Inhibition of Plasminogen Activator Inhibitor-1
Its Mechanism and Effectiveness on Coagulation and Fibrosis
Objective— Serine protease inhibitors (serpin) play a central role in various pathological processes including coagulation, fibrinolysis, malignancy, and inflammation. Inhibition of serpins may prove therapeutic. As yet, however, only very few small molecule serpin inhibitors have been reported. For the first time, we apply a new approach of virtual screening to discover novel, orally active, small molecule serpin inhibitors and report their effectiveness.
Methods and Results— We focused on a clinically important serpin, plasminogen activator inhibitor-1 (PAI-1), whose crystal structure has been described. We identify novel, orally active molecules able to enter into the strand 4 position (s4A) of the A β-sheet of PAI-I as a mock compound. In vitro they specifically inhibit the PAI-1 activity and enhance fibrinolysis activity. In vivo the most effective molecule (TM5007) inhibits coagulation in 2 models: a rat arteriovenous (AV) shunt model and a mouse model of ferric chloride–induced testicular artery thrombosis. It also prevents the fibrotic process initiated by bleomycin in mouse lung.
Conclusions— The present study demonstrates beneficial in vitro and in vivo effects of novel PAI-1 inhibitors. Our methodology proves to be a useful tool to obtain effective inhibitors of serpin activity.
Serine protease inhibitors (serpin) play a central role in the regulation of a variety of pathological processes including coagulation, fibrinolysis, malignancy, and inflammation.1 Their inhibition may prove therapeutic.
Serpins consist of a β-sheets-rich body. They contain an exposed mobile reactive center loop (RCL)2 which, once cleaved by a target serine protease, inserts its N terminus part into the strand 4 position (s4A) of the A β-sheet, triggering its antiprotease activity. Small molecule compounds able to enter into the s4A position of the A β-sheet as a mock molecule may thus prevent the biological activity of the serpin.
Up to now, however, only very few small molecule serpin inhibitors have been described, none of which are in clinical use. Most have been discovered by high-throughput random screening (HTS) of a large chemical library,3–5 a rather inefficient strategy. In the present study, a new approach of virtual screening based on the 3-dimensional structure of a serpin, (PAI-1), was used to discover novel, orally active, small molecule compounds able to inhibit the target molecule.
PAI-1 regulates the plasminogen activation system through inhibition of its target serine proteases, tissue-type and urokinase-type plasminogen activator (tPA and uPA).1 Studies in humans and animals have demonstrated that PAI-1 expression is enhanced in various disorders such as thrombosis, fibrotic diseases, atherosclerosis, radiation damage, and cancer progression.6 PAI-1 has been linked with fibrin deposition evolving into organ fibrosis and atherosclerosis, or with striking alterations of cell adhesion and migration mediating cancer progression.7 The absence of PAI-1 in PAI-1 knockout mice markedly attenuates these pathological processes.8–12 Inhibition of PAI-1 by a neutralizing antibody13 provides similar promising results in animal experiments. Small molecule PAI-1 inhibitors, active orally, should prove useful to treat not only thrombotic disorders but also fibrotic processes and cancer.2,14 They should be more efficient than the thrombolytic agents in present use such as streptokinase and recombinant tPA, both of which are protein-based and require intravenous administration to obtain a rapid onset action.
We relied on the 3-dimensional structure of PAI-1 and on the virtual screening method to identify novel orally bioavailable molecules able to fit into the s4A position of PAI-1 as a mock compound. These compounds inhibit PAI-1 activity in vitro and coagulation in vivo in two rodent models of vascular thrombosis. Furthermore, we demonstrate for the first time that they also prevent the fibrotic process induced in mice lung by bleomycin, and thus that, in this model, PAI-1 is not a surrogate marker of fibrosis but rather its main cause.
