Characterization of a Novel Class of Polyphenolic Inhibitors of Plasminogen Activator Inhibitor-1

Abstract

Plasminogen activator inhibitor type 1, (PAI-1) the primary inhibitor of the tissue-type (tPA) and urokinase-type (uPA) plasminogen activators, has been implicated in a wide range of pathological processes, making it an attractive target for pharmacologic inhibition. Currently available small-molecule inhibitors of PAI-1 bind with relatively low affinity and do not inactivate PAI-1 in the presence of its cofactor, vitronectin. To search for novel PAI-1 inhibitors with improved potencies and new mechanisms of action, we screened a library selected to provide a range of biological activities and structural diversity. Five potential PAI-1 inhibitors were identified, and all were polyphenolic compounds including two related, naturally occurring plant polyphenols that were structurally similar to compounds previously shown to provide cardiovascular benefit in vivo. Unique second generation compounds were synthesized and characterized, and several showed IC₅₀ values for PAI-1 between 10 and 200 nm. This represents an enhanced potency of 10–1000-fold over previously reported PAI-1 inactivators. Inhibition of PAI-1 by these compounds was reversible, and their primary mechanism of action was to block the initial association of PAI-1 with a protease. Consistent with this mechanism and in contrast to previously described PAI-1 inactivators, these compounds inactivate PAI-1 in the presence of vitronectin. Two of the compounds showed efficacy in ex vivo plasma and one blocked PAI-1 activity in vivo in mice. These data describe a novel family of high affinity PAI-1-inactivating compounds with improved characteristics and in vivo efficacy, and suggest that the known cardiovascular benefits of dietary polyphenols may derive in part from their inactivation of PAI-1.

Introduction

Plasminogen activator inhibitor type 1 (PAI-1)³ is the primary physiologic inhibitor of uPA and tPA with a well characterized role in fibrinolysis (1). PAI-1 also plays a role in many physiologic processes, including angiogenesis, wound healing, and cell migration (2–6), and has been implicated in fibrotic diseases of the kidney and lung, and in tumor metastasis (7–11). More recently, PAI-1 has been linked to obesity and metabolic syndrome (12–16), and to the development of vascular diseases such as venous thrombosis and atherosclerosis (17–19). The prospect that PAI-1 may play a direct role in the early development of a variety of diseases has made it an attractive target for drug development (20, 21). However, the structural complexity of PAI-1 has made the identification and development of PAI-1 inhibitors challenging. This is due in part to the metastable structure of PAI-1, which can adopt several different conformations, including active, latent, cleaved, and protease complexed (1). These different forms of PAI-1 provide conformational control of PAI-1 interactions and dictate its localization to either matrix or the cell surface and control its activity in cell signaling events (22, 23).

Active PAI-1 inhibits protease targets and is associated with vitronectin in plasma or the extracellular matrix. In contrast, PAI-1-protease complexes shift affinity from vitronectin to receptors of the low density lipoprotein receptor family, transferring PAI-1 from vitronectin to the cell surface (22). Active PAI-1 is inherently unstable and undergoes a spontaneous conformational change that results in inactivation of PAI-1 to a latent form that does not bind either vitronectin or low density lipoprotein receptor family members with high affinity (22, 24). The flexible structure of PAI-1, the lack of a rigid active site, and its multiple functions all contribute to the difficulties in identifying and designing small-molecule PAI-1 inactivators. Despite these obstacles, several small-molecule PAI-1 inhibitors have been described (25–36); however, each has significant limitations that have reduced their potential for further drug development.

One of the best characterized compounds is PAI-039, also known as tiplaxtinin, which has been shown to reduce physiologic PAI-1 activity and to be efficacious in animal models of disease (3, 37–39). However, PAI-039 has relatively low affinity for PAI-1, and does not inhibit vitronectin-bound PAI-1 (32, 40). To develop better PAI-1 inactivators, we screened a library of known compounds for high affinity PAI-1 inhibitors with improved solubility and activity against vitronectin-bound PAI-1. A high throughput screen of the MicroSource SPECTRUM library identified five novel PAI-1 inactivating compounds. Two of the molecules identified were related natural polyphenolic compounds, which suggested a potential structure-activity relationship. Second generation compounds were designed and synthesized to probe this structure-activity relationship and tested for their ability to block PAI-1 activity in both purified systems and in vivo.

