Vitronectin Governs the Interaction between Plasminogen Activator Inhibitor 1 and Tissue-type Plasminogen Activator*

The “serpin” plasminogen activator inhibitor 1 (PAI-1) is the fast acting inhibitor of plasminogen activators (tissue-type (t-PA) and urokinase type-PA) and is an essential regulatory protein of the fibrinolytic system. Its P1-P1’ reactive center (R346 M347) acts as a “bait” for tight binding to t-PAIurokinase-type PA. In vivo, PAI-1 is encountered in complex with vitronectin, an interaction known to stabilize its activity but not to affect the second-order association rate constant (kl) between PAI- 1 and t-PA. Nevertheless, by using PAI- 1 reactive site variants (R346M, M3475, and R346M M347S), we show that the binding of vitronectin to the PAI- 1 mutant proteins improves plasminogen activator inhibition. In the absence of vitronectin the PAI-1 R346M mutants are virtually inactive toward t-PA (kl C1 X lo3 M” s-’). In contrast, in the presence of vitro- nectin the rate of association increases about 1,000-fold (kl of 6-8 X 10‘ M” s-‘). This inhibition coincides with the formation of serpin-typical, sodium dodecyl

The key reaction constitutes the formation of a tight bond between the serine residue of the catalytic triad of the protease and the P1 residue of the inhibitors' reactive center, located close to its carboxyl terminus. This reaction generates an equimolar, SDS-stable, inactive complex. Like the other inhibitory serpins, PAI-1 actually functions as a pseudosubstrate since the P1-P1' peptide bond, being R346-M347 ( 5 ) , acts as a "bait" and mimics the R560-V561 peptide bond of plasminogen (lo), the substrate for the plasminogen activators t-PA and urokinase-type PA. Although the reactive center P1-P1' residues of a serpin clearly are an important determinant for its target specificity, obviously other distinct amino acids are required for the highly selective interaction between a target serine protease and its serpin. This is best illustrated by comparing the distinct target specificity of PAI-1 with a2-antiplasmin (11,12), both harboring identical P1 (R) and P1' (M) residues ( 5 , 13). As yet, the identity of other specificity-determining amino acid residues has not been elucidated. In contrast, data are available on the areas on t-PA, apart from the serine residue (S478) of the catalytic triad, which contribute to the interaction with PAI-1. Notably, it has been shown that three basic amino acids (K296, R298, and R299) in the carboxyl-terminal B chain of t-PA are of crucial importance for the inhibition by PAI-1 (14). Furthermore, competition has been described between intact t-PA and an isolated domain ( i e . "kringle 2" domain) of the aminoterminal A chain of t-PA for the interaction with PAI-1, a finding that is indicative for a role of the kringle 2 domain in the inhibition reaction (15).
Recently it has become evident that a detailed study of the interaction between PAI-1 and t-PA (or urokinase-type PA) requires the participation of the multifunctional protein vitronectin (16). Vitronectin has been identified as the PAI-1binding protein both in plasma (17,18) and in the subendothelial matrix (19). Moreover, both PAI-1 and vitronectin are found in the a-granules of platelets, and complexes of these components are released upon platelet activation (20). These findings may indicate that the complex between PAI-1 and vitronectin represents the physiologically relevant form of the inhibitor. Indeed, in the presence of vitronectin the activity of PAI-1 is substantially preserved (18). Moreover, we have recently presented additional evidence for the functional interaction between vitronectin and PAI-1 (21). Specifically, vitronectin endows PAI-1 with thrombin-inhibitory properties, enhancing in a dose-dependent manner the formation of SDS-stable complexes between the inhibitor and thrombin.
