Structural Basis for Recognition of Urokinase-type Plasminogen Activator by Plasminogen Activator Inhibitor-1

Abstract

Plasminogen activator inhibitor-1 (PAI-1), together with its physiological target urokinase-type plasminogen activator (uPA), plays a pivotal role in fibrinolysis, cell migration, and tissue remodeling and is currently recognized as being among the most extensively validated biological prognostic factors in several cancer types. PAI-1 specifically and rapidly inhibits uPA and tissue-type PA (tPA). Despite extensive structural/functional studies on these two reactions, the underlying structural mechanism has remained unknown due to the technical difficulties of obtaining the relevant structures. Here, we report a strategy to generate a PAI-1·uPA(S195A) Michaelis complex and present its crystal structure at 2.3-Å resolution. In this structure, the PAI-1 reactive center loop serves as a bait to attract uPA onto the top of the PAI-1 molecule. The P4–P3′ residues of the reactive center loop interact extensively with the uPA catalytic site, accounting for about two-thirds of the total contact area. Besides the active site, almost all uPA exosite loops, including the 37-, 60-, 97-, 147-, and 217-loops, are involved in the interaction with PAI-1. The uPA 37-loop makes an extensive interaction with PAI-1 β-sheet B, and the 147-loop directly contacts PAI-1 β-sheet C. Both loops are important for initial Michaelis complex formation. This study lays down a foundation for understanding the specificity of PAI-1 for uPA and tPA and provides a structural basis for further functional studies.

Introduction

The urokinase-type plasminogen activator (uPA)² system is composed of uPA, its cognate receptor (uPAR), and two specific inhibitors, plasminogen activator inhibitor (PAI)-1 and PAI-2 (1). Extensive studies have shown that the uPA system plays a pivotal role in cell adhesion, migration, invasion, and tissue remodeling (2–5). As a serine protease, uPA specifically catalyzes the conversion of inactive plasminogen to active plasmin, which can degrade a number of proteins in the extracellular matrix. Malignant tumor cells utilize this system to invade locally and to eventually spread to distant sites (for reviews, see Refs. 6 and 7). Studies on mice with specific gene deletions support a causative role for the uPA system in cancer spread (3, 8). In human malignant tumors, the levels of uPA, PAI-1, and uPAR are significantly higher than those in the corresponding normal tissues (1, 9). Moreover, tumor uPA and PAI-1 levels are currently recognized as being among the most extensively validated biological prognostic factors in breast cancer (10, 11).

The activity of uPA is regulated by its cell-surface receptor (uPAR), which can concentrate uPA and its catalytic activity onto pericellular regions. Another major physiological regulator is PAI-1, which reacts rapidly with both free uPA and uPAR-bound uPA. PAI-1 enhances the cellular internalization of the uPA·uPAR complex through interaction with certain members of the low density lipoprotein receptor family. Once internalized, the PAI-1·uPA complex is degraded, whereas uPAR is recycled back to the cell surface (12).

As an important member of the serine protease inhibitor (serpin) family, PAI-1 contains three β-sheets, nine α-helixes, and a flexible reactive center loop (RCL). PAI-1 inhibits its target proteases following the classical serpin mechanism of protease inhibition, including the initial (Michaelis) complex formation and the second step of covalently linked complex formation.

