A regulatory hydrophobic area in the flexible joint region of plasminogen activator inhibitor-1, defined with fluorescent activity-neutralizing ligands. Ligand-induced serpin polymerization.

We have characterized the neutralization of the inhibitory activity of the serpin plasminogen activator inhibitor-1 (PAI-1) by a number of structurally distinct organochemicals, including compounds with environment-sensitive spectroscopic properties. In contrast to latent and reactive center-cleaved PAI-1 and PAI-1 in complex with urokinase-type plasminogen activator (uPA), active PAI-1 strongly increased the fluorescence of the PAI-1-neutralizing compounds 1-anilinonaphthalene-8-sulfonic acid and 4,4'-dianilino-1,1'-bisnaphthyl-5,5'-disulfonic acid. The fluorescence increase could be competed by all tested nonfluorescent neutralizers, indicating that all neutralizers bind to a common hydrophobic area preferentially accessible in active PAI-1. Activity neutralization proceeded through two consecutive steps as follows: first step is conversion to forms displaying substrate behavior toward uPA, and second step is to forms inert to uPA. With some neutralizers, the second step was associated with PAI-1 polymerization. Vitronectin reduced the susceptibility to the neutralizers. Changes in sensitivity to activity neutralization by point mutations were compatible with the various neutralizers having overlapping, but not identical, binding sites in the region around alpha-helices D and E and beta-strand 1A, known to act as a flexible joint when beta-sheet A opens and the reactive center loop inserts as beta-strand 4A during reaction with target proteinases. The defined binding area may be a target for development of compounds for neutralizing PAI-1 in cancer and cardiovascular diseases.

Plasminogen activator inhibitor-1 (PAI-1) 1 is a fast and specific inhibitor of the serine proteinases urokinase-type (uPA) and tissue-type plasminogen activator (tPA) and, as such, an important regulator of extracellular proteolysis in turn over of extracellular matrix and in fibrinolysis (for reviews see Refs. 1 and 2). PAI-1 binds with high affinity to vitronectin (for reviews see Refs. 3 and 4) and may regulate cell migration and adhesion by inhibition of vitronectin binding of integrins and the uPA receptor (5-10). The PAI-1 level in malignant tumors is one of the most informative biochemical markers of a poor prognosis (for reviews see Refs. 11 and 12), and PAI-1 seems to be causally involved in tumor invasion and angiogenesis (13). A high PAI-1 level in blood plasma is a risk factor for ischemic cardiovascular disease and venous thromboembolism (for review see Ref. 14). PAI-1 is therefore a potential target for both anti-cancer and anti-thrombotic therapy.
PAI-1 belongs to the serpin superfamily. Serpins are composed of 3 ␤-sheets and 9 ␣-helices. Serpins and their target proteinases form stable complexes by interaction of the active site of the proteinases with the reactive center peptide bond (P 1 -P 1 Ј) in the solvent-exposed, ϳ20-amino acid long peptide loop, the reactive center loop (RCL) (for reviews see Refs. 2 and 15-17). There is both structural and biochemical evidence that complex formation is associated with the P 1 -P 1 Ј bond being cleaved, the active site Ser of the proteinase linked to the carboxyl group of the P 1 residue by an ester bond, the Nterminal part of the RCL inserted as strand 4 of the large central ␤-sheet A (s4A), and the proteinase translocated across the plane of ␤-sheet A, toward the other pole of the molecule (18 -27). Under specific in vitro conditions, however, the ester bond is hydrolyzed, and the serpin exhibits substrate behavior. This also leads to insertion of the N-terminal part of the RCL as s4A (for a review see Ref. 2). RCL insertion also takes place during transition to the inactive, latent state, occurring spontaneously only in PAI-1 (28). Latent PAI-1 can be reconverted to the active form by denaturation and refolding (29).
During RCL insertion, the so-called small serpin fragment, consisting of s1A, s2A, s3A, ␣-helix F (hF) and the loop connecting hF and s3A (the hF/s3A-loop), moves relative to the rest of the molecule, the large serpin fragment, to make space for the RCL between s3A and s5A. The regions around hD and hE were proposed to form a flexible joint during RCL insertion (30) (Fig. 1). The idea of flexibility of this region was supported by the observation that it contains a cluster of peptide bonds with differential susceptibility to nontarget proteinases in active, latent, and reactive center-cleaved PAI-1 (31). Comparison of the x-ray crystal structures of latent, active, and reactive center-cleaved PAI-1 (28,(32)(33)(34) demonstrated directly the conformational changes. Vitronectin binds near the flexible joint region (35-37) (see Fig. 1) and delays the conversion of PAI-1 to the latent state (for reviews see Refs. [1][2][3]. A number of compounds have been found to inhibit the reaction of PAI-1 with uPA and tPA (38 -50). Among these are some organochemicals, including 1-dodecyl sulfuric acid (42,43), the diketopiperazine derivative XR5118 (49), and the anthranilic acid derivative AR-H029953XX (50). The flexible joint region was implicated in the PAI-1 neutralizing activity of the latter, by the demonstration that Glu substitutions of three basic residues in hD and hE (see Fig. 1) protected PAI-1 against activity neutralization (50).
