The Contributions of Integrin Affinity and Integrin-Cytoskeletal Engagement in Endothelial and Smooth Muscle Cell Adhesion to Vitronectin*

The serine proteinase inhibitor, plasminogen activator inhibitor type-1 (PAI-1), binds to the adhesion protein vitronectin with high affinity at a site that is located directly adjacent to the vitronectin RGD integrin binding sequence. The binding of PAI-1 to vitronectin sterically blocks integrin access to this site and completely inhibits the binding of purified integrins to vitronectin; however, its inhibition of endothelial and smooth muscle cell adhesion to vitronectin is at most 50-75%. Because PAI-1 binds vitronectin with ∼10-100-fold higher affinity than purified integrins, we have analyzed the mechanism whereby these cells are able to overcome this obstacle. Our studies exclude proteolytic removal of PAI-1 from vitronectin as the mechanism, and show instead that cell adhesion in the presence of PAI-1 is dependent on integrin-cytoskeleton engagement. Disrupting endothelial or smooth muscle cell actin polymerization and/or focal adhesion assembly reduces cell adhesion to vitronectin in the presence of PAI-1 to levels similar to that observed for the binding of purified integrins to vitronectin. Furthermore, endothelial cell, but not smooth muscle cell adhesion to vitronectin in the presence of PAI-1 requires both polymerized microtubules and actin, further demonstrating the importance of the cytoskeleton for integrin-mediated adhesion. Finally, we show that cell adhesion in the presence of PAI-1 leads to colocalization of PAI-1 with the integrins αvβ3 and αvβ5 at the cell-matrix interface.

Cell adhesion receptors play important roles in maintaining cell anchorage and polarity in addition to promoting migration and differentiation. Many of the known adhesion receptors belong to the integrin family of noncovalent heterodimeric transmembrane proteins that are expressed by most cell types. The integrin family consists of at least 18 ␣and 8 ␤-subunits that combine to form as many as 24 different ␣␤ dimers (1). The extracellular domains of both integrin subunits are required for binding to adhesion proteins. In addition, integrins need physiological concentrations of divalent cations such as calcium or magnesium. Integrins have relatively short cytoplasmic tails that interact with an intracellular protein complex, termed focal contacts, which include adaptor proteins and kinases. The focal contacts control the linkage between the cytoskeleton and integrins and mediate intracellular signaling. Integrins direct forces generated by the cytoskeleton onto the extracellular matrix (ECM) 2 to produce the traction necessary for cell adhesion and migration (2). Understanding the mechanism and control of integrin-mediated cell adhesion and migration is of special interest because of their importance in processes such as wound healing and angiogenesis as well as pathologies associated with these processes.
In general, integrins recognize short linear peptide sequences on adhesion proteins, the most prevalent being arginine-glycine-aspartic acid (RGD), which is found in many ECM proteins such as fibronectin, collagens, and vitronectin (3). Although many integrins require additional sequences for effective adhesion, the vitronectin receptor integrins ␣v␤3, and to some extent ␣v␤5, are fairly unique in their ability to bind most proteins with a simple RGD sequence, including non-ECM proteins such as thrombin and transforming growth factor ␤1-latency-associated protein (LAP) (4,5). These integrins have been the subject of intense study because their expression is up-regulated by migratory cells and their activities have been linked to pathological processes, such as tumor metastasis (6). The specificities of ␣v␤3 and ␣v␤5 are not limited to vitronectin; however, this interaction is of special interest because vitronectin can accelerate the migration of vascular cells such as smooth muscle cells as compared with many other adhesion proteins (7,8).
Vitronectin is a 75-kDa plasma glycoprotein that plays a role in regulating blood coagulation, complement action, fibrinolysis, and wound healing. Initially vitronectin was called "serum spreading protein," because it was identified as the primary factor responsible for adhesion and spreading of many cell types * This work was supported in part by National Institutes of Health Grants grown in tissue culture. Besides being an abundant plasma protein (0.2-0.3 mg/ml), it is also found in the extravascular ECM, particularly at sites of injury or remodeling and it specifically associates with fibrin clots (9). In circulation vitronectin is in a closed conformation and does not interact with integrins, however, upon binding to a surface such as collagens at sites of vascular injury, vitronectin undergoes a conformational change that exposes the RGD integrin binding site as well as a high affinity binding site for PAI-1 (10).
PAI-1 is a 50-kDa proteinase inhibitor found mostly in platelet ␣-granules. It is released upon platelet activation at sites of vascular injury resulting in local concentrations that can be high relative to normal plasma levels. PAI-1 is a member of the serine proteinase inhibitor (serpin) superfamily, and is the main inhibitor in vivo of urokinase-type plasminogen activator (uPA) and tissue-type plasminogen activator. Unlike most serpins, the active conformation of PAI-1 is unstable and converts to an inactive latent conformation with a half-life of ϳ1 h under physiological conditions. Vitronectin binds PAI-1 with high affinity (K d ϳ 1 nM) and stabilizes its active conformation but does not affect the ability of PAI-1 to inhibit plasminogen activators (PAs). Likewise, PAI-1 influences the activity of vitronectin by blocking the RGD integrin binding site, which can influence cellular adhesion and migration in vitro and in vivo (8,11). Importantly, the high affinity PAI-1h as for vitronectin is lost upon its reaction with a proteinase such as uPA (12).