Materials and Methods
Virtual screening was performed by the software system MOE (Molecular Operating Environment).15 The initial chemical library, of about 2 240 000 entries, merges various commercially available chemicals and deletes redundant entries. In the first step, 2 different filters based on molecular descriptors were used for coarse screening of compounds. The first filter was calculated from drug molecules that are clinically used in Japan. It was represented by the specific distributions of the molecular descriptors for these molecules and represents their general drug-likeness. The second filter was calculated from distributions of the molecular descriptors common to known reference inhibitors16 and from the inhibitory reactive-center loop peptide N-Ac-TVASS-NH217 that binds PAI-1. It represents the specific lead-likeness of the inhibitory molecules. Through the 2 filters we efficiently reduced the number of compounds to about 3000.18 The second step involved docking simulation of the selected compounds to the PAI-1 molecule by the program Ph4Dock.19 The crystal structure (1A7C) of the complex between PAI-1 and the inhibitory reactive-center loop peptide was obtained from the Protein Data Bank20 and used as the target structure for the docking study. After removal of water molecules and binding peptide, hydrogen atoms were added in accord with the standard protonation states of acidic and basic residues in proteins and their positions were optimized. Because the cleft that is occupied by β-strand 4A in the latent form of PAI-1 likely plays a critical functional role, we focused on it for docking simulation. This target site was characterized by the alpha site finder function21 in MOE for the subsequent docking experiments. The MMFF94s22 force field was used in the docking simulation. In the final step, the positions of hydrogen atoms in PAI-1 and all atoms in small molecules were optimized. Docking results were sorted by the following interaction energy.
Uele and Uvdw are the electrostatic and van der Waals interactions between the protein and the small molecule, respectively, and Uligand is the conformation energy of the small molecule. The molecules with the lowest Utotal values are considered as candidate molecules for biological assay.
In Vitro Assays, Toxicity, and Pharmacokinetics
Please see the supplemental data section, available online at http://atvb.ahajournals.org, for detailed Methods.
Arteriovenous Shunt Model
Animal experiments were performed in accordance with the Animal Experimentation Guidelines of Tokai University School of Medicine. Thrombus formation in arteriovenous (AV) shunts was achieved in male CD rats (Charles River Japan Inc, Kanagawa, Japan) by a previously described method.23 Before the study, TM5007 (300 mg/kg), warfarin (1.2 mg/kg), or ticlopidine (500 mg/kg), suspended in 0.5% carboxymethyl cellulose sodium salt (CMC) solution, was given by gavage, or tPA (275000 IU/kg) was administered intravenously by a bolus injection (n=7 for each group). Control rats were given 0.5% CMC solution only (n=7). Blood was allowed to circulate through the shunt for 30 min. The wet weight of the thrombus covering the silk thread was eventually measured.
Ferric Chloride–Induced Thrombosis Model and Visualization of Thrombi
Male mice were anesthetized with an intraperitoneal injection of 12 mL/kg ketamine-xylazine. Rhodamine 6G (Sigma; 0.1 mL of 0.1%) was injected intravenously. A testicular artery (100 to 150 μm in diameter) was carefully exposed for ferric chloride (FeCl3) treatment. A cotton thread (0.2 mm in diameter) saturated with 0.25 mol/L FeCl3 was applied to the adventitial surface of the testicular artery. After 5 minutes, the cotton thread was replaced by a saline solution in the wound. Thrombus formation in the testicular artery was subsequently monitored through 3-dimensional imaging using an ultrafast laser confocal microscope equipped with a piezo-electric motor control unit as previously reported.24 The time from endothelial damage by FeCl3 to occlusion of testicular arteries by large thrombi was measured. Mice were pretreated either by gavage of TM5007 (200 mg/kg), twice a day for 4 days. The antiplatelet glycoprotein (GP) IIb/IIIa agent, tirofiban (0.13 mg/kg, Wako), was single administered in a single intravenous injection before the injury.