In plasma, PAI-1 circulates at a nanomolar concentration (13), much less compared with other serpins, e.g. antithrombin (2.3 μm) (14). PAI-1 inhibits uPA very rapidly with a second-order rate constant of 4.8 × 10⁶ m⁻¹ s⁻¹ and inhibits tissue-type PA (tPA) even faster (2.6 × 10⁷ m⁻¹ s⁻¹) (15). In addition, PAI-1 also inhibits thrombin (1.1 × 10³ m⁻¹ s⁻¹), plasmin (2.7 × 10⁴ m⁻¹ s⁻¹), and trypsin (15), albeit at much slower rates. The ability of PAI-1 to specifically and quickly recognize its target proteases has long been a subject of extensive research. During the past 2 decades, the structures of free uPA (16) and various conformations of PAI-1 have been well established (including active (17, 18), latent (17, 19), and cleaved (21–23) conformations and in complex with an inhibiting pentapeptide (24) or with the somatomedin B domain of vitronectin (25)), and >600 PAI-1 variants have been constructed to elucidate the structure/function relationship; however, no direct structural information about the PAI-1/uPA interaction has been disclosed. A major obstacle for the structural study of PAI-1 is its instability and propensity to aggregate: active PAI-1 spontaneously converts into the inactive latent form under physiological conditions with a half-life of 1–2 h (26).

Here, we report a strategy to generate a stable PAI-1·uPA Michaelis complex and present its crystal structure at 2.3-Å resolution. The structure reveals extensive active-site interactions between the PAI-1 RCL and uPA catalytic sites and also demonstrates that the exosite interactions involving the 37- and 147-loops of uPA are pivotal for the initial Michaelis complex formation. This structure also provides a model to understand the specificity of PAI-1 for tPA.

EXPERIMENTAL PROCEDURES

Protein Expression and Purification

The catalytically inactive form of the human uPA protease domain (Ile-16–Glu-244 in chymotrypsin numbering) containing the S195A mutation was expressed in Pichia pastoris yeast strain X-33 and purified as described previously (27). The recombinant stable PAI-1 mutant 14-1B containing four point mutations (N150H, K154T, Q319L, and M354I) (28) and a hexahistidine tag was expressed in Escherichia coli using the expression vector pT7-PL and C41(DE3) cells (29, 30). To prepare the serpin·protease Michaelis complex, the PAI-1 cells was incubated with a 2∼5-fold molar excess of uPA(S195A) in buffer A (50 mm KHPO₄ (pH 6.0) and 1 m NaCl). The cells were then lysed by sonication, followed by ultracentrifugation. The subsequent supernatant was applied to a nickel-nitrilotriacetic acid column (Qiagen). The column was stepwise washed with buffer A containing 40 and 80 mm imidazole. The PAI-1·uPA complex was subsequently eluted with buffer A containing 300 mm imidazole and dialyzed against 20 mm Tris-HCl (pH 8.0) and 150 mm NaCl. The complex was then further purified using a Superdex-75 column (GE Healthcare). Fractions shown by SDS-PAGE to contain the complex were pooled and concentrated to 10 mg/ml.

Crystallization and Data Collection

Crystals of the PAI-1·uPA noncovalent complex were grown at 20 °C with the hanging drop method by mixing equal volumes of protein solution and precipitant solution (1.4 m ammonium sulfate and 0.1 m Tris-HCl (pH 7.4)). Crystals grew to a size of 60 × 75 × 60 μm after 4 months. For x-ray data collection, a single crystal was transferred to a cryoprotectant solution containing 30% glycerol, 1.6 m ammonium sulfate, and 0.1 m Tris-HCl (pH 7.4) and was flash-frozen in a liquid nitrogen stream. Data collection was performed at 100 K and a wavelength of 1.0 Å at beam line BL17U of the Shanghai Synchrotron Radiation Facility (Shanghai, China). The data were processed and scaled using the HKL2000 suite (31). These crystals have hexagonal shape and belong to the P3₂21 space group with one complex in the crystallographic asymmetric unit.

Phasing and Refinement

The structure of the PAI-1·uPA complex was solved by molecular replacement using MolRep (32) with structures of the active stable variant of PAI-1 (Protein Data Bank code 1DVM) (17) and the uPA catalytic domain (code 2O8T) as searching models. To put two models in the same crystallographic origins, the translational function of PAI-1 was fixed during the search of the translational function of uPA. The model from the molecular replacement solution packed well in the unit cell without intra- and intermolecular conflicts and was subjected to rigid body refinement, giving an R-factor of 39.9% and an R_free of 38.6% at a resolution range of 84.6 to 4.0 Å. The structure was refined with iterative cycles of manual model building using COOT (33) and restrained refinement with REFMAC5 (34). The structure was validated with PROCHECK (35) and analyzed by PyMOL (36). 96.4% of the residues are in the most favored regions of the Ramachandran plot, with 2.6% in the additionally allowed region and 1% in the disallowed region. The x-ray data collection and final crystallographic refinement statistics are presented in Table 1.