We have now studied the mechanism of action of a few PAI-1-neutralizing organochemicals, including the negatively charged AR-H029953XX and 1-dodecyl sulfuric acid and the positively charged XR5118. Basic biochemical knowledge about the mechanism of PAI-1 neutralization is necessary for development of clinically applicable inhibitors of PAI-1 function in cancer and cardiovascular diseases. We have characterized neutralizer binding sites and neutralizer-induced molecular changes of PAI-1.
Natural human PAI-1 was purified in the latent form from serumfree conditioned medium of dexamethasone-treated HT-1080 cells by immunoaffinity chromatography (43). Wild type (wt) human PAI-1 was expressed in the Escherichia coli strain TG1 by the use of the expression vector pAlter®-Ex1 and purified from bacterial lysates by ion exchange chromatography on a CM-50 Sephadex column (51), followed by immunoaffinity chromatography (43). Amino acid sequencing of the purified protein gave the expected, slightly modified N-terminal sequence MHVHPPSYVAHL. Expression of recombinant wt and mutant human PAI-1 in the yeast Pichia pastoris and its purification from the conditioned medium in the latent form were performed as described (52). Human wt and PAI-1 R97E/R136E/R139E were expressed in CHO cells as before (50). Latent PAI-1 was converted to the active form by denaturation in 4 M guanidinium chloride and refolding by extensive dialysis against phosphate-buffered saline (PBS; 0.01 M sodium phosphate, pH 7.4, 0.15 M NaCl) at 0°C. Active PAI-1 was stored at Ϫ80°C. Reactive centercleaved PAI-1 (form C) and uPA⅐PAI-1 complex were prepared as described (31).
uPA-Human uPA was purchased from Wakamoto Pharmaceutical Co., Tokyo, Japan. Low molecular weight uPA (LMW-uPA), with N terminus at Lys-136, was prepared as described (53). The molar concentrations of active native and LMW-uPA were determined by active site titration with p-nitrophenyl-guanidinobenzoate.
Fluorescence Measurements-Fluorescence emission spectra for ANS and bis-ANS in the absence and presence of PAI-1 were recorded with a SFM 25 spectrofluorimeter (Kontron Instruments), using excitation wavelengths of 386 and 395 nm, respectively. The emission was recorded over the range of 400 -600 nm. The change in fluorescence was measured about 10 min after mixing PAI-1 with the fluorescent ligands in a buffer of 0.1 M Tris-HCl, pH 8.1, at 14°C, unless otherwise indicated. The observed fluorescence intensities, F obs , were corrected for the dilution effect of the added ANS and bis-ANS solutions, for the low background fluorescence of ANS and bis-ANS in buffer alone, and for the inner filter effect of the varying concentrations of ANS and bis-ANS. The correction for the inner filter effect was performed using the equation F ϭ F obs ⅐((2.303⅐⑀ Ex ⅐[ANS or bis-ANS] T )/(1-10 Ϫ⑀Ex⅐[ANS or bis-ANS]T )), where ⑀ Ex is the molar extinction coefficient of ANS and bis-ANS at the excitation wavelength; [ANS or bis-ANS] T is the total concentration of ANS or bis-ANS; and F and F obs are the corrected and the observed fluorescence intensities, respectively (54).
