Expression of the membrane receptor uPAR induces profound changes in cell morphology and migration, and its expression correlates with the malignant phenotype of cancers. To identify the molecular interactions essential for uPAR function in these processes, we carried out a complete functional alanine scan of uPAR in HEK293 cells. Of the 255 mutant receptors characterized, 34 failed to induce changes in cell morphology. Remarkably, the molecular defect of all of these mutants was a specific reduction in integrin-independent cell binding to vitronectin. A membrane-tethered plasminogen activator inhibitor-1, which has the same binding site in vitronectin as uPAR, replicated uPAR-induced changes. A direct uPAR–vitronectin interaction is thus both required and sufficient to initiate downstream changes in cell morphology, migration, and signal transduction. Collectively these data demonstrate a novel mechanism by which a cell adhesion molecule lacking inherent signaling capability evokes complex cellular responses by modulating the contact between the cell and the matrix without the requirement for direct lateral protein–protein interactions.
The urokinase plasminogen activator (uPA) and its receptor (uPAR) play important roles in physiological processes such as wound healing, inflammation, and stem cell mobilization, as well as in severe pathological conditions such as HIV-1 infection, tumor invasion, and metastasis (Alfano et al., 2002; Sidenius and Blasi, 2003; Gyetko et al., 2004; Selleri et al., 2005; Lund et al., 2006). Besides uPAR's well-established role in the regulation of pericellular proteolysis, it also modulates cell adhesion, migration, and proliferation through interactions with proteins present in the extracellular matrix, including vitronectin (Vn) (Wei et al., 1994; Kjøller and Hall, 2001; Blasi and Carmeliet, 2002). The absence of transmembrane and intracellular domains renders uPAR signaling incompetent, and it is generally believed that signal transduction originating from this receptor must involve lateral interactions with transmembrane proteins. In accordance, uPAR-mediated processes have been proposed to require its interaction with membrane proteins including members of the integrin family (Wei et al., 1996), chemokine receptors (Resnati et al., 2002), and receptor tyrosine kinases (Liu et al., 2002).
The important role of uPAR in tumor cell adhesion, migration, invasion, and proliferation makes this receptor an attractive drug target in cancer treatment; however, this is complicated by the extent of the published uPAR “interactome”. With such a goal in mind, the most important question becomes which of the many molecular interactions are really essential to mediate uPAR function. Recently, the crystal structures of uPAR in complex with a peptide antagonist (Llinas et al., 2005) and with the N-terminal fragment of uPA (Barinka et al., 2006; Huai et al., 2006) were presented, providing the first rational basis toward understanding how uPAR may organize its multiple molecular interactions. Attempts to identify the regions in uPAR involved in the specific interaction with Vn and integrins have been published (Li et al., 2003; Degryse et al., 2005; Chaurasia et al., 2006; Wei et al., 2007); however, the results from these experiments are largely incongruent and remain to be confirmed in independent studies.
In this study we have applied a genetic approach, based on exhaustive mutagenesis and complementation experiments, to determine the essential direct molecular interactions required for uPAR to induce changes in cell morphology and migration. This was achieved in a comprehensive and unbiased way through a complete functional alanine scan of human uPAR in human embryo kidney (HEK) 293 cells. Of the 255 alanine substitutions analyzed, 34 were found to completely, or partially, impair uPAR-induced changes in cell morphology. The molecular defect of all mutants was subsequently shown to be an impaired binding to the somatomedin-B (SMB) domain of Vn. Although the RGD motif in Vn was dispensable for uPAR-dependent cell binding to Vn, it was implicated in subsequent signal transduction and changes in cell morphology, underscoring the importance of integrin receptors downstream of uPAR in these processes. Although integrins are clearly involved in the cellular processes initiated by uPAR expression, we found strong evidence against any functional relevance of a direct molecular interaction between these molecules. First, all the mutants identified in the complete alanine scan have the same molecular defect (impaired Vn binding), suggesting that this is also the only required interaction. Second, alanine substitutions of the published integrin interaction sites in uPAR had no effect on receptor function. Finally, a recombinant GPI-anchored plasminogen activator inhibitor-1 (PAI-1) molecule (PAI-1/GPI), which also binds Vn, mimicked the cellular effects of uPAR expression.
We conclude that a direct Vn interaction is both necessary and sufficient to initiate uPAR-induced changes in cell morphology, migration, and signaling independently of direct lateral protein–protein interactions. The importance of the uPAR–Vn interaction was not cell type–specific, as very similar data were obtained in CHO cells.
Overexpression of uPAR in 293 cells induces signal transduction and changes in cell morphology and migration
It has been previously reported that expression of uPAR modulates the adhesion and motility of 293 cells through interactions with Vn, integrins, and G protein–coupled receptors (Wei et al., 1994, 1996, 1999, 2001; Degryse et al., 2005; Gargiulo et al., 2005; Chaurasia et al., 2006;). In accordance, we find that expression of human uPAR in HEK293 Flp-In T-REx (293) cells induces changes in cell morphology, migration, and signaling (Fig. 1). The morphological changes include a general flattening of the cells, reduced cell–cell contact, disappearance of membrane ruffles, and formation of extensive lamellipodia (Fig. 1 A, left panels), as well as a complete reorganization of the matrix-proximal F-actin cytoskeleton (Fig. 1 A, right panels). The changes in cell morphology reflect the cell motility induced by uPAR expression in these cells (Fig. 1 B and Video 1). Migratory (Nguyen et al., 1999) and proliferative (Aguirre Ghiso et al., 1999) signaling downstream of uPAR involves activation of the Ras/MAPK signaling pathway, and in accordance we found that uPAR-expressing cells had approximately threefold increased ERK1/2 activation as compared with mock-transfected cells (Fig. 1 C).