The PAI-1 binding site on vitronectin has been localized to the N-terminal somatomedin-B domain, which is adjacent to the RGD sequence. Although the RGD and the PAI-1 binding sites are thought to be functionally independent, their close proximity results in apparent competition between PAI-1 and purified integrins in solid phase binding assays. The reported IC 50 value for PAI-1 inhibition of integrin binding to vitronectin is ϳ1 nM, which is close to its K d for vitronectin. The efficient inhibition by PAI-1 is not surprising given that affinity of purified ␣v␤3 for RGD containing ligands has been reported to be on the order of 10 -100 nM (13,14). However, in this study we show that significantly higher concentrations of PAI-1 are needed to inhibit cellular integrin binding to vitronectin (IC 50 ϳ 20 -60 nM) compared with purified integrins, and more importantly, we show that PAI-1 can only maximally inhibit cell adhesion ϳ50 -75%, whereas it completely blocks purified integrin binding.
To analyze this discrepancy between PAI-1 inhibition of purified versus cell-associated integrins, we performed in vitro adhesion studies using endothelial cells (EC), and smooth muscle cells (SMC) together with specific PAI-1 and vitronectin mutants. These data indicate that both the EC and SMC adhere to the vitronectin RGD sequence regardless of whether PAI-1 is present or not, and that adhesion in the presence of PAI-1 requires the association of integrins with the cellular cytoskeleton. We present a model to illustrate this mechanism that suggests that the association of integrins with the cytoskeleton provides the rigidity necessary to "nudge" PAI-1 aside without removing it to gain access to the RGD.
Native human vitronectin and RGE mutant vitronectin in pMEL-Bac vectors were a gift from Dr. D. Sane (Wake Forest, NC). Human vitronectin wild type and RGE mature mutant coding sequences were amplified from the pMELBac plasmids. The forward and reverse primers were 5Ј-CCGGCCGGCCC-GCTCGGGGACCAAGAGTCATGC-3Ј and 5Ј-CGGCGGC-GCGGGTTTAAACTCATTAATGGTGATGGTGATGATG-CAGATGGCCAGGAGC-3Ј, respectively, with engineered AvaI and PmeI restriction sites following the His 6 tag. After amplification, the PCR products were digested with AvaI/PmeI and cloned into the Drosophila Expression System vector pMT/ BiP/V5-His C (Invitrogen) under the metallothionein-inducible promoter. The constructs with corresponding sequences were selected by DNA sequencing with pMTBiP forward and reverse primers provided by Invitrogen. Stable transfectants were generated for each Drosophila expression plasmid encoding the vitronectin mutants by co-transfection of S2 cells with a vector carrying the blasticidin resistance gene (pCoBlas, Invitrogen), according to the manufacturer's protocol. Stably transfected cell lines were cultivated at 28°C in Drosophila serum-free medium supplemented with 20 mM L-glutamine, penicillin/streptomycin, Fungizone, and blasticidin at 25 g/ml (Invitrogen). Vitronectin expression was induced by adding 0.5 mM CuSO 4 (final concentration) to SFM growth medium at a cell density of ϳ3.0 ϫ 10 6 cells/ml and conditioned media was collected on day 5 of induction. The vitronectin mutants were purified from conditioned media, depleted in Cu 2ϩ ions with chelating resin Chelex 100, in one-step affinity chromatography on BD Talon Metal Affinity Resin. Radiolabeling of HMK-PAI-1 with [␥-32 P]ATP was performed essentially as described (12).
Solid Phase Binding Assays-Competition binding assays of PAI-1 and integrins were performed essentially as described (8), with some modifications. Briefly, 96-well microtiter plates were coated with native human vitronectin for 1 h at 37°C in TBS followed by blocking with 1% BSA in TBS. Purified ␣v␤3 or ␣v␤5 (100 nM) in Dulbecco's modified Eagle's medium containing 1% BSA, 0.1% octyl glucoside, and 0.5 M MnCl 2 were allowed to attach to vitronectin in the presence of increasing concentrations of PAI-1 (141b) for 16 h at 4°C, after which unbound integrins were washed away and the plates incubated with either goat anti-human integrin ␤3 or ␤5 antibodies. After washing with TBS containing 0.1% Tween 20, plates were incubated with alkaline phosphatase-conjugated rabbit anti-goat antibody and developed.
Adhesion Assays of Endothelial Cells and Smooth Muscle Cells-Adhesion assays were performed using 24-well plates, which were coated with 1 g/ml of purified human native vitronectin for 1 h at 37°C, after which unoccupied sites were blocked with 1% BSA in TBS for 1 h at 37°C or overnight at 4°C. Endothelial cells and smooth muscle cells were detached using trypsin/EDTA, followed by centrifugation in the presence of 1 ⁄ 3 volume of 10% fetal bovine serum in Dulbecco's modified Eagle's medium to neutralize the trypsin. The cells were resuspended in serum-free Dulbecco's modified Eagle's medium containing 1% BSA and seeded onto 24-well plates at densities between 10 3 and 10 5 cells/well. The cells were allowed to attach and spread for 45-60 min in the presence or absence of PAI-1 mutants and other effectors indicated in the figure legends. After incubation, non-attached or weakly attached cells were removed by three washes with TBS along with vigorous agitation. Firmly attached cells were quantified by measuring intracellular acid phosphatase activity with para-nitrophenyl phosphate as described (17) and plotted as the percentage of acid phosphatase activity compared with the control. The stable PAI-1 mutant 141b (18) was used exclusively unless otherwise indicated.

