Importance of the hinge region between alpha-helix F and the main part of serpins, based upon identification of the epitope of plasminogen activator inhibitor type 1 neutralizing antibodies.

The serpin plasminogen activator inhibitor type 1 (PAI-1) is an important protein in the regulation of fibrinolysis and inhibits its target proteinases through formation of a covalent complex. In the present study, we have identified the epitope of two PAI-1 neutralizing monoclonal antibodies (MA-33H1F7 and MA-55F4C12). Based upon differential cross-reactivity data of these monoclonals with PAI-1 from different species and on a sequence alignment between these PAI-1s, combined with the three-dimensional structure, we predicted that the residues Glu(128)-Val(129)-Glu(130)-Arg(131) and Lys(154) (at the hinge region between alpha-helix F and the main part of the PAI-1-molecule) might form the major site of interaction. Therefore a variety of alanine mutants were generated and evaluated for their affinity toward both monoclonal antibodies. The affinity constants of MA-55F4C12 and MA-33H1F7 for PAI-1 were 2.7 +/- 1.6 x 10(9) M(-1) and 5.4 +/- 1.7 x 10(9) M(-1), respectively, but decreased between 13- and 270-fold upon mutation of Lys(154) to Ala(154) or Glu(128)-Val(129)-Glu(130)-Arg(131) to Ala-Ala-Ala-Ala. The combined mutations (PAI-1-EVER/K), however, resulted in an absence of binding to either of the antibodies. Both antibodies bound to PAI-1-wt/t-PA complexes with a similar affinity as to PAI-1-wt (K(A) = 4-5 x 10(9) M(-1)). The epitope localization reveals the molecular basis for the neutralizing properties of both monoclonal antibodies. In addition, it provides new insights into the validity of various models that have been proposed for the serpin/proteinase complex, excluding full insertion of the reactive-site loop.

PAI-1 inhibits its target proteinases by the formation of a stable, inactive complex. After the formation of an initial, reversible Michaelis-like complex, the proteinase cleaves the active site of PAI-1 and forms a stable, covalent complex resulting in the inactivation of the proteinase (6,7). The structure of a covalent complex between PAI-1 and t-PA in particular, or between a serpin and its target proteinase in general, is presently unknown. However, two different models have been proposed. According to one model, the proteinase moves after the initial attack to the opposite pole of the serpin, thereby resulting in a complete insertion of the N-terminal side of the reactive-site loop (7)(8)(9). Alternatively, it was hypothesized that movement of the proteinase following the initial proteinase/ serpin interaction is less extended, yielding a complex in which the N-terminal side of the reactive-site loop is only partially inserted (10,11), accompanied by a distortion of the catalytic triad of the proteinase (6) and stabilized by multiple interactions between serpin and proteinase.
Although PAI-1 is synthesized as an active molecule, it converts spontaneously to an inactive, latent form that can be partially reactivated by denaturing agents (12). In this latent conformation, the active site is inaccessible for the target proteinases as a result of the insertion of the N-terminal side of the reactive-site loop in ␤-sheet A of the PAI-1 molecule (13). In addition, a third conformation with substrate properties has been identified (14 -16). This form of PAI-1 reacts with t-PA or u-PA, resulting in a cleavage of the active site of PAI-1 but without the formation of a covalent complex (17).
Previously, we have characterized a panel of monoclonal antibodies that neutralize PAI-1 activity by converting the active pathway into the non-inhibitory substrate pathway (18). For two of these antibodies, MA-55F4C12 and MA-33H1F7, the binding region was found to be located remote of the reactivesite loop, within a region comprising residues at positions 128 -156 (19). Within the three-dimensional structure of PAI-1, these residues cover ␣-helix F (residues 128 -145) and part of the turn connecting ␣-helix F and ␤-strand s3A (residues 146 -156).