Bleomycin-Induced Pulmonary Fibrosis
Male C57BL/6J (CLEA Japan Inc.) mice weighing 19 to 21 g were anesthetized with intraperitoneal pentobarbital and their trachea exposed by a cervical incision. Ten animals served as controls. Ten animals received an intratracheal instillation of bleomycin (Nippon Kayaku) dissolved in saline (1.5 U/kg), and 10 animals received in addition by gavage TM5007 (200 mg/kg) suspended in 0.5% CMC, twice a day for 14 days. Lung tissue was obtained for histological analysis and measurement of hydroxyproline content. Hydroxyproline was measured in tissue hydrolysates by the method of Kivirikko et al.25 Tissue sections were stained with hematoxylin and eosin and pulmonary fibrosis was scored on a scale of 0 to 8 using a previously described method.26 Azan stain for collagen was also used.
All data are expressed as the mean±SE. Differences among groups were assessed by Kruskal-Wallis test. The statistical significance was determined by 2-tailed Mann Whitney U test. Values are considered significant at P<0.05. All statistical analyses were performed on the statistical package SPSS for Windows (Version 14.0, SPSS).
Our library, encompassing 2 240 000 chemicals, was virtually screened, as described in methods. The use of 2 different filters reduced the number of compounds to about 3000, the first filter assessing drug-likeness and the second evaluating specific lead-likeness of the inhibitory molecules.
Docking simulation was then undertaken by a program Ph4Dock19 for the remaining compounds. The crystal structure (1A7C) of the complex of PAI-1 with its inhibitory RCL peptide20 was used, knowing that the 14-aa peptide corresponding to the N terminus of the RCL of PAI-1 inhibits the in vitro activity of PAI-1.27 The program evaluated whether the compound is able to fit within the PAI-1 cleft. Virtual screening by a combination of 2 filters and by the docking method identified 95 candidate compounds with high binding affinity to the s4A position of PAI-1. The simulated binding mode of 1 novel candidate compound (TM5001) is illustrated (please see supplemental Figure I). The compound binds tightly within the cleft, in the s4A position.
Docking simulations were also undertaken for PAI-1 inhibitors previously identified by HTS3,5 to understand their mechanism of action. Both tiplaxtinin and ZK4044 bind at almost the same site as TM5001 and TM5007 (Figure 1). The similarity of these characteristics suggests that the 4 inhibitors share a common binding region at the s4A position, despite completely different chemical structures.
In Vitro Assessment
We purchased or synthesized 28 of the candidate compounds discovered by virtual screening and tested their biological activities in vitro by three different assays. Inhibition of PAI-1 activity was measured by tPA-dependent hydrolysis of peptide substrate. The 2 most effective candidate compounds, N, N′-bis (3,3′-carboxy-4,4′-phenyl-2,2′-thienyl) hexanedicarboxamide (TM5001) and N, N′-bis [3,3′-carboxy-4,4′-(2,2′-thienyl)-2,2′- thienyl] hexanedicarboxamide (TM5007) had an efficacy comparable to that of tiplaxtinin (IC50 for TM5001, TM5007, and tiplaxtinin 28.6±7.3, 29.2±4.2, and 40±7 μmol/L, respectively). TM5001 and TM5007 share a common binding mode with TM5001 as illustrated in Figure 1. It suggests a similar molecular mechanism for their activity. Neither TM5001 nor TM5007 (up to 250 μmol/L) modified other serpin/serine protease systems (ie, α1-antitrypsin/trypsin and α2-antiplasmin/plasmin): their PAI-1 inhibitory activity appears thus specific (please see supplemental Figure II).
On SDS-PAGE, PAI-1 formed a covalent complex with tPA whereas no PAI-1/tPA complex formation was observed when PAI-1 was preincubated with our compounds (exemplified for TM5007 in Figure 2).
Finally the inhibition of fibrinolysis was tested on a fibrin plate (exemplified for TM5007). The area of tPA-induced fibrinolysis was decreased by PAI-1. Preincubation of PAI-1 with our compounds prevented this effect (please see supplemental Figure III).
Toxicity and Pharmacokinetics
Cytotoxicity of TM5001 and TM5007 was assessed with HeLa cells as the LDH activity released into the culture medium after 24-h incubation. Results are expressed as percentage of the LDH release induced by the lysis of all cells. TM5001 (100 μmol/L) and TM5007 (250 μmol/L) did not significantly raise maximum LDH activity above unstimulated controls (30.8±7.1 and 26.6±2.2%, respectively, versus 23.8±1.6%; P=0.077 and 0.137).