Fluorescent Microscopy of Endothelial Cells and Smooth
Muscle Cells-Coverslips were coated with purified vitronectin (1 g/ml) for 45-60 min at 37°C and blocked using 1% BSA in TBS for 1 h. EC and SMC were allowed to adhere to the coverslips for 1 h in the presence or absence of PAI-1 (141b stable mutant, unless indicated otherwise) at 37°C, 5% CO 2 , after which cells were fixed with 10% formalin and permeabilized with TBS containing 1 mM CaCl 2 , 1 mM MgCl 2 , and 0.1% Triton X-100. Cells were stained for actin filaments using rhodaminelabeled phalloidin and focal contacts were stained using antipaxillin monoclonal antibody followed by rabbit anti-mouse Alexa Fluor 568.
Integrins and PAI-1 were visualized after cells were allowed to adhere to vitronectin-coated coverslips for 45-60 min as described above. Plates were rinsed with 50 mM HEPES, 100 mM NaCl with 5 mM CaCl 2 and 1 mM MgCl 2 , pH 7.5, followed by cross-linking with 0.4 g/ml bis(sulfosuccinimidyl)suberate in 50 mM HEPES, 100 mM NaCl with 5 mM CaCl 2 and 1 mM MgCl 2 , pH 7.5, for 15 min at room temperature. After cross-linking, plates were washed with TBS to inactivate bis(sulfosuccinimidyl)suberate and incubated with uPA (0.1 mM in TBS) for 30 min at room temperature in an attempt to remove unreacted PAI-1 from the immobilized vitronectin. Plates were then rinsed with TBS buffer followed by fixation with 10% buffered formalin for 2 min after which plates were washed with TBS containing 0.1% Triton X-100 and proteinase inhibitor mixture. Anti-human PAI-1 and either LM609 or P1F6 (0.1 g/ml) were added in 0.1% TBS/BSA at room temperature for 1 h followed by washes and addition of secondary antibodies, goat anti-mouse (Alexa Fluor 488) and goat anti-rabbit (Alexa Fluor 568), followed by washing 3 times in 50 mM TBS buffer. Slides were mounted with Prolong Gold medium.
For laser scanning confocal microscopy, images were acquired using a Radiance 2100 laser scanning system (Bio-Rad) using the Laser Sharp 2000 version 4.1 software. All confocal images were acquired under the Kalman mode with iterations at a minimum laser power wattage set based on negative control slides. The following settings were used for the experiments: ϫ100 oil immersion objective lens, 8-bit, 1024 ϫ 1024 pixel resolution at CCD camera, Look-Out Table (LOT) set to green for 488 nm emission and red for 568 nm emission. The dichroic mirrors were set to 560 DCLPXR for green and 650 DCLPXR for red. The emission filters were set to HD515/30 for green and HQ600/40 for red. Finally, laser scanning speed was set to 160 lines/s with sequential recording of two channels. No special effect or filter (i.e. reduce noise, blur, sharpen, etc.) was used. We chose to examine the horizontal z-sections that had the most intense integrin staining (green), which were ϳ0.5 m from the surface of the slide and compared that to PAI-1 staining (red) in the same sections. The analysis of digital images was performed by Volocity (version 2.6.3; Improvision, UK) (19) using the colocalization session algorithm that determined the overlapped portion of integrin and PAI-1 staining. The classifier thresholds for each color channel were determined based on corresponding non-immune IgG control slides and the sample areas were subsequently measured. Due to size differences among cells, colocalization areas were determined as the percentage area of the whole cell.
Co-precipitation of PAI-1 and Integrins-Culture dishes (100 mm) were coated with vitronectin (1 g/ml) for 1 h at 37°C and blocked using 1% BSA in TBS for 1 h. EC and SMC were allowed to adhere to the plates for 1 h in the absence or presence of N-terminal biotinylated 141B-PAI-1 (final concentration 1 M). The dishes were rinsed twice with 50 mM HEPES, pH 7.5, 100 mM NaCl, 5 mM CaCl 2 , 1 mM MgCl 2 followed by cross-linking with 0.4 mM 3,3Ј-dithiobis(sulfosuccinimidylpropionate) in 50 mM HEPES, pH 7.5, 100 mM NaCl, 5 mM CaCl 2 , 1 mM MgCl 2 for 15 min at room temperature. After cross-linking, the buffer was replaced with TBS containing 100 nM uPA for 15 min. Plates were then rinsed with PBS and flash frozen. Cells were lysed in TBS containing 1% Triton X-100, 0.1% SDS, and Complete Protease Inhibitor Mixture. Lysates were sheared using a syringe fitted with a 25-gauge needle. Insoluble debris was pelleted by centrifuging lysates for 5 min at 10,000 ϫ g. Biotinylated proteins and conjugates were pulled down using immobilized streptavidin and incubated overnight at 4°C with end over end rotation. The immobilized protein complexes were pelleted by centrifuging for 2-3 min at 3000 ϫ g. The beads were washed 4 times with TBS containing 1% Triton X-100 and 0.1% SDS. Proteins were eluted with NuPAGE LDS Sample Buffer containing 5% ␤-mercaptoethanol by boiling the samples for 5 min. The proteins were separated by SDS-PAGE on 4 -20% gradient gels and transferred to polyvinylidene difluoride. Integrin ␣v was detected with rabbit anti-human ␣v antiserum and horseradish peroxidase-conjugated donkey anti-rabbit IgG. The blots were stripped and reprobed for PAI-1 using sheep anti-human PAI-1 and horseradish peroxidase-conjugated donkey anti-sheep IgG. SuperSignal West Pico Chemiluminescent Substrate and film were used for detection.