To study the mechanism of action of the three participants, PAI-1, vitronectin, and plasminogen activators, in more detail the Interaction between PAI-1 and t-PA 10701 we have chosen the following approach. First, we have eliminated the dominating contribution of the PAI-1 reactive center P1 residue (R346) to the efficacy and specificity of the inhibitor-protease interaction. For that purpose, we have used site-directed mutagenesis to replace the arginine by a methionine residue. The choice of a methionine at the P1 position was inferred from the observation that another serpin, dantitrypsin, having methionine at P1 (22), does not inhibit t-PA or urokinase-type PA.2 As anticipated, the resulting mutant protein (PAI-1 R346M) is virtually inactive toward t-P A or urokinase-type PA. Second, we analyzed the effect of the presence of vitronectin on the reaction of PAI-1 R346M with t-PA and urokinase-type PA. Third, t h e potential contribution of the non-protease part (i.e. the amino-terminal A chains) of t-PA and urokinase-type PA, respectively, to t h e efficacy of the interaction with PAI-1 R346M was investigated. From the data of these experiments, we conclude that vitronectin promotes in a dose-dependent manner the formation of SDS-stable, inactive complexes between the proteases and PAI-1 R346M. At optimal concentrations vitronectin enhances the second-order association rate constant between the (mutant) inhibitor and the protease about 1,000fold. The competence of vitronectin to promote this reaction is substantially higher if the non-protease A chains of t-PA or urokinase-type PA are present, indicating a role for area(s) other than the catalytic triad of the proteases in the tripartite interaction. Human fibrinogen (grade L) and the chromogenic substrate S2288 (H-D-isoleucyl-prolyl-arginyl-p-nitroanilide) were obtained from KabiVitrum (Stockholm, Sweden). Fibrinogen was treated with 1 mM diispropyl fluorophosphate for 18 h at room temperature to inactivate potentially contaminating proteases and was incubated simultaneously with 1 ml of a lysine-Sepharose suspension to remove traces of plasminogen. The Sepharose beads were removed by centrifugation, and the fibrinogen was dialyzed at room temperature against 50 mM Tris-HC1 (pH 7.4), 500 mM NaC1. Subsequently, traces of vitronectin in the fibrinogen preparation were depleted by incubating twice for 2 h at room temperature with an oligoclonal anti-vitronectin antiserum coupled to Sepharose. The eluate was collected, dialyzed against 50 mM Tris-HC1 (pH 7.4), 150 mM NaC1, and aliquots were stored at -70 "C until use. After this treatment the fibrinogen was free of vitronectin (data not shown). The oligoclonal anti-vitronectin-sepharose was made by coupling five different monoclonal antibodies directed against vitronectin3 to cyanogen bromide-activated Sepharose (2.5 mg of protein/ml of beads), as indicated by the supplier.

Materials
High and low molecular weight urokinase-type PA were from Calbiochem. Two-chain Bowes melanoma t-PA (850,000 IU/mg) was obtained from Biopool (Umea, Sweden). "S-Labeled t-PA was immunopurified from conditioned medium of Bowes melanoma cells (24), metabolically labeled with [35S]methionine and [35S]cysteine, as described (25). Human recombinant t-PA (rt-PA) and t-PA del.FEKlK2 were purified from stably transfected mouse L-cells as described (25). The apparently homogeneous single-chain preparations were fully converted into two-chain forms by incubation with plasmin-Sepharose, as determined by silver staining (26)  polyacrylamide gel electrophoresis (SDS-PAGE, 27). Glutamic acidplasminogen was purified from fresh human plasma and activated to plasmin as described (28). Plasmin-Sepharose was made by coupling plasmin to cyanogen bromide-activated Sepharose (3 mg of protein/ ml of beads). Hybridoma cells, producing the murine anti-human factor VI11 monoclonal antibody CAg69 (291, were kindly provided Dr. J. A. van Mourik (Dept. of Blood Coagulation of this institute).
These CAg69-producing cells were grown as described,' and monoclonal antibodies were purified from the conditioned media using protein A-Sepharose as indicated by the supplier. CAg69-Sepharose was made by coupling monoclonal antibody CAg69 to cyanogen bromide-activated Sepharose (4.5 mg of protein/ml of beads) as described by the manufacturer. Vitronectin was purified from human plasma to apparent homogeneity as described previously (31). The protein concentration of the purified preparation was spectrophotometrically determined by using an extinction coefficient of E (1% at 280) = 13.0 and a molecular weight of 75,000 (32). Collagenase from Achromobacter iophagus was from Boehringer Mannheim. Collagenase-Sepharose was made by coupling collagenase to cyanogen bromide-activated Sepharose (5 units of collagenase/ml of beads) as described by the supplier, using 100 mM sodium borate (pH 8.21, 500 mM NaCl as coupling buffer. All buffers used for the activity of the metalloprotease collagenase were supplemented with 5 mM CaCl,. Contaminating serine proteases were inactivated by incubating the collagenase-Sepharose for 3 h at room temperature with 1 m M diisopropyl fluorophosphate. Oligonucleotides were made using an automated DNA synthesizer (Applied Biosystems model 318A). Nucleic acid-modifying enzymes were either from New England Biolabs (Beverly, MA) or from GIBCO-Bethesda Research Laboratories and were used as indicated. Sequenase DNA polymerase was obtained from U.

