Mutational and Immunochemical Analysis of Plasminogen Activator Inhibitor l*

We have undertaken a structural and functional analysis of recombinant plasminogen activator in- hibitor type 1 (PAI-1) produced in Escherichia coli using site-directed mutagenesis and immunochemistry. Expression of recombinant PAI-yielded an inhibitor Mutations in both the reactive center and residues (Arg-Met) and a secondary binding site for

Tissue-type plasminogen activator (tPA)' and urokinase (UPA) are serine proteases that catalyze the proteolytic cleavage of the inactive precursor plasminogen to the active protease plasmin. Plasminogen activation is regulated through several mechanisms, including the controlled synthesis and secretion of plasminogen activators and the modulation of their enzymatic activity by specific inhibitors (Gerard and Meidell, 1989). Both plasminogen activator inhibitor 1 (PAI-1) and plasminogen activator inhibitor 2 (PAI-2) are members of the serine proteinase inhibitor (serpin) superfamily (Carrel1 et al., 1987;Ny et al., 1986;Sprengers and Kluft, 1987 Pannekoek et al., 1986;Ye et al., 1987). By analogy with other serpins (Carrel1 and Travis, 1985), the best current model for the inhibition of plasminogen activators by PAI-postulates that a loop of amino acid residues in the serpin molecule (termed the reactive center) resembling the normal substrate binds to the active site of the protease. However, instead of proteolytic cleavage and release of the products, a covalent bond forms between the catalytic serine residue of the plasminogen activator and the Pl residue of PAI-and irreversibly inactivates the enzyme.
Members of the chymotrypsin family of serine proteases demonstrate selectivity toward their cognate inhibitors that is determined in part by surface loops of amino acid residues which interact with the inhibitor (Bode et al., 1989;Carrel1 et al., 1987;Madison et al., 1989Madison et al., , 1990." Proteins of the serpin superfamily are most divergent in sequence within the reactive center (Ye et al., 1987), suggesting that these residues are important determinants of the specificity toward their cognate proteases. In support of this hypothesis, a change in the Met-Ser reactive site of al-antitrypsin to Arg-Ser alters the specificity of the serpin for elastase and converts it into an effective inhibitor of thrombin and trypsin-like enzymes (Carrell and Travis, 1985). Similarly, mutagenesis of n2-antiplasmin by deletion of the reactive center Pl arginine residue to yield a Met-Ser reactive site increases its rate of inhibition of elastase (Holmes et al., 1987). While PAI-and PAI-share 38% amino acid sequence identity and probably have similar tertiary structures (Carrel1 et al., 1987) sequence divergence in their reactive centers may account for rate constants of inhibition of both uPA and tPA by PAI-that are lo-loofold faster than those of PAI- (Sprengers and Kluft, 1987). Consequently, PAI-is the principal plasminogen activator inhibitor in plasma, even in the presence of high levels of PAI- (Jorgensen et al., 1987). In the present study, the functional significance of the reactive center Pl and Pl' residues of PAI-was assessed by mutational analysis. A "second site" on PAI-which contains the sequence Asp1"~-Leu""'-Lys'"4 was also chosen for mutagenesis. By alignment of the primary structures of PAI-and al-antitrypsin and examination of the three-dimensional structure of the antitrypsin molecule (Loebermann et al., 1984), these residues have been localized to one end of the molecule between sheet 2A and helix E (Ye et al., 1987). Although located in the nonhomologous carboxyl-terminal extension of c&antiplasmin, an identical tripeptide sequence corresponds to one of three potential binding sites for the kringle region of plasmin (Sasaki et al., 1986;Hortin et al., 1988). Since plasminogen activators also contain kringles, second site substitution mutants in the Asp"" and LYsI"~ residues of the PAI-protein were constructed in an attempt to disrupt any potential charged-pair interactions between these residues and those in plasminogen activator. To assess the functional significance of the reactive center and second site sequences as determinants of the inhibitory activity and protease specificity of  wild-type and mutant PAI-molecules have been expressed in Escherichia coli and used both in enzyme inhibition assays to measure the rate constants of inhibition of the serine proteases tPA, uPA, and thrombin, and in complementary immunochemical experiments using anti-peptide antibodies. The results of these studies show that the reactive center sequences of PAI-are a primary determinant of inhibitor specificity and activity and suggest that other undefined determinants outside of the Pl and Pl' residues play an important role in the inhibition of serine protease activity. per liter, 6.0 g of Na2HPOI, 3.0 g of KH,PO,, 0.5 g of NaCl, 1.0 g of NH,Cl, 0.5 g of MgS04. 7H20, 5.0 g of Casamino acids, 10.0 g of glucose, 10 ml of glycerol, 1 mg of thiamine-HCl, and 25 mg of ampicillin.