In the present study, we hypothesized that evaluation of the differential species reactivity of these antibodies, combined with a linear sequence alignment of the region from 128 to 156 in PAI-1 of various species and with the relative three-dimensional localization of these residues, could provide unambiguously which amino acids are directly involved in the interaction with the neutralizing monoclonal antibodies. The obtained data demonstrate that the epitope is composed of five residues at maximum. The precise nature and the location of this nonlinear epitope have important implications for the structural basis of the neutralizing mechanism of the antibodies and for the model of the complex between serpins and their target * This study was supported by Grant OT/98/37 from the Research Fund Katholieke Universiteit Leuven and by Grant 7.0034.98 from the Fund for Scientific Research (FWO-Vlaanderen). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes were obtained from Amersham Pharmacia Biotech (Uppsala, Sweden), from Roche Molecular Biochemicals or from Stratagene (La Jolla, CA). T4 DNA ligase and alkaline phosphatase were purchased from Roche Molecular Biochemicals. T4 kinase was from U. S. Biochemical Corp., Klenow polymerase was from Amersham Pharmacia Biotech, PfuTurbo TM DNA polymerase was purchased from Stratagene. Synthetic oligonucleotides (for mutagenesis and DNA sequencing) were synthesized by Amersham Pharmacia Biotech.
The oligo-directed mutagenesis system using the pMa/c plasmids and the Escherichia coli strains WK6 and WK6 mutS (20) were kindly provided by Corvas (Ghent, Belgium). pMc-PAI-1 was constructed as described before (21). The expression vector pIGE20 containing a heatinducible promotor, the pAcI plasmid encoding a thermolabile repressor, as well as the E. coli strains DH1 and MC1061, for cloning and expression respectively, were kindly provided by Innogenetics (Ghent, Belgium). M13K07 helper phage was obtained from Promega (Leiden, The Netherlands).
General DNA Techniques-DNA manipulation techniques were carried out according to standard procedures and following the instructions of the manufacturers. Plasmid DNA was isolated using Nucleobond ® cartridges (Macherey-Nagel). DNA was sequenced with the Autoread Sequencing ® kit and the Automated Laser Fluorescent ALF ® apparatus (both from Amersham Pharmacia Biotech). PCR was performed using the GeneAmp ® 2400 (Perkin-Elmer).
Construction of PAI-1 Mutants-The following synthetic oligonucleotides were designed to introduce the desired mutations by site-directed mutagenesis using the pMa/c method (22): A, 5Ј-CACGGCTCCG-GCGCCAAGCAAGTTGC-3Ј was used to mutate Lys 154 into Ala and simultaneously introduced a HaeII restriction site (PuGCGCPy); B, 5Ј-GAATCTGGCTGCCGCCGCGGCTGAAAAGTCC-3Ј was used to mutate Glu 128 -Val 129 -Glu 130 -Arg 131 into Ala-Ala-Ala-Ala and simultaneously introduced a SacII restriction site (CCGCGG). The introduction of the additional restriction site allowed the confirmation of the desired mutation by restriction enzyme analysis. pMa-PAI-1-K, pMa-PAI-1-EVER, and pMa-PAI-1-EVER/K were constructed using oligonucleotides A, B, and AϩB, respectively. Randomly selected clones containing the pMa-PAI-1 construct were evaluated for the presence of the mutated sequence. Therefore, small scale DNA preparations from the pMa-PAI-1 constructs were analyzed for the presence of the mutated sequences by restriction enzyme analysis.
For all mutants, large scale DNA preparations were made and subjected to nucleotide sequencing for confirmation of the mutations.
Expression and Purification of PAI-1-wt and PAI-1 Mutants-MC1061 E. coli competent cells were cotransformed with pAcI and pIGE20-PAI-1-wt or one of the pIGE20-PAI-1 mutant expression constructs. Clonal isolates were grown at 28°C in LB medium with 50 g/ml kanamycin and 10 g/ml tetracycline on an orbital shaker at 300 rpm. Overnight cultures were diluted 1:100 and grown to an absorbance (A 580 nm ) of 0.2. Then, PAI-1 expression was induced by increasing the temperature to 42°C. Cells were harvested after 3 h by centrifugation for 20 min at 4000 ϫ g at 4°C and resuspended in 50 mM sodium acetate, pH 5.5, containing 2 mM glutathione, 0.5 g/ml leupeptin, 0.05 mM phenylmethanesulfonyl fluoride, 1 mM dithiothreitol, 0.7 g/ml pepstatin A, 1 mM benzamidine-HCl, and 1 g/ml antipain. The cell suspension was disrupted in a French ® pressure cell (SLM Instruments Inc., Rochester, NY), and the cell lysate was cleared by ultracentrifugation for 20 min at 40,000 ϫ g at 4°C. The PAI-1-containing supernatant was collected and immediately subjected to purification.