EXPERIMENTAL PROCEDURES

Primary Screen

In conjunction with the Center for Chemical Genomics (CCG) at the University of Michigan, we developed a PAI-1 activity assay to screen for compounds with anti-PAI-1 activity in the MicroSource SPECTRUM compound collection. This collection consists of known drugs, compounds approved for agricultural use, natural products, and other bioactive compounds. A chromogenic assay was used with a 2:1 molar ratio of PAI-1 to uPA. We selected uPA because it is considerably more active toward low molecular weight substrates than tPA, allowing for more than 10-fold lower concentrations of uPA and PAI-1 in this screen (5 nm uPA and 10 nm PAI-1) compared with assays using tPA (70 nm tPA and 140 nm PAI-1) (41). The screen was performed in 384-well microtiter plates in the CCG lab as follows: recombinant active human PAI-1 (final 10 nm) was incubated for 60 min at 23 °C either with or without 10 μm of each compound in high throughput screen buffer (10 mm HEPES, pH 7.4, 150 mm NaCl, 0.005% Tween 20), uPA was added (final 5 nm) to each reaction well, and the incubation continued for an additional 30 min at 23 °C. Residual uPA activity in each reaction mixture was then determined with p-Glu-Gly-Arg p-nitroanilide chromogenic substrate (Sigma) (final 0.25 mm) measured spectrophotometrically at 405 nm after 60 min. Compounds that inactivated PAI-1 were identified by the restoration of uPA activity. The extent of uPA activity restoration was determined by comparing each drug-containing sample to wells with untreated PAI-1 (100% PAI-1 activity) and to wells with uPA only (0% PAI-1 activity). The data from this screen were then uploaded to the CCG informatics system and positive hits were identified as any compound that increased uPA activity by more than 3 S.D. above control and compound wells on each plate. Using these selection criteria, the primary screen of 2000 compounds yielded an initial total of 23 compounds deemed positive hits. Each of these hits was then re-assayed by dose-response testing using the same chromogenic assay with the compounds at the following concentrations (0.1, 0.32, 1, 3.2, 10, 32, and 100 μm) in duplicate by the CCG. In this secondary analysis 19 of the 23 compounds were deemed positive; however, 3 of these compounds were known to have significant toxicity and therefore were not analyzed further. Samples of the 16 remaining compounds were then obtained from the CCG for further analysis in our laboratory. These more detailed analyses first investigated whether each compound had intrinsic absorbance at 405 nm that would give false positive absorbance readings, or was not completely soluble in the assay buffer system used because insolubility and compound precipitation could likewise lead to false positive absorbance readings. Each compound was also tested for its ability to directly block PAI-1 complex formation with uPA by SDS-PAGE analysis. For this latter analysis each compound was incubated at 10 μm with 1 μg of PAI-1 for 15 min at 23 °C followed by the addition of 1 μg of uPA for an additional 5 min at 37 °C. Approximately half of the 16 compounds either had intrinsic absorbance at 405 nm or insolubility in the buffer system. Of the remaining compounds, 5 directly inhibited PAI-1 activity.

Enzymatic Assays

Recombinant nonglycosylated or glycosylated active human PAI-1 (PAI-1 and PAI-1_glyco, respectively) or recombinant murine PAI-1 (mPAI-1) was incubated at 2 nm for 15 min at 23 °C with increasing concentrations of each compound in assay buffer (40 mm HEPES, pH 7.8, 100 mm NaCl, 0.005% Tween 20, 0.1% Me₂SO), followed by the addition of uPA (Molecular Innovations) or tPA (Genentech) to 3 nm and further incubation for 30 min at 23 °C. At each drug concentration, parallel control reactions without PAI-1 were assembled. Residual enzymatic activity was determined by addition of an equal volume of 100 μm Z-Gly-Gly-Arg-AMC (Calbiochem) fluorogenic substrate for uPA or Pefafluor tPA (Centerchem) for tPA, and the rate of AMC release monitored at 23 °C (excitation 370 nm and emission 440 nm). The percent change in PAI-1 activity was determined according to Equation 1,

where E_i is the enzyme activity at drug concentration i; P_i is the enzyme activity in the presence of PAI-1 at drug concentration i; E₀ is the enzyme activity in the absence of drug; and P₀ is the enzyme activity in the presence of PAI-1 in the absence of drug. The effect of the compounds on 2 nm anti-thrombin III in the presence of 3 units/ml of heparin was also determined using 3 nm α-thrombin. The reactions were assembled as above except that 10% Me₂SO was included in the assay buffer to ensure compound solubility at the higher concentrations used. Residual α-thrombin activity was measured using an equal volume of 100 μm benzoyl-Phe-Val-Arg-AMC (Calbiochem).

Synthesis of New Inhibitors

Synthetic procedures and spectroscopic data for CDE compounds are provided in supplemental “Methods”.

Surface Plasmon Resonance (SPR) Analysis

Direct binding of PAI-1 treated with vehicle or inhibitor to anhydrotrypsin (Molecular Innovations) was monitored using a Biacore 2000® optical biosensor. Bovine anhydrotrypsin was immobilized to CM5 SPR chips at a level of ∼2000 response units in 10 mm sodium acetate, pH 5.0. The reference flow cell surface was left blank to serve as a control. Remaining binding sites were blocked by 1 m ethanolamine, pH 8.5. All binding reactions were performed in assay buffer. PAI-1 at 2 nm was first incubated with the indicated concentrations of inhibitor in assay buffer for at least 15 min at 23 °C. Binding of PAI-1 to anhydrotrypsin was then monitored at 25 °C at a flow rate of 30 μl/min for 2.5 min, followed by 2 min of dissociation. Chip surfaces were regenerated with a 1-min pulse of 10 mm glycine, pH 1.5, followed by a 1-min wash of assay buffer. Injections were performed using the Wizard Customized Application program in automated mode. Binding experiments were performed in duplicate and corrected for background and bulk refractive index by subtraction of the reference flow cell, and data were analyzed with BIAevaluation 3.1 (Biacore) by linear fitting of the initial association phase. Compound-induced alterations in PAI-1 binding to anhydrotrypsin were determined by comparing the initial slopes of the association phases because there is a linear relationship between the slope and the concentration of available active PAI-1 (supplemental Fig. S1 and Ref. 32). These data were then fit to an exponential association equation to determine the apparent affinity between PAI-1 and compound.