Cell extracts were prepared from 50.ml cultures by a modification of the published method  as follows. Cells were pelleted by centrifugation, washed in 20 ml of 50 mM Tris-HCl, pH 8.0, 1 mM EDTA, recentrifuged, and resuspended in 3.6 ml of the same buffer on ice. Extraction was accomplished by the addition of 0.2 ml of 10 mg/ml lysozyme for 20 min, 0.08 ml of 10% nonionic detergent Nonidet P-40 for 10 min, and 0.2 ml of 5 M NaCl for 10 min. The cells were briefly disrupted in a sonifier/celI disruptor (Branson) to reduce the viscosity before centrifugation at 15,000 x g for 30 min at 4 "C. Glycerol was added to the clarified cell lysates to a concentration of lo%, and the lysates were stored in aliquots at -80 "C until use. rPAI-1 was purified as previously described for use as an antigen   (Hekman and Loskutoff, 1985). Inhibition of Plasminogen Actiuators-The details of kinetic inhibition assays for all enzymes are given in Table I. Because PAIactivity is labile at 37 "C, all preincubations of enzymes and inhibitors were performed at 23-25 "C to stabilize the inhibitor activity. The inhibition of tPA by rPAI-1 in crude extracts of E. coli was studied under pseudo-first order conditions. The inhibition of uPA was measured under second order conditions with an equimolar concentration of inhibitor.
The enzyme inhibitor mixes were preincubated in microtiter plate wells for different time periods before 5-fold dilution by the addition of the substrate mixture. The microtiter plates were then incubated at 37 "C, and the absorbance at 405 nm was monitored over time to determine residual enzyme activity.  were constructed in the reactive center and in a second site in order to assess the functional significance of these two structural regions. Specifically, the Pl arginine and the Pl' methionine residues (Arg.'4"-Met14') in the reactive center of rPAI-1 were changed to Arg-Ser to resemble the thrombin inhibitor, antithrombin III, to Met-Ser to resemble the elastase inhibitor, n,-antitrypsin, to Arg-Val to resemble the substrate plasminogen, and to Lys-Met to resemble a lyserpin. In the second site, the Asp"'" residue was mutated to lysine, and Lys""' was replaced by glutamic acid. The conservative mutation from Lysio4 to arginine was also constructed.
Wild-type and varient PAI-proteins were expressed in E. coli TG-1 cells grown under tryptophan-limited conditions. Cell lysates were prepared by a cleared lysate procedure and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis. Coomassie Blue staining (Fig. 1A) revealed a major protein band with an apparent molecular weight of 43,000, that comigrated with a purified human rPAI-1 standard. This band was absent from lysates prepared from cultures of E. coli TG-1 cells that had not been transformed with the rPAI-1 expression vector.
The identity of this band as rPAI-1 was verified by Western blot analysis using polyclonal anti-PAI-antibody. As shown in Fig. lB, a single immunoreactive band was evident in each lane that coincided with the M, 43,000 band detected on the Coomassie Blue-stained gel. No immunoreactive band was detected in the TG-1 bacterial cell lysate. Densitometric scanning of the gel shown in Fig. 1A and comparison to the sample of purified rPAI-1 of known concentration indicated that the cell lysates contained 0.2-0.5 mg/ml rPAI-1. This suggests that rPAI-1 represents 5-10% of the total extractable protein in the cell lysates. The crude lysates were adjusted to contain equivalent amounts of rPAI-I before analysis of their inhibitory activity.