All purification steps were performed at 4°C and were carried out as described previously (22,26) with minor modifications. Briefly, the supernatant was diluted (1:4) with 0.15 M KH 2 PO 4 -Na 2 HPO 4 , pH 6, and 2 mM glutathione (buffer P i ) including 0.2 M NaCl and applied to a SP-Sepharose column (1.5 ϫ 25 cm) equilibrated with buffer P i containing 0.2 M NaCl. The column was washed with the same buffer containing 0.3 M NaCl, and bound proteins were eluted with buffer P i with a linear sodium chloride gradient (0.3-1.3 M). Fractions were evaluated for the PAI-1 content and purity by sodium docecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). PAI-1-containing fractions were pooled, dialyzed against P i , and applied onto a heparin-Sepharose column (1.2 ϫ 5 cm) equilibrated with buffer P i . The column was washed with buffer P i containing 0.2 M NaCl. Bound proteins were eluted with buffer P i with a linear sodium chloride gradient (0.2-1.3 M). Fractions containing PAI-1 were pooled and the concentration was determined spectrophotometrically at 280 nm using an absorbance coefficient A 1 cm 1% of 10. Determination of the Conformational Distribution of PAI-1-wt and PAI-1 Mutants-Samples of PAI-1-wt and PAI-1 mutants were diluted to a final concentration of 80 g/ml in 0.045 M KH 2 PO 4 -Na 2 HPO 4 , pH 6, containing 0.2 M NaCl and incubated with a 2-fold molar excess of t-PA or u-PA at 37°C for 30 min. The reaction products were analyzed by SDS-PAGE using 10 -15% gels followed by Coomassie Brilliant Blue staining. Quantitation of the formed reaction products (complexed, non-reactive, and cleaved, corresponding to the presence of active, latent, and substrate conformations, respectively; Ref. 22) was done by densitometric scanning of the gels using Imagemaster® (Amersham Pharmacia Biotech).
Determination of the Stability of PAI-1-wt and PAI-1 Mutants-Samples of PAI-1-wt and PAI-1 mutants were diluted to a final PAI-1 protein concentration of 30 -235 g/ml with the appropriate diluent to obtain a buffered solution with 0.045 M KH 2 PO 4 -Na 2 HPO 4 , 70 mM NaCl, pH 7.2-7.4. Samples were incubated at 37°C, and aliquots were removed at various times and assayed for their inhibitory activity against t-PA using an immunofunctional assay, adapted to a method described before (27) with minor modifications. Briefly, the samples were incubated with a 2-fold molar excess of t-PA at 37°C for 30 min. Subsequently, the amount of PAI-1⅐t-PA complex formed was analyzed in an enzyme-linked immunosorbent assay, using microtiter plates coated with MA-21F7 (directed against human PAI-1) for capture and horseradish peroxidase-conjugated MA-51H8 (directed against t-PA) for tagging. The half-lives for inactivation were calculated with the program Graph Pad Prism™ using "one phase exponential decay" according to the equation: Y ϭ span*e ϪK*X ϩ plateau. The half-life of the decay is then equal to 0.693/K.