S. Biochemical Corp.
Construction of Expression Plasmids-Standard recombinant DNA techniques were performed as described (33). Plasmids encoding human PAI-1, which are suitable for the expression of biologically functional rPAI-1 in E. coli, were constructed as follows. Expression plasmid pMBL11, containing the E. coli tryptophan (trp) promoteroperator, was digested with the restriction endonucleases XhoI and BamHI to allow insertion of the PAI-1 in an open reading frame. The entire PAI-1 coding sequence was isolated from plasmid pUC8/PAIex DNA (4) using the restriction endonucleases ApaLI and BglII and inserted in the vector preceded by an XhoI-ApaLI linker (sequence of the upper strand) 5'TCGAGGTAAAAAAGAAGACTTCGACAT-

CTACGACGAAGACGAAAACCAGTCTCCGATCGGTCCGGCT-
GGTCCGGCTG3').This linker encodes the amino acid sequence of an epitope (E) of human factor VI11 for monoclonal antibody CAg69 (amino acids KKEDFDIYDEDENQSP) (29), followed by a cleavage site (c) for collagenase (amino acids IGPAJGPA), together preceding the amino terminus of mature PAI-1. This construct results in the PAI-1 expression plasmid designated pMBLll/EcPAI-I. The codons used to create the linker were based on the preferred codon usage in E. coli (34). The nucleotide sequence, containing the open reading frame encoding PAI-1, was established by DNA sequencing. The presence of the epitope on the PAI-1 fusion protein allows immunoaffinity purification of any PAI-1 variant. After purification the epitope can be removed from the protein by cleavage with collagenase.
Construction of PAZ-I Mutants-Site-directed mutagenesis was used to construct the PAI-1 variants encoding the mutants PAI-1 R346M, PAI-1 M347S, and PAI-1 R346M M347S and performed as described (35). Both strands of a linker (sequence of the upper strand

Vitronectin and the Interaction between PAI-1 and t-PA
corresponding fragment of the expression plasmids pMBLll/EcPAI-PA1 R346M M347S. 1 to yield the plasmids pEc-PA1 R346M, pEc-PA1 M347S, and pEc-Expression and Purification of PAZ-1 and PAI-1 Mutants-Transformed cells were grown in 2 X YT (33), containing ampicillin (150 mg/liter) and L-tryptophan (250 mg/liter), to a density of 4 X 10' cells/ml. The cells were centrifuged for 10 min at 10 000 X g, washed in M9 minimal medium (33), and again collected by centrifugation, resulting in the removal of tryptophan and derepression of the promoter-operator. To express PAI-1 (variant) proteins, the cells were resuspended in 250 ml of M9 medium containing ampicillin (150 mg/ ml) and 0.25% (w/v) acid-hydrolyzed casamino acids, devoid of tryptophan, and grown for 5-6 h. For metabolic labeling of the proteins with ['"S]methionine, the casamino acids were replaced by 1 X RPMI amino acids, lacking tryptophan and methionine, and 1 mCi/liter [:"SS]methionine. Subsequently, the cells were pelleted by centrifugation, washed in 20 mM Tris-HC1 (pH 7.0), 100 mM NaC1, and finally resuspended in 25 ml of 20 mM Tris-HCl (pH 7.0), 100 mM NaCI, 10 mM EDTA, and 0.1% (v/v) Tween 80. After sonication, three times for 45 s, debris was pelleted by centrifugation for 20 min a t 45 000 X g and discarded. The supernatant was dialyzed against 20 mM Tris-HC1 (pH 7.0), 25 mM NaCI, and 0.1% (v/v) Tween 80 (69-binding buffer) and subsequently added to the monoclonal murine antibody CAg69 coupled to Sepharose beads and incubated overnight a t 4 "C by end-over-end rotation. After washing the Sepharose three times with 69-binding buffer and twice with 69-binding buffer supplemented with 100 mM NaC1, PAI-1 was eluted from the monoclonal CAg69-Sepharose by an end-over-end incubation for 1 h in 69-binding buffer supplemented with 1 M NaCI. Subsequently, the CAg69 epitope and the collagenase cleavage sequence could be removed by adding CaCI2 to a final concentration of 5 mM to the eluate and incubating for 1 h at 37 "C with collagenase-Sepharose. Finally, the PAI-1-containing solution was dialyzed against 20 mM Tris-HC1 (pH 7.2), 200 mM NaCl, 0.1% (v/v) Tween 80. Then, Q-Sepharose Fast Flow beads, washed with the same buffer, were added. After a 30-min incubation, the supernanant was taken and dialyzed against 20 mM Tris-HC1 (pH 7.8), 150 mM NaC1, 0.1% (v/v) Tween 80, resulting in an apparently homogeneous PAI-1 preparation, as determined by silver staining after SDS-PAGE. Purified endothelial cell-derived PAI-1 (ECCM PAI-1) was obtained as described previously (36). ECCM PAI-1 and the nonmutated rPAI-1 proteins from E.