Inhibition of uPA by Wild-type and Mutant rPAI-I-Cornparison of the inhibitory functions of wild-type and mutant rPAI-1 required an accurate quantitation of the concentration of active rPAI-1 in each lysate sample. This was accomplished by preincubating increasing quantities of the rPAI-1 extracts with uPA at 23 "C for 45 min. As a control, an equivalent amount of non-PAI-I producing E. coli TG-1 lysate was preincubated with uPA to measure any inherent uPA inhibitory activity. The fraction of uPA remaining active was then determined using the indirect chromogenic assay. Non-PAIl-producing E. coli TG-1 cell lysates lack uPA inhibitory activity, whereas lysates prepared from cells harboring either the wild-type or mutant rPAI-1 expression vectors inhibited uPA in a dose-dependent manner (Fig. 2). Under the experimental conditions used, 5 ng of total wild-type rPAI-I resulted in a 50% inhibition of uPA activity. The reactive center mutants rPAI-1 (Rs4"-+K), rPAI-1 (M'4'-S), and rPAI-1 (M347+V), as well as the second site mutants rPAI-1 (K""+ R) and rPAI-1 (K""'+E) appeared to be almost as active as wild-type rPAI-1 in inhibiting uPA. However, replacement of the second site aspartic acid (D1"2) residue with lysine resulted in a variant that appeared to be about 40-fold less active than wild-type rPAI-1). The PI-PI' double mutant rPAI-I (R"46,M'147+M,S) failed to inhibit uPA under the conditions used and was indistinguishable from the non-PAI-producing E. coli lysate. I'rtnc~l A, Coomassie Blue staining pattern ol'crutle bacterial cell lysates analyzed on a 12.5% polyacrylamide :'el hy sodium doclecyl sull'ate-pol?racr?ilamide gel electrophoresis. Pnnc/ 14, Western blot analysis of cell lysate proteins f'ractionated on a 12.5% polyacrylamide gel and transf'erred to nitrocellulose membrane. PAI-antigen was detected using polyclonal rabbit anti-PAI-as f'irst antibody and the Rio-Rad immunoblot alkaline phosphatase detection system. The nrrows indicate the rPAI-1 protein.
The apparent differences in the anti-uPA activities illustrated in Fig. 2 can be attributed either to kinetic differences in the rate of interaction of the different rPAI-1 proteins with uPA, or to differences in the fraction of active rPAI-1 present in each lysate. To resolve this question, uPA was preincubated with increasing concentrations of rPAI-1 protein at 23 "C for periods up to 20 h, rather than 45 min, to allow the uPA/ PAI-interaction to go to completion. The fraction of active uPA remaining was determined and the active concentration of rPAI-1 present in the different extracts was obtained by extrapolation to 100% inhibition on a plot of percent inhibition versus the total amount of rPAI-1 protein preincubated with uPA. From this analysis, it was calculated that only about 13% of the total wild-type rPAI-1 was present in an active form. For all other samples except rPAI-1 (D'"'+K) and rPAI-1 (R"",M"'+M,S), the fraction of active rPAI-1 present in the lysates tested was at most 2-fold less than that for wild-type rPAI-1. However, only 0.3% of the rPAI-1 (D""+K) was present in an active form, which is about 40fold lower than that observed for wild-type rPAI-1. The PI-P,' variant rPAI-1 (R"",M"'+M,S) could not be titrated since no inhibition of uPA was achieved at any rPAI-1 concentration used. The apparent differences in uPA inhibitory activities illustrated in Fig. 2, therefore, reflect differences in the fraction of active PAI-present in each lysate rather than differences in the rates of interaction of uPA with wild-type and mutant rPAI-1. The low percentage of active inhibitor in the variant rPAI-1 (D"'"-+K) is likely a result of the improper folding of this protein. Preincubation of uPA with equivalent active concentrations of wild-type and mutant rPAI-1s resulted in the same extent of inhibition of uPA. To verify this result, the kinetics of inhibition of uPA by wild-type and mutant rPAI-1 were measured under second order conditions (Table II). Consistent with the data in Fig.  2, wild-type and all mutant rPAI-1s (except rPAI-1 (R'"',M:"'-+M,S)) appear to be equally rapid inhibitors of uPA.
