Termination of Bleeding by a Specific, Anti-catalytic Antibody to Plasmin

Summary

Background:

Excessive, plasmin-mediated fibrinolysis augments bleeding and contributes to death in some patients. Current therapies for fibrinolytic bleeding are limited by modest efficacy, low potency and off-target effects.

Objectives:

To determine whether an antibody directed against unique loop structures of the plasmin protease domain may have enhanced specificity and potency for blocking plasmin activity, fibrinolysis and experimental hemorrhage.

Methods:

The binding specificity, affinity, protease cross-reactivity, and antifibrinolytic properties of a monoclonal plasmin inhibitor antibody (Pi) were examined and compared to epsilon aminocaproic acid (EACA), a clinically used fibrinolysis inhibitor.

Results:

The Pi specifically recognized loop 5 of the protease domain and did not bind to other serine proteases or nine other non-primate plasminogens. By comparison to EACA, the Pi was ~7 logs more potent for neutralizing plasmin cleavage of small molecular substrates and >3 logs more potent for quenching fibrinolysis. When compared to α2-antiplasmin, the most potent covalent inhibitor of plasmin, the Pi was similarly effective for blocking catalysis of a small molecular substrate and was a more potent fibrinolysis inhibitor. Fab or chimerized Fab fragments of the Pi were equivalently effective. In vivo, in a humanized model of fibrinolytic surgical bleeding, the Pi significantly reduced bleeding more than a clinical dose of EACA.

Conclusions:

A monoclonal antibody directed to unique loop sequences in the protease domain, is a highly specific, potent, competitive plasmin inhibitor that significantly reduces experimental surgical bleeding in vivo.

Introduction

Bleeding is a serious or fatal complication of many conditions including surgery, injury and coagulation factor deficiency. Hemorrhage is the most common cause of death due to injury and excessive degradation of the fibrin thrombus (fibrinolysis) by plasmin contributes to as many as 40% of deaths from trauma [1]. Plasmin’s enzymatic activity is inhibited by the fast-acting serpin α2-antiplasmin and, the less potent inhibitor α2-macroglobulin [2-5]. Plasmin activity is also governed by substrate binding interactions with lysine residues on fibrin that are mediated by plasmin kringles, as well as by alterations to fibrin induced by activated factor XIII and by thrombin-activatable fibrinolysis inhibitor. Unregulated plasmin not only dissolves thrombi, but degrades clotting factors (fibrinogen, factor V, factor VIII), which impairs coagulation, thereby enhancing bleeding risk. Not surprisingly, severely injured patients with excessive fibrinolysis have been shown to have more severe coagulopathy, to receive more transfusions and to have a three-fold higher mortality [6]. A large study of lower risk trauma patients showed that current anti-fibrinolytic therapy caused modest but statistically significant reductions in death (5.7% to 4.9%) if given within three hours of trauma [7].

Two types of small molecule fibrinolytic inhibitors have been used for more than 50 years, lysine analogs and active site inhibitors. The lysine analogs, epsilon amino caproic acid (EACA) and tranexamic acid, interact with plasmin kringles to unfold the molecule and block its interactions with fibrin. Bovine pancreatic trypsin inhibitor (BPTI, aprotinin), is a competitive active site inhibitor of numerous serine proteases [8], which has both anticoagulant and antifibrinolytic effects [9]. BPTI is more effective at stopping bleeding and in preventing plasmin generation, than the lysine analogs [10-12]. Both the lysine analogs and BPTI accumulate in renal failure and cross the blood brain barrier. [10, 13-17] Both classes of agents are non-specific and have off-target effects; for example, tranexamic acid may cause seizures and BPTI is no longer available in some countries because of safety concerns. By comparison to BPTI, the lysine analogs have the undesirable pro-fibrinolytic properties of promoting non-fibrin-dependent plasminogen activation and preventing plasmin from being neutralized by its major inhibitor α2-antiplasmin.