To monitor the inhibition of vitronectin-bound PAI-1, human vitronectin purified under non-denaturing conditions was coupled to a CM5 sensor chip to a surface density of ∼1000 response units (32). Five nm PAI-1 was injected over the chip at a rate of 20 μl/min at 25 °C for 4 min, followed by assay buffer alone or 100 nm CDE-066 in assay buffer for 10 min, and then 100 nm uPA for 5 min. After injections of PAI-1 or CDE-066, the chip was washed with assay buffer for ∼4 min. Results were corrected for background and bulk refractive index in BIAevaluation 3.1.

SDS-PAGE/Western Blotting

Human PAI-1 at 2 nm was incubated with the indicated concentrations of the compound for 15 min at 23 °C in assay buffer, followed by a 30-min incubation with 3 nm uPA or tPA. Samples were analyzed via reducing SDS-PAGE with 10% Tris-HCl gels (Bio-Rad) and transferred onto polyvinylidene difluoride overnight. PAI-1 was detected using polyclonal high-titer sheep anti-human PAI-1 antibody (Molecular Innovations), horseradish peroxidase-conjugated donkey anti-sheep IgG (Jackson ImmunoResearch Laboratories), and Pierce ECL Western blotting substrate (Thermo Scientific).

Reversibility Assay

The reversibility of PAI-1 inactivation by the compounds was performed essentially as described (32). PAI-1 (2 nm) was incubated with each compound at 3–5-fold its IC₅₀ value as determined using uPA, followed by serial 2-fold dilutions into assay buffer and further incubation to allow dissociation of the compounds from PAI-1. UPA (3.5 nm) was then added and PAI-1 activity was determined. The PAI-1 activity in compound-containing samples recovered at each dilution was calculated as a percentage of the activity in PAI-1 dilutions carried out in parallel without compound. Initial concentrations of compounds were 150 nm CDE-008, 75 nm CDE-031, 400 nm CDE-034, 300 nm CDE-056, 50 nm CDE-066, and 50 nm CDE-082.

Inhibition of mPAI-1 in ex Vivo Plasma

Murine PAI-1 was added to PAI-1-depleted murine plasma (Molecular Innovations) at 5000 pg/ml. Ten microliters of increasing concentrations of compound in assay buffer containing 10% Me₂SO and 10 μl of mPAI-1-reconstituted plasma were incubated for 15 min at 23 °C in a filter plate (Millipore), followed by the addition of 25 μl of SeroMAP beads (Luminex) coupled to uPA (2500 beads/well), and further incubated in the dark on a microtiter plate shaker for 2 h. The plate was vacuum washed 3 times with wash buffer (PBS, pH 7.4, 0.05% Tween 20), then 50 μl of PBS, pH 7.4, 1% bovine serum albumin, and 50 μl of 4 μg/ml of biotin-labeled rabbit anti-mPAI-1 (Molecular Innovations) was added to each well and the plate incubated at 23 °C in the dark on a microtiter plate shaker for 1 h. After vacuum washing 3 times, 50 μl of PBS, pH 7.4, 1% bovine serum albumin, and 50 μl of 4 μg/ml of streptavidin-R-phycoerythrin conjugate (Molecular Probes) was added to each well and incubated with shaking at 23 °C for 30 min in the dark. After washing another 3 times, 100 μl of sheath fluid (Luminex) was added to each well, shaken for 5 min in the dark at 23 °C, and read on a Luminex¹⁰⁰ (median setting, 50-μl sample size, 100 events/bead). Mean fluorescence intensities of unknown samples were converted to picograms/ml of base on a standard curve of mPAI-1 in mPAI-1-depleted plasma using a five-parameter regression formula (Masterplex QT version 4.0, Miraibio).

Plasma Enzymatic Assay

Citrated blood was collected from the inferior vena cava (IVC) of C57Bl6J mice that were either PAI-1 null or vitronectin/PAI-1 null and plasma were prepared by centrifugation (15 min at 1500 × g). The plasma was treated with 10 μg/ml aprotinin (Roche Applied Science) for 15 min at 23 °C before reconstituting with 20 nm PAI-1. Plasma (10 μl, with or without PAI-1) was placed in microtiter wells with 80 μl of CDE-066 or PAI-039, synthesized as described (40) in assay buffer containing 10% Me₂SO and incubated for 15 min at 23 °C, followed by addition of 10 μl of 25 nm uPA, and a further incubation for 30 min. Residual enzymatic activity was monitored as above using the fluorogenic uPA substrate, and PAI-1 activity was determined using Equation 1.

Inhibition of PAI-1 in Vivo

Transgenic mice heterozygous for murine PAI-1 overexpression (10) were weighed, then anesthetized with isoflurane for the duration of the experiment. The IVC was isolated and 50 μl of citrated blood was collected as pre-treatment samples. The syringe was replaced with a syringe containing vehicle or CDE-066 (in lactated Ringers) and 100 μl was injected for doses of 3, 10, and 30 mg/kg. After 1 h, 300 μl of citrated blood was collected via IVC, after which the mice were euthanized. Plasma was isolated by centrifugation at 1500 × g for 15 min at 23 °C. All animal experiments were approved by the Institutional Animal Care and Use Committee of Unit for Laboratory Animal Medicine at the University of Michigan. To determine active murine PAI-1 levels in the plasma, 10 μl of plasma, diluted in PAI-1-depleted murine plasma (Molecular Innovations), 10 μl of buffer (PBS, pH 7.4, 1% bovine serum albumin), and 25 μl of uPA-coupled SeroMAP beads were added to a filter plate and incubated by shaking overnight at 4 °C in the dark, and the reactions were analyzed as above in the ex vivo plasma assay.