Affinity of MA with PAI-1-wt and PAI-1 Mutants-Affinity constants for the binding between MA and PAI-1-wt or PAI-1 mutants were determined using the BIAcore™ 3000 analytical system equipped with the CM5 sensor chip (BIAcore AB) as described previously (28). In brief, the monoclonal antibodies were covalently coupled to 2000 resonance units (using a concentration of 10 g/ml in 10 mM acetate buffer, pH 4.5) using the automatic Wizard mode. Subsequently, PAI-1 was injected at a concentration between 20 and 800 nM (PAI-1 antigen) at a flow rate of 20 l/min (injection volume is 40 l). After each cycle the chip was regenerated using 10 l of a 15 mM HCl solution. The analyses of the association and dissociation phases were made with the software of the BIAcore™ 3000 (local fit, Langmuir binding).
Preparation of PAI-1⅐t-PA Complex-PAI-1⅐t-PA complex was prepared as described before (29). Briefly, t-PA was incubated with an excess active purified PAI-1 for 25 min at 37°C and applied on a Sepharose 4B column to which a monoclonal antibody against t-PA (MA-62E8) was coupled. The column was washed with phosphate-buffered saline, and bound PAI-1⅐t-PA complex was eluted with 3 M KSCN. PAI-1⅐t-PA containing fractions were pooled and dialyzed against 50 mM Tris, pH 8.3 containing 250 mM arginine. The purity and concentration of the preparation was evaluated by SDS-PAGE and an enzymelinked immunosorbent assay using microtiter plates coated with MA-21F7 (directed against PAI-1) for capture and horseradish peroxidaseconjugated MA-51H8 (directed against t-PA) for tagging. (Table I), it can be deduced that MA-55F4C12 reacts similarly with human, rat, and murine PAI-1, but exhibits no affinity for porcine and rabbit PAI-1. For MA-33H1F7 a similar pattern was observed, except toward porcine PAI-1 for which some, but strongly reduced, reactivity (100-fold lower versus human PAI-1) was observed. Both antibodies react with active, latent, and substrate forms of PAI-1 with comparable affinity (data not shown), excluding the possibility that differences in species reactivity (or reactivity toward any of the mutants studied, see below) would arise from differences in relative amounts of the three conformations.

Affinity of MA-55F4C12 and MA-33H1F7 with PAI-1 from Different Species and Amino Acid Sequence Alignment-From the affinity constants
Sequence alignment between amino acid 128 and 156 of the PAI-1 s from the latter five species (Table I) shows that residues 128 -130 and the residue at position 154 are conserved among human, rat, and murine PAI-1, whereas these particular residues differ from these in porcine and rabbit PAI-1. Consequently, we hypothesized that residues 128 -130 and residue 154, or a combination of both, play a major role in the interaction between PAI-1 and the antibodies. Localization of these residues within the three-dimensional structure revealed that all are exposed at the surface and therefore are potential candidates to contribute to an epitope. Three-dimensional analysis in the region revealed that the residue at position 131 might also contribute to the possible epitope. It was therefore decided to produce a variety alanine mutants within this region of human PAI-1. Either all residues at position 128 to 131 and/or at position 154 were mutated, or single residues at positions 128, 129, 130, and 131, respectively, were mutated in the absence or presence of the mutation at position 154.
Expression and Characterization of the PAI-1 Mutants-All mutants exhibited inhibitory activity toward t-PA, although to a different extent and with different ratios between the various conformations (Table II, A). PAI-1-K exhibited functional activities similar to these of PAI-1-wt, whereas PAI-1-EVER and PAI-1-EVER/K acted more as a substrate. The activity of the single alanine mutants (PAI-1-E 128 , PAI-1-V, PAI-1-E 130 , and PAI-1-R) varied between 32% and 64%. The activity of the double alanine mutants (single mutants combined with the K 154 mutation) varied between 40% and 51% (data not shown). Incubation of PAI-1-wt and PAI-1 mutants at 37°C for 24 h resulted in a complete loss of inhibitory activity, predominantly due to conversion of the active conformation to the latent conformation, whereas the substrate conformation remained stable (Table II, B) (data not shown for "single" and "double" mutants). The rate of the active to latent transition was similar   for PAI-1-K as for human PAI-1. The mutants PAI-1-EVER and PAI-1-EVER/K were significantly less stable (p Ͻ 0.0001, versus PAI-1-wt) ( Table III).