A more potent, specific plasmin inhibitor, which doesn’t cross the blood brain barrier and placenta may reduce morbidity and mortality in patients with severe, life-threatening hemorrhage. Creating highly specific catalytic inhibitors of plasmin is challenging, because the enzymatic active site has significantly homology with other trypsin-like serine proteases [18]. Even α2-antiplasmin, the most potent and specific inhibitor of plasmin, also rapidly inhibits and interacts with other non-kringle containing serine proteases. However, the surface-exposed loops surrounding the active site of plasmin are structurally unique and they mediate highly specific interactions of plasmin with substrates and other molecules [19, 20]. In this study, we tested the hypothesis that an antibody specifically directed against the unique loop structures in the plasmin protease domain, would reduce plasmin’s enzymatic activity, fibrinolysis and surgical bleeding.

Methods

Recombinant proteins.

Plasmin protease domain mutants were generated in which the protease loop structures of plasmin were swapped with those of Factor D, a homologous serine protease, and the mutants were recombinantly expressed in E. coli as described [19]. Loop amino acid sequences in the plasmin protease domain were chimerized with the corresponding loop sequences of factor D as indicated (loop 3: TRFGQ changed to LNGA; loop 5: AHCLEKSPRPSSY changed to AHCLEDAADGKV; loop 6: AHQEVNLEPHV changed to AHSLSQPEPSK; loop 7: EPTRKD changed to HPDSQPDTIDHD). The cDNA sequence of the Pi was isolated from hybridoma cells by PCR cloning with redundant primers. A humanized chimeric version of the Pi was expressed in HEK293 cells by transient cell transfection (Creative Biolabs, Shirley NY). After 96 hours, the suspension culture was collected. Chimeric Pi was purified by HiTrap rProteinG FF and passed through a 0.2 um filter.

Saturation binding assay.

Saturation binding assays were performed by ELISA in 96-well plates coated with 2ug/ml of plasmin or plasminogen (Haematologic Technologies, Essex Junction, VT) at room temperature for 1 hour. Wells were washed with PBS-T: (10 mM phosphate buffer pH 7.4, 150mM NaCl, 0.05% Tween 20) six times. Non-specific protein binding sites were blocked with 1% BSA for 1 hour at room temperature. After washing the plate as above, various concentrations of antibody (0.002 to 100ug/ml) were added to the plate in duplicate and for 1 hour at room temperature. The plate was washed and horseradish peroxidase (HRP) conjugated goat anti-mouse secondary antibody (1:8000, Santa Cruz, Dallas, TX) was added to the plate. After 1 h incubation at room temperature, the plate was washed and TMB was added. The absorbance was then read at 370nm.

The binding of the Pi to human plasmin (113 nM) was also assessed using surface plasmon resonance using a Pioneer FE with a COOH5 chip in which the different forms of Pi were immobilized as described by the manufacturer (Fortebio, Fremont, CA).

Specificity of Pi binding.

The binding of the Pi to different serine proteases was assessed by ELISA in 96-well plates coated with human plasminogen (Haematologic Technologies, Essex Junction, VT), mouse plasminogen (Innovative research, Novi, MI), tPA (Genentech, South San Francisco, CA), trypsin (Sigma, St. Louis, MO) or chymotrypsin (Sigma, St. Louis, MO) (10 ug/ml) and blocked with 1% BSA. After incubation and washing, Pi or control monoclonal antibody supernatants (anti-digoxin) were added to the wells for 1 h in triplicate. Then goat anti mouse antibody conjugated with HRP (1:5000, Sc-2005, Santa Cruz, Dallas, TX) was then added. The plate was washed after incubation for 1h. TMB was added to the wells and absorbance at 370 nm was recorded.

Plasmin inhibition analysis.

Human plasmin (Haematologic Technologies, Essex Junction, VT) (2.8 nM) was mixed with S2251 substrate (Molecular Innovations, Novi, MI) and Pi (1x10⁻⁹ to 3.3x10⁻⁸M) or EACA (Sigma, St. Louis, MO) (3.9x10⁻³ to 0.25x10⁻¹M) or no inhibitor, and the rate of substrate cleavage was monitored by the release of paranitroanilide product at A405 nm for 20 min. The percent residual activity of plasmin as a function of the log concentration of inhibitor was determined.