Data and Statistical Analysis

Data were analyzed and IC₅₀ values were calculated using Grafit 5. Apparent K_D values for the binding of compounds to PAI-1 were determined using GraphPad Prism 4. Data were analyzed for significance with a Student's t test using non-diluted samples in the reversibility assays and 0 mg/kg of CDE-066 treatment in the in vivo assays as the control groups, with p < 0.05 considered significant.

RESULTS

High Throughput Screen

The MicroSource SPECTRUM compound library was screened under stringent conditions such that PAI-1 was present at a 2-fold molar excess over uPA, and each compound was tested at a concentration of 10 μm. The statistical criteria of 3 S.D. above the control and compound means on each plate resulted in 23 hits. These compounds were further tested by dose-response analysis, and 19 remained positive in this secondary screen. Of these, 16 were deemed safe and subjected to further study including SDS-PAGE analysis of complex formation between PAI-1 and uPA. Based on these analyses, 5 compounds were confirmed as PAI-1 inhibitors in both enzymatic and SDS-PAGE assays, yielding a final hit rate of 0.25%. The structures and IC₅₀ values of these 5 compounds along with two related compounds are shown in Fig. 1.

FIGURE 1.

IC₅₀ values of PAI-1 inactivating compounds from high throughput screen and related compounds. The two-dimensional structures of the five hits from the screen (A, C, D, E, and G) and two related compounds (B and F) are shown with IC₅₀ values from the enzyme assays. Recombinant active human PAI-1 (final 2 nm) was incubated for 15 min at 23 °C with increasing concentrations of each compound in assay buffer. Next uPA (final 3 nm) was added to each reaction well and incubated for an additional 30 min at 23 °C. Activity of uPA in each reaction mixture was determined with the Z-Gly-Gly-Arg-AMC fluorogenic substrate (final 50 μm). UPA activity was measured fluorometrically (excitation 370 nm and emission 440 nm) for 15 min. The IC₅₀ values were calculated using Grafit IC₅₀ fit and the mean ± S.E. are based on three independent experiments. The asterisk indicate compounds identified in the original screen, and the dagger indicates related compounds not identified in the original screen.

Each of these five compounds contain polyphenolic moieties, and three of them, tannic acid (TA), epigallocatechin-3,5-digallate (EGCDG), and sennoside A, are naturally occurring plant polyphenols with reported biological activities (42–46). The former two compounds, TA and EGCDG, have highly related structures that both contain galloyl or gallo-galloyl moieties suggesting the possibility of a structure-activity relationship between polyphenols in general, and more specifically gallic acid moieties and PAI-1 inactivation. We therefore examined two additional galloyl-containing compounds, epigallocatechin monogallate (EGCG) and gallic acid (Fig. 1, B and F). Monomeric gallic acid was 1000-fold less active toward PAI-1 than TA, whereas EGCG inhibited PAI-1 only ∼10-fold less well than TA. Thus, each of the galloyl-containing compounds was able to inhibit PAI-1, but the efficacy of inhibition appears dependent on the number of galloyl units in each compound and their relative orientation or context.

Synthetic Compounds

The IC₅₀ value obtained with 7 nm TA was ∼1000-fold lower than our previously reported IC₅₀ value for the PAI-1 inactivator, PAI-039 (32), and is markedly lower than any previously reported small-molecule PAI-1 inactivating compound (25–36). Likewise the IC₅₀ values obtained with EGCG and EGCDG were also significantly better than PAI-039 and most other PAI-1 inactivators, suggesting that galloyl-containing compounds may represent a potent new family of PAI-1 inactivating compounds. However, TA is not an ideal drug candidate as its molecular mass of nearly 2000 daltons is considered too large, and it was subject to aggregation at micromolar concentrations. Therefore, we synthesized a series of novel compounds containing different numbers of galloyl moieties in different structural configurations and compared their activity against PAI-1 to determine a potential structure-activity relationship between galloyl-containing compounds and PAI-1 inhibition. In addition, to make this analysis sensitive to inactivators with low nanomolar IC₅₀ values, the PAI-1 concentration in the assay was lowered from 10 to 2 nm. Using these optimized assay conditions, we were able to accurately determine IC₅₀ values for several novel compounds. Six of these compounds, four digallates, one trigallate, and one pentagallate, are shown in Fig. 2. Comparison of the IC₅₀ values of these 6 compounds demonstrated IC₅₀ values ranging from 10 to 174 nm for inactivation of PAI-1 (Fig. 2 and Table 1). The activity of each compound against glycosylated human PAI-1 (PAI-1_glyco) and murine PAI-1 (mPAI-1) was also compared with nonglycosylated recombinant human PAI-1 (PAI-1) (Table 1). In general the compounds inhibited PAI-1_glyco as well as the nonglycosylated form; however, most inhibited mPAI-1 less well than human PAI-1. The two exceptions were the pentagallate, CDE-066, and TA, which inhibited all forms of PAI-1 equally well.

FIGURE 2.

Structures of six synthetic compounds. A, the two-dimensional structures of the 6 synthetic polyphenolic compounds are pictured. The inhibition curves of each compound against PAI-1 in the presence of either uPA (B) or tPA (C) are shown. The data were plotted using the Grafit IC₅₀ fit and are based on three independent experiments, points represent the mean ± S.E. For comparison PAI-039 has a reported IC₅₀ in a similar assay system of ∼10 μm (32).

TABLE 1.