PAI-1 Inhibition of Cellular Integrin Binding to Vitronectin Is
Less Efficient than with Purified Integrins-Integrins and the cytoskeleton form a functional unit that provides traction and directionality through controlled assembly and disassembly of cell-ECM contacts (20,21). Understanding the mechanisms of how adhesion proteins and the cytoskeleton influence integrin function is critical for our understanding of cell migration and we believe that the competition between PAI-1 and integrins for vitronectin binding provides a well defined, simple and physiologically relevant system to examine integrin properties, such as affinity modulation.
The PAI-1 binding site on vitronectin has been well studied by site-specific mutations that identified several amino acids in the somatomedin-B domain between residues 4 and 40 as being vital for high affinity PAI-1 binding (22) and interestingly all of these residues are conserved across mammalian species (23). The RGD sequence in vitronectin is at position 45-47, and because of this proximity, PAI-1 is thought to inhibit the binding of integrins by steric hindrance. The somatomedin-B domain and the RGD function independently, with PAI-1 being unable to block cell adhesion to a vitronectin fragment having the RGD but lacking the somatomedin-B (24). However, the PAI-1 binding site and the RGD are in close proximity because solid-phase binding assays, such as shown in Fig. 1A, demonstrate competitive binding of PAI-1 and purified integrins ␣v␤3 and ␣v␤5 to vitronectin. The IC 50 of the PAI-1 inhibition of ␣v␤3 and ␣v␤5 was 2.5 and 5 nM, respectively, which is consistent with previous reports on the PAI-1 inhibition of purified ␣v␤3, ␣v␤5, and ␣IIb␤3 binding to vitronectin (8,25). Fig. 1B shows that PAI-1 was also a competitive inhibitor of cell adhesion to vitronectin, although it was less efficient, yielding an IC 50 of 31 and 60 nM for EC and SMC, respectively. Additionally, whereas PAI-1 inhibition of the purified integrin was competed down to baseline levels (Fig. 1A), the inhibition of cell adhesion to vitronectin was not complete, generally reaching a maximum of around 50 -75% inhibition. These data are consistent with earlier studies (26), showing that a 10-fold higher concentration of active PAI-1 is required to inhibit ϳ50% of HEp-2 cells adhesion to vitronectin in comparison to purified integrins. These results demonstrate that PAI-1 inhi-bition of cellular adhesion to vitronectin conflict with the results using purified integrins. To examine whether cell adhesion occurs through a mechanism that is not affected by PAI-1 we analyzed the morphology of EC and SMC, with respect to their intracellular adhesion machinery and effects of specific inhibitors.
Shown in Fig. 2A are representative fields of EC and SMC that adhered either to vitronectin alone or to vitronectin in the  presence of increasing concentrations of PAI-1 and stained for paxillin. In the absence of PAI-1, the cells demonstrated punctate paxillin staining, characteristic of organized focal contacts that were mostly observed in extended podia. In the presence of 0.1 M PAI-1, the cell to surface contact area, or "footprint," of both EC and SMC was substantially reduced, which could be due to a decrease in available vitronectin binding sites. Although the cell footprints were retracted, paxillin staining still displayed a punctate pattern indicating that focal contacts remained organized. At 1 M PAI-1 the vitronectin-adhering cells were retracted further, and the paxillin staining, although still punctate, was mostly observed in short podia extending from a very compact cell body, demonstrating that although focal contacts were less intense they were still present. Because punctate paxillin staining is indicative of integrin-mediated clustering and cell adhesion, (27,28) we conclude that the EC and SMC were adhering to the vitronectin via integrins, despite the presence of PAI-1.
Confocal analysis of the cells stained with TRITC-phalloidin showed that with increasing concentrations of PAI-1, cells adopted vertical and elongated structures with progressively smaller footprints. Although the cells appeared distressed, they did not undergo apoptosis associated with loss of cell anchorage because they were viable and did not show nuclear fragmentation as evaluated by Hoechst staining and, furthermore, returned to normal morphology if PAI-1 was removed from vitronectin with mild elastase treatment (data not shown).