The effect of Pi on the velocity of substrate cleavage by human plasmin was examined. Human plasmin (100 nM, Hematologic Technologies, Inc.) was mixed with or without Pi (50 nM) in the presence of the substrate S2251 (0.1 to 4 mM). The initial rate of substrate cleavage (6 min.) was monitored at A405. Data were plotted and analyzed using Michaelis-Menten kinetics by the Graphpad Prism Program.

The inhibition of human plasmin (Haematologic Technologies, Essex Junction, VT) by a2AP and Pi was compared. Human plasmin (200 nM) was mixed with a2AP (12.5 to 200 nM) or Pi (15.625 to 500 nM) or no inhibitor and the cleavage of S2251 was monitored at A405 in duplicate.

The effects of mouse or chimeric Fab on plasmin activity were examined. Plasmin (25 nM) was mixed with various concentrations (0 – 200 nM) of mouse Fab or chimeric Fab (cFab) and S2251 (0.5 mM) in TBS buffer pH 7.4. Plasmin activity was measured by the release of paranitroanilide product at A405 over time.

The effect of Pi on streptokinase-induced plasmin activity was examined in cynomolgus plasma. Streptokinase (Sigma, St. Louis, MO) (15 U, 10 uL) was added to cynomolgus plasma (10 ul) containing S2251 (10 ul, final 0.5mM), TBS buffer (10 ul, pH 7.4), Pi (0-2 uM, 10 ul). The cleavage of S2251 at 37°C was measured for 20 min. and the percent of activity was determined by comparison to samples without Pi.

Specificity of plasmin inhibition.

The effect of the Pi on different animal plasmins was examined after addition of the plasminogen activator urokinase. Various citrated animal plasmas (5ul; Equitech-Bio, Kerrville, TX, USA) were mixed in microtiter plates in the presence of Pi (1 uM, 10 ul) or buffer, S2251 (10 ul, 0.5 mM final), urokinase (100 or 200 U) in a total volume of 100 ul with PBS. The plates were incubated at 37° C. and the rate of plasmin generation was monitored by substrate cleavage at 405 nm for 1 h.

Factor V cleavage.

Human factor V (2 ug, IHFV, Innovative Research, Novi, MI) was mixed with plasmin (0.05 ug) for 1 hr at 21° C in the presence of EACA (1 mM), Pi (5 ug) or a control monoclonal antibody (digoxin, 5 ug). Samples were then subjected to SDS-PAGE under reducing conditions on 7.5% gels. Protein bands were detected by staining with Coomassie blue dye.

Fibrinolytic assays in vitro.

The effects of EACA vs. Pi on fibrinolysis were examined in duplicate in pooled human plasma clotted with CaCl2 (10 mM) and thrombin (1 U/ml) in the presence of trace amounts of ¹²⁵I fibrinogen for 1 h at 37° C. After washing in 1 ml of TBS pH 7.4, tPA (5 nM) was added in the presence of various amounts of Pi (62.5 to 1000 nM), EACA (0.125 mM to 2mM) or no inhibitor. After 2.5 h the amount of fibrinolysis was determined by measuring the amount of ¹²⁵I-fibrin dissolved in the supernatant.

The effects of a2AP and Pi on fibrinolysis were examined in clots were formed as above in the presence of human plasminogen (2 uM) and a2AP (1 uM) or Pi (1 uM) and various amounts of tPA (0-10 nM). Clot dissolution was monitored in duplicate over 2 h in microtiter plate wells at A405 nm. The amount of fibrinolysis was determined by the relative decrease in A405 as described [21].

The ability of Pi to inhibit primate clot lysis was also examined in a microtiter plate assay. Human, baboon or African green monkey plasma (Primate Biologicals, Miami, FL) (5 ul) was added to 96-well plates in the presence of various concentrations of Pi (0-2000nM), CaCl2 (2mM), thrombin (0.125U/ml) and human fibrinogen (2 mg/ml) and urokinase (10 U). The wells were incubated at 37°C, clot lysis was recorded kinetically for 120 min. and the time required for complete clot lysis was recorded.