coli extracts exhibited identical second-order association rate constants for the reaction toward t-PA (data not shown).
Determination of PAI-1 Actiuity-Initially the activity of purified PAI-1 and PAI-1 mutants was determined by SDS-PAGE followed by reverse-fibrin autography (37). A more detailed analysis of the activity of PAI-1 was done as follows. PAI-1 was activated by dialysis for 1.5 h a t room temperature against 6 M guanidinium chloride containing 0.1% (v/v) Tween 80. The denaturant was removed by overnight dialysis at 4 "C against 20 mM Tris-HC1 (pH 7.8), 150 mM NaC1, 0.1% (v/v) Tween 80 (TBST buffer). The activity of the resulting PAI-1 preparations was determined by titration against two-chain Bowes melanoma t-PA. To that end, 2 nM t-PA was incubated in the presence or absence of 12 nM vitronectin a t 37 "c in a final volume of 25 pl of TBST buffer with increasing concentrations of PAI-1 or PAI-1 mutants. After a 1-h incubation, the reaction was stopped by the addition of 150 pl of 100 mM Tris-HC1 (pH 8.4), 0.1% (v/v) Tween 80, and 50 pl of 5 mM S2288. The residual t-PA activity was determined by continuously measuring substrate conversion a t 405 nm in a Titertek Twinreader (Flow Laboratories). Under these conditions, conversion of the substrate by t-PA as well as by urokinase-type PA was linear for a t least 6 h. To determine whether inhibition of t-PA activity coincides with the formation of proteaseinhibitor complexes, activated PAI-1 (variants) were incubated for 1 h a t 37 "C in TBST buffer with 3 nM "S-labeled t-PA in the presence or absence of 12 nM vitronectin. The formation of SDS-stable protease-inhibitor complexes was analyzed by SDS-PAGE followed by fluorography. Determination of PI-PI' by Amino Acid Sequencing-Amino acid sequence analysis was used to determine the amino-terminal amino acid of the carhoxyl-terminal postcomplex peptide of PAI-1 variants having an altered P1 residue. For that purpose a complex was formed in the presence of vitronectin between PAI-1 R346M M347S and two-chain Bowes melanoma t-PA for 2 h at 37 "C. After the addition of an SDS-loading huffer without a reducing agent, the sample was immediately n~~h i e r t~d tn SDS-PAGE (Mini-Protean 11, Bio-Rad) for 50 min at ?On 1 ' .lnrl .1 C' and whsequently blotted (40 min a t 100 V and 4 "C ~n lo:! " 1~ ~IV(.:IIP, 2.3 msq Tris-HCl (pH 8.3) in 35% (v/v) methanol) onto a polyvinylidene difluoride membrane (Immobilon-P, Millipore, Bedford, MA). After an incubation of 2 min a t room temperature in 1 M NH,OH to dissociate the complex, the filter was quickly washed three times with distilled water and air dried. After staining the filter with Coomassie Brillant Blue R-250, the band representing the complex between the mutant PAI-1 R346M M347S protein and t-PA was excised and directly sequenced by automated Edman degradation (Applied Biosystems model 477A) (38). The liberated phenylthiohydantoins were identified by means of on-line high performance liquid chromatography (Applied Biosystems model 120A).