uPAR-induced changes in cell morphology occur through Vn and require integrin-dependent RGD engagement
As previously described (Wei et al., 1994, 1996; Gargiulo et al., 2005), expression of uPAR in 293 cells resulted in a strong increase in cell adhesion to Vn, whereas adhesion to other ECM proteins including fibronectin, type-1 collagen, and laminin was unaffected (Fig. 2 A). The binding sites for uPAR and integrins on Vn are distinct, with uPAR recognizing the SMB domain (Deng et al., 1996a), and integrins the adjacent RGD motif. To address the relative importance of uPAR versus integrin binding to Vn, we performed adhesion assays using a recombinant N-terminal fragment of Vn (Vn(1–66)) that includes both the SMB domain and the RGD motif. In addition, we monitored adhesion to two variants of this fragment where either most of the SMB domain had been removed (Vn(40–66), representing essentially RGD alone), or the RGD sequence had been mutated into RAD (Vn(1–66)RAD) (Fig. 2 A). Adhesion of 293/uPAR cells was supported as long as the SMB domain was present and was not affected by mutating the RGD motif. Mock-transfected 293 cells adhered poorly to all of these substrates and not at all to Vn(1–66)RAD (Fig. 2 A). Profiling of Vn receptor expression by the 293 cells and antibody inhibition experiments (Fig. S1, A and B) indicated that the weak RGD-dependent adhesion of mock-transfected 293 cells to Vn was mediated by the αvβ5 integrin. Collectively, the adhesion data demonstrate that uPAR promotes cell adhesion to Vn through a direct interaction with the SMB domain, and not by inducing integrin binding to the RGD motif.
Although the RGD motif in Vn is dispensable for adhesion of 293/uPAR cells, it is at least partially required for the subsequent changes in cell morphology and signal transduction (Fig. 2, B and C). Although 293/uPAR cells seeded on Vn(1–66) undergo marked changes in morphology, including cell flattening and extensive lamellipodia formation, these changes are absent when seeded on Vn(1–66)RAD, where only a round adhesion patch can be observed under the cell body. Under the same conditions, mock-transfected cells retain a rounded cell body, fail to form lamellipodia, and only extend thin membrane protrusions with limited matrix contact as evidenced also by the weak adhesive strength of these cells (Fig. 2 A). Mock-transfected cells remained completely round on Vn(1–66)RAD in accordance with the adhesion data. When plated on Fn, no differences in cell adhesion and morphology were observed comparing 293/mock and 293/uPAR cells (unpublished data). The level of ERK1/2 activation was higher (Fig. 2 C) when 293/uPAR cells were seeded on Vn(1–66) as compared with Vn(1–66)RAD, suggesting that also uPAR-induced downstream signaling is at least partially integrin dependent.
In summary, these data show that the ability of uPAR to induce changes in 293 cell morphology is a two-step process in which uPAR, through the direct interaction with the SMB domain of Vn, triggers cell attachment to the matrix. Subsequently, this initial adhesion is followed by engagement of integrins, possibly αvβ5, with matrix Vn, triggering changes in cell morphology, migration, and signal transduction.
A direct Vn interaction is fundamental for the ability of uPAR to induce changes in cell morphology
In addition to the direct interaction with matrix Vn, extensive data suggests that the ability of uPAR to modulate cell adhesion, migration, and proliferation requires a complex network of lateral interactions with a variety of membrane proteins including integrins, receptor tyrosine kinases, and chemokine receptors. To address the nature of these lateral uPAR interactions in a comprehensive and unbiased fashion, we next conducted a complete functional alanine scan of uPAR in 293 cells. Receptor-induced changes in 293 Flp-In T-REx cells were chosen for this purpose for two reasons: (1) as shown in Fig. 1, uPAR expression in these cells causes very evident changes in cell morphology which can be scored easily by phase-contrast microscopy; and (2) in addition, the Flp-In system generates pools of isogenic transfectants carrying a single copy of the expression cassette, thus eliminating potential artifacts caused by clonal differences or heterogeneous expression levels.
We generated 255 single amino acid substitution mutants of uPAR by site-directed mutagenesis (all residues of mature uPAR excluding 28 cysteines) and expressed these in 293 Flp-In T-REx cells. As expected, the cell surface expression of all mutant receptors was similar and comparable to that of wild-type uPAR as evaluated by FACS analysis and immunoblotting (Table I; Fig. S1, C and D; and Table S1 for a complete listing).
Of the 255 receptor variants analyzed, 34 were found to fully or partially impair the ability of uPAR to induce changes in cell morphology (listed in Table I). None of these residues coincided with previously identified interaction sites for Vn or integrins (Li et al., 2003; Degryse et al., 2005; Chaurasia et al., 2006; Wei et al., 2007), but several are known to be involved in uPA binding (Gardsvoll et al., 2006). As the 293 cells do not express uPA (Wei et al., 1994), the molecular defect of these mutants could not be explained by available evidence. Five of the mutants were found to suffer from severe folding problems as evidenced by the presence of high molecular weight covalent aggregates in immunoblotting experiments, and were not analyzed further (Table I and Fig. S1 D).