Inhibition of tPA by Wild-type and Mutant rPAI-l-As with uPA inhibition, except for rPAI-1 (R"",M""+M,S), wild-type and mutant rPAI-1 inhibited tPA in a dose-dependent manner and the titration curves were almost identical to those shown in Fig. 2 (data not shown).
Pseudo-first order kinetics of tPA inhibition at 23 "C were measured for each inhibitor and the data are summarized in Table II. Consistent with previously published results (Hekman and Loskutoff, 1988b), the rates of inhibition of tPA by wild-type and mutant rPAI-1s are almost an order of magni-
Neither the reactive center nor the second site mutants in the PAI-protein significantly affected the rate of inhibition of tPA by rPAI-1, with the exception of the double pointmutation (R"46,M"4'+M,S) in the reactive center which totally abolished inhibitory activity.

Comparison of Wild-type rPAI-1 to Endothelial
Cell PAI-I-Determinations of the rate constants of inhibition of both tPA and uPA by rPAI-1 and human endothelial cell PAIrevealed no significant difference between the recombinant protein and the glycosylated product (Table III).
Kinetics of Inhibition of Thrombin by Wild-type and Mutant rPAI-l-To examine the effect of alterations within the reactive center and the second site on the specificity of the inhibitory activity of PAI-1, we have compared the abilities of the wild-type rPAI-1, the rPAI-1 variants, and antithrombin III to inhibit thrombin.
The inhibition of thrombin was measured in both the presence and absence of heparin (Table  IV). Wild-type rPAI-1 is only a slow inhibitor of thrombin (+ heparin kl = 8 X 10" M-' s-', -heparin K1 = 1 X 10' M-' s-l) by comparison to antithrombin III (+heparin kl = 3 x l@ M-l s-' , -heparin kl = 2 X lo4 M-' s-l). Although the rate enhancement of antithrombin III activity was low by comparison to what is usually obtained with high quality preparations of heparin, the effect was significant and was repeatedly observed. Interestingly, the rates of inhibition of thrombin by the rPAI-1s were also significantly increased by heparin, although the magnitude of this effect was slightly less than that observed for antithrombin III. PAI-contains several positively charged residues (ArgT6, Ly?', and Lys"') that may account for binding and stimulation by heparin that are localized to helix D of the proposed heparin-binding site of antithrombin III (Huber and Carrell, 1989). Substitution of the P, arginine residue with lysine (as in rPAI-1 (R"46-+K)) or methionine (as in rPAI-1 (R:'46,M"4'+M,S)) abolished thrombin inhibition. Replacement of the Pl' serine of antithrombin III with a residue containing a bulky side chain can significantly reduce the rate of complex formation with thrombin (Stephens et al., 1988). In agreement with this observation, the rPAI-1 (M"'+ V) variant showed a lo-fold lower rate of thrombin inhibition than wild-type rPAI-1. The rPAI-1 (M"474S) variant, which contains the same reactive center pair as antithrombin III, showed a slight increase in the rate of thrombin inhibition compared to wild-type rPAI-1, although this rate was still two orders of magnitude slower than that for antithrombin III. The rate of thrombin inhibition by the second-site variants showed that the conservative replacement of the wild-type Lys'"' residue with arginine (as in rPAI-1 (K'04+R)) had no effect on the antithrombin activity as compared with wildtype rPAI-1, whereas replacement of the same residue with glutamic acid (rPAI-1 (K'""+E)) decreased the antithrombin activity about lo-fold.
Inhibition of Trypsin and Elastase-The inhibition of trypsin and elastase by wild-type and mutant rPAI-1 was tested at 23 "C! under pseudo-first order conditions. Even when a lofold molar excess of inhibitor was incubated with the enzyme for periods up to 18 h, none of the rPAI-1 lysates inhibited either trypsin or elastase (data not shown). Even rPAI-1 (R'4",M'4'--tM,S), which contains the same reactive center pair of residues as ocl-antitrypsin, showed no inhibitory activity.