TABLE 1.

Data collection and refinement statistics

Values in parentheses are for the highest resolution shell. r.m.s.d., root mean square deviation.

Data collection
Space group	P3221
Cell dimensions	a = b = 97.7, c = 171.9 Å; α = 90.00°, β= 90.00°, γ = 120.00°
Resolution (Å)	99.0-2.3 (2.34-2.30)
Rmergea	0.141 (0.321)
I/σI	21.5 (3.4)
Completeness (%)	98.7 (90.1)
Redundancy	5.9 (2.0)
Refinement
No. of reflections	80,462
Rworkb	0.22
Rfreec	0.27
No. of atoms
Non-H protein	4928
Water	314
B-factors
Protein	28.1
Water	34.3
r.m.s.d.
Bond lengths (Å)	0.0088
Bond angles	1.26°

^a R_merge = Σ|I_i − 〈I〉|/ΣI_i, where I_i is the intensity of the ith observation, and 〈I〉 is the mean intensity of the reflections.

^b R_work = Σ|F_o − F_c|/Σ|F_o|, where F_o and F_c are the observed and calculated structure factor amplitudes, respectively.

^c R_free was calculated using 5% of the data excluded from refinement.

RESULTS

Strategy for Crystallization of the PAI-1·uPA Complex

In contrast to the large majority of other serpins, wild-type PAI-1 in its active form is conformationally unstable. Under physiological conditions, active PAI-1 spontaneously converts to the inactive latent form with a half-life of 1–2 h at 37 °C (26). The first PAI-1 structure to be solved was indeed the latent PAI-1 form (19). A major breakthrough in the structural study of the PAI-1 structure was the generation of a PAI-1 quadruple mutant, N150H/K154T/Q319L/M354I (28), which extends the half-life of the active form from 2 to 145 h. Study of this protein led to the first x-ray structure of active PAI-1 (17, 18, 37). In addition, PAI-1 is prone to form aggregates (38, 39). High salt and low pH conditions are typically needed to keep PAI-1 stable in vitro (17, 40), but these conditions cause difficulty in protein crystallization using the typical vapor diffusion crystallization methods. In this study, we employed a novel strategy to avoid the polymerization of PAI-1 and to increase the yield of complex formation. Specifically, we added excess purified active site-mutated uPA(S195A) protein directly to the E. coli cells overexpressing the stable variant of PAI-1 prior to cell lysis. This PAI-1 variant contains a His tag, and thus, a nickel-nitrilotriacetic acid column was used to pull out the PAI-1·uPA complex. A gel filtration chromatography step was used to remove a small amount of the aggregated PAI-1. This procedure enabled us to obtain a high yield of the PAI-1·uPA complex. The PAI-1·uPA complex from this procedure produced little aggregate and was stable for >4 months at room temperature, even in low salt and neutral pH conditions, and finally yielded diffraction-quality protein crystals.

Overall Structure

The crystal structure of the PAI-1·uPA complex was determined at 2.3-Å resolution by molecular replacement using the structures of the active stable variant of PAI-1 (Protein Data Bank code 1DVM) (17) and the uPA catalytic domain (code 2O8T) as searching models. The structure has been refined to an R-factor of 22%, an R_free of 27%, and good geometry (Table 1). The final refined model (Fig. 1a), including the RCL, is well defined by the electron density map and comprises residues 5–379 of PAI-1 14-1B, residues 16–244 of uPA(S195A), and 314 water molecules. In the structure of the PAI-1·uPA Michaelis complex, the RCL of PAI-1 serves as bait to attract uPA onto the top of the PAI-1 molecule (above β-sheets B and C of PAI-1) (Fig. 1, a and b), forming a large interface of 1393 Ų.