The IC₅₀ values of TA and six synthetic compounds against various forms of PAI-1 or anti-thrombin III using the indicated target enzymes

IC₅₀ values (nm) ± S.E. were determined using the Grafit IC₅₀ fit. Values are based on three independent experiments.

Compound	PAI-1		PAI-1glyco	mPAI-1	Hep:anti-thrombin IIIa
	uPA	tPA	uPA	uPA	α-Thrombin
TA	6.6 ± 1.1	8.0 ± 0.3	4.8 ± 1.2	4.1 ± 1.2	11,800 ± 300
CDE-008	44 ± 5	53 ± 4	28 ± 2	162 ± 27	>10,000b
CDE-031	20 ± 1	28 ± 1	18 ± 1	132 ± 14	>10,000b
CDE-034	116 ± 11	174 ± 26	169 ± 21	644 ± 53	>300,000
CDE-056	74 ± 4	86 ± 8	152 ± 28	758 ± 26	>300,000
CDE-066	10 ± 1	12 ± 2	13 ± 2	10 ± 1	>300,000
CDE-082	14 ± 1	18 ± 1	56 ± 2	79 ± 4	15,400 ± 4,400

^a Values represent measured IC₅₀ values or the highest concentration of compound tested.

^b <20% of Hep:anti-thrombin III was inactivated at the highest compound concentration used.

The inactivation of PAI-1 by the polyphenolic compounds was specific, because only TA and CDE-082 (IC₅₀ > 10 μm) showed any inhibition of the related serpin anti-thrombin III. Some of the gallate-containing compounds tested did show an apparent inhibition of tPA in assays with a chromogenic or fluorogenic substrate; however, little inhibition of tPA by these compounds was seen when the physiologic substrate of tPA, plasminogen, was used (supplemental Fig. S2), suggesting that the compounds may be interacting with the low molecular weight tPA substrates.

It is also apparent from these data that although a single gallate (gallic acid, 6.6 μm) is a relatively poor inhibitor of PAI-1, a minimum of two galloyl units translates into significant anti-PAI-1 activity (20–116 nm, Fig. 2 and Table 1). Compound CDE-008 was compared with several similar digallates with linkers of different lengths between the gallate moieties, and CDE-008 was found to have the optimal distance between the galloyl units (data not shown). To further explore structural requirements for digalloyl compound inhibition of PAI-1, we examined 1,2-disubstituted galloyl units on different ring structures to determine whether cis (CDE-031), trans (CDE-034), or planar (CDE-056) relationships between galloyl units inhibited PAI-1 more effectively. All of these compounds were active against PAI-1 with the cis form (CDE-031) being ∼2-fold more active against PAI-1 than the acyclic CDE-008. These data demonstrate that the relative orientation of the gallates is important for anti-PAI-1 activity, with the cis form inhibiting PAI-1 ∼4-fold better than the planar form and ∼6-fold better than the trans form (Table 1).

SPR Analysis

To establish binding constants for the drugs to PAI-1, an indirect approach using SPR was employed. Varying concentrations of each drug were preincubated with PAI-1 in solution and then passed over immobilized anhydrotrypsin, and the loss of PAI-1 binding to anhydrotrypsin was quantified. We have previously shown for PAI-1 binding to vitronectin (32) that the slope of the association phase of PAI-1 binding to an immobilized ligand has a linear relationship with the concentration of available active PAI-1 in solution. This relationship is also true for PAI-1 binding to immobilized anhydrotrypsin (supplemental Fig. S1). Thus, when the slopes of the association phase are plotted as a percent of control PAI-1 binding in the absence of compound versus the concentration of the compound, an IC₅₀ can be calculated for the drug-induced inhibition of PAI-1 interaction with anhydrotrypsin (Fig. 3). From a transformation of these data, the apparent K_D values for all six compounds binding to PAI-1 can be calculated (Table 2). The apparent K_D values range from 3.1 to 67 nm and are significantly tighter than the previously reported values for other PAI-1 inactivators (28, 32, 33, 40). These values are also similar to the IC₅₀ values calculated in PAI-1 inhibition assays (Table 1). These data indicate that drug binding interferes with the initial association of PAI-1 with the protease and can block formation of the PAI-1-protease Michaelis-like complex.

FIGURE 3.

Apparent affinity between PAI-1 and synthetic inhibitors assessed by SPR. Two nm PAI-1 was incubated with the concentrations indicated of each synthetic compound and the mixtures injected over an anhydrotrypsin-conjugated CM5 sensor chip. The compound-dependent change in the initial association rates for PAI-1 binding to anhydrotrypsin, which is directly proportional to the amount of free PAI-1 in the analyte, is plotted against the compound dose to determine the apparent K_D values of each compound for PAI-1. Data are based on two independent experiments; points represent the mean ± S.E. For comparison PAI-039 has a reported affinity for PAI-1 of ∼15 μm in a direct SPR binding assay system (32).

TABLE 2.

Affinity between PAI-1 and synthetic compounds as measured by SPR

The data from Fig. 3 were fit to an exponential association curve in GraphPad Prism 4 to calculate the apparent K_D. Shown are the mean ± S.E. of two independent experiments.