Because the initial step in uPAR-induced signaling and cell morphology changes is triggered by the direct binding to the SMB domain of Vn (Fig. 2), we next analyzed the adhesive properties of cells expressing the remaining 29 mutant receptors with impaired ability to change cell morphology (Fig. 3). Remarkably, all of the mutant receptors displayed an impaired ability to promote cell binding to Vn(1–66)RAD (Fig. 3 A). There was a strong correlation between the ability to induce changes in cell morphology and the reduction in uPAR binding to the SMB domain of Vn (Fig. 3 B), demonstrating a direct link between these parameters. Receptor mutants that did not impair the ability of uPAR to induce changes in cell morphology (S56A) as well as receptors containing mutations in published integrin interaction sites (E134A/E135A, S245A, H249A, and D262A) all displayed normal adhesion to the same substrate. All cell lines adhered equally well to Fn (Fig. S2 A). Ligand binding has been demonstrated to enhance uPAR-dependent cell adhesion to Vn, and we therefore also performed the adhesion experiments in the presence of exogenously added pro-uPA (Fig. 3 C). Under these conditions the adhesion of most of the mutants was restored to that of wild-type uPAR, suggesting that ligand binding may compensate for the molecular defect of most of these mutants. The rescued Vn adhesion was also associated with restored uPAR cell morphology when cells were cultured in the presence of pro-uPA (Fig. S2 B). For some mutants (in particular W32A and R91A) cell adhesion and uPAR morphology was only partially recovered by the addition of pro-uPA, suggesting that these may be key residues for the uPAR–Vn interaction.
Identification of the Vn interaction site in uPAR
The location of the mutated residues on the published crystal structure of uPAR combined with the relatively small size of the SMB domain suggest that only part of the identified residues actually engage SMB directly. To identify this subset we next generated and expressed soluble variants of the mutant receptors and assessed their binding to immobilized Vn (Sidenius et al., 2002). As we were seeking residues in the direct molecular interaction interface, the analysis was restricted to the 19 mutant receptors where the substituted amino acid exhibits a surface exposure area on the published crystal structure above 15Å2. In these binding assays, five mutants (W32A, R58A, I63A, R91A, and Y92A) were found to display greater than fivefold reduction in binding to immobilized Vn (Fig. 3 D) while still having a normal binding to immobilized pro-uPA (unpublished data), suggesting that these residues are involved in the direct interaction with Vn. The impaired Vn binding of these mutants was observed also in the presence of excess pro-uPA, suggesting (Sidenius et al., 2002) that the defect of these mutants is not related to impaired uPAR dimerization (unpublished data).
On a molecular surface representation of the uPAR structure (Fig. 3 E), W32 and R91 define a composite epitope with residues located in both domain 1 and 2 of uPAR. The residues affected by the R58A, I63A, and Y92A substitutions are all located close to these residues, suggesting that they may also engage Vn directly. The Vn-binding epitope is located distal to the membrane anchorage position (indicated in cyan) and lies on the “top-back” of the molecule with respect to the central uPA-binding cavity. Although we currently do not know the precise molecular explanation for the failure of the remaining mutant receptors to bind to Vn in the absence of pro-uPA, it should be stressed that from a functional point of view their defect is identical to that of W32A and R91A, i.e., impaired Vn binding.
The binding to matrix Vn is the only direct uPAR interaction required to trigger changes in cell morphology, migration, and signaling
The fact that all the mutants with impaired ability to induce changes in cell morphology share the same molecular defect in RGD-independent Vn binding suggests that this may also be the only important uPAR interaction required to trigger changes in cell morphology, migration, and signaling. This is particularly intriguing because the ability of uPAR to modulate cell adhesion and migration has been shown previously to require direct uPAR–integrin interactions (Wei et al., 1996, 2007; Degryse et al., 2005; Gargiulo et al., 2005; Chaurasia et al., 2006). Individual (not depicted) or combined alanine substitutions (E134A/E135A/S245A/H249A/D262A, termed uPAR/Int−) in the published integrin binding sites in uPAR had no effect on the ability of uPAR to induce changes in cell morphology and F-actin cytoskeleton (Fig. 4 A), Vn-adhesion (Fig. 4 B), signaling (Fig. 4 C), and cell migration (Fig. S3 A). On the other hand, a single alanine substitution in the Vn-binding epitope (W32A) completely abolished all these cellular effects of uPAR expression. Although these data argue strongly against any functional relevance of direct integrin interaction(s) for the biological activity of uPAR in our cell system, the single amino acid substitution strategy does not allow us to exclude the existence of additional and functionally redundant binding sites for integrins in uPAR.
To directly address the possibility that no lateral interactions are required for the observed biological effects of uPAR expression, we conducted a genetic complementation experiment. PAI-1 shares no sequence or structural homology with uPAR; however, both molecules bind the SMB domain of Vn (Deng et al., 1996a). The Vn-binding site in PAI-1 is located distal from the C terminus to which we attached the GPI-anchoring signal of uPAR (Fig. S3 B). The resulting membrane-tethered PAI-1 (termed PAI-1/GPI) was efficiently expressed on the cell surface (Fig. S3 C) and induced strong cell binding to Vn(1–66)RAD (Fig. 4 B). Remarkably, the expression of PAI-1/GPI in 293 cells induced changes in cell morphology, adhesion, and ERK1/2 activation comparable to those induced by uPAR (Fig. 4, A–C). The ability of PAI-1/GPI to induce these changes in cell morphology strictly required its interaction with Vn, as the introduction of a triple alanine substitution that efficiently disrupts its interaction with Vn (R103A/M112A/Q125A, termed PAI-1/GPI/Vn−) (Jensen et al., 2004) failed to induce any of these changes (Fig. 4, A–C).