Even though most single mutations between residues 128 and 131 (Table IV, C) had only a marginal effect on the affinity, the affinity of MA-55F4C12 for PAI-1-E 130 and PAI-1-R was decreased 35-fold. In contrast, combination of any single mutation in the region from 128 to 131 with the K154A mutation resulted in a pronounced decrease of affinity for either of the antibodies studied (Table IV, D). Indeed, MA-55F4C12 does not bind to PAI-1 if the mutation K154A is combined with the mutation E128A, E130A, or R131A. In addition the affinity of MA-55F4C12 for PAI-1-V/K is 20-fold reduced compared with that for PAI-1-K and 1500-fold compared with that for PAI-1-wt.
The affinity of MA-33H1F7 is lost completely only by a combination of the mutations E130A and K154A. Combination of K154A with mutation R131A results in 4-fold decrease of affinity compared with that for PAI-1-K. In contrast, combination of mutation E128A or V129A with mutation K154A does not result in a further decrease of the affinity compared with mutation K154A only.
Complex formation of PAI-1 with t-PA had no effect on the affinity of the antibodies for PAI-1 (Table IV, E). DISCUSSION In this study the epitope of MA-55F4C12 and MA-33H1F7 was identified within a stretch of 29 amino acids in PAI-1 (region Glu 128 -Ala 156 ), mainly based on the differential reactivity of these antibodies for PAI-1 from different species. In combination with the three-dimensional structure of PAI-1, this approach allowed to designate five amino acids (Glu 128 , Val 129 , Glu 130 , Arg 131 , and Lys 154 ) that are the major determinants for the interaction between PAI-1 and the antibodies. Thus, the epitope of the antibodies does not cover the complete ␣-helix F and turn connecting ␣-helix F and ␤-strand s3A, but is restricted to the hinge region between ␣-helix F and the main part of the PAI-1 molecule (Fig. 1).
To check the importance of the five residues, eleven alanine mutants were produced and characterized. The replacement of the bulky or charged residues (Glu 128 , Val 129 , Glu 130 , Arg 131 , and Lys 154 ) into small, non-charged alanines was predicted to destroy the epitope. This approach ("alanine-scan") has already  been used before to detect epitopes or functional determinants (30,31). It should be noted that mutation of residue Arg 131 was included primarily based on its location in the three-dimensional structure.
Comparison of the affinity constants in Table IV, part D versus those in part C and those for PAI-1-K, indicate that for MA-55F4C12 all five residues act cooperatively in the binding to the antibody and constitute the epitope. A similar comparative analysis for MA-33H1F7 reveals that the residues Glu 128 and Val 129 most likely do not contribute significantly, thereby suggesting that the epitope of MA-33H1F7 is predominantly composed of three residues (Lys 154 /Glu 130 /Arg 131 ), positioned virtually linearly in the three-dimensional structure.
It has been suggested previously that in serpins ␣-helix F, together with ␤-strands s1-3A, can move relative to the remainder part of the molecule and that this movement is a prerequisite for the inhibitory activity (32). The localization of the epitope within the three-dimensional structure of PAI-1 is thus likely to form the molecular basis for the PAI-1 neutralizing properties of these antibodies. Indeed, binding of MA-55F4C12 or MA-33H1F7 to this particular region can be expected either to restrict the conformational flexibility of ␣-helix F and ␤-strands s1-3A or to reposition this region relative to the remainder part of the molecule. Either of these effects will disturb the shutter-like movement in ␤-sheet A required to accommodate properly the new ␤-strand s4A during interaction between PAI-1 and t-PA and to lock subsequently the serpinproteinase complex. Consequently, the rate or the extent of insertion of the reactive-site loop into ␤-sheet A will be reduced, resulting in a shift from the inhibitory to the substrate pathway (18). This is in accordance with the current view that the kinetics and/or extent of insertion of the reactive-site loop plays an important role in the inhibition mechanism of the serpins (6, 11, 17, 32, 33).