Acute toxicity of TM5001 and 5007 was evaluated in vivo in mice. Various single doses of up to 2000 mg/kg of both compounds elicited no symptoms up to 2 wks later.
Subacute toxicity of TM5007 was assessed in rats at 2 different doses (300 mg/kg/d for 1 wk or 2000 mg/kg/d for 2 week) given daily. Neither blood pressure nor body weight was modified. No biochemical abnormalities were noted in plasma and urine including bleeding time, APTT, PT, TT, and red blood cell count (please see supplemental Tables).
Plasma Tmax, Cmax, and T1/2 were calculated in rats given orally 50 mg/kg of each of the 2 compounds. They reached 18 h, 32 μmol/L, and 54 h for TM5001, and 18 h, 8.8 μmol/L, and 124 h for TM5007.
In Vivo Assessment
TM5007 was chosen for further investigations, as it proved more effective than TM5001. Its in vivo anticoagulant effectiveness was assessed by weighing the blood clot obtained in a rat AV shunt model (Table). Blood clot weight was significantly lower in rats given 300 mg/kg of TM5007 (54.8±0.8 mg) than in vehicle-treated rats (74.3±3.2 mg) (P<0.01). This effect, obtained at a plasma concentration of TM5007 of 5.2±0.7 μmol/L, was equivalent to that of warfarin (1.2 mg/kg) and ticlopidine (500 mg/kg), and superior to that of tPA (275000 IU/kg). PAI-1 activity was significantly reduced by TM5007 (0.8±0.1 versus 1.0±0.1 ng/mL in vehicle treated rat) but not by other agents, whereas APTT and PT were not modified (Table).
The anticoagulant effectiveness of TM5007 was also evaluated in FeCl3-treated mouse testicular arteries. Growth of thrombi, followed by 3-dimensional imaging, led to complete arterial occlusion within 12.7±2.7 min (n=29) in vehicle-treated mice versus only 56.3±8.9 min in TM5007-treated mice (n=15, P<0.01 compared to the control group). This effect was obtained at a TM5007 plasma concentration of 4.6±0.6 μmol/L. Tirofiban had a similar effect (58±12.4 min; n=5, P<0.01 compared to the control group).
The in vivo antifibrotic effect of TM5007 was tested in a mouse model of bleomycin induced pulmonary fibrosis. Bleomycin increased significantly the lung hydroxyproline content above that of control mice (232.9±8.5 versus 140.2±4.8 μg/lung, P<0.001). TM5007 significantly lowered the bleomycin-induced lung hydroxyproline content (204.2±9.5 μg/lung, P<0.05). Bleomycin raised plasma PAI-1 activity above that of control mice (1.7±0.2 versus 0.8±0.1 ng/mL, P<0.001). TM5007, at a plasma concentration of 9.2±0.2 μmol/L, lowered significantly the bleomycin-induced PAI-1 rise (1.2±0.1 ng/mL, P<0.05). Histological evidence of pulmonary fibrosis was markedly improved by TM5007 (Figure 3A through 3C). Azan staining disclosed the accumulation of collagen in bleomycin-treated lungs (Figure 3D through 3F). Bleomycin raised the fibrosis score above control (4.7±0.37 versus 0.5±0.17, P<0.001), a rise that was partially prevented by TM5007 (2.9±0.38, P<0.01), in good agreement with the results of plasma PAI-1 activity.