Together these data demonstrate that a direct uPAR–Vn interaction is both required and sufficient to trigger changes in cell morphology, migration, and signal transduction. Other crucial molecular events downstream of this interaction, including integrin engagement of the RGD motif in Vn, do not require any additional direct uPAR interactions.
A direct Vn interaction is required and sufficient for uPAR to induce changes in cell morphology, adhesion, and signaling in CHO cells
The central importance of the Vn interaction in the biology of uPAR is not cell type specific. When seeded at low density, CHO cells form colonies with an epithelial-like morphology characterized by tight cell–cell contact and a highly defined colony contour (Fig. 5 A, left panels). In sharp contrast, CHO cells expressing high levels of uPAR form sparse colonies with no cell–cell contact. The expression of uPAR is also associated with marked changes in the organization of matrix-proximal F-actin cytoskeleton (Fig. 5 A, right panels). In analogy to the 293 cell line, uPAR expression was associated with a strong and specific increase in RGD-independent cell adhesion to Vn (Fig. 5 B) and caused increased ERK1/2 activation (Fig. 5 C).
Despite the obvious differences between the morphological features of CHO and 293 cells, the direct Vn interaction was found to be required for the cellular effects of uPAR also in this cell line (Fig. 5). Alanine substitution in the Vn binding site (W32A) abolished the ability of uPAR to induce scattering of cells, changes in F-actin cytoskeleton, and ERK1/2 activation. As observed with the 293 cells, the exogenous addition of pro-uPA rescued the uPAR cell morphology in a mutant-dependent manner (Fig. S4). Furthermore, single alanine substitutions (not depicted) or combinations of substitutions in the published integrin interaction sites had no effect on uPAR function. Finally, the PAI-1/GPI receptor replicated uPAR-induced changes in a Vn-dependent manner also in this cell line.
A direct uPAR–Vn interaction is required for uPA-induced cell changes
All the mutants identified in this study fail completely or partially to induce cell adhesion to Vn in the absence of pro-uPA (Fig. 3 A), yet their adhesive properties are in most cases restored to normal levels in the presence of exogenous pro-uPA (Fig. 3 C). As most of these mutants bind pro-uPA normally (unpublished data), they provide an important tool to dissect the importance of the direct Vn interaction in signal transduction induced by uPA binding as well as for the analysis of the cellular processes occurring subsequent to Vn engagement of the receptor. In the absence of exogenous pro-uPA expression of W32A and T54A mutant receptors both fail to induce cell binding to Vn (Fig. 6 A) and both display a low level of ERK1/2 activation similar to that of mock-transfected cells (Fig. 6 B). In the presence of the pro-uPA, cells expressing the T54A receptor, but not the W32A receptor, display a fully restored Vn binding (Fig. 6 A) and ERK1/2 activation state (Fig. 6 B). For comparison the Vn binding and ERK1/2 activation is constitutively high in cells expressing wild-type uPAR and only moderately affected by the addition of uPA. As the wild-type, W32A and T54A receptors all bind pro-uPA equally well these data document that uPA signaling to ERK1/2 occurs through the induction of uPAR binding to Vn.
The recovery of Vn binding and ERK1/2 activation induced by pro-uPA on cells expressing uPAR/T54A is paralleled by rapid changes in cell morphology and focal contact turnover as illustrated by the translocation of paxillin-GFP into newly formed lamellipodia (Fig. 6 C and Video 2) and by the scattering of preformed CHO colonies (Fig. 6 D and Video 3), which resembles a bona fide epithelial–mesenchymal transition. None of these uPA-induced changes were observed in cells expressing uPAR/W32A, underscoring the fundamental requirement for a direct uPAR–Vn interaction in these cellular processes.
Using a systematic and unbiased approach, we demonstrate that a direct interaction between cell surface uPAR and the extracellular protein Vn is required for the ability of uPAR to modulate changes in cell morphology, migration, and signal transduction in two different cell lines. This conclusion is based on two central observations. All of the alanine substitutions that affect the ability of uPAR to induce changes in cell morphology and migration directly affect uPAR-mediated cell binding to Vn independently of integrin binding to the RGD motif of Vn. Furthermore, a GPI-anchored PAI-1 molecule, which shares only the binding site in the SMB domain of Vn with uPAR, induces cellular changes that are virtually identical to those induced by uPAR.
It has been extensively documented that the ability of uPAR to modulate cell adhesion, migration, and proliferation occurs through the regulation of integrin signaling (Wei et al., 1996; Degryse et al., 2005; Gargiulo et al., 2005; Chaurasia et al., 2006). Our data are fully in accordance with this view as we observe that the RGD motif in Vn is involved in uPAR-mediated downstream signal transduction and changes in cell morphology. Nevertheless, our data argue firmly against a functionally relevant direct interaction between uPAR and integrins. Alanine substitutions in the previously published integrin interaction sites, alone or in combination, have no effect on uPAR function in both cell lines analyzed. We found no mutants with normal RGD-independent Vn binding and impaired ability to induce changes in cell morphology. Finally, an engineered PAI-1/GPI molecule, which recapitulated the membrane–ECM interaction induced by uPAR, was sufficient to induce similar changes. Our data do not rule out direct uPAR–integrin interactions; however, they do allow us to conclude that if these interactions occur they are of little or no functional importance with respect to uPAR-induced changes in cell morphology, migration, and signaling, at least in 293 and CHO cells.