Even though some of the mutants in the EVER region (Table  II) exhibit an increased substrate behavior, further supporting the role of this region in the inhibitor versus substrate properties of serpins, we, as well as others (34), have no straightforward explanation for these modified biochemical properties. It could be speculated that replacement of the charged residues by alanines and subsequent loss of interaction within the Nterminal portion of ␣-helix F, might result in a repositioning of ␣-helix F relative to the main ␤-sheet A (e.g. closer to ␤-sheet A, closer to the s3A-s5A opening) thereby impairing the shutterlike movement in ␤-sheet A, required for the inhibitory properties of a serpin.
Localization of the epitope has also implications for the mechanism of complex formation between PAI-1 and its target proteinase. Structures of the serpin/proteinase complexes are not elucidated yet, but several models have been proposed for such covalent complexes (7)(8)(9)(10)(11). According to one model (Fig.  2C), the proteinase moves 180°after the initial complex formation to form a stable covalent complex in which the reactive-site loop is completely inserted (7)(8)(9). It is obvious that the epitope of MA-55F4C12 and MA-33H1F7 is not accessible in this model (Fig. 2D). According to another model ( Fig. 2A), the proteinase moves only to a limited extent such that the reactive-site loop is only partially inserted (10,11). In the latter model the epitope clearly remains accessible since the proteinase is located rather remote from the epitope (Fig. 2B). According to the affinity constants (Table IV), the antibodies MA-55F4C12 and MA-33H1F7 bind to PAI-1 in complex with t-PA with an affinity similar to that for free PAI-1, demonstrating that the epitope is neither disturbed nor covered by complex formation. Taken together these findings rule out the model in which full insertion is proposed, but are compatible with the models in which partial insertion is proposed ( Fig. 2A) (10, 11). Interestingly, this conclusion is also in agreement with a recent study FIG. 2. Models of serpin/proteinase complexes. A and B, partial insertion of the reactive-site loop; C and D, full insertion of the reactive-site loop. The major ␤-sheet of the serpin is indicated in green; the reactive-site loop is indicated in red; the ␣-helix F and turn connecting ␣-helix F and ␤-strand s3A are indicated in yellow; the residues of the epitope are indicated in orange. The dark, oval-shaped figure represents the proteinase. The Y-shaped figure represents the antibody. (35) in which localization of a neoantigenic epitope in antithrombin-III/thrombin complexes indicates that in this complex at the most, residue P14 to P6 of the reactive-site loop can be inserted in ␤-sheet A.
To the best of our knowledge, this is the first report describing the elucidation of an epitope of a serpin-neutralizing monoclonal antibody down to five residues, being non-linear ("conformational") and remote of the active site. The identified epitope is different from the epitopes of other monoclonal antibodies with an inhibitory effect on a serpin. Asakura et al. (36) identified an epitope in the reactive-site loop of anti-antithrombin III (P8 -P13), whereas Keijer et al. (37) suggested that the binding region of an inhibitory antibody was situated between amino acid residues 320 and 379 of PAI-1, a region containing the reactive-site loop. According to Björquist et al. (38), a panel of PAI-1-inhibiting antibodies can be divided into at least three groups, representing three non-overlapping epitopes. None of the suggested epitopes were located at the hinge region of ␣-helix F, although the epitope of CLB-2C8 is located near this area (residues 128 -145) (39). Interestingly, other antibodies have been described with an epitope partially covering the same area (residues 125-132), that did not exert any inhibitory effect on PAI-1 (40). This further confirms that the functional properties of the currently described antibodies are due to the simultaneous interaction with residues Glu 128 -Arg 131 and with residue Lys 154 .
In conclusion, this study describes the localization of a conformational epitope of two anti-PAI-1 inhibitory antibodies, thereby providing a molecular explanation for their neutralizing mechanism. The precise localization of the epitope has also implications for the model of the complex between serpins and their target proteinases, i.e. only a model in which the Nterminal side of the reactive-site loop of serpins is partially inserted is compatible with the current observation.