Inhibition Kinetics-The second-order association rate constants (k,) for the inhibition of two-chain t-PA, two-chain full-length rt-PA, or of a two-chain rt-PA mutant (t-PA del.FEKlK2), lacking the carboxyl-terminal B-chain, with PAI-1 and PAI-1 variants was determined as follows. 1.5 nM of either full-length rt-PA or t-PA del.FEKlK2 was incubated a t 37 "C in TBST buffer with 2 nM active PAI-1. The concentrations of active inhibitor were titrated in the presence of vitronectin as described above. At various incubation periods ranging from 7.5 s to 10 min, the reaction was stopped by a 40-fold dilution of the reaction mixture with 1. Binding of PAI-1 to Fibrin Matrices-The preparation of fibrin matrices in 24-well tissue culture plates was done essentially as described (39). To prevent aspecific binding, wells were coated for a t least 16 h a t 4 "C with 2% (w/v) bovine serum albumin in 50 mM sodium carbonate (pH 9.4). After coating, the wells were washed twice with phosphate-buffered saline containing 0.1% (v/v) Tween 80 and 0.1% (w/v) bovine serum albumin (PTB buffer). To each well 250 pl of fibrinogen (0-1 mg/ml), completely devoid of plasminogen and vitronectin, was added in PTB buffer to 25 pl of human thrombin (10 units/ml). The fibrin matrix was then air dried at 37 "C. Prior to use, the matrix was incubated for 30 min at 37 "C with 300 pl of PTB buffer and subsequently washed three times with PTB buffer. '"S-Labeled PAI-1 was added to 200 p1 of P T B buffer in a fibrin-coated well and incubated for 2 h a t 37 "C. The supernatant was then quantitatively removed, and after the addition of 200 pl of 2% (w/v) SDS, the radioactivity was determined by liquid scintillation counting. The wells were rapidly washed twice with 1.5 ml of P T B buffer. The bound PAI-1 was eluted a t room temperature by gently shaking with 200 pl of 1% (w/v) SDS. After transfer of each eluate to a scintillation vial, the well was washed with 100 pl of 1% SDS, which was subsequently transferred to the corresponding vial to determine the amount of radioactive PAI-1.

RESULTS
Declerck et al. (18) have reported that vitronectin stabilizes the active conformation of PAI-1. Our interpretation of this finding is that the binding of PAI-1 to vitronectin allows a more stable exposure of distinct amino acids of PAI-1 leading toward a tight interaction with t-PA or urokinase-type PA.
We hypothesize that these amino acid residues are not necessarily those that constitute the reactive center P1-Pl'(R346"347).
Consequently, to study the role of vitronectin in the interaction between PAI-1 and t-PA, the PAI-1 reactive site residues were replaced. The choice for those novel residues is based on the corresponding amino acids of alantitrypsin, a serine protease inhibitor (serpin) that does not inhibit t-PA or urokinase-type PA. The P1 arginine (R346) of wild-type PAI-1 was replaced by a methionine (M346) and/ or the PI' methionine (M347) by a serine (S347). The constructed PAI-1 mutants PAI-1 R346M, PAI-1 R346M M347S, and PAI-1 M347S were expressed in E. coli, purified to homogeneity, and analyzed by reversed-fibrin autography (Fig.  1). In accord with the consensus on the dominating role of the P1 residue for target protease specificity (7-9), we observe  in the Pl position are inactive. In contrast, PAI-M347S exhibits a specific activity comparable to that of wild-type PAI-1.
It is well documented that PAI-from conditioned medium of cultured vascular endothelial cells or from extracts of E. co/i transformed with PAI-cDNA is encountered as a mixture of active and latent PAI- (4,11,40). The latent form can be activated by treatment with e.g. guanidinium chloride (41-43). In addition, we recently reported that the active form of PAI-specifically binds to fibrin whereas a much lower binding to fibrin was observed with the latent form (44). Thus, monitoring the capacity of PAI-and its variants to bind to fibrin can be taken as a parameter for the ability of the different PAI-preparations to be activated. Indeed, upon activation the binding of the inactive mutants PAI-R346M and PAI-R346M M347S to fibrin is similar to that of wildtype PAI-whereas there is hardly any binding of the latent forms (Fig. 2). From these data we conclude that the various PAI-derivatives differ in inhibitory activity although their abilities to be activated are identical.