Based on binding assays using purified proteins and the presence of pro-uPA, we were able to identify five residues (W32, R58, I63, R91, and Y92) that, when mutated into alanine, impair the ability of uPAR to interact with Vn. Noticeably, when these five mutants were tested for their ability to mediate cell attachment to Vn in the presence of uPA, only two of them (W32 and R91) were still significantly impaired. These two residues are therefore likely to represent the most important direct interaction site for Vn in uPAR. That we have identified the bona fide Vn-binding epitope in uPAR is supported by several structural, functional, and evolutionary observations: the binding epitope is located distal to the GPI anchor, as would be expected for a membrane receptor that interacts with a component of the ECM. The location of the binding site is different from the location of the uPA-binding site, allowing for the simultaneous binding of these two molecules. The composite epitope contains residues both in domain 1 (W32, R58, and I63) and domain 2 (R91 and Y92), explaining why proteolytic cleavage between these two domains abolishes binding (Høyer-Hansen et al., 1997; Sidenius and Blasi, 2000). All the residues in the identified epitope are exposed on the crystal structure of uPAR. Finally, all the residues are highly conserved between different species.
How do we explain the mutants with impaired cell binding to Vn but normal binding when purified proteins are used? In contrast to cell binding assays, the binding of soluble uPAR to immobilized Vn cannot be measured in the absence of uPA (Sidenius et al., 2002). Several of the mutants (R53, L55, Y57, L113, E120, L123, R145, and G146) are located in the area of contact between domain 1 and 2 of uPAR, suggesting that the correct “docking” of these two domains is critical for Vn binding. Binding of uPA, which engages extensive areas of both domain 1 and 2 of uPAR (Gardsvoll et al., 2006; Huai et al., 2006), may compensate for this defect by enforcing the correct Vn binding competent alignment of these two domains. In addition, the N52 and T54 mutants suggest that in the absence of uPA, glycosylation of uPAR at N52 is required for Vn binding. Particularly intriguing is the finding that several of the mutant residues (L31, R53, L55, Y57, T64, L66, and E68) have their side chains exposed inside the uPA-binding cavity of uPAR. The fact that these residues were identified in a screen where uPA is absent suggests that something else, which is required for the ability of uPAR to bind Vn, is engaging this cavity in the absence of uPA and that this engagement is required for uPAR binding to Vn. The identity of this interaction has yet to be established, but based on the finding that uPAR binding to Vn involves receptor dimerization (Sidenius et al., 2002; Cunningham et al., 2003), it is tempting to speculate that this interaction partner may be uPAR itself.
Interestingly, part of the Vn binding site identified in this study (R91 and Y92) maps to the previously identified chemotactic epitope SRSRY located in the linker region connecting domain 1 and 2 of uPAR (Fazioli et al., 1997). However, it seems unlikely that this chemotactic activity is important in our cell systems, as most of the critical residues identified in this study do not map to this region, yet they do share the same molecular defect. Furthermore, PAI-1 does not contain an SRSRY sequence, yet does it mimic uPAR function. Interestingly, peptides covering this region (uPAR aa 84–95) have been shown to inhibit uPAR-dependent cell adhesion to Vn (Liang et al., 2003).
Although PAI-1/GPI closely mimic uPAR function with respect to the changes in cell adhesion, morphology, and signal transduction presented in this work, this molecule should be used as a uPAR mimic with caution, as it has cellular effects which are not seen with uPAR. These effects include a pronounced tendency of the CHO cells to “lose pieces”, possibly because of an impaired lamellipodia retraction, as well as an impaired cytokinesis in the 293 cells (unpublished data). These effects are likely due to the fact that PAI-1 binds with much higher affinity to Vn than uPAR (Okumura et al., 2002).
In line with several other studies (Jo et al., 2003; Gargiulo et al., 2005; Chaurasia et al., 2006; Mazzieri et al., 2006), we observe that overexpression of uPAR leads to increased ERK1/2 activation in both cell lines analyzed. A direct uPAR–Vn interaction is required for this activation, as Vn binding–deficient uPAR mutants display levels of active ERK1/2 comparable to those of mock-transfected cells. Interestingly, uPA binding to uPAR also leads to ERK1/2 activation in different experimental systems (Nguyen et al., 1998; Aguirre Ghiso et al., 1999), suggesting that both overexpression of the receptor and ligand binding induces the same signal transduction pathway(s), possibly through a common molecular mechanism. In line with other studies we have used systems with high-level expression of uPAR (∼3 × 106 binding sites per cell in both cell lines as determined by binding assays using Eu3+-labeled pro-uPA; unpublished data), and under these conditions cell adhesion to Vn is independent of pro-uPA binding to uPAR (Wei et al., 1994, 1996; Cunningham et al., 2003). At physiological expression levels, uPAR-dependent cell adhesion to Vn requires uPA binding (Waltz and Chapman, 1994; Sidenius and Blasi, 2000; Sidenius et al., 2002). Consequently, it appears likely that uPA binding may actually induce “Vn signaling” by stimulating uPAR binding to matrix Vn. In support of this possibility we show that there is a strict correlation between the ability of pro-uPA to promote Vn binding and to induce ERK1/2-activation and changes in cell morphology (Fig. 3, Fig. 6, Fig. S2 B, and Fig. S4).