Estimation of the parameters for bis-ANS-PAI-1 binding was initiated by determination of the fluorescence intensity per 1 M PAI-1-bound bis-ANS, F M , as the slope of the line relating the fluorescence intensity at 480 nm to the bis-ANS concentration at 6 M PAI-1 and bis-ANS concentrations below 6 M. The concentrations of PAI-1-bound bis-ANS at any PAI-1 and bis-ANS concentrations were then calculated from the fluorescence intensity measured at these conditions by Equation 1,   (33). Please notice that the RasMol program uses the same signature for ␣-helices and the short 3 10 -helix found in the hF/s3A loop. izers was done by competition studies, in which 0.5 or 1 M PAI-1 was preincubated with ANS (100 M), and the fluorescence intensity at 470 nm was then recorded 10 min after addition of nonfluorescent competitors in various concentrations. The fluorescence intensity at max was expressed relative to that obtained in the absence of nonfluorescent competitor (RFI 2 ). Assuming competition of fluorescent and nonfluorescent ligands for binding to one site, an excess of nonfluorescent ligand I over PAI-1, and simple binding equilibria for both ANS and I, the following Equation 4 is true. Measurements of the Effects of Neutralizers on Specific Inhibitory Activity of PAI-1-The specific inhibitory activity of PAI-1, i.e. the fraction of the total amount of PAI-1 forming a stable complex with uPA at the conditions used, was measured by titration of PAI-1 against uPA in a peptidyl anilide assay, in the presence of several concentrations of each neutralizing compound. Shortly, PAI-1 was serially diluted with a buffer of 0.1 M Tris-HCl, pH 8.1, 0.25% gelatin at 37°C, corresponding to concentrations between 0.02 and 40 g/ml and a volume of 100 l. To each dilution series, a particular neutralizing compound was added in a particular concentration and incubated with PAI-1 for 10 min at 37°C. Portions of uPA solutions of the same temperature and in the same buffer were then added, corresponding to a final uPA concentration of 0.25 g/ml, a final volume of 200 l, and final PAI-1 concentrations between 0.01 and 20 g/ml. Incubation was then continued until the process of inhibition of uPA had come to an end. Control experiments showed that this was in all cases achieved in less than 2 min. The remaining uPA enzyme activity was determined by incubation with the peptidyl anilide substrate S-2444 at 37°C and measurement of the increase in absorbance at 405 nm. The specific inhibitory activity of PAI-1 was calculated from the amount of PAI-1 that had to be added to inhibit 50% of the 0.25 g/ml uPA. The IC 50 values for the compounds being tested, i.e. the concentrations causing a 50% reduction in the PAI-1 specific inhibitory activity, were determined from plots of the specific inhibitory activity against the concentration of the compounds in the final assay mixture.
Measurement of Time Course of Changes in the Specific Inhibitory Activity of PAI-1-PAI-1 was incubated at various concentrations between 0.33 and 330 g/ml in 0.1 M Tris-HCl, pH 8.1, 0.25% gelatin, at 0 or 37°C, without or with neutralizers. After various times, samples were removed for assay of specific inhibitory activity, which in this case was performed by diluting all samples to a final PAI-1 concentration of 0.33 g/ml, using the same buffer and the same temperature as in the original incubation. uPA was added to a final concentration equivalent to 90% of the activity of PAI-1 in the absence of neutralizers. The uPA solutions were without or with BSA in a concentration corresponding to a final concentration, in the assay, of 1%. After incubation for a time sufficient for the process of inhibition of uPA to come to an end (2 min), the remaining uPA activity was estimated with S-2444 and used to calculate the specific inhibitory activity of PAI-1 in the samples.
Analysis of Functional Behavior of PAI-1 by Reaction with LMW-uPA and SDS-PAGE-After incubation at various conditions, samples of 5 g PAI-1 were mixed with 10 g LMW-uPA. After 2 min, the samples were precipitated with trichloroacetic acid and subjected to SDS-PAGE in gels with 6 -16% polyacrylamide. The gels were stained with Coomassie Blue. In this analysis, inhibitory active PAI-1 will be recovered as a complex with LMW-uPA. PAI-1 exhibiting substrate behavior will be recovered as the large N-terminal fragment produced by reactive center cleavage, whereas the only 33 amino acid long C-terminal fragment will not be recovered by the gel system used here. Inert PAI-1, for instance the latent form, will be recovered in the position of native full-length PAI-1 (45). Control experiments (not shown) ensured that no further changes in the reaction between PAI-1 and LMW-uPA occurred by prolonging the incubation time beyond the routinely used 2 min.
Gel Filtration-PAI-1 was analyzed by fast protein liquid chromatography gel filtration on a Superdex 200 HR10/30 column (Amersham Pharmacia Biotech) in 0.1 M Tris-HCl, pH 8.1, 0.5 M NaCl at 4°C, using a flow rate of 0.4 ml per min. The following marker proteins were used: BSA (M r 67,000), murine IgG (M r 150,000), and ␤-galactosidase (M r 540,000). The void volume was determined with blue dextran. Before the chromatography, the samples were passed through a layer of 200 l of BSA-coupled Sepharose 4B and a filter with a pore size of 0.22 m, resulting in removal of more than 98% of the organochemicals present in the samples. A quantitative estimation of the distribution of PAI-1 between polymer and monomer peaks was obtained by cutting out the areas representing the peaks on the recordings of A 280 and weighing them.
Native Gel Electrophoresis-The electrophoresis was performed at 4°C in gradient gels with 5-15% polyacrylamide, in the absence of SDS (55). Coomassie Blue was added to the cathode buffer to a final concentration of 0.02% (56). The following marker proteins were used: BSA (M r 67,000), beef heart lactate dehydrogenase (M r 140,000), beef liver catalase (M r 232,000), horse spleen ferritin (M r 440,000), and hog thyroid thyroglobulin (M r 669,000).
Molecular Graphics-RasMol version 2.6 (Roger Sayle, Glaxo Research and Development, Greenford, Middlesex, UK) was used to display three-dimensional protein structures. The coordinate (Protein Data Bank) files corresponding to the x-ray crystal structure of active PAI-1 (33) were kindly provided by Dr. R. J. Read, Cambridge, UK.