To study the potential effect of vitronectin on the interaction between PAI-(variants) and t-PA or urokinase-type PA, we employed an end point assay with the direct chromo- In accordance with the analysis by reversed-fibrin autography (Fig. lC), we observed efficient inhibition both with wild-type PAI-and with PAI-M347S whereas no measurable inhibition of t-PA activity was seen with either PAI-R346M or PAI-R346M M347S (Fig. 3) can also endow PAIL1 with thrombin-inhihitory properties (45). Furthermore, we presented evidence indicating that the mechanism of these cofactors in promoting the association hetween 1'AI-l and thromhin is quite different (21, 45). Here, we show that heparin, in contrast to vitronectin, does not promote the activity of the "inactive" mutants PAI-R346M and PAI-R346M MB47S (Fig. 8). This ohservation provides additional evidence for a different mechanism of interaction between PAIL1 and either heparin or vitronectin.
To evaluate the extent of the vitronectin-induced inhibition of t-PA by the various PAI-mutants, we determined the second-order association rate constants (k,) in the presence and in the absence of vitronectin.
As shown in Table I results in a rat.e of t-PA inhibition which is only 40-%)-fold slower than t-PA inhibition by wild-type PAIeither with or without vitronectin.
The mechanism of vitronectin-promoted inhibition of t-PA hy the "Pl" mutants of PAI-was investigated further. For that purpose we monitored the formation of SDS-stable complexes, a characteristic feature of a serpin-target protease interaction, either in the presence or in the absence of vitronectin.
Analysis was performed using SDS-PA(;E (Fig. 4). Clearly, in t.he presence of vitronectin, t-PA inhibition by PAI-R346M coincides with the formation of equimolar, SDS-stable complexes of t-PA and PAI-RB46M. Identical results were obtained with the mutant PAI-I R346M M347S. In contrast, no complex formation is observed in the absence of vitronectin.
To establish whether the M346-M/S347 peptide bond of the PAI-variants is attacked by the catalytic center of t-PA, analogous to the R:34fj-M:147 peptide bond of wild-type PAI-1, we determined the amino-terminal amino acid sequence of the postcomplex peptide of one of the variants. For that purpose, a complex was formed between PAI-R346M M347S and t-PA in the presence of vitronectin.
The complex was isolated, subsequently dissociated on a filter by treatment with NH.,OH, and the filter was directly used for automated amino acid sequence analysis. Each cycle of the l<drnan Debradation resulted in the identification of four amino a('iti derivatives in equimolar amounts. These residues represent the known amino termini of PAIL1 (51, of the A chain 01' I-PA, and of the H chain of t-PA (46) and an addit ional forrrt h sequence. The latter amino acid sequence is recorded as: sequences that are present in equimolar amounts (data not shown). In agreement with ohscrvat ions on co~nplr~x formation hetween wild-type I'AI-I and I-I'A (C), the dissociation of the complex and concomitant rele;lsca of the postcomplex peptide were effectively prclmoted t)y t rest merit 01' the complex with NH,OH and resulted in equimcllar iIm(l\lnts of all four amino-terminal sequences. It should he noted that the amino-terminal sequence of the postcom~~lex pept ide with or without treatment of the complex with NH ,OH is identical.
We conclude that the novel M346-M/S347 peptide bond of the PAI-1 variants functions as the P1-P1' bond for t-PA, in a manner similar to the R346-M347 peptide bond of wildtype PAI-1.
The vitronectin-dependent inhibition with the inactive mutants PAI-1 R346M and PAI-1 R346M M347S was analyzed further by using another target protease of PAI-1, i e . high molecular weight urokinase-type PA. Results similar to those for the inhibition of t-PA were obtained, namely that inhibition of urokinase-type PA by the PAI-1 mutants is only observed in the presence of vitronectin. Surprisingly, however, low molecular weight urokinase-type PA, containing essentially only the protease domain, is inhibited to a significantly lower extent than high molecular weight urokinase-type PA (data not shown). This unexpected finding may indicate an involvement of the non-protease part of urokinase-type PA, and possibly of t-PA, in the vitronectin-dependent inhibition. To investigate this option in more detail, we compared the inhibition of either full-length recombinant t-PA and a recombinant t-PA mutant (t-PA del.FEKlK2) which contains only the protease domain ( i e . "B" or "light" chain) by the mutant proteins PAI-1 R346M and PAI-1 R346M M347S. As a control, it is shown that wild-type PAI-1 inhibits t-PA and t-PA del.FEKlK2 to the same extent both in the presence or in the absence of vitronectin (Fig. 5). Interestingly, both in the end point assay and in the determination of the second-order association rate constants ( k l ) , the presence of vitronectin causes a significant difference between the k1 of the PAI-1 mutants with either t-PA or t-PA del.FEKlK2. For the inhibition by PAI-1 R346M, the full-length protein displays a k, value of 8 X lo5 M" s" whereas t-PA del.FEKlK2, lacking the amino-terminal A ("heavy") chain, displays a 6-fold lower kl (1.

