape) and the reorganization of actin cytoskeleton that plainly reflects this motile morphology (Fig.  5). These results suggest that besides the Jak/Stat pathway, D2A can activate other downstream signaling molecules such as small GTP-binding proteins that are known to regulate the organization of actin cytoskeleton (23). This conclusion would be in keeping with previous observations in fibroblasts (19,22). Taken together, these observations show that D2A has signaling capacities, acting through ␣v␤3-dependent and not uPARcontrolled pathways, and that the Jak/Stat pathway is directly involved in the regulation of D2A-induced cell migration.
Based on these considerations, we propose that D2A contains two types of information, including sequences that mediate its binding to integrins ␣v␤3 or ␣5␤1, and sequences for the activation of integrin-dependent signaling pathways involved in the regulation of cell migration. The dissociation between these two sets of properties was observed when we introduced muta- tions into the sequence of D2A. Our attention on a particular epitope (i.e. GEEG) resulted from the observation that although peptide D2B had a sequence that was the reverse of D2A, it was equally active in chemotaxis, and both peptides contain the same GEEG sequence. Thus, we modified GEEG into a GAAG epitope and tested the chemotactic activity of this new peptide named D2A-Ala. This peptide was identical to D2A except for the substitution of two alanines for the two glutamic acid residues. Peptide D2A-Ala was not chemotactic and had no signaling activity (Figs. 4, 6, and 7), demonstrating that the GEEG sequence is the chemotactically active epitope harbored by peptides D2A and D2B. Despite this, D2A-Ala could still interact with ␣v␤3 and ␣5␤1, as indicated by its inhibition of the suPAR-integrin co-precipitation (Fig. 6). Strikingly, the introduction of the GAAG epitope converted D2A into a powerful inhibitor of VN-induced cell migration with an extremely low IC 50 , about 10 -20 fM (Fig. 7). In addition, D2A-Ala also inhibited Stat1 activation as well as the appearance of the motile cell morphology and actin cytoskeleton reorganization promoted by VN. Moreover, D2A-Ala inhibited migration induced by other extracellular matrix proteins such as FN and LN, suggesting that it can block other integrins. Therefore, the inhibitory ability of D2A-Ala might reside in its ability to block the formation of and/or disrupt uPAR-integrin complexes such as uPAR-␣v␤3 and uPAR-␣5␤1. Thus, the mutation introduced in the GEEG epitope of D2A destroyed its signaling information without affecting its ability to interact with ␣v␤3 integrin. When investigating the effects of D2A and D2A-Ala on cells that do not express uPAR but express ␣v␤3 and ␣5␤1 (27), we found that peptide D2A failed to stimulate migration of HEK-293 cells, whereas D2A-Ala succeeded in inhibiting VN-induced chemotaxis in the same cells. Thus, the induction of cell migration by D2A requires the presence of uPAR on the cell surface, whereas the inhibition of VN chemotaxis just requires the binding of D2A-Ala to the integrin. These data reveal subtle differences in the agonistic and antagonistic mechanisms of peptides D2A and D2A-Ala and suggest that binding of uPAR can modify the affinity and/or the avidity of integrins.
When considered altogether, the agonistic effects of D2A and the dominant negative behavior of D1HD3-uPAR suggest that uPAR can exert both a positive and a negative regulation on integrins. However, it seems likely that besides the D2A epitope, other sites of interaction between uPAR and integrins exist. These sites might be located in domain 3 because of the apparent dominant negative activity of the D1HD3 variant and because the D1D2 cells were less sensitive to low doses of VN (Fig. 1). Moreover, D2A and D2A-Ala do not share with the ␣325 peptide the property of inhibiting uPAR-mediated cell adhesion on VN (Table III). Therefore, D2A does not represent the binding site of the ␣ integrin subunit onto uPAR. This is a further argument for suggesting that uPAR and the integrins interact through at least two sites.
A model based on our data might explain the activating effects of uPAR on integrins; uPAR can first "anchor" (or contact) integrins through a still undefined epitope, which would allow the subsequent interaction of the D2A sequence with the integrin, reinforcing the interaction and allowing signaling. This is a real possibility now that the uPAR x-ray structure has been elucidated. Indeed, the D2A region lies outside of the site of interaction with uPA and is thus accessible to interaction with other molecules (63). Integrin ligands, such as VN, could bind to the "activated" integrin and induce signaling, although it is possible that VN binding to integrin may be required for the uPAR-integrin interaction. In our model, the D2A region has two functions, integrin binding and signaling, and these functions can be dissociated, since in the context of the entire D2A peptide the GEEG sequence is required for signaling but not for binding (also D2A-Ala can inhibit suPAR-integrin coimmunoprecipitation) (Fig. 6).
Under certain conditions, uPAR may negatively regulate the integrin, as in the case of the D1HD3-uPAR variant. When uPAR contacts the integrin only through binding site(s) located within domain 1 or 3 of uPAR (mutant D1HD3), the integrin appears to be locked in a signaling-inactive state. This block is bypassed by the addition of the D2A peptide.
In summary, in this study we have identified a new chemotactic sequence located in domain 2 of uPAR. Peptide D2A, the synthetic peptide bearing that sequence, binds to integrins and acts through these receptors activating integrin-dependent signaling pathways. The two glutamic acid residues in the GEEG epitope are crucial for signaling, and the GEEG sequence itself has full chemotactic activity. In addition, we have also identified a form of uPAR, D1HD3, that behaves as a dominant negative mutant. These data suggest that uPAR may induce positive or negative regulation of integrin function by inducing different conformations through different multiple interactions. The positive regulation would be achieved by binding to multiple sites, including D2A located on domain 2, and by extrapolation from the D1HD3 results, on domain 3 of uPAR. The negative regulation would rather be obtained by interactions involving site(s) located only on domain 3 (and/or domain 1). Finally, by introducing mutations into the D2A sequence, we have generated a powerful integrin inhibitor, D2A-Ala. This inhibitor is extremely interesting and might be effective against the many physiological and pathological processes in which uPAR and integrins are involved, such as cancer and inflammatory diseases. Further studies will have to validate these points.