Statistical Analysis-Data were evaluated by Student's t test, and differences in results with a p value below 0.005 were considered statistically significant. Nonlinear regression analysis, by the method of least squares, were performed by the SigmaPlot program (Jandel Scientific Software, San Rafael, CA).

Neutralization of the Inhibitory Activity of PAI-1 by Various
Compounds-We measured the IC 50 values for the effect on the specific inhibitory activity of PAI-1 of a variety of amphipathic compounds, including the negatively charged PAI-1 neutralizers 1-dodecyl sulfuric acid and AR-H029953XX (42,43,50) and the positively charged PAI-1 neutralizer XR5118 (49). About half of the compounds tested were found to neutralize PAI-1 in the concentration ranges used, with IC 50 values between 0.6 and 300 M (Fig. 2).
Routinely, the assays were performed with guanidinium chloride-activated HT-1080 PAI-1, but the compounds also neutralized guanidinium chloride-activated P. pastoris PAI-1, guanidinium chloride-activated CHO cell PAI-1, and spontaneously active E. coli PAI-1 (see also below). Routinely, we performed the experiments at pH 8.1, but the same results were obtained at pH 7.4 (data not shown).
ANS, AR-H029953XX, bis-ANS, 1-dodecyl sulfuric acid, and XR5118 were selected for further characterization, ANS and bis-ANS because the fluorescence of these compounds generally increases at transfer from a hydrophilic to a hydrophobic environment (57) and they therefore can be used for fluorimetric binding assays, AR-H029953XX, bis-ANS, 1-dodecyl sulfuric acid, and XR5118 because of their relatively low IC 50 values.
Fluorimetric Analysis of Neutralizer-PAI-1 Binding-In the presence of 1 M active PAI-1, the fluorescence emission intensities of ANS in concentrations above 10 M and of bis-ANS in concentrations above 0.1 M were strongly increased over those attained in the absence of protein. The wavelengths with maximal fluorescence intensity ( max ) shifted from 490 to 470 nm for ANS and from 490 to 480 nm for bis-ANS. In contrast, only a small fluorescence increase was observed with 1 M latent, 1 M reactive center-cleaved, and 1 M uPA-complexed PAI-1 (Fig. 3). These observations are consistent with ANS and bis-ANS binding to a hydrophobic area preferentially accessible in active PAI-1.
At PAI-1 concentrations above 5 M and bis-ANS concentrations below 5 M, the fluorescence intensity of bis-ANS at 480 nm was independent of the PAI-1 concentration, showing that practically all added bis-ANS was bound to PAI-1 at these conditions. Therefore, the fluorescence intensity per 1 M bound bis-ANS, F M , could be determined from the bis-ANS concentration dependence of the fluorescence intensity at these conditions (Fig. 4A). It was then possible to use fluorescence intensity measurements to estimate the concentrations of bound and free bis-ANS in the presence of lower PAI-1 concentrations (0.125-1 M), at which there is an equilibrium between bound and free ligand. By analysis of the relationship between bound and free bis-ANS, we found a K d of 0.57 Ϯ 0.29 M and a number of binding sites per PAI-1 molecule of 1.66 Ϯ 0.62 (n ϭ 12) (Fig. 4B and Table I).
With ANS, only a small fraction of the ligand added to M concentrations of PAI-1 was bound to the protein, rendering calculations of the stoichiometry impossible. However, an esti-mate of the K d was obtained by studying the ANS concentration dependence of the fluorescence intensity in the presence of PAI-1 (Table I).
To measure the dissociation constants for the binding to PAI-1 of the nonfluorescent neutralizers, PAI-1 was preincubated with 100 M ANS, and the fluorescence intensity was recorded after addition of increasing concentrations of the nonfluorescent compounds. AR-H029953XX, 1-dodecyl sulfuric acid, or XR5118 caused a decrease in the ANS fluorescence intensity down to background levels. In contrast, two nonneutralizing, nonfluorescent compounds, 1-nonanesulfonic acid and 2-propylpentanoic acid, caused very little inhibition of the fluorescence in concentrations up to 1 mM (Fig. 5). K i values for the nonfluorescent compounds were determined by analysis of the displacement curves (Table I).
Similar competition experiments were done with bis-ANS. The results obtained (not shown) were in full agreement with those obtained with ANS.
The fluorimetrically determined dissociation constants for ANS, bis-ANS, 1-dodecyl sulfuric acid, and XR5118 (Table I)  were not significantly different from the corresponding IC 50 values (Fig. 2). Only in the case of AR-H029953XX, the dissociation constant was somewhat lower than the IC 50 value. This disagreement may be related to the irreversible reaction of PAI-1 with uPA leading to dissociation of AR-H029953XX⅐PAI-1 complex during the determination of PAI-1specific inhibitory activity.