Compound	Apparent KD
	nm
CDE-008	23 ± 1
CDE-031	31 ± 2
CDE-034	67 ± 3
CDE-056	51 ± 6
CDE-066	3.1 ± 0.2
CDE-082	5.3 ± 0.2

SDS-PAGE

Each compound was also tested for its ability to block complex formation between PAI-1 and PAs, and examples of CDE-008, CDE-066, and CDE-082 are shown in Fig. 4. For these studies each compound was incubated with PAI-1, then either uPA or tPA was added and the formation of SDS-stable complexes was monitored by SDS-PAGE. The concentrations of PAI-1 and PAs were the same as those used in the enzyme assays (see Fig. 1 and Table 1), and we observed comparable IC₅₀ values between the two techniques. Inhibition of covalent complex formation also closely mirrored inhibition of PAI-1 binding to anhydrotrypsin (see Fig. 3 and Table 2). An increase in PAI-1 cleavage was also noted with each compound primarily at compound concentrations just below the IC₅₀; however, this was modest compared with the near complete loss of covalent complex, and much less cleavage was observed at compound concentrations above the IC₅₀. Together with the SPR studies, these data suggest that the principal mechanism of PAI-1 inactivation by these compounds is the inhibition of the PAI-1 Michaelis-like complex formation with PAs, but that at concentrations near the IC₅₀ some increase of PAI-1 substrate behavior may be induced.

FIGURE 4.

CDE compounds inhibit complex formation between PAI-1 and uPA or tPA. PAI-1 (2 nm) was incubated with 10-fold dilutions of CDE-008 (A), -066 (B), and -082 (C) for 15 min at 23 °C in assay buffer. Then uPA (left panels) or tPA (right panels) was added (3 nm final) and complexes were formed at 23 °C for 30 min. Samples were analyzed by reducing SDS-PAGE followed by transfer to polyvinylidene difluoride membranes and immunoblotting for PAI-1. SDS stable complexes (asterisk), unreacted PAI-1 (open arrowhead), and cleaved PAI-1 (closed arrowhead) were detected.

Inactivation of PAI-1 Is Reversible

To test whether the inhibition of PAI-1 by the synthetic polyphenols was reversible, PAI-1 and each synthetic compound was incubated at a concentration where most of the anti-proteolytic activity of PAI-1 was abolished. The mixtures were then serially diluted to reduce the compound concentrations, incubated for an additional 30 min, and the mixtures tested for restoration of PAI-1 activity. Fig. 5 shows that for each synthetic polyphenol, PAI-1 activity increased upon dilution, indicating that PAI-1 inactivation by the compounds is reversible. The extent of PAI-1 activity recovered with each compound was slightly less than predicted, suggesting that the rate of dissociation between PAI-1 and these novel compounds is relatively slow and the samples may not have reached a new equilibrium after 30 min. Consistent with this mechanism, incubation of PAI-1 for various times with CDE-066, the most potent synthetic compound, demonstrated that the IC₅₀ remained unchanged from 15 min until termination of the experiment at 4 h (supplemental Fig. S3). These data indicate that the compounds do not irreversibly modify PAI-1, and are consistent with a high affinity reversible mechanism of action.

FIGURE 5.

Inactivation of PAI-1 by the synthetic inhibitors is reversible. PAI-1 (2 nm) was incubated with the compounds shown at 3–5-fold excess concentrations over the IC₅₀ of each compound for 15 min, then serially diluted 1:1 three times and further incubated for 30 min. PAI-1 activity was determined as described under “Experimental Procedures” and is shown as a percentage of control activity without compound. The data represent the mean ± S.E. of at least three independent experiments and were evaluated against the activities of the undiluted samples using a Student's t test (*, p < 0.05; **, p < 0.01).

Inhibition of mPAI-1 in Plasma

Each of the new compounds was tested for anti-PAI-1 activity in ex vivo plasma. This tests the ability of the drugs to inhibit mPAI-1 in the presence of plasma proteins, including vitronectin. None of the newly generated digallate compounds were active against mPAI-1 in the plasma activity assay (Fig. 6). This was likely due to high nonspecific protein binding of these digallates in plasma because the digallates were also ineffective against mPAI-1 in buffers containing high concentrations of bovine serum albumin (data not shown). In contrast, all of the compounds with at least 3 galloyl moieties inhibited mPAI-1 in the plasma, including the trigallate (CDE-082), pentagallate (CDE-066), and TA. Overall, TA had the lowest IC₅₀ against mPAI-1 in plasma (data not shown) but it was less specific than the novel polyphenols as it also inhibited normal plasma clotting, whereas CDE-066 and CDE-082 did not (supplemental Fig. S4). CDE-066 exhibited the lowest IC₅₀ of the new compounds in plasma, and was also significantly more specific than TA, therefore CDE-066 was used in further studies in plasma and in vivo.

FIGURE 6.

Inhibition of mPAI-1 by synthetic compounds in ex vivo plasma. Murine plasma depleted of PAI-1 was reconstituted with 5000 pg/ml of mPAI-1 and treated with each compound, and residual active mPAI-1 detected by Luminex. Curves were generated with the Grafit IC₅₀ fit and the IC₅₀ ± S.E. are indicated, NI indicates no detectable inhibition. The data are based on three independent experiments performed in duplicate.

Our previous studies (32) with PAI-039, the most widely studied PAI-1 small-molecule inhibitor, indicated that it is unable to inhibit PAI-1 bound to vitronectin, and one of the main objectives of the current study was to identify compounds that could inhibit PAI-1 in the presence of vitronectin. This was examined by adding a known amount of PAI-1 to murine plasma from either PAI-1 null mice or from mice doubly null for PAI-1 and vitronectin. After incubating the PAI-1 in these plasmas, samples were incubated with dilutions of either CDE-066 or PAI-039 and then tested for PAI-1 inhibition of uPA. Fig. 7 demonstrates that unlike PAI-039, which is only inhibitory in plasma that lacks vitronectin, CDE-066 inhibited PAI-1 equally well in plasma with or without physiologic vitronectin.