DISCUSSION
The nature of the P1 amino acid residue undoubtedly is of crucial importance for the specificity of a particular serine protease inhibitor (8). This is best illustrated by the pathological manifestations of a patient who presented a bleeding diathesis caused by the substitution for the P1 residue (M) of al-antitrypsin (al-AT) by an arginine (R) residue (al-AT "Pittsburgh"), being the P1 residue of e.g. antithrombin I11 (48). As a result, the patient had an acquired increased antithrombin activity and a simultaneous lack of elastase-inhibitory activity. Our data on the properties of PAI-1 mutants having an altered P1 residue are in agreement with the view that this residue is a prime determinant for the target specificity of a serpin. Replacement of the P1 arginine residue (R346) of PAI-1 by a methionine virtually inactivates the inhibitor. However, as will be discussed in a following paragraph, apart from the P1 residue, clearly other amino acid residues also perform an essential function in the specific interaction with the target serine proteases t-PA and urokinase-type PA.
The importance of the nature of the P1' residue of a serpin for the target specificity is less obvious than that of the P1 residue. In some cases the alteration of the P1' residue, notably the replacement of the P1' serine (S) residue of c y Iantitrypsin by alanine (A) (49) and the conversion of the antithrombin I11 P1' serine (S) residue to leucine (L) (50), results in mutant inhibitors with decreased efficiency toward their respective target proteases elastase and thrombin. In general, however, the nature of the P1' residue seems to be less critical than that of the P1 residue although limitations have been proposed with regard to the size and the hydrophobicity of the side chains (51). This view is supported further by the properties of the P1' mutant protein PAI-1 M347S described here. Clearly, neither the specificity nor the rate of association with t-PA is affected by this alteration. Similarly, another mutant, PAI-1 M347V, displays a kl similar to that of wild-type PAI-1.6 From these observations we conclude that the nature of the P1' residue of PAI-1 is relatively insignificant for the efficacy and specificity of this inhibitor.
The difference in the rate of association between full-length t-PA and t-PA lacking the A chain (and high and low molecular weight urokinase-type PA) toward the PAI-1 variants R346M and R346M M347S suggests a role for the respective non-protease chain in the inhibition reaction. Recently, similar observations have been made on the involvement of in particular kringle 2 on the A chain of t-PA in the interaction with PAI-1 (15,52). First, it was demonstrated that PAI-1 binds with low affinity to isolated kringle 2 and that the isolated kringle 2 partly abolishes the inhibition of intact t-PA by PAI-1 (15). Second, using reaction conditions different from those employed here it was reported that the presence of kringle 2 within the t-PA molecule slightly decreases the rate of association between t-PA and PAI-1 (52). It should be noted, however, that in our hands in the absence of vitronectin the second-order association rate of PAI-1 with either t-PA or t-PA del.FEKlK2 is not significantly different. We assume that the apparent discrepancy with the aforementioned report is a result of the quite distinct conditions during the inhibition reaction. At face value our data and those of the others mentioned above coincide in the notion that the non-protease A chain of t-PA and urokinase-type PA is implicated in the interaction with PAI-1. At present, the available data are insufficient to describe the interaction of the non-protease part of t-PA with PAI-1 in more detail.