The results of the fluorimetric analyses are most readily explained by the hypothesis that all the tested neutralizers have overlapping binding sites.
Analysis of Activity Neutralization by Site-directed Mutagenesis-The five neutralizers were tested on a series of PAI-1 variants, expressed in P. pastoris or CHO cells (Table II). The IC 50 values for neutralization of P. pastoris and CHO PAI-1 wt were not significantly different from those for neutralization of HT-1080 PAI-1. The susceptibility to the neutralizers was changed by substitution of specific amino acids in the flexible joint region. The P94K substitution, in hD (see Fig. 1), caused a 10-fold increase in the IC 50 for XR5118 but had no effects on the IC 50 for the negatively charged neutralizers. The double substitution R97K/H98K was without effect. The triple substitution R97E/R136E/R139E (see Fig. 1) increased the IC 50value for AR-H029953XX from about 5 M to more than 20 M and increased the IC 50 values for ANS-, bis-ANS-, and 1-dodecyl sulfuric acid neutralization about 3-fold, but did not affect the IC 50 value for XR5118 neutralization. Conclusively, the IC 50 values for all neutralizers were sensitive to mutations in the flexible joint region, consistent with the conclusion from the fluorimetric measurements of these compounds having overlapping binding sites. Still, the substitutions in the flexible joint region affecting the susceptibility to the negatively charged neutralizers had no effect on the susceptibility to the positively charged neutralizer and vice versa.
Competition between Vitronectin and PAI-1 Neutralizers-  Table I.

TABLE I Dissociation constants for binding of neutralizers to PAI-1
The dissociation constants were determined either as K d values by analysis of the fluorescence increase associated with the binding of ANS and bis-ANS to HT-1080 PAI-1 or as K i values by analysis of the competition of ANS binding to HT-1080 PAI-1 by the nonfluorescent compounds (AR-H029953XX, 1-dodecyl sulfuric acid, 1-nonanesulfonic acid, 2-propylpentanoic acid, and XR5118). Further details are given in the text. The data are given as means, S.D., and numbers of experiments.  Table I Table I. Preincubation of PAI-1 with vitronectin decreased its susceptibility to all five neutralizers but to a variable extent. Thus, the IC 50 value for bis-ANS increased about 50-fold, to 27 Ϯ 9 M (n ϭ 3), but the IC 50 value for 1-dodecyl sulfuric acid increased only about 7-fold, to 110 Ϯ 4 M (n ϭ 4). The IC 50 value for XR5118 increased to more than 250 M and that for ANS to more than 1000 M.
Time Course of Changes of PAI-1-specific Inhibitory Activity-The PAI-1-specific inhibitory activity was determined after incubations for various times at 0 or 37°C, at various PAI-1 concentrations, without or with neutralizers in concentrations of severalfold the IC 50 values and severalfold the PAI-1 concentrations. To distinguish between reversible and irreversible neutralization, we took advantage of the fact that none of the five neutralizers affected PAI-1 in buffers with 1% BSA (data not shown), in agreement with the ability of BSA to bind several molecules of a variety of hydrophobic compounds (58,59). Accordingly, the assays of the specific inhibitory activity were performed as follows: 1) in the presence of a neutralizer concentration equal to that used in the incubation (reversible plus irreversible neutralization); and 2) in the presence of 1% BSA, the BSA being added at the same time as uPA, to remove free and reversibly bound neutralizers (irreversible neutralization).
Without neutralizers, active PAI-1 lost its activity, due to conversion to the latent form, with a half-life of 44 Ϯ 9 min (n ϭ 6) at 37°C and of more than 5 days at 0°C, without or with BSA in the assay and independently of the PAI-1 concentration (data not shown). Fig. 6 shows representative experiments on the time course of neutralizer-induced activity loss, with 1-dodecyl sulfuric acid. By using assays without BSA, more than 85% of the PAI-1 activity was lost in less than 5 min at 0 as well as 37°C. By using assays with BSA, the activity loss was slower and less complete. The PAI-1 activity measured in the assays with BSA did not change by increasing the BSA concentration from the routinely used 1% to 5% or by prolonging the incubation time with BSA, before the addition of uPA, from 0 to 10 min, indicating that the large excess of BSA rapidly removed both free and PAI-1-bound neutralizer. A relatively fast and a relatively slow activity loss could therefore be distinguished, the fast one requiring the continued presence of neutralizer and the slow one being irreversible upon removal of neutralizer. The rate of the irreversible activity loss increased with increasing temperature and increasing PAI-1 concentration. Thus, the rate-limiting step of the irreversible activity loss seemed to be bi-or multimolecular.
The time courses of the activity losses induced by the other negatively charged neutralizers had the same characteristics. Only, the rate of the ANS-induced irreversible activity loss did not depend on the PAI-1 concentration, indicating that the rate-limiting step was monomolecular in this case (data not shown).