The effects of Pi IgG, Fab or chimeric Fab were also compared in a clot lysis assay in microtiter plates. Human plasma (5 ul) was mixed with Pi (IgG or Fab, 0- 400nM), fibrinogen (Enzyme Research Laboratories) (2mg/ml), CaCl2 (2 mM), thrombin (Sigma, St. Louis, MO) (0.125U/ml) and urokinase (Sigma, St. Louis, MO) (10 U) in a total volume of 80 ul. Clot dissolution at 37° C. was monitored at 405nm. The percent clot dissolution at 30 min. was determined as the percent change in peak turbidity from baseline = 100% x (peak absorbance–30 min. absorbance)/ (peak absorbance – baseline absorbance).

Fibrinolytic bleeding assays in vivo.

Animal studies were approved by the Institutional Animal Care and Use Committee. Isoflurane-anesthetized C57Bl6 female mice (6-10 weeks old) were treated in a randomized, blinded fashion with saline (Control), Pi (100 nmole/kg) or a clinical dose of EACA (0.51 mmole/kg) by intravenous infusion via the jugular vein [22]. Ten minutes later, human plasminogen (100 nmole/kg) and plasminogen activator (40,000 IU of streptokinase/kg) were given over 60 min. Tails were pre-warmed for 5 mins in 3 mL of 37°C saline as described and 40 min. after initiation of the infusion tail transection was performed (10 mm from tip) and monitored for 30 min [23, 24].

Monoclonal antibody generation.

Monoclonal antibodies were generated by somatic cell fusion as we have described using splenocytes from mice immunized four times over eight months with human plasmin protease domain coupled to aprotinin agarose [25]. Direct binding assays were performed to identify mice with the highest titer of antibody binding to the human plasminogen protease domain. Mice were hyper-immunized intraperitoneally with the human plasminogen protease domain (50 ug) pre-activated with urokinase (100 U, 30 min.). After euthanasia, immune splenocytes were fused with SP2/0 cells with polyethylene glycol as we have described [25]. Fused cells were cultured in hypoxanthine aminopterin and thymidine selection medium at 37°C in 5% CO₂ atmosphere. Clones were selected by a radioimmunoassay in which microtiter plate were coated with goat-anti-mouse Fab antibody (5 ug/ml, 25 ul) for 1 hr. Wells were blocked with 1%bovine serum albumin. Hybridoma supernatants were added and plates were incubated for 1 hr at room temperature (RT). After washing, wells were incubated with ¹²⁵I-human plasmin (50,000 cmp/25 ul) for 1 hr. After washing, bound ¹²⁵I-human plasmin was detected by gamma scintigraphy. The positive hybridomas were subcloned by limiting dilution and stored in liquid nitrogen.

Monoclonal antibody purification.

Hybridoma supernatant was applied to an anti-mouse IgG agarose column (Immunechem, Burnaby, BC, Canada) with a flow speed at 1ml/min. The column was washed with PBS pH 7.4. The antibody was eluted with 0.2M glycine, pH 2.9, followed by dialysis in PBS pH 7.4 at 4°C overnight. The antibody was concentrated with Amicon Ultra-15 centrifugal filters (Millipore, Billerica, MA) according to the manufacturer. The protein concentration was determined by BCA method.

Statistical analyses.

Data are shown as mean ± standard error. Saturation binding and velocity vs. substrate curves were analyzed by nonlinear regression using a one-site binding hyperbole and Michaelis-Menten kinetic models with the aid of GraphPad Prism. In vivo bleeding data was log-transformed to achieve normality as assessed by Shapiro-Wilk testing followed by one-way ANOVA with Holm-Sidak correction for multiple comparisons. A p<0.05 was considered significant.

Results

Specific binding to extracellular loops in the plasmin protease domain.