Neutralization of rPAI-1 Inhibitory Activity with Antibodies-Polyclonal rabbit antibodies raised against purified PAI-1 completely neutralized the inhibitory activity of a crude extract of wild-type rPAI-1 (Fig. 3). Control rabbit IgG showed no neutralization of PAI-activity even at a concen- tration 10 times higher than that required for complete neutralization by polyclonal anti-PAI-antibody. Titration of the neutralizing capacities of the anti-peptide antibodies against wild-type rPAI-1 revealed that only the anti-reactive center peptide antibody neutralized PAI-activity while the anti-second-site peptide antibody behaved identically to the control. In subsequent experiments in which a fixed amount of IgG was preincubated with increasing amounts of rPAI-1 and its variants, 10 Kg of IgG was used to ensure antibody excess.
Two types of results were obtained in these experiments and representative sets of data for each type are shown in panels A and B of Fig. 4. The data presented in panel A are representative of titrations of rPAI-1 proteins which possess a wild-type reactive center sequence and include the wild-type rPAI-1 and the variants rPAI-1 (Kln4*E), rPAI-1 (K""+R), and rPAI-1 (D""+K).
As before, the polyclonal anti-PAI-IgG showed complete neutralization of the inhibitory activity of the rPAI-1 extracts, whereas the control IgG showed no neutralization of these preparations. Examination of the curves generated using the anti-peptide IgGs again demonstrated that only the anti-reactive center IgG neutralized these inhibitor preparations, while antibody directed against the second-site peptide was indistinguishable from control IgG.
However, a different set of data were obtained for the variant rPAI-1 preparations rPAI-1 (R""6+K), rPAI-1 A. (M347-S), and rPAI-1 (M"'-+V) which contain mutations in the reactive center and an example is shown in Fig. 4, panel B. Although polyclonal anti-PAI-antibody still showed complete neutralization of rPAI-1 activity, the anti-reactive center peptide antibody no longer neutralized these variant inhibitor preparations. DISCUSSION In this study, the functional significance of the reactive center and a second site on the PAI-molecule has been assessed by both mutational analysis and anti-peptide antibody neutralization experiments.
In order to minimize the possibility of improper folding and aberrant tertiary structure, we have used substitution mutants instead of deletion mutants. A high level prokaryotic expression system was used for rPAI-1 protein production . Under optimal consitions about 1.0-2.5 mg of rPAI-1, constituting 510% of total extractable protein, were obtained per 50 ml of bacterial culture. The recombinant PAI-produced in the prokaryotic expression system was recognized by human endothelial cell PAI-antiserum in Western blot analysis and its rate constants of inhibition for both tPA and uPA were similar to those observed for authentic, human endothelial cell PAI-1. Both these results and those described previously  suggest that glycosylation of PAI-is unnecessary for inhibitory activity. As demonstrated previously for PAI-made by endothelial cells (Hekman and Loskutoff, 1985), PAI-produced in E. cob exists in two forms, one inherently active and a latent form that can be reactivated by denaturation and renaturation. By the criteria of activity, stability, and specificity, guanidine-reactivated PAI-is indistinguishable from the native, active form of the protein (Hekman and Loskutoff, 1988a). The latent form of rPAI-1 can also be reactivated with guanidine hydrochloride.' The expression of PAI-in bacteria therefore seems to be a suitable method for the production of authentic PAI-activity.
Our data on the inhibition of a panel of serine proteases confirm that PAI-is an extremely specific inhibitor of plasminogen activators.
The rate constants for the interaction between wild-type rPAI-1 and either tPA or uPA ( 106-lo7 M-' s-l) are several orders of magnitude faster than that observed for the inhibition of thrombin (8.0 X 10' M-' s-l). Additionally, neither wild-type rPAI-1 nor any of its variants inhibited trypsin or elastase in our assays, even when a large excess of inhibitor and long incubation times were used. Our results do not agree with previously published studies in which purified, guanidine-activated bovine PAI-was shown to inhibit trypsin in a dose-and time-dependent manner with a second order rate constant of 7.0 x lo6 M-' s-l at 37 "C! (Hekman and Loskutoff, 1988b). The reasons for this discrepancy are unclear but may reflect sequence differences between the human and bovine PAI-proteins, our use of crude rPAI-1 preparations, or differences in the incubation temperature (23 versus 37 "C).
One feature of Pl specificity in several different families of serine protease inhibitors is that replacement of lysine for arginine at the PI position usually maintains inhibitor strength and specificity.