The time course of the activity loss induced by the positively charged XR5118 was not significantly different in assays without and with BSA, showing that the XR5118-induced activity loss was totally irreversible. The rate of XR5118-induced activity loss was strongly temperature-dependent. We found no evidence that the rate of XR5118-induced neutralization in-

PAI-1 Neutralization
creased with increasing PAI-1 concentration, suggesting that the neutralization process proceeded monomolecularly (data not shown).

Analysis of the Functional Behavior of Neutralizer-induced Forms of PAI-1 by Reaction with LMW-uPA and SDS-PAGE-
The effect of the neutralizers on the functional behavior of PAI-1 was analyzed by reacting neutralizer-treated PAI-1 with a molar excess of LMW-uPA and separating the reaction products by SDS-PAGE. A representative experiment, with bis-ANS, is shown in Fig. 7. Without incubation with neutralizers, most of the PAI-1 formed a stable complex with LMW-uPA, whereas minor fractions were latent or exhibited substrate behavior, respectively (Fig. 7, lane to the left). Complex formation was totally abolished within 2 min of incubation of PAI-1 with bis-ANS at 37°C. An inactive form with substrate behavior predominated for the first 2 min, after which time PAI-1 was gradually converted to an inert form. At 0°C, complex formation was abolished more slowly, and substrate behavior predominated for at least 2 h (data not shown). Identical observations were done with ANS, AR-H029953XX, and 1-dodecyl sulfuric acid. Neutralization by the positively charged XR5118 was at 0°C associated with a fast conversion to a form with a substrate behavior, followed by conversion to an inert form, but at 37°C, there was no detectable increase in substrate behavior, but an immediate increase in the amount of the inert form (data not shown). HT-1080 PAI-1 and E. coli PAI-1 reacted in the same manner to all neutralizers (data not shown).
Characterization of Neutralizer-induced PAI-1 Forms by Gel Filtration-PAI-1 was preincubated at 0 or 37°C in the absence or the presence of neutralizers, for practical reasons with a relatively high PAI-1 concentration (330 g/ml), before being analyzed by gel filtration. Without neutralizers, both latent and active PAI-1 migrated as one major symmetrical peak, in a position which by comparison to the migration of the marker proteins was that expected for monomeric PAI-1 (Fig. 8).
Preincubation of active but not latent PAI-1 with ANS, AR-H029953XX, bis-ANS, or 1-dodecyl sulfuric acid resulted in conversion of PAI-1 to larger species, the sizes of which increased with increasing incubation time, with a strongly temperature-dependent rate. Eventually, about 85% of the material migrated in the void volume ( Fig. 8 and data not shown). Five hundred M of the negatively charged, but non-neutralizing 1-nonanesulfonic acid and 2-propylpentanoic acid did not induce polymerization (data not shown). We concluded that the negatively charged neutralizers were able to induce PAI-1 polymerization.
In contrast, the positively charged XR5118 did not cause polymerization (Fig. 8). Moreover, preincubation with XR5118 inhibited the polymerization induced by bis-ANS (Fig. 9) or any of the other negatively charged neutralizers (data not shown). Analyzing the XR5118 concentration dependence of the inhibition under the assumption of competition between XR5118 and the negatively charged neutralizers for a single binding site, a K i value could be calculated that agreed well with the fluorimetrically determined dissociation constant for XR5118 binding ( Fig. 9 and Table I).
A representative analysis of the functional behavior of PAI-1 in individual peaks of the gel filtration profiles is shown in Fig.  10. PAI-1 was incubated with 1-dodecyl sulfuric acid for 30 min at 0 or 37°C before gel filtration, both incubations leading to complete neutralization when assayed in the continued presence of neutralizer. The polymeric forms in the gel filtration profile had little or no inhibitory activity, and with increasing size, they contained an increasing fraction of inert PAI-1. The monomeric peak seen after preincubation at 0 o C had regained an activity similar to that of PAI-1 not exposed to neutralizers, whereas the small monomeric peak left after incubation at 37°C was devoid of activity. It should be noticed that at least a fraction of the inert material in the monomeric peak is latent PAI-1 contaminating the preparation from the start.
Characterization of Neutralizer-induced Forms of PAI-1 by Native Gel Electrophoresis-PAI-1 was subjected to native polyacrylamide gel electrophoresis after incubation at a concentration of 330 g/ml for 30 min at 37°C without or with neutralizers (Fig. 11). Without neutralizers, PAI-1 migrated mainly as a monomer in the expected position relative to the M r markers. With the negatively charged neutralizers, but not with XR5118, most PAI-1 was converted to distinct, slower migrating bands. By comparison to the migration of the markers, the PAI-1 bands seem to represent dimers, trimers, tetramers, etc. The gel system did not allow resolution of polymer species with an M r above ϳ300,000. We concluded that PAI-1 neutralization by the negatively charged neutralizers, but not XR5118, is associated with formation of distinct PAI-1 polymers.