FIGURE 7.

CDE-066 but not PAI-039 inhibits PAI-1 in the presence of vitronectin. Plasma collected from PAI-1 null or PAI-1/vitronectin null mice were reconstituted with 20 nm PAI-1, and then vehicle or PAI-1 inactivators, CDE-066 (A) or PAI-039 (B) were added at the concentrations indicated, and the samples incubated. Residual PAI-1 activity was determined using uPA and Z-Gly-Gly-Arg-AMC as described under “Experimental Procedures.” The data are shown as the mean ± S.E. and are based on three independent experiments performed in duplicate.

The ability of CDE-066 to inactivate PAI-1 bound to purified vitronectin was verified in vitro via BIAcore. To be certain that the PAI-1 was in complex with vitronectin, PAI-1 was injected over immobilized vitronectin and complex formation was detected by changes in relative response units. These data demonstrate that as expected active PAI-1 binds vitronectin with high affinity and dissociates very slowly from immobilized vitronectin; however, upon reaction with uPA, PAI-1 affinity for vitronectin is reduced by several orders of magnitude (22, 24) and the PAI-1 rapidly dissociates from vitronectin, (Fig. 8, large dots), In contrast, PAI-1 bound to vitronectin and then exposed to 100 nm CDE-066 does not dissociate from the immobilized vitronectin following the uPA injection (Fig. 8, small dots). These data indicate that CDE-066 blocks the association of uPA with PAI-1 even when in complex with vitronectin, and are consistent with the hypothesis that the primary mechanism by which CDE-066 inactivates PAI-1 is to prevent non-covalent complex formation with target proteases. These data also demonstrate that CDE-066 is not inducing PAI-1 cleavage as a substrate, or latency, because both cleaved and latent PAI-1 also exhibit low affinity for vitronectin (22, 24), and would likewise result in loss of PAI-1 signal from the chip. Finally, consistent with the reversibility studies shown in Fig. 5, these data indicate that the dissociation of CDE-066 from PAI-1 is relatively slow because even after 240 s of wash PAI-1 is still inhibited by CDE-066.

FIGURE 8.

Inactivation of vitronectin-bound PAI-1 by CDE-066 assessed by SPR. PAI-1 (5 nm) was injected over a vitronectin-conjugated CM5 chip, followed by 100 nm CDE-066 (small dots) or vehicle (large dots). Residual vitronectin-bound PAI-1 activity was assessed by injection of 100 nm uPA, with active PAI-1 binding to the uPA and rapidly dissociating from the chip resulting in loss of surface response units, whereas CDE-066-inactivated PAI-1 remains on the chip surface after the uPA injection. The starts of injections and washes are indicated by black arrows.

PAI-1 Inactivation in Vivo

Finally, to examine whether CDE-066 inhibits murine PAI-1 in vivo, mice overexpressing PAI-1 were treated acutely with either vehicle or increasing concentrations of CDE-066. Plasma samples were removed from each mouse before treatment and then 1 h following intravenous infusion of CDE-066 at the indicated concentrations (Fig. 9). Plasma samples were then tested for active PAI-1 levels. Although a small increase in active PAI-1 was observed in the vehicle-treated animals, a dose-dependent decrease in active PAI-1 was observed after 1 h of treatment with CDE-066. These data indicate that CDE-066 can significantly inhibit PAI-1 in vivo.

FIGURE 9.

CDE-066 reduces endogenous active PAI-1 in mouse plasma. Citrated blood was removed via the IVC from mice overexpressing PAI-1 before, and 1 h following treatment with the indicated dose of CDE-066. Active murine PAI-1 was measured by Luminex assay and compared with standards of known murine PAI-1 concentrations. The data are expressed as a percentage of active PAI-1 present in the plasma relative to active PAI-1 at time 0 for each mouse. The data represent the mean ± S.E., n = 5 at each dose, and were evaluated against the 0 mg/kg treatment using a Student's t test (*, p < 0.05; **, p < 0.01).

DISCUSSION

PAI-1 is thought to play a role in several chronic “lifestyle” diseases, including cardiovascular and fibrotic diseases, and metabolic syndrome. These pathologic associations make PAI-1 an ideal drug target; however, its metastable structure has made it a difficult candidate for drug design and study. To date most small-molecule inhibitors of PAI-1 lack high affinity for PAI-1 and are unable to inhibit PAI-1 in the presence of its plasma binding protein, vitronectin. To identify higher affinity inhibitors with better drug development potential, a high stringency screening assay was performed and a class of polyphenolic compounds was identified with anti-PAI-1 activity. A subset of these with the highest anti-PAI-1 activity contained galloyl moieties, and one, TA, demonstrated the lowest IC₅₀ of any small-molecule PAI-1 inhibitor yet reported. One other study has identified members of the acylphloroglucinol class of polyphenols, sideroxylonals A–C, as potential PAI-1 inactivating compounds (27). However, the reported IC₅₀ values of these compounds (3.3–5.3 μm) are 2–3 orders of magnitude higher than TA and the novel synthetic polyphenols described here and are comparable with the IC₅₀ of the simplest gallate compound in the current study, gallic acid (6.6 μm). This suggests that many polyphenolic compounds may share PAI-1 inactivating activity, but that the galloyl moiety may be a critical determinant in polyphenols for potent anti-PAI-1 activity.