In this paper we report that the presence of vitronectin results in the conversion of virtually inactive P1 mutants of PAI-1 (PAI-1 R346M and PAI-1 R346M M347S) into potent inhibitors of plasminogen activators. Furthermore, we demonstrate that in the presence of vitronectin the peptide bond M346-M/S347 of the PAI-1 mutant proteins is attacked by t-PA. In the absence of vitronectin, no complex formation is detected, and attack of the M346-M/S347 peptide bond does not occur. Hence, we conclude that in the presence of vitronectin, residues M346 and MIS347 perform the function of P1 and P1' residues, similar to the genuine P1 (R346) and P1' (M347) of wild-type PAI-1 both in the presence and absence of vitronectin. In view of the concept of a serpin acting as a pseudosubstrate (7,8), our finding that the novel M346 residue of the mutant PAI-1 proteins effectively functions as the P1 reactive center residue is surprising. According to this concept and supported by the altered phenotype of CUIantitrypsin "Pittsburgh" mentioned before (48), the P1 residue should be identical to the P1 residue of the substrate, being arginine (R560) of plasminogen (10) in the t-PA/PAI-1 system. However, although vitronectin increases the association rate between PAI-1 R346M (R346M M347S) and t-PA about 1,000-fold, the interaction between wild-type PAI-1 and t-PA is still 40-50-fold faster, illustrating the important contribution of the arginine at the P1 position of PAI-1.
To explain the role of vitronectin in the inhibition of plasminogen activators by PAI-1, we attempt to fit our observations in the best current model for inhibition of serine proteases (E) by their cognate serpin (I) (7). This model distinguishes the generation of a reversible complex (EI) followed by conversion to an SDS-stable tight complex (ET) and, finally, dissociation of the protease ( E ) and the cleaved inhibitor (I'). The inhibition of t-PA by PAI-1 is characterized by the following pertinent observations. ( a ) In uiuo, PAI-1 is encountered as a complex bound to its binding protein vitronectin (17-20); ( b ) the addition of t-PA to vitronectin. PAI-1 complexes results in the formation of t-PAePAI-1 complexes devoid of vitronectin (18); (c) no cleavage of PAI-1 R346M or PAI-1 R346M M347S by t-PA is observed in the absence or presence of vitronectin, as is the case for wild-type PAI-1; and ( d ) there is no evidence for a direct interaction between t-PA and vitronectin; neither by studies using immobilized vitronectin nor by immunoprecipitation techniques could association between t-PA and vitronectin be detected (data not shown). In the absence of vitronectin, the Er complex between t-PA and the PAI-1 R348 M mutants is not formed, as illustrated by the lack of SDS-stable complexes. Apparently, the altered P1 residue of PAI-1 is not positioned appropriately, and proteolytic attack on the P1-P1' peptide bond does not occur. However, the formation of the reversible (EI) complex cannot be excluded, and arguments can be advanced which support this possibility. The formation of the EI complex may be mediated, at least in part, by the positively charged amino acid residues (K296, R298, R299) of t-PA which are distinct from the catalytic triad but are crucial for t-PA inhibition by PAI-1 (14). It is conceivable that the counterparts of these basic t-PA residues on PAI-1 do not comprise the reactive center that interacts with the catalytic triad of t-PA but are located elsewhere on the protein. From the observation that the mutant proteins bind equally well to fibrin as wild-type PAI-1 (Fig. 2) we infer that the overall structure of these proteins is comparable. Consequently, we deduce that amino acids of the PAI-1 mutants, supposed to interact with the basic region of t-PA to form the reversible EI complex, interact in a manner similar to those of wildtype PAI-1.
In the presence of vitronectin efficient generation of the SDS-stable Er complex can be observed without a subsequent cleavage reaction. Consequently, with the use of PAI-1 reactive site mutants as a model system, we deduce the following alternatives for the function of vitronectin in the inhibition reaction. (a) Vitronectin may promote the formation of the reversible complex (EI) or decrease the decay of such complexes, and/or ( b ) vitronectin may increase the formation of the subsequent tight SDS-stable complexes (Er). For both options we conclude that vitronectin exerts its function by interacting with PAI-1. We are currently investigating the different alternatives for the action of vitronectin on the inhibition reaction. It is anticipated that such studies may provide additional insight in the role of vitronectin in the inhibition of t-PA by wild-type PAI-1.
In general, the data reported here indicate that vitronectin substantially alleviates a suboptimal inhibitory condition for t-PA or urokinase-type PA. Although in this study such conditions are created by using PAI-1 P1 mutants, we envision that in uiuo suboptimal inhibitory conditions may occur as well. Specifically, it has been reported that phosphorylated urokinase-type PA is partly refractory to inhibition by PAI-1 (53). Furthermore, it has been shown that t-PA complexed with a2-macroglobulin, but fully able to activate plasminogen, is inhibited by PAI-1 at an approximately 1,000-fold slower rate (30). Possibly, under those conditions, vitronectin may restore full inhibition of the respective plasminogen activators by PAI-1.