A Two-step Neutralization Mechanism-Comparing the time course of the change of PAI-1-specific inhibitory activity (Fig.  6), the time course of the changes in the functional behavior (Fig. 7), the time course of polymerization (Fig. 8), and the functional behavior of individual peaks of the gel filtration profiles (Fig. 10), we concluded that PAI-1 neutralization follows variations over a basic two-step mechanism, by which neutralizer-complexed PAI-1 (NϳPAI-1) is rapidly converted to a form exhibiting substrate behavior (NϳPAI-1 S ) and subsequently to an inert form (NϳPAI-1 I ): NϳPAI-13 NϳPAI-1 S 3 NϳPAI-1 I With the negatively charged neutralizers AR-H029953XX, bis-ANS, and 1-dodecyl sulfuric acid, the first step is reversible upon removal of the neutralizer, as evidenced by the recovery of specific inhibitory activity in assays with BSA (Fig.6) and the reappearance of inhibitory activity in the monomer peak of the gel filtration profile (Fig. 10), whereas the second step consists of an irreversible polymerization, as evidenced by the absence of inhibitory activity of the polymer peaks of the gel filtration profiles and the PAI-1 concentration dependence of the rate of loss of specific inhibitory activity in assays with BSA (Fig. 6). The fact that the time course of the irreversible neutralization by ANS was independent of the PAI-1 concentration suggests the existence of an intermediate, irreversibly inactivated monomeric form of ANS-complexed PAI-1, which subsequently polymerizes. XR5118-induced neutralization seemed to proceed by two irreversible conversions, neither of which involves polymerization.

DISCUSSION
On the basis of the fluorimetric binding assays and the site-directed mutagenesis studies reported here, we concluded that four negatively charged (ANS, AR-H029953XX, bis-ANS, and 1-dodecyl sulfuric acid) and one positively charged (XR5118) organochemical PAI-1 neutralizers have overlapping but not identical binding sites in a hydrophobic area in the flexible joint region of PAI-1. Neutralization proceeds through two steps as follows: first conversion to forms with substrate behavior and second to inert forms. With the negatively charged neutralizers, but not the positively charged one, the second step was associated with PAI-1 polymerization.
Besides the five neutralizers studied in detail, a number of other amphipathic organochemicals also neutralized PAI-1 (Fig. 2). Some of the compounds, including 1-dodecyl sulfuric acid and deoxycholic acid, are generally used as detergents. However, the neutralizing effect of the five compounds studied in detail did not seem to depend on their detergent properties. First, the critical micelle concentration for 1-dodecyl sulfuric acid is 8 mM (60,61), much higher than its 15 M IC 50 value for PAI-1 neutralization. Second, although structurally very different compounds were able to neutralize PAI-1 with a relatively low IC 50 value, a certain specificity was observed. For instance, among the two carboxylic acids with unbranched aliphatic side chains, the length of the side chain appeared to be decisive for activity (Fig. 2). Third, the observations that XR5118 inhibited the polymerization induced by the negatively charged neutralizers (Fig. 9) are difficult to reconcile with a detergent mechanism. Fourth, the well defined PAI-1 polymers observed by native gel electrophoresis (Fig. 11) are in contrast to the expectancies from a detergent-induced, nonspecific aggregation.
Rather, our findings point to the five neutralizers exerting their effect by binding to overlapping sites in a hydrophobic area with a relatively low binding specificity. This notion is supported by the observation that all nonfluorescent neutralizers competed the binding of ANS and bis-ANS (Fig. 5) and by the largely good agreement between the IC 50 values (Fig. 2) and the fluorimetrically determined dissociation constants (Table I). Likewise, our bis-ANS binding studies were compatible with a single class of binding sites with respect to binding affinity. Although the binding studies led to determination of the number of bis-ANS-binding sites per PAI-1 molecule to a value between 1 and 2 and therefore did not exclude the existence of two independent bis-ANS-binding sites, the induction of about 80% polymerization of 6 M PAI-1 by 5 M bis-ANS (Fig.  8) is a strong argument for a single binding and effector site for bis-ANS. Furthermore, the observed XR5118 inhibition of the polymerization induced by the negatively charged neutralizers (Fig. 9) is most readily explained by XR5118 and the negatively charged neutralizers having overlapping binding sites. Thus, it seems most likely that both the reversible and irreversible neutralizer effects are caused by binding of one neutralizer molecule per PAI-1 molecule. The possibility of several effector sites in PAI-1, each causing neutralization by a specific mechanism, seems remote.