Our mutagenesis experiments are consistent with this observation and also suggest a requirement for a basic amino acid at the Pl position of the reactive center of PAI-for effective inhibition of plasminogen activators. By contrast, replacement of the Pl arginine by lysine in rPAI-1 (R""6+K) or by histidine in antithrombin III Glasgow (Owen et al., 1988) results in a complete loss of inhibitory activity toward thrombin and suggests an absolute require-ment for arginine at the Pl position. Examination of the three-dimensional structure of a-thrombin cocrystallized with a synthetic inhibitor shows that this stringency is due to the insertion of 9 residues into a surface loop of thrombin which forms part of the recognition site on the enzyme for the Pl residue and effectively limits both substrate and inhibitor specificity (Bode et al., 1989). By alignment of serine proteases of the chymotrypsin family, the corresponding loop in both tPA and uPA is relatively short' and may account for proteolytic cleavage after both arginine and lysine residues.
It has been shown that the residue in the Pl' position in most inhibitor families is less constrained; virtually any residue except proline is tolerated (Laskowski and Kato, 1980). Mutagenesis of the Pl' methionine residue in the reactive center of PAI-to valine or serine did not affect the rate of interaction with uPA or tPA, which suggests both the tolerance of other residues and the lack of an absolute requirement for Met in the Pl' position for PAI-to effectively inhibit plasminogen activators. A Pl' methionine is also not required for the inhibition of plasmin by a,-antiplasmin (Holmes et al., 1987), and oxidation of methionine residues with Nchlorosuccinimide has no effect on the association rate of the inhibitor with a variety of serine proteases (Shieh and Travis, 1987). A Pl' serine is not required for the inhibition of thrombin, although extensive mutagenesis of the Pl' residue of antithrombin III has shown that there are apparently both a size optimum and hydrophobicity effects of the side chain on the rate of interaction (Stephens et al., 1988). Our results on the inhibition of thrombin by rPAI-1 and its Pl' mutants also support this conclusion.
Previous results have demonstrated that even a single amino acid change in the Pl site of the reactive center of CY~antitrypsin (Owen et al., 1983;Jallat et al., 1986;Heeb et al., 1990), a?-antiplasmin (Holmes et al., 1987), or cri-antichymotrypsin (Rubin et al., 1990) can dramatically increase the rate of inhibition of other, non-cognate serine proteases by the variant serpins. Even though variants such as rPAI-1 (R"46,M'14'+M,S) and rPAI-1 (MX4?+S) contain the same reactive center Pl-Pl' residues as al-antitrypsin and antithrombin III, respectively, contrary to our prediction there was little if any increase in either the rate or specificity of inhibition of trypsin, elastase, or thrombin by these variant inhibitors.
We suggest that this result is another aspect of the rigid specificity of PAI-for plasminogen activators. It seems that PAI-lacks structural characteristics aside from the reactive center Pl-Pl' residues that are essential for the efficient inhibition of other serine proteases. Other results' have demonstrated a functional role for the P4' and P5' residues of PAI-in the interaction with tPA, and it is possible that residues at these positions are at least partly responsible for the strength and specificity of serine protease inhibition.
The results of anti-peptide antibody neutralization of PAI-1 inhibitory activity suggest: (i) antibody directed against the reactive center of PAI-can sterically block the access of this region to the active site of plasminogen activator; (ii) antibody binding to the second site located some distance from the reactive center does not block the inhibitory function of PAI-1; (iii) antibody directed against the wild-type sequence in the reactive center will not neutralize the inhibitory activity of a reactive-center variant, indicating that this anti-peptide an-tibody is highly specific for the wild-type reactive center sequence. Interestingly, only half of the rPAI-1 activity was usually neutralized by the anti-reactive center antibody, although this may simply reflect a low affinity of the antipeptide antibody for the native protein.
Polyclonal antibody inhibition of PAI-activity is complete, even on reactive center variants of rPAI-1. One possibility is that the polyclonal antibody can still recognize and neutralize the reactive-center variant molecules due to high affinity binding. Alternatively, other epitopes on PAI-which are recognized and bound by the polyclonal antibody could account for the more effective neutralization of activity. The nature of these epitopes is currently unknown, although experiments are underway to test these alternatives.