Although the core protease structure is highly conserved among serine proteases, the external loop structures of the plasmin protease domain (Fig. 1A) differ significantly from other enzymes and have been shown to mediate specific interactions of plasmin with substrate modifiers (such as streptokinase), substrates and inhibitors [19]. A monoclonal antibody plasmin inhibitor (Pi) was generated that bound specifically to the protease domain of plasmin (microplasmin), as indicated by immunoblotting (Fig. 1B). Binding studies with different plasmin mutants, in which the loop structures of plasmin were altered, showed that by comparison to a polyclonal antibody against plasmin, the Pi bound to an epitope in the protease domain (Fig. 1A) that required native loop 5, but did not require the native sequences of loops 3, 6 or 7 (Fig. 1C).

Fig. 1. Binding specificity and avidity of Pi for human plasmin.

A) Ribbon diagram of microplasminogen showing the loops of the plasmin protease domain (PDB entry 1QRZ) visualized with iCN3D (https://www.ncbi.nlm.nih.gov/Structure/icn3d/icn3d.html). B) Pi binds specifically to the protease domain of reduced human plasmin by immunoblotting. Relative migration of molecular standards (kDa) is shown. C) Binding of Pi to various plasmin protease loop mutants. Loop amino acid sequences in the plasmin protease domain were chimerized with the corresponding loop sequences of factor D as indicated (loop 3: TRFGQ changed to LNGA; loop 5: AHCLEKSPRPSSY changed to AHCLEDAADGKV; loop 6: AHQEVNLEPHV changed to AHSLSQPEPSK; loop 7: EPTRKD changed to HPDSQPDTIDHD). Wells of a microtiter plate were coated with native (wild-type, WT) or various microplasmin loop mutants (5 ug/ml). After blocking, Pi, a polyclonal rabbit anti-plasmin (plasmin Ab) or a non-reactive mouse (anti-digoxin, control) antibody (1 ug/ml) were added to wells. The bound antibodies (Pi, control) were detected by a secondary anti-mouse peroxidase (1:10, 000) and the polyclonal plasmin antibodies were detected by anti-rabbit peroxidase antibody (1:5, 000) followed by TMB substrate with monitoring at A370.

The Pi is a potent competitive inhibitor of plasmin catalytic function with small substrates.

The Pi potently neutralized plasmin cleavage of a small tripeptide paranitroanilide substrate (S2251) with ~ 7 logs greater potency than EACA (Fig. 2B). The Pi markedly reduced the K_m for plasmin cleavage of S2251 by 14-fold (from 2.45 to 0.17 mM), but it had little effect on the V_max (0.011 vs. 0.014 A405/6 min.), suggesting that it acted largely as a competitive inhibitor of plasmin activity (Fig. 2B). The Pi had similar potency for inhibiting plasmin cleavage of S2251 as did α2-antiplasmin, the most potent, covalent inhibitor of plasmin (Fig. 2C). To determine whether the Fc region of the MAb resulted in steric hindrance or affected the inhibitory function of the Pi, we examined the effects of Fab fragments of the Pi, which lacked the Fc domain. Both the mouse Fab and chimeric Fab forms of the Pi potently inhibited plasmin activity with the tripeptide substrate (Fig. 2D), indicating that potential steric effects by the larger IgG molecule did not contribute significantly to blocking plasmin activity. Various forms of the Pi bound tightly to human plasminogen as assessed by saturation binding studies (Fig. 2E). The Pi forms also bound tightly to human plasmin as assessed by surface plasmon resonance experiments (Table 1). The K_d for mouse IgG and chimeric IgG forms ranged between 2.4-4.5 x 10⁻⁹M; the K_d for mouse and chimeric Fab ranged between 2.8-4.2 x 10⁻⁹M.

Fig. 2. Comparative inhibition of plasmin activity by Pi EACA and a2AP.

A). Human plasmin (2.8 nM) was mixed with S2251 substrate and Pi (1x10⁻⁹ to 3.3x10⁻⁸M) or EACA (3.9x10⁻³ to 0.25x