Because paxillin and other proteins comprising the focal contacts serve as anchoring points for the organization and assembly of the actin cytoskeleton, we examined staining of polymerized actin in EC and SMC. Fig. 2B shows representative images of TRITC-phalloidin staining of EC and SMC that adhered to vitronectin either alone or in the presence of 0.1 or 1 M PAI-1. Fig. 2B shows that both cell types formed organized actin filaments when allowed to adhere and spread on vitronectin, but with addition of increasing concentrations of PAI-1, the cells became progressively more compact with the shorter and less branched actin filaments. This staining pattern indicates that organization of polymerized actin was sensitive to PAI-1 and is in agreement with the paxillin staining shown in Fig. 2A, indicating that both the focal adhesions and the actin fibers of the cells were severely perturbed by PAI-1 treatment. However, these cellular components were still functional and able to mediate adhesion to vitronectin, even in the presence of PAI-1. The results, however, do not rule out the possibility of other binding sites on vitronectin besides the RGD. Integrin recognition sites are not limited to a linear RGD sequence. Other motifs have been identified and some integrins require additional binding sites for efficient adhesion, as has been demonstrated for ␣5␤1 and the leukocyte integrin ␣M␤2 (29,30). Although ␣v␤3 and ␣v␤5 recognize RGD sequences in vitronectin as their primary binding site, some exceptions have been demonstrated (31).
To test whether the cells are adhering to other sites on vitronectin, we performed adhesion assays using a vitronectin mutant that has the cell binding sequence RGD mutated to an RGE (VnRGE) (32). This mutant was purified in its native single chain form under non-denaturing conditions and was indistinguishable from native vitronectin with respect to SDS-PAGE, PAI-1 binding, and reactivity to the panel of monoclonal antihuman vitronectin antibodies (data not shown). The results of adhesion assays are shown in Fig. 3A and demonstrate that neither ECs nor SMCs (A and B, respectively) were able to attach to the VnRGE mutant. Incubation of cells with VnRGE, up to 2 h, demonstrated that no significant cell adhesion occurred in the absence of an intact RGD integrin binding site (data not shown). This is in agreement with previous studies using VnRGE (32) and suggests that like the cells used in those previous studies, EC and SMC require the RGD. Additionally, these results also preclude involvement of other cell surface proteins such as uPAR, which has been shown to bind and facilitate cell adhesion to the somatomedin-B domain of vitronectin and not the RGD (24,33).
Although these results do not rule out a role for other sites that might be involved in post-adhesion events, they simplify the experimental system considerably, because the data suggest that immobilized vitronectin is a monovalent integrin ligand. Therefore, to further examine the integrin requirements, we performed adhesion assays on native vitronectin in the presence of a panel of integrin inhibitors. The adhesion of EC and SMC is shown in Fig. 3, C and D, respectively, and demonstrates that RGD peptides, but not RGE peptides, inhibited adhesion of both cell types to vitronectin by 80 -90%, which supports the previous results demonstrating the importance of RGD for adhesion. Furthermore, Fig. 3 shows that an anti-functional antibody for ␣v␤3 (LM609) partially inhibited the binding of EC and SMC to vitronectin, whereas an antifunctional antibody against ␣v␤5 (P1F6) produced less inhibition. The inhibition by LM609, but not P1F6, was enhanced by addition of PAI-1, indicating that there are other cellular integrins besides ␣v␤3 involved in the adhesion to vitronectin. Regardless of contributions by other integrins on adhesion, these cells need to overcome the inhibition by PAI-1 and adhere to the vitronectin RGD sequence, because they do not adhere to the VnRGE mutant. However, it is also possible that the cells produce proteinases that cleave PAI-1 and subsequently expose the vitronectin RGD site.

PAI-1 Is Not Inactivated by Adhering Cells-Many cells have been shown to express and secrete a variety of proteinases,
including uPA and tissue-type plasminogen activator, which are the in vivo targets of PAI-1, and it is likely that EC and SMC in culture express either PAs or matrix metalloproteinases. PAs efficiently remove PAI-1 from vitronectin because complex formation of PAI-1 with proteinases induces a rapid conformational change in PAI-1 resulting in an approximate 1000-fold reduction of its affinity for vitronectin and an equally dramatic increase in PAI-1 affinity for endocytic receptors of the lowdensity lipoprotein receptor family (12). This leads to rapid repartitioning of PAI-1 from vitronectin in the ECM to clearance receptors on the cell surface, and subsequent endocytosis and degradation of the PAI-1⅐proteinase complex.
To examine whether EC and SMC express proteinases that cleave PAI-1, we performed adhesion assays in the presence of specific PAI-1 mutants that are either defective in vitronectin binding, proteinase specificity, or inhibitory ability. Fig. 4, A  and B, show the adhesion of EC and SMC, respectively, to vitronectin in the absence or presence of several PAI-1 mutants (1 M). PAI-1R346A (PAI-1 A ) is an efficient inhibitor of pancreatic elastase, but not PAs and binds vitronectin with wildtype affinity (16). PAI-1R101A,Q123K (PAI-1 AK ) inhibits PAs as effectively as WT-PAI-1 but does not bind vitronectin with measurable affinity (15). PAI-1T333R,A335R (PAI-1 R ) is not a proteinase inhibitor but binds vitronectin with wild-type affinity, which is decreased after proteinase cleavage but not completely abrogated as observed with cleaved WT-PAI-1 (34). As seen in Fig. 4, PAI-1 mutants that bind vitronectin with high affinity regardless of proteinase specificity inhibited cell adhesion to a similar degree. Conversely, PAI-1 AK , which is a pro-teinase inhibitor but lacks vitronectin binding, did not inhibit cell adhesion.