Despite the low IC₅₀ of TA and its ability to inhibit PAI-1 in the high protein environment of plasma (data not shown), it is not an ideal drug candidate due to its molecular mass of nearly 2000 daltons and its relative promiscuity, interacting with other proteins as well as itself at low- to mid-micromolar concentrations. Nonetheless, the inhibition of PAI-1 by TA and other gallate-containing molecules (EGCG, EGCDG, and gallic acid) formed the basis for development of follow-up compounds with improved properties compared with these naturally occurring polyphenols. Smaller di-, tri-, and pentagallates were designed with improved solubility in physiologic buffers and greater specificity toward PAI-1. These studies determined that although two galloyl moieties were sufficient to provide potent anti-PAI-1 activity, a minimum of 3 galloyl groups was required for efficacy in plasma. This suggests the relationship between specificity for PAI-1 and nonspecific bulk protein binding is complex and is not dependent on only the number of galloyl subunits.

The synthetic polyphenolic derivatives demonstrate clear advantages over previous pharmacologic inactivators of PAI-1. For example most of the existing PAI-1 inhibitors exhibit IC₅₀ values in the low- to mid-micromolar range in comparable in vitro assays, which is several orders of magnitude less potent than the best novel synthetic polyphenolic derivatives described here (25–27, 29, 30, 32, 34–36). Another class of PAI-1 inhibitors based on diketopiperazine derivatives have been described with in vitro IC₅₀ values reported in the 0.2–1 μm range; however, these compounds suffered from considerable physicochemical problems such as insolubility in physiologic buffer systems and were not subject to further development (47). CDE-066, in contrast, is soluble in physiologic saline solution at concentrations greater than 10 mm without loss of anti-PAI-1 activity (data not shown). Two other PAI-1 inactivators have been described with IC₅₀ values reported in the mid-nanomolar range; however, these compounds are ineffective against vitronectin-bound PAI-1, the predominant form of PAI-1 in plasma and the extracellular matrix (28, 33). Likewise several compounds with micromolar IC₅₀ values are also ineffective against vitronectin bound PAI-1 (26, 32). The resistance of vitronectin-bound PAI-1 to these inhibitors is thought to be due to the location of the binding site for these compounds, in a hydrophobic cavity on PAI-1 that is defined by α-helices D and E and β-strands 1A and 2A, and directly adjacent to the vitronectin-binding site (26, 28, 32). In contrast, the CDE-066 compound shows vitronectin-independent anti-PAI-1 activity in a purified system, in ex vivo plasma, and in vivo in PAI-1 transgenic mice.

The primary mechanism of action by which CDE-066 and the other synthetic polyphenols inactivate PAI-1 appears to be by binding to PAI-1 in a reversible manner and preventing stabilization of the non-covalent Michaelis complex with target proteases. This is demonstrated in Fig. 3 wherein preincubation of PAI-1 with each of the compounds inhibits its binding to the inactive protease, anhydrotrypsin. Identical data were also obtained in similar experiments using an inactive mutant of tPA (data not shown), indicating that the effect of the compounds on the initial association of PAI-1 with a protease is independent of the target protease. The SDS-PAGE analysis shown in Fig. 4 suggests that the polyphenolic compounds can also promote substrate behavior in PAI-1. However, in contrast to the loss of Michaelis complex formation (Fig. 3) and the loss of covalent complex formation (Fig. 4) the extent of cleavage observed is not dose dependent with the compounds added and varies with compound and target enzyme. It is possible that the extent of cleavage may be overestimated in these experiments due to complex dissociation during SDS-PAGE. Note, for example, that even in the absence of any compound, cleaved PAI-1 is apparent under experimental conditions where the stoichiometry of inhibition is near 1 (SI = 1.06, data not shown). Finally, consistent with the primary mechanism of action being inhibition of PAI-1:protease association, SPR experiments demonstrated that no CDE-066-dependent PAI-1 cleavage was detected when PAI-1 bound to vitronectin was reacted with active uPA (Fig. 8). This suggests that the combination of compounds and denaturants during SDS-PAGE may alter how PAI-1 is observed to behave.

The identification of naturally occurring polyphenols as a class of PAI-1 inhibitors is intriguing because such compounds, especially polyphenols derived from teas, fruits, and cocoa, have been suggested in recent years to provide benefits against pathologies such as chronic inflammation, neurodegeneration, cancer, and cardiovascular disease (48–50). Several mechanisms of action have been proposed for dietary polyphenols, characterizing these compounds as antioxidants, antiplatelet agents, and anti-inflammatory agents. Of particular relevance to PAI-1 are the proposed mechanisms by which dietary polyphenols may regulate hemostasis and prevent cardiovascular disease. In ex vivo and cell culture studies, dietary polyphenols have been shown to reduce tissue factor expression (51), increase plasminogen activator levels (52), and decrease PAI-1 via changes in gene expression (53). These effects are observed at micromolar concentrations of the compounds, a dose range that is well within the effective concentrations of the polyphenols identified in our study. Thus, it is possible that a previously unrecognized direct inactivation of PAI-1 may contribute to the complex pro-fibrinolytic and cardioprotective effects associated with dietary polyphenols. Future studies will focus on improving the specificity and activity of this class of synthetic polyphenolic compounds against PAI-1 as well as clarifying the role that direct PAI-1 inactivation may play in the healthful benefits derived from dietary polyphenols.