FIGURE 1.

Overall structure of the PAI-1·uPA Michaelis complex and comparison with the free active form of PAI-1. a, stereo view of the crystal structure of the complex between PAI-1 (blue) and uPA (magenta). The RCL is colored yellow. b, the ribbons of the PAI-1·uPA Michaelis complex are colored according to its temperature factors (blue to red for B-factors of 11.0 to 62.5 Ų) to illustrate the mobile regions in the structure. c, superposition of PAI-1 of the PAI-1·uPA complex with the free active form of PAI-1 (Protein Data Bank code 1B3K; colored in green, with the RCL in red). The P1 residue Arg-I346 of the two molecules is shown in ball-and-stick representation. The red arrows point out the structural differences between free PAI-1 and complexed PAI-1, including α-helixes (h) H and G and β-strands (s) 1B and 2B. The inset is a close-up view with a +45° rotation along the perpendicular axis to show the RCL reposition starting at the topological switch point at P16 Ser-I331 (shown as sticks) and a maximal RCL movement of 28 Å upon complex formation.

Structural Transition Associated with PAI-1·uPA Michaelis Complex Formation

Compared with the structure of uncomplexed active PAI-1 (18), major conformational changes upon complex formation are observed in the region from P16 (Ser-I331) to P5′ (Glu-I351) of the RCL. (“I” represents PAI-1 residues, and “E” represents uPA residues. The number after “P” indicates the position of the residue N-terminal to the scissile bond; the prime indicates residues C-terminal to the scissile bond.) This region adopts a conformation that is very different from the uncomplexed active PAI-1 observed previously (Fig. 1c) (17, 18). In the free PAI-1 structures, the RCL wraps above β-sheet B of PAI-1 and adopts an extended conformation. Such a conformation is consistently observed regardless of the crystal packing (Protein Data Bank code 1B3K) (18) and crystallization conditions (code 1DVM) (17). Once having formed a complex with uPA, the PAI-1 RCL moves away from β-sheets B and C and adopts a circular conformation (Fig. 1c). It should be mentioned that this conformation also does not result from the crystal packing. The topological switch point for this transition is observed at P16 (Ser-I331). The main chain dihedral angles of this residue change from −24° and 67° to −22° and −73° upon complex formation, leading to a nearly 180° flip of the RCL peptide chain orientation and a maximal RCL movement of 28 Å (Fig. 1c). Comparison of the cleaved form of PAI-1 (code 3CVM) (22) with the current complexed PAI-1 reveals that the topological switch point for RCL to insert into β-sheet A upon scissile bond cleavage is also at P16 (data not shown).

Besides the RCL, other movements of PAI-1 upon complex formation are relatively small: 1) α-helix G moves toward uPA by ∼2.0 Å, whereas α-helix H moves in the opposite direction by ∼1.8 Å; and 2) the turn connecting strands 1B and 2B (Thr-I214–His-I219) moves closer to uPA (Fig. 1c, red arrows). Notably, there are no conformational differences in these areas when the Michaelis complex is compared with cleaved or latent PAI-1, suggesting that the conformation of PAI-1 has been primed to a latent conformation upon forming the Michaelis complex, at least in these regions. On the uPA side, little change is observed upon the formation of the initial encounter complex (with a root mean square deviation of 0.6 Å for all 980 main chain atoms compared with uPA (Protein Data Bank code 1OWE)) (41)). Taken together, the large shift in the PAI-1 RCL shows the intrinsic conformational flexibility of the RCL and indicates an induced fit mechanism for the formation of the Michaelis complex.