To directly examine the fate of PAI-1 in these adhesion assays, we used an HMK-PAI-1 mutant, which can be radiolabeled with high specific activity with 32 P, without loss of function (12). In these experiments, radioactivity was not redistributed from immobilized vitronectin to the supernatant with addition of cells as compared with the no cell controls. Furthermore, a significant amount of cleaved or proteinase-complexed PAI-1 was not detected by autoradiography of SDS-PAGE (data not shown). Together, these results suggest that cellular proteinase activity is not responsible for the cell adhesion to vitronectin in the presence of PAI-1 and that they employ other mechanisms to overcome the PAI-1 inhibition.
It is possible that there is "targeted removal" of PAI from vitronectin, e.g. cells are only removing and endocytosing PAI-1 from discreet number of sites, allowing for limited attachment. However, because most of the adhesion assays are performed with excess PAI-1 in the media, the cells would have to be continually inactivating PAI-1, which would likely be detected. On the other hand, if cellular integrins were effectively competing PAI-1 from vitronectin in a stoichiometric fashion, for example, by increasing their affinity for vitronectin, it is unlikely that it would be detected in the above experiments.
EC and SMC Adhesion to Vitronectin and PAI-1 Is Not Mediated by High Affinity Integrins-Given that the above data indicate that EC and SMC neither removed PAI-1 from vitronectin by proteinase activity nor bypassed the vitronectin RGD site, it is possible that they competed for PAI-1 binding. To effectively compete in these assays, however, their affinities must be elevated from the submicromolar range of purified integrins, to the nanomolar range. But because the primary integrin-binding site on vitronectin is a linear tripeptide, which in terms of protein-protein interactions is a relatively small surface with little structure, it is difficult to envision that these interactions could result in higher affinity for vitronectin than PAI-1, which has a more expansive contact area.
Nevertheless, many studies have reported that affinities of integrins, including ␣v-containing integrins, can be modulated by mechanisms that alter their conformations to generate either high or low affinity competent structures. High and low affinity binding of integrins is not thought to be an innate property, but is elicited in response to specific signals originating either from extracellular or intracellular triggers. These signals, termed "outside-in" and "inside-out" signaling, respectively, have been demonstrated using various methods, but do not represent mutually exclusive models of integrin binding (35,36).
Outside-in signaling is thought to originate by structural changes in the integrin extracellular domains leading the proteins to assume either high or low affinity conformations. These structural changes, functionally integrin activation, are either generated upon ligand binding or mediated by divalent metal ions such as manganese (Mn 2ϩ ), which initially was shown to enhance ligand binding of the purified fibronectin receptor integrin (37). The integrin ␣-subunit has a metal binding site, which is thought to bind divalent cations such as Ca 2ϩ and Mg 2ϩ , but many studies use Mn 2ϩ because it is believed to activate most integrins regardless of ligand occupancy (38).
A case for intracellular factors mediating integrin affinity modulation, or inside-out signaling, has been proposed for focal contact proteins, such as talin, that bind integrin tails and induce conformational changes in their extracellular domains (39). Talin, like paxillin, is an actin-binding focal adhesion protein involved in connections between integrins and the cytoskeleton, especially the initial integrin-cytoskeleton bonds (40). The importance of the actin cytoskeleton for cell adhesion has been amply demonstrated in cells expressing integrins with truncated cytoplasmic domains that do not connect to the actin cytoskeleton and do not support cell adhesion (41). Additionally, disrupting the actin cytoskeleton with specific antagonists such as cytochalasin-D (CD) can affect cell migration and has been shown to inhibit ␣IIb␤3 function (42). These and other results argue that intracellular factors also have an active role in integrin function.
Based on many studies, including the ones mentioned above, well defined procedures have been identified to analyze whether integrin affinity is responsible for the binding properties of purified integrins versus intact cells binding to vitronectin. Therefore, we treated EC with either Mn 2ϩ and/or CD and the results shown in Fig. 5 demonstrate that PAI-1 at 1 M was less efficient at inhibiting cell adhesion in the presence of Mn 2ϩ , whereas EC treated with CD were more sensitive to PAI-1. Furthermore, including Mn 2ϩ did not substantially improve adhesion of cells treated with CD (Fig. 5A). Similar results were obtained with SMC (data not shown), indicating that Mn 2ϩ allows cells to adhere more effectively to vitronectin with PAI-1 present, which would be consistent with a high affinity integrin conformation that efficiently competes for vitronectin binding. On the other hand, the data also demonstrate that disrupting the actin cytoskeleton with CD rendered the cells more sensitive to PAI-1 inhibition, even in the presence of Mn 2ϩ (Fig. 5A).