Relying on virtual screening and the 3-dimensional structure of the complex of PAI-1 with its inhibitory peptide, we have identified 2 novel, orally bioavailable, small molecule PAI-1 inhibitors, TM5001 and TM5007. Both are stable, nontoxic, and devoid of cellular toxicity as demonstrated in vitro by their inability to raise LDH levels in the medium of cultured HeLa cells. The absence of acute and subacute toxicity is confirmed in vivo in mice given a single dose of up to 2000 mg/kg, or in rats fed 300 mg/kg for 1 week or 2000 mg/kg for 2 weeks. The in vivo effectiveness of TM5007 is demonstrated in animal models of either acute vascular thrombosis or of chronic lung fibrosis, without deleterious effects on blood pressure or bleeding, in good agreement with previous results in PAI-1 deficient mice and humans.8,14
The specificity of the effect of TM5007 on PAI-I was further documented in other serpin/serine protease systems (ie, α1-antitrypsin/trypsin and α2-antiplasmin/plasmin) by a chromogenic assay with synthetic substrates. TM5007 exhibited no inhibitory activity against any of the closely related serpins or serine proteases at a concentration of 250 μmol/L. This concentration is approximately 10 times above the IC50 (29 μmol/L) for TM5007 against PAI-1 in the in vitro study and approximately 25 to 50 times greater than the peak plasma levels of this compound in vivo in mice and rats (5 to 10 μmol/L). The activity of TM5007 appears thus specific for PAI-I.
Inhibition of thrombus formation has been demonstrated for all known PAI-I inhibitors, both in vitro and in vivo in acute thrombotic models.3–5,28,29 Delayed effects, by contrast, on subsequent tissue remodeling were demonstrated only for tiplaxtinin in a murine model of angiotensin II–induced hypertension.11 We confirm that TM5007 is a powerful antithrombotic agent, which does not prolong bleeding time, PT, and APTT. Its effect, at a dosage of 300 mg/kg, was equivalent to that of warfarin (1.2 mg/kg) and ticlopidine (500 mg/kg) and superior to that of tPA (275000 IU/kg).
The antithrombotic effect of TM5007 was also demonstrated in a FeCl3-induced testicular artery thrombosis mouse model. Oral pretreatment with TM5007 (200 mg/kg twice daily for 4 days) was as efficacious as a single intravenous injection of an antiplatelet drug, Tirofiban (0.13 mg/kg).
Of note, the IC50 of TM5007 against PAI-1 calculated in vitro (29.2 μmol/L) exceeds the peak plasma levels (5 to 10 μmol/L) of this compound observed in vivo in mice and rats. This discrepancy probably reflects the differences in the experimental systems. In vitro, PAI-1 inhibition is measured directly. In vivo, by contrast, its effect on thrombus formation is complex as it involves several factors other than PAI-1/tPA.
We demonstrate for the first time in the present study that PAI-I inhibition prevents the fibrotic process initiated in the lung by bleomycin. Eitzman et al, studying mice overexpressing or lacking the PAI-1 gene, have demonstrated a strong correlation between PAI-1 expression and collagen accumulation in lung tissue.9 The inhibition of lung fibrosis by TM5007 establishes that PAI-1 is not a mere surrogate marker of fibrosis but rather its main cause. This observation is of potential importance. Fibrotic changes are indeed associated with the failure of several other organs, including the heart, vessels, liver, and kidney. Their prevention might transform the fate of numerous diseases such as cardiovascular disease, liver cirrhosis, renal disease, and radiation injury.
A few small molecule PAI-I inhibitors have been discovered by the rather inefficient HTS screening of a chemical library.3–5 Gerlatova et al30 demonstrated that a binding epitope for tiplaxtinin is adjacent to a previously identified interaction site for vitronectin, thereby suggesting that the antiserpin activity of this drug is mediated by an interaction between PAI-I and vitronectin. By contrast, we focused on the s4A position as a target site of PAI-I inhibition. Our docking simulation demonstrated that TM5001 and TM5007 preferentially bind to this site, suggesting that our compounds exert inhibitory activity through blocking the s4A position. On SDS-PAGE, no PAI-1/tPA complex formation was indeed observed when PAI-1 was preincubated with our compounds (Figure 2). It thus appears from the results that the inhibitory mechanism of our compound is not identical with that of tiplaxtinin. Interestingly, despite their completely different chemical structures, tiplaxtinin as well as ZK4044 potentially associate with the s4A position on our docking simulation, in good agreement with a previous assumption by Gerlatova et al30 that tiplaxtinin inhibits the PAI-1 at multiple mechanisms. Altogether, these findings confirm the critical role of the s4A position as a target site of PAI-I inhibition.