What is the molecular mechanism of uPAR-induced “Vn signaling”? The TIR-FM images and adhesion assays clearly demonstrate that expression of uPAR dramatically increases the contact between the cell and the ECM. Although uPAR does so passively by a physical interaction with matrix Vn, it will invariably bring all matrix receptors present in the plasma membrane in closer contact with their extracellular ligands. In the 293 cell line the uPAR-induced increase in cell/matrix contact triggers subsequent changes in cell morphology and signaling that are at least partially mediated by an RGD-dependent integrin, possibly αvβ5. Although it is possible that uPAR and integrins bind contemporarily to the same Vn molecule, this does not seem to be required because PAI-1, which is known to block integrin binding to Vn (Stefansson and Lawrence, 1996), induces similar changes when tethered to the cell membrane.
Because the Vn-binding form of uPAR is associated with membrane structures known as lipid rafts (Cunningham et al., 2003), the engagement of Vn by uPAR will bring these membrane domains in contact with the matrix. Lipid rafts are enriched in signal transduction molecules (Simons and Toomre, 2000) and the uPAR–Vn interaction will thus promote the tight encounter between the ECM, its different cell surface receptors, and the signal transduction machinery. As different cells have different repertoires of membrane receptors and different matrices have different repertoires of ligands, our data suggest that the single interaction between uPAR and Vn may be responsible for many of the proteolysis-independent biological effects initiated by uPAR.
It is important to underscore that the conclusions of this work cannot be directly extrapolated to uPAR functions different from its ability to induce changes in cell morphology, basal cell migration, and ERK1/2 activation. For example, we have not assayed the ability of the different uPAR mutants to modulate directional cell migration and cell proliferation. Consequently, our data does not exclude nor confirm the requirement for additional direct uPAR interactions in these processes. Nevertheless, the identification of the Vn binding site presented here provides the first tool to directly address the importance of the Vn interaction in any experimental system where uPAR is expressed by transfection and/or infection. The single alanine substitution W32A efficiently disrupts the interaction with Vn while maintaining normal pro-uPA binding.
Materials And Methods
CHO Flp-In and HEK293 Flp-In T-REx cells, expression vectors pcDNA5/FRT/TO and pOG44, zeocin, blasticidin S HCl, and Ham's F12 medium were from Invitrogen. Dulbecco's modified eagle medium (DME) and αMEM were from BioWhittaker. PBS, trypsin, glutamine, penicillin, and streptomycin were obtained from EuroClone, and fetal bovine serum (FBS) was from HyClone. Non-tissue culture plates were from Falcon Becton Dickinson. Tetracycline, poly-l-lysine, laminin (from Engelbreth-Holm-Swarm murine sarcoma), phalloidin-FITC, and CHO protein-free culture medium were from Sigma-Aldrich. FuGENE 6, fibronectin, and Hygromycin B were from Roche. Type-1 collagen from rat tail was from BD Biosciences. Urea-purified Vn was obtained from Promega. Pro-uPA was provided by Dr. Jack Henkin (Abbott Laboratories, Abbott Park, IL). Antibodies against total and phosphorylated ERK1/2 were from Cell Signaling Technology. Blocking antibodies against αvβ3 (LM609) and αvβ5 (P1F6) integrins were from Immunological Sciences. Monoclonal antibody against β1 integrin (mAb 13) was from BD Biosciences. Monoclonal antibodies against human uPAR and the PAI-1 polyclonal antibody were provided by Dr. Gunilla Høyer-Hansen (Finsen Laboratory, Copenhagen, Denmark) and Dr. Peter Andreasen (Århus University, Århus, Denmark), respectively. Goat anti–mouse and goat anti–rabbit Ig F(ab′)2-FITC were from Jackson ImmunoResearch Laboratories.
Expression vector construction
The expression vector encoding full-length uPAR was generated by amplifying (oligos uPARu and uPARd) the human uPAR cDNA (Roldan et al., 1990) and cloned in the pcDNA5/FRT/TO vector using Kpn1/Not1. The expression vector encoding a GPI-anchored PAI-1 molecule was generated by uniting the complete PAI-1 coding region (aa −22 to 380), a short linker sequence (AlaGlyAlaGlyAlaGlyLys), and the GPI-anchoring signal sequence of uPAR (aa 276–313). The PAI-1 coding region and the linker region were generated by amplifying a HT1080 cDNA with oligos PAIu and PAId, and the GPI-anchor signal by amplifying the uPAR cDNA with oligos GPIu and uPARd. The PAI-1/linker fragment (Kpn1–HindIII) and the GPI fragment (HindIII–Not1) were assembled into Kpn1–Not1 digested pcDNA5/FRT/TO. The Vn(1–66)/Fc fusion protein was generated by attaching the first 66 amino acids of mature human Vn to a constant region of a human IgG. Primers and templates were as follows: a vector containing part of the Vn cDNA (pTrx-Vn(1–97) [Deng et al., 1996b]) amplified with oligonucleotides Vn(1–66)u and Vn(1–66)d and a vector containing an Fc region of human IgG (pIG-1 [Fawcett et al., 1992]) amplified with primers FcU and FcD. The Vn(1–66) (Kpn1–Xho1) and Fc (Xho1–Not) fragments were assembled in Kpn1–Not1 digested pcDNA5/FRT-TO. Variants of this construct where the SMB domain was deleted (Vn(40–66)/Fc) or where only the signal peptide was retained (Fc) were generated as above, substituting the oligo Vn(1–66)u with oligos Vn(40–66)u and SigU, respectively. The correct sequences of the complete coding regions of all constructs were verified by sequencing. All oligonuclotide sequences can be found in Table S1 . The expression vector encoding paxillin-GFP was provided by Dr. Alan F. Horwitz (University of Virginia, Charlottesville, VA; Laukaitis et al., 2001).