To further examine how Mn 2ϩ and CD affected EC and SMC integrin binding to vitronectin we performed adhesion assays in the presence of varying concentrations of PAI-1. Fig. 5, B and C, show the effects of increasing concentrations of PAI-1 on EC and SMC adhesion to vitronectin, respectively, in the presence of either Mn 2ϩ or CD alone or together. In the presence of Mn 2ϩ (closed circles), the PAI-1 IC 50 values for EC and SMC were 28 Ϯ 3 and 45 Ϯ 6 nM, respectively, comparable with those obtained from Fig. 1B (31 and 60 nM for EC and SMC, respectively). However, there was a clear difference in the extent of PAI-1 inhibition of adhesion, with adhesion inhibited only ϳ10 -20% at the highest concentrations of PAI-1, compared with 50 -60% seen in Fig. 1B. CD (closed squares) has the opposite effect on cell adhesion, with adhesion almost totally inhibited by PAI at the highest concentrations used. Interestingly, the PAI-1 IC 50 values of CD-treated cells were 3.6 Ϯ 0.5 and 5.2 Ϯ 1.1 nM for EC and SMC, respectively, which is lower than those in Fig. 1B but comparable with purified integrins. When Mn 2ϩ is added together with CD (open circles), cell adhesion to vitronectin in the presence of PAI-1 was partially restored. There was a shift in the IC 50 (11.5 and 15 nM for EC and SMC, respectively), which might indicate that a Mn 2ϩ -mediated integrin conformational change is responsible for the increase in the IC 50 , however, PAI-1 still inhibited cell adhesion ϳ80 -90% at the highest concentrations.
These data suggest that binding of EC and SMC to vitronectin in the presence of PAI-1 is not solely mediated by increased integrin affinity. Additionally, the fact that the solid phase binding assays of purified ␣v␤3 and ␣v␤5, shown in Fig. 1A, were performed in the presence of Mn 2ϩ further supports the notion that Mn 2ϩ -induced integrin conformational changes alone do not result in affinities for vitronectin that are higher than PAI-1. These results are supported by previous observations showing that Mn 2ϩ results in ϳ4-fold higher IC 50 values for a fibronectin fragment competing for integrin binding to fibronectin (37), and although Mn 2ϩ increased PAI-1 IC 50 values approximately 2-3-fold in EC and SMC treated with CD, their adhesion to vitronectin and PAI-1 appeared to be more dependent on integrin contacts with actin filaments. These results also apply to adherent cells because untreated EC and SMC that were allowed to attach and spread on vitronectin for 16 h were efficiently detached when treated with CD and PAI-1, either in the presence or absence of Mn 2ϩ , but this treatment did not affect EC and SMC bound to fibronectin (data not shown).
To further analyze the role of cytoskeleton in EC and SMC, we examined the effects of phenylarsine oxide (PAO) on adhesion to vitronectin in the presence of PAI-1. PAO is a tyrosine phosphatase inhibitor and has been used to demonstrate that phosphorylation of focal adhesion proteins, such as paxillin, plays an important role in the regulation of adhesions and organization of the actin cytoskeleton (43)(44)(45). Additionally PAO does not cause cell activation or interfere with integrin function (46). We performed adhesion assays in the presence of PAO (0.1 M) and found that this inhibited adhesion of both EC and SMC to vitronectin in the presence of PAI-1, similar to that seen with CD, further suggesting the importance of the cytoskeleton for their adhesion (data not shown).
Taken together these data suggest that the connection between integrins and the actin cytoskeleton is the primary factor mediating the adhesion of EC and SMC to vitronectin in the presence of PAI-1, perhaps by generating traction on the substrate. Interestingly, our results are similar to those obtained from studies of ␣IIb␤3 function, which demonstrated that actin cytoskeleton assembly with CD and PAO inhibited integrin fibrin clot retraction but not adhesion (47). Additionally, many cell types including EC, SMC, fibroblasts, and monocytes exhibit matrix retraction following adhesion, and although the exact mechanism behind matrix retraction is poorly understood, it does depend on force generation that is conveyed to the integrins via cytoskeleton fibers (48,49) and could also be related to fibronectin fibrillogenesis, which requires integrins, actin, and microtubules (50). Therefore, to further analyze the role of intracellular filaments in the adhesion of EC and SMC to vitronectin in the presence of PAI-1, we examined possible contributions of other cytoskeletal elements.
Role of Microtubules in Adhesion of EC and SMC to Vitronectin in the Presence of PAI-1-Besides actin filaments, there are at least two other types of cytoskeletal fibers, microtubules and intermediate filaments. These fibers are thought to act as scaffolds that control cell shape, support vesicular trafficking, and provide tensile strength to counter external mechanical stresses; however, it has also been demonstrated that intermediate filaments and microtubules contribute to integrin adhesion in some cells either directly or through actin (51)(52)(53). Depolymerization of microtubules by treatment with colchicine impacts actin assembly and causes cell rounding in fibroblasts (54) and collapse of EC tube formation (55). The possible interactions between paxillin with microtubules were also demonstrated in human T lymphoblasts (56).
Because of these results, we examined the effect of colchicine on the adhesion of EC and SMC to vitronectin in the presence of increasing concentrations of PAI-1. The results are shown in Fig. 6 and demonstrate that EC were very susceptible to colchicine, with a PAI-1 IC 50 of 8.2 Ϯ 1.6 nM (open circles), whereas PAI-1 inhibition of SMC adhesion (closed circles) was not greatly affected (IC 50 ϭ 74 Ϯ 2.1 nM). Furthermore, PAI-1 inhibited colchicine-treated EC adhesion around 80% at the highest concentration of PAI-1, whereas the colchicine-treated SMC seemed unaffected. Vinblastine, which is another microtubule disruptor, had the same effect on EC as colchicine, but the intermediate filament disruptor, 3,3Ј-iminodipropionitrile, did not affect EC or SMC adhesion to vitronectin in the presence of PAI-1 (data not shown).