Interaction of the PAI-1 RCL with the uPA Active Site

The whole RCL in the current structure is in an extended conformation. The residues from P5 (Ile-I342) to P5′ (Glu-I351) are observed with well defined electron density (Fig. 2a), whereas the residues from P15 (Gly-I332) to P6 (Ile-I341) are quite mobile, as demonstrated by a much higher average temperature factor (65 Ų) of this segment compared with the average temperature factor of the entire PAI-1 (25 Ų) (Fig. 1b), even though these residues are also well resolved by the electron density map contoured at 1.0σ. This segment locates near the RCL switch point (P16). Such a high mobility of this RCL segment can be favorable for the quick insertion of the RCL into β-sheet A once the scissile bond is cleaved.

FIGURE 2.

Stereo views of the interaction between the PAI-1 RCL and the uPA active site. a, stereo diagram showing the interaction between RCL (P5–P5′) and uPA and the electron density map of the RCL (contoured at 1.0σ). The catalytic domain of uPA shown as a surface is in the standard orientation. The catalytic pockets are marked as S1, S2, and S4, respectively. The electrostatic potential is colored from −63 to +63 kilotesla/charge. b and c, stereo diagrams showing the detailed interactions between the PAI-1 RCL (yellow backbone) and uPA (magenta backbone). b, detailed interactions between the RCL P1–P3 residues and uPA in the complex. c, detailed interactions between the RCL P′ residues (P1′–P5′) and uPA in the complex.

The interaction of the PAI-1 RCL with uPA involves the residues from P4 (Val-I343) to P5′ (Glu-I351). The residues from P4 to P3′ (Pro-I349) form a tight loop within the uPA catalytic site and contact extensively with the active site of uPA: on the N-terminal side (P side) of the scissile bond, the P1 residue (Arg-I346) occupies the S1 specificity pocket (Fig. 2, a and b) and has a strong interaction with uPA by making a total of 10 hydrogen bonds with the surrounding uPA residues (three main chain hydrogen bonds from the carbonyl to Ala-E195 and oxyanion hole residues Gly-E193 and Asp-E194, two main chain hydrogen bonds from the nitrogen atom to His-E57 and Ser-E214, three hydrogen bonds from the side chain guanidine group to Ser-E190 and Gly-E219, and two charged parallel hydrogen bonds from the side chain guanidine group to the carboxylate of Asp-E189). Ala-I345 at P2 fits into the small S2 pocket formed by the imidazole side chain of His-E57 and His-E99. It makes a hydrogen bond with the side chain of Gln-E192. Ser-I344 at P3 occupies the S4 pocket and interacts with two loops of uPA: the main chain oxygen makes two hydrogen bonds with 217-loop residues Trp-E215 and Gly-E216, and the side chain hydroxyl group forms two hydrogen bonds with the 97-loop partners His-E99 and Leu-E97b. The interaction between the P4 residue (Val-I343) and uPA is relatively weak: it makes only a single van der Waals contact (4.5 Å) with Gly-E216.

On the P′ side of the RCL, Met-I347 at P1′ participates in two hydrogen bonds with Gly-E193 and His-E57 (Fig. 2c). The P2′ residue Ala-I348 makes a hydrogen bond with Val-E41. Pro-I349 at P3′ forms a van der Waals contact (4.0 Å) with Gln-E192. The P4′ residue Glu-I350 makes two bifurcated hydrogen bonds with the side chain oxygen of Tyr-E60b in the 60-loop region.

Taken together, the interactions between PAI-1 and the uPA active-site region (involving PAI-1 RCL residues P4–P3′) account for ∼64% of the total contact area. This extensive interaction is critical to compensate for the entropy penalty resulting from the loss of RCL flexibility during complex formation.