Alanine substitutions were generated by site-directed mutagenesis according to the QuikChange protocol (Stratagene). Alanine residues present in the uPAR sequence were substituted with glycine. Multiple rounds of mutagenesis were used to generate constructs carrying multiple substitutions. Expression vectors encoding soluble uPAR variants were generated by a second round of site-directed mutagenesis changing codon 284 of uPAR into a stop-codon (oligo pair A284Stop).
Parental HEK293 Flp-In T-REx cells were cultured in DME supplemented with 10% FBS, 100 U/ml penicillin, 100 U/ml streptomycin, 5 mM glutamine, 15 μg/ml blasticidin, and 100 μg/ml zeocin at 37°C in 5% CO2. Parental CHO Flp-In cells were cultured in Ham's F12 supplemented as above, but without blasticidin. Transfections were performed with a 1:10 ratio of pcDNA5/FRT/TO-based expression vector and pOG44 using FuGENE 6, and stable HEK293 Flp-In T-Rex and CHO Flp-In transfectants were selected in medium lacking zeocin using 150 μg/ml and 300 μg/ml hygromycin B, respectively. Expression in HEK293 Flp-In T-REx was induced by adding 1 μg/ml tetracycline to the medium overnight.
Expression and purification of recombinant proteins
For the production of soluble uPAR mutants, semi-confluent CHO Flp-In stably transfected with the relevant expression vectors was washed with PBS and incubated for 7–10 d in CHO protein-free medium. The concentration of suPAR in the conditioned media was typically higher than 100 nM as determined by ELISA. The conditioned medium was used for in vitro binding assays without further purification. Recombinant Fc-fusion proteins were expressed in the same way, but were purified by standard Protein A affinity chromatography.
Cell surface expression of integrins, uPAR, and PAI-1/GPI were analyzed by flow cytometry. αvβ3, αvβ5, and β1-integrins were detected using primary antibodies (10 μg/ml). PAI-1/GPI was detected using a polyclonal antibody (2 μg/ml) and uPAR was detected using the monoclonal antibodies R2 and R4, as well as a polyclonal rabbit anti-uPAR antibody (all 2 μg/ml). Cells were stained with appropriate secondary FITC-labeled antibody (diluted 1:100) and analyzed by flow cytometry (FACSCalibur; BD Biosciences).
Scoring of cell morphology
For the scoring of cell morphology, changes induced by the different mutant receptors in semi-confluent cells were inspected by phase-contrast microscopy by three to five independent, trained observers. The observers were asked to score the morphology of the cells in comparison to uPAR and mock-transfected cells analyzed in parallel. A cumulative score of “uPAR” or “mock” morphology was assigned to clones where unanimous scoring by the individual observers was achieved. A cumulative score of “intermediate” morphology was assigned to clones where the majority (two or more) of the observers gave an “intermediate” score.
Adhesion assays were performed as described previously (Cunningham et al., 2003). In brief, cells were harvested, counted, and allowed to adhere in the presence or absence of 10 nM pro-uPA for 30 min at 37°C to 96-well plates (3 × 104 cells/well) coated with the different substrates as indicated. After washing, the adherent cells were fixed, stained with crystal violet, and quantified by measuring the absorbance at 540 nm. Coatings were as follows: 100 μg/ml poly-l-lysine, 10 μg/ml fibronectin, 10 μg/ml type-1 collagen, 20 μg/ml laminin, and 1 μg/ml vitronectin (Vn). Vn(1–66), Vn(1–66)RAD, Vn(40–66), and Fc were all coated at 1 μg/ml.
Total internal reflection fluorescence microscopy (TIR-FM)
Cells were plated on glass coverslips and allowed to adhere overnight in the presence (293 cells) or absence (CHO cells) of 1 μg/ml tetracycline. Cells were washed in PBS, fixed in 4% paraformaldehyde, quenched with 30 mM NH4Cl, permeabilized with 0.1% Triton X-100, 0.2% BSA in PBS, and blocked with 2% BSA in PBS. Labeling of F-actin was performed using 0.2 μg/ml phalloidin-FITC, 1% BSA in PBS. TIR-FM imaging of cells was performed using a Biosystem TIR-FM workstation (Olympus) based on the Cell^R Imaging System (Olympus). A 488-nm argon laser was coupled in an inverted epifluorescence motorized microscope (IX81; Olympus) and focused at an off-axis position of the objective back focal plane. Cells plated on glass coverslips were viewed through a high-aperture 60× objective lens (UIS2 60× TIRFM PlanApo N, NA 1.45; Olympus) with an additional 1.6× lens. Images (16-bit depth) were acquired using an Orca-ER (C4742-80) Cooled CCD digital camera (Hamamatsu). Coverslips were directly inserted into an Attofluor Cell chamber (Invitrogen), rinsed with PBS (to maintain the requested difference of refractive indexes) and subjected to TIRF analysis. Time-lapse TIR-FM imaging was performed as above, with the exception that fixation/permeabilization/staining was omitted and that the cells were maintained at 37°C in normal growth medium throughout the recordings (every 15 s for a total of 15 min). Adjustment of brightness/contrast and smoothening of images was done using ImageJ 1.38i and always applied to the entire image.