These results further suggest that integrins in EC and SMC require the support of the cytoskeleton for adhesion to vitronectin in the presence of PAI-1 and imply differences in the organization and cross-talk of the cytoskeletal fibers of these cells. But these results still do not provide a mechanism to explain how the traction or rigidity of the integrin-cytoskeleton connections enables them to gain access to the vitronectin RGD that is sterically blocked by PAI-1.
Colocalization of EC and SMC Integrins with PAI-1-Given the sizes of PAI-1 (ϳ50 kDa) and integrin dimers (ϳ250 kDa) as well as the proximity of their two binding sites on vitronectin, it would seem likely that a ternary complex of these proteins could not exist without considerable steric hindrance; however, if PAI-1 and integrins do concurrently bind to vitronectin, then they must be in very close proximity. To examine whether both proteins are accommodated on vitronectin, we performed chemical cross-linking studies. EC and SMC adhering to vitronectin in the presence of PAI-1 (1 M) were treated with cross-linker and stained for PAI-1 and ␣v␤3 or ␣v␤5. Fig. 7, A-C, show representative staining of EC adhering to vitronectin in the presence of PAI-1 and stained for ␣v␤3 (panel A, green) and PAI-1 (panel B, red). Combining PAI-1 and ␣v␤3 staining (Fig. 7C) shows that there was colocalization of the two proteins, especially at the cell periphery. To quantify the colocalization, we performed confocal studies that demonstrate that both ␣v␤3 and ␣v␤5 integrins on EC (Fig. 7D) and SMC (Fig. 7E) colocalized with PAI-1 on vitronectin, but colocalization was more pronounced between ␣v␤3 integrins and PAI-1 than for ␣v␤5 and PAI-1 in both cell types (30 versus 18% for EC and 47 versus 26% for SMC, respectively). It is unlikely that specific binding occurs between PAI-1 and these integrins because colocalization of the two proteins was not observed in the absence of chemical cross-linking (data not shown). These results suggest that ␣v␤3 might be more dominant in mediating this adhesion and support the results in Fig. 3C, showing that PAI-1 was more effective at inhibiting cell adhesion in the presence of the LM609 than P1F6.
The close proximity of PAI-1 and ␣v integrins was further confirmed from pull down cross-linking experiments in which EC or SMC adhering to vitronectin in the absence or presence of N-terminal biotinylated PAI-1 (1 M) were treated with the membrane impermeable cross-linker 3,3Јdithiobis(sulfosuccinimidylpropionate). As is seen in Fig. 7F, ␣v integrins co-precipitated with biotinylated PAI-1 in both SMC and EC. Interestingly, the predominant form of cross-linked ␣v integrins in ECs was 150 kDa indicating that it was unprocessed. In contrast, nearly all of the crosslinked ␣v integrins were processed in SMCs resulting in detection of the 25-kDa form (57). These results also provide a mechanism for the EC and SMC abilities to bind vitronectin despite PAI-1. The binding of EC and SMC integrins is not due to higher affinity of the integrins alone, but is due to the ability of an integrin backed by the cytoskeletal rigidity to nudge PAI-1, without removing it, to fully access the RGD. To illustrate our results, we present a working model on the binding of cellular integrins to vitronectin and the fate of the bound PAI-1 (Fig. 8). In this   Alexa Fluor 568, B). C shows the overlaid images from A and B. (Note that the diffuse red background seen in B and C is residual PAI-1 remaining on vitronectin, not nonspecific background.) D and E show quantitation of integrin and PAI-1 colocalization in EC and SMC, respectively, by confocal analysis. F shows co-precipitation of PAI-1 cross-linked to integrin ␣v in both EC and SMC. The inset bar in A denotes 10 m. Each field and bar graph represents a typical result of 3-5 experiments (5-10 fields analyzed per experiment). Asterisk denotes p Ͻ 0.05. model, PAI-1 binds vitronectin and sterically blocks purified integrins and cellular integrins that are uncoupled from the focal contacts and the cytoskeleton (Fig. 8A), leading to efficient inhibition. If, however, the cellular integrins are rigidly held by the cytoskeleton and focal adhesions, the forces generated by these connections are able to give the integrins enough rigidity to access the RGD without removing the PAI-1 (Fig. 8B).
In addition to influencing cell adhesion by blocking nontethered integrins, the proteinase inhibitory properties of PAI-1 also play a role by facilitating endocytosis and recycling of integrins through covalent complexes with PAs its high affinity for members of the low density lipoprotein receptor family (33,58). Because PAI-1 expression can also influence the cellular integrin repertoire (59), its upstream signaling can impact cell migration at the transcriptional level and complement the effects PAI-1 has on the integrin cytoskeletal engagement.
Further demonstration of the importance of the cytoskeleton in integrin binding to vitronectin is the effect that the microtubule disruptor, colchicine, has on EC, but not SMC adhesion in the presence of PAI-1. This function of the microtubules in EC adhesion is currently being studied, but may be related to a cell-specific ability of EC to integrate their cytoskeleton more extensively as a mechanism for maintaining endothelial integrity because of shear stresses generated by the circulation (60). Alternatively, because their sensitivity to colchicine is similar to that demonstrated by lymphocytes, these cells may share mechanisms or linker proteins that make the microtubules and the actin cytoskeleton more interdependent (52,53).