Interaction of PAI-1 with uPA Exosites

Exosite interaction between uPA and PAI-1 (defined here as contacts outside RCL residues P4–P3′) account for about 36% of the total contact surface area of the complex. In the current structure, almost all uPA exosite loops are involved in the interaction with PAI-1. Exosite loops of uPA tend to be disordered in the structures of uncomplexed uPA, e.g. the 37- and 97-loops in Protein Data Bank codes 2NWN (27) and 1OWE (41). However, most of them, especially the 37- and 147-loops, are well ordered in the current complex structure, as illustrated in Fig. 1b, in which the ribbons are colored by the temperature factors. Such lower B-factors for the 37- and 147-loops than for any other uPA exosite loops highlight their roles in complex formation.

Notably, the uPA 37-loop displays a highly positive electrostatic potential to interact with a corresponding negatively charged area on the surface of PAI-1 β-sheet B (Fig. 3a). This strong electrostatic interaction is conferred by four positively charged residues in the 37-loop (residues 34–39, YRRHRGGSVT). Of the residues of the 37-loop, Arg-E37a plays the most important role and provides a contact surface area of ∼43 Ų to PAI-1 β-sheet B: the side chain of Arg-E37a forms two hydrogen bonds with the side chain of Glu-I212; the NH1 atom of Arg-E37a participates in a cation-π interaction with the aromatic side chain of Tyr-I220 (3.6 Å) and also possibly a cation-π interaction with the aromatic side chain of Tyr-I241 (5 Å); and the carbonyl atom of Arg-E37a also makes a hydrogen bond with the hydroxyl group of Tyr-I241 (Fig. 3a). Both Tyr-I220 and Tyr-I241 partially shield the above hydrogen bonds from solvent and thus enhance the hydrogen-bonding strength. In addition, Arg-E35 forms long-distance bifurcated hydrogen bonds (4.4 Å) with Glu-I350 and is likely to be important in maintaining the uPA 37-loop in the current extended conformation by interacting with uPA His-E37 and Tyr-E60b through cation-π interactions.

FIGURE 3.

Exosite interactions between PAI-1 and uPA. a, interactions between the uPA 37-loop and nearby PAI-1. The surface of the PAI-1·uPA complex is in an orientation identical to that in Fig. 1a. The electrostatic potential is colored from −63 to +63 kilotesla/charge from red to blue. A stereo close-up view of the contact is shown on the right. b, stereo view of the interactions between the uPA 147-loop (magenta) and β-sheet C of PAI-1 (blue).

Besides the 37-loop, the 147-loop (also named the autolysis loop) is another important loop recognized by PAI-1. This loop makes contacts with the region between strands 1C and 4C of PAI-1 (Fig. 3b), accounting for 22% of the total contact surface. The carbonyl atom of Thr-E147 is hydrogen-bonded to the hydroxyl group of Ser-I183, whereas Tyr-E149 forms two hydrogen bonds with Arg-I187 and Glu-I351, respectively. In addition, Tyr-E151 can make a short-range (3.2 Å) van der Waals contact with the P2′ residue Ala-I348. Other exosite loops (the 60-, 97-, and 217-loops) also have direct interactions with PAI-1. These uPA exosite interactions with PAI-1, together with the active-site interactions, determine the specificity of PAI-1 for uPA.

DISCUSSION

In contrast to the large majority of other serpins, PAI-1 exists at a low circulating level (nanomolar concentration) (13) and a short half-life under physiological conditions (active PAI-1 spontaneously converts to its inactive latent form within 1–2 h at 37 °C) (26). Thus, the ability of PAI-1 to specifically and quickly recognize its target protease has long been a subject of extensive research. Here, we have presented the structure of the noncovalent Michaelis complex between PAI-1 and uPA and revealed the detailed mechanism at the atomic level.