Utilization of the 3-dimensional structure of PAI-1 has allowed not only an understanding of the molecular events leading to PAI-1 inhibition but also the virtual screening of new inhibitors for clinical use. Their potential applications include thrombotic disorders (arterial and venous), fibrotic diseases, amyloidosis, obesity, and type 2 diabetes mellitus.31,32 The availability of a specific inhibitor should also offer a potentially important pharmacological tool to investigate the role of PAI-1 in these processes.
Finally, elucidation of the 3-dimensional structure of other serpins together with the use of virtual screening, should allow the identification of other small molecule serpin inhibitors, to curtail their harmful effects.
We thank H. Saito, K. Okada, O. Matsuo, H. Oishi, and K. Nakazato for their expertise.
Sources of Funding
This study was supported by a grant from the New Energy and Industrial Technology Development Organization in Japan (to T.M., N.H., and K.K.).
Original received October 8, 2007; final version accepted January 10, 2008.
Carrell RW, Boswell DR. Serpins: the superfamily of plasma serine protease inhibitors. In: Barrett, Salvesen, eds. Proteinase Inhibitors. Elsevier Science. 1986; 403–420.
Eitzman DT, Westrick RJ, Xu Z, Tyson J, Ginsburg D. Plasminogen activator inhibitor-1 deficiency protects against atherosclerosis progression in the mouse carotid artery. Blood. 2000; 96: 4212–4215.
Weisberg AD, Albornoz F, Griffin JP, Crandall DL, Elokdah H, Fogo AB, Vaughan DE, Brown NJ. Pharmacological inhibition and genetic deficiency of plasminogen activator inhibitor-1 attenuates angiotensin II/salt-induced aortic remodeling. Arterioscler Thromb Vasc Biol. 2005; 25: 365–371.
MOE (Molecular Operating Environment), version 2004.04; Chemical Computing Group Inc.: Montreal, Quebec, Canada, 2004.
Egelund R, Einholm AP, Pedersen KE, Nielsen RW, Christensen A, Deinum J, Andreasen PA. A regulatory hydrophobic area in the flexible joint region of plasminogen activator inhibitor-1: defined with fluorescent activity-neutralizing ligands. J Biol Chem. 2001; 276: 13077–13086.
Xue Y, Bjorquist P, Inghardt T, Linschoten M, Musil D, Sjolin L, Deinum J. Interfering with the inhibitory mechanism of serpins: crystal structure of a complex formed between cleaved plasminogen activator inhibitor type 1 and a reactive-centre loop peptide. Structure. 1998; 6: 627–636.
Edelsbrunner H, Facello M, Fu P, Liang J Measuring Proteins and Voids in Proteins. Proceedings of the 28th Hawaii International Conference on Systems Science. 1995;256–264.
Ashcroft T, Simpson JM, Timbrell V. Simple method of estimating severity of pulmonary fibrosis on a numerical scale. J Clin Pathol. 1988; 41: 467–470.
Friederich PW, Levi M, Biemond BJ, Charlton P, Templeton D, van Zonneveld AJ, Bevan P, Pannekoek H, and ten Cate JW. Novel low-molecular-weight inhibitor of PAI-1 (XR5118) promotes endogenous fibrinolysis and reduces postthrombolysis thrombus growth in rabbits. Circulation. 1997; 96: 916–921.
Gorlatova NV, Cale JM, Elokdah H, Li D, Fan K, Warnock M, Crandall DL, Lawrence DA. Mechanism of inactivation of plasminogen activator inhibitor-1 by a small molecule inhibitor. J Biol Chem. 2007; 282: 9288–9296.
Smith LH, Dixon JD, Stringham JR, Eren M, Elokdah H, Crandall DL, Washington K, Vaughan DE. Pivotal role of PAI-1 in a murine model of hepatic vein thrombosis. Blood. 2006; 107: 132–134.