Phase-contrast and time-lapse imaging
Phase-contrast and time-lapse live-cell imaging was performed at 37°C, 5% CO2 with an inverted microscope (IX70; Olympus) equipped with an incubation chamber (Solent Scientific). Cells were always plated in serum-containing growth medium unless otherwise stated, and viewed through 20× (LCPlanFl, NA 0.4 Ph1, Olympus) or 40× (UIS2 40× UPlanFLN, NA 0.75 Ph2; Olympus) objective lenses with an additional 1.5× lens. All time-lapse acquisitions were performed using the 20× objective. The acquisition system includes a digital camera (Sensys; Roper Scientific) and System Control Software Metamorph 7.0r4 (Universal Imaging). Adjustment of brightness/contrast and smoothening of images was done using ImageJ 1.38i and always applied to the entire image. Cell migration speed was quantified with ImageJ 1.38i using the plug-in “manual tracking”. In each experiment, 20 randomly chosen cells were tracked and their average migration speed throughout the experiment was calculated.
Binding assays of purified proteins
In vitro binding assays were performed essentially as described previously (Sidenius et al., 2002) using Nunc Maxisorb black-well plates for the detection of bound suPAR with an Eu3+-labeled polyclonal anti-uPAR antibody (0.5 μg/ml) followed by the measurement of time-resolved fluorescence. Binding to immobilized Vn was measured using an excess of suPAR (80 nM) in the presence of a limiting concentration of pro-uPA (20 nM). All measurements were done in triplicate, and the specific binding calculated by subtraction of the nonspecific binding to BSA-coated wells.
Cell lysis and Western blotting
Cells were washed and lysed directly on the culture dish in ice-cold lysis buffer (25 mM Tris, pH 7.6, 150 mM NaCl, 1% Triton X-100, protease inhibitor cocktail [Complete-EDTA-free], 1 mM PMSF, 1 mM EDTA, 1 mM NaF, and 1 mM Na3VO4). After clarification by centrifugation (16,000 rcf, 15 min, 4°C) the total protein content was determined using the DC-Protein assay (Bio-Rad Laboratories) with BSA as standard. Equal amounts of total protein were separated by SDS-PAGE and probed as indicated.
The significance of differences in cell adhesion and ERK1/2 activation states were established using t test (paired, two-tailed) after log-transformation of the data from independent experiments.
The surface-accessible area of the residues of uPAR was calculated based on the A-chain of pdb-entry 1YWH (Llinas et al., 2005). Calculations were performed using the program AREAIMOL from the CCP4 suite of programs (Collaborative Computational Project, Number 4, 1994) with a probe-radius of 1.4 Å.
Online supplemental material
Fig. S1 (A and B) shows the profiling of Vn receptor expression and functionality in 293 cells. Fig. S1 (C and D) provides an example of comparable expression levels of the different mutants expressed in 293 cells. Fig. S2 A demonstrates that adhesion to Fn is unaffected by the expression of Vn-deficient uPAR mutants in 293 cells. Fig. S2 B and Fig. S4 show the rescue of morphological changes induced by exogenous pro-uPA for some uPAR mutants in 293 and CHO cells, respectively. Fig. S3 A indicates the migration speed of 293 cells expressing either Vn or integrin-deficient uPAR mutants. Fig. S3 B illustrates the crystal structure of the PAI-1–SMB complex and the position of the attached GPI anchor. Fig. S3 C illustrates the comparable cell surface expression level of uPAR and PAI-1. Table S1 lists information on all the mutants generated in this study, as well as all sequences of all oligonucleotides used in the study. Video 1 illustrates the increase in cell migration upon uPAR expression in 293 cells. Video 2 shows the redistribution of paxillin-GFP upon uPA-mediated rescue of uPAR/Vn binding in CHO cells. Video 3 shows the scattering effect upon uPA-mediated rescue of uPAR/Vn binding in preformed CHO colonies.
We thank Danika Trautmann for help with mutagenesis and Thomas R. Schneider for help with the evaluation of accessible surface areas. Massimiliano Garrè, Dario Parazzoli, and Mario Faretta from the Imaging Unit at IFOM are acknowledged for their assistance with microscopy. Dr. Francesco Blasi is acknowledged for helpful discussions.
This work was supported by grants from the European Union to Francesco Blasi (Integrated Project Cancerdegradome: CT 2003-503297 and Network of Excellence MAIN: CT 2003-502935) and from the Italian Association for Cancer Research (AIRC) and the Cariplo Foundation to N. Sidenius.
Abbreviations used in this paper: GPI; glycosylphosphatidylinositol, PAI-1, plasminogen activator inhibitor-1; SMB, somatomedin-B domain; uPA, urokinase plasminogen activator; uPAR, urokinase plasminogen activator receptor; Vn, vitronectin.