As an initial step of the serpin inhibitory mechanism, the interaction between the PAI-1 RCL and the uPA active site plays a crucial role in the final covalent complex formation. In the present Michaelis complex structure, we observed that PAI-1 RCL residues P4–P5′ directly contact the uPA catalytic site, accounting for about two-thirds of the total contact area. Such an interaction with a long extended P′ region of serpin was also observed on the other Michaelis complexes paired with multispecific serpins (42), e.g. P4–P6′ in the antithrombin·S195A Factor Xa complex (14) and P4–P5′ in antithrombin·thrombin·heparin ternary complexes (43). Among those residues, the P1 residue Arg-I346 serves the most important role in the interaction with uPA by making a total of 10 hydrogen bonds with the surrounding residues in the S1 pocket of uPA. This extensive hydrogen-bonding network highlights the importance of P1 Arg-I346 for target specificity (44).

Besides the active sites, the interactions between PAI-1 and uPA exosites are also important for the formation of the initial Michaelis complex. The present structure shows that almost all uPA exosite loops, including the 37-, 60-, 97-, 147-, and 217-loops, are involved in the interaction with PAI-1. The 37- and 147-loops are the most important exosite loops as shown by their extensive interactions with PAI-1 as well as their lower B-factors compared with any other uPA exosite loops.

Compared with other serine protease such as trypsin (45), Factor Xa (14), and thrombin (43), a significant difference in uPA is its long 37-loop, illustrated by superposition of various serpin·protease Michaelis complexes (supplemental Fig. 1). This long 37-loop permits uPA to be close to PAI-1 and to form a strong interaction that includes electrostatic and cation-π bonds between the 37-loop and β-sheet B of PAI-1. Notably, the major involvement of PAI-1 β-sheet B in exosite interactions is also unique to the PAI-1·uPA complex: in most other serpin·protease Michaelis complexes, it is β-sheet C and the surrounding residues of serpin that mediate the interaction with the protease exosites (42). These structural features constitute the foundation of PAI-1 specificity for uPA.

Besides uPA, PAI-1 also recognizes tPA with high affinity and specificity. To further clarify the specificity of PAI-1 for uPA and tPA, we constructed a structural model of the Michaelis complex between tPA (Protein Data Bank code 1BDA) and PAI-1 based on the present PAI-1·uPA structure (supplemental Fig. 2). This model suggests a similar electrostatic interaction between the tPA 37-loop and PAI-1 β-sheet B. In addition, an alignment of the primary sequences of uPA and tPA shows that the tPA 37-loop (residues 34–39, FAKHRRSPGER) has even more positively charged residues than the uPA 37-loop (residues 34–39, YRRHRGGSVT), rendering extensive interactions between PAI-1 and the tPA 37-loop compared with the PAI-1·uPA complex. For example, Arg-37b is likely involved in the interaction with PAI-1 Tyr-I220, and Arg-E39, which is a threonine in uPA, is likely to form bifurcated hydrogen bonds with Glu-I351 (supplemental Fig. 3). These additional interactions may explain the faster association of PAI-1 with tPA compared with uPA and are consistent with the published mutagenesis data showing that the charge reversal mutation of tPA Arg-E39, R39E, dramatically reduces the rate of interaction between tPA and PAI-1 (46) and that the alanine substitution of PAI-1 Glu-I351 reduces the rates of Michaelis complex formation (k_a) with tPA by ∼5.4-fold, whereas the corresponding mutations have only a small effect on binding to uPA (∼1.2-fold) but no effect on binding to trypsin (47). This structural model, together with the present PAI-1·uPA complex structure, thus suggests that the protease 37-loop is one of the most important determinants of the inhibition specificity of PAI-1 for uPA and tPA.

In summary, this study have provided the structural basis for further functional studies concerning the importance of individual amino acids in the initial reaction between the plasminogen activators and PAI-1 using site-directed mutagenesis and biophysical methods. Also, the results presented here facilitate new studies of the translocation of the plasminogen activators from their position in the Michaelis complex to their position in the final complex, across the plane of β-sheet A. Before this translocation can take place, the exosite interactions have to be broken (20, 47). The strategy used in this study to generate the PAI-1·uPA Michaelis complex can also be used for further structural studies of PAI-1 and maybe other serpins.