It was previously suggested that the flexible joint region of PAI-1 contains a regulatory, hydrophobic, ligand-binding area (50). The x-ray crystal structures of active PAI-1 and ␣ 1 PI are in agreement with a hydrophobic cavity in this area (33, 34, 62, The fraction of polymerized material at a given concentration of XR5118 (P x ) was corrected for the background of polymerized material (P 0 ) and divided by the fraction of polymerized material after preincubation with bis-ANS alone, likewise corrected (P b Ϫ P 0 ). The (P x Ϫ P 0 )/(P b Ϫ P 0 ) values were plotted versus the XR5118 concentration. The K i value for XR5118 inhibition of polymerization could be calculated from the XR5118 concentration causing half-maximal inhibition of polymerization (56 M), the K d value for bis-ANS binding (0.57 M), and the equation ) (which is analogous to Equation 4 in the text). In the experiment shown, the calculated K i was 2.8 M. 63). Our present findings are consistent with the hypothesis that the five neutralizers bind in this cavity. First, amino acid substitutions in hD and hE changed the IC 50 values. Second, the stronger binding of ANS and bis-ANS to active than to latent, reactive center-cleaved and uPA-complexed PAI-1 is in agreement with the expectancies from the x-ray crystal structures, showing that the distance between s2A and hD and hE decreases upon insertion of RCL as s4A (28,(32)(33)(34)40). The volume of the hydrophobic cavity in active PAI-1 was estimated to 676 Å 3 (63). Each of the bulky and and in most cases rigid neutralizer molecules would therefore be expected to occupy a large fraction of the volume of the cavity. Thus, the binding competition between the negatively and the positively charged neutralizers is not in conflict with the observation that the amino acid substitutions in hD and hE, respectively, affected the IC 50 values for the negatively and the positively charged neutralizers differently. This observation is, in fact, in full agreement with their different structures and the charge changes associated with the substitutions. We therefore suggest that the variation in neutralization kinetics, in the induced molecular changes, and in the differential response to vitronectin is due to different neutralizers occupying different subsites within the same hydrophobic area. In this way, the different neutralizers may have different effects on the conformation of ␤-sheet A and thereby different effects on the movements of the RCL and the tendency to polymerization. This situation is reminiscent of estrogen receptor binding of estrogen agonists, antagonists, and partial agonists, which have overlapping binding sites, but induce different conformational changes in the receptor protein (64 -66).
Induction of serpin polymerization by small organochemical ligands is a novel finding. Previously described serpin polymerization all implicate the RCL of one serpin molecule forming an additional strand in ␤-sheet A or ␤-sheet C of another molecule. There is evidence for three different modes of loopsheet polymerization. 1) Heating or mutations in the so-called shutter region can lead to polymerization by the RCL of one molecule inserting as s4A of another molecule (for reviews see Refs. 2 and 17). 2) The RCL of one molecule can insert as s1C in another molecule, in which the intrinsic s1C has been extracted (67)(68)(69).
3) The RCL of one molecule can hydrogen-bond to s6A of another molecule and thus form an s7A (33,34). It is striking that the fluorescence observed by addition of ANS and bis-ANS to active PAI-1 remained high after incubation under conditions leading to polymerization, whereas the fluorescence induced by addition of these compounds to latent, reactive center-cleaved and uPA-complexed PAI-1 was much lower. This observation argues against the polymerization involving insertion of the RCL as s4A and favors the other modes of polymerization. The mode of the polymerization described here is clearly different from the recently reported PAI-1 polymerization induced in both latent and active PAI-1 at the strongly acidic pH of 4 (70).
An important perspective is the possibility of utilizing the hydrophobic area as a target for anti-cancer and anti-thrombotic drugs. To do so, strategies must be developed to circumvent certain problems that have become apparent from the studies described here. First, the strong binding to serum albumin common to all the compounds studied here must be avoided. Second, since PAI-1 is expected to be bound to FIG. 10. Analysis of 1-dodecyl sulfuric acid-induced PAI-1 forms by gel filtration and SDS-PAGE. HT-1080 PAI-1 (330 g/ml) was incubated with 100 M 1-dodecyl sulfuric acid for 30 min at 0 or 37°C, as indicated. Two hundred-g portions were then subjected to gel filtration. The absorbance at 280 nm is shown versus the elution time. Five-g portions of PAI-1, either before gel filtration or from individual peaks of the gel filtration profile, were reacted with 10 g of LMW-uPA for 2 min at 37°C and then subjected to SDS-PAGE. The positions of LMW-uPA⅐PAI-1 complex, native/inert PAI-1, reactive center-cleaved PAI-1 (RCC-PAI-1) and LMW-uPA are indicated to the left. vitronectin in vivo, pharmacologically potentially interesting molecules must have a high affinity not only to free PAI-1 but also to PAI-1 in its vitronectin-associated state. Third, other serpins have similar hydrophobic areas, and the specificity of PAI-1 neutralizers of potential interest for in vivo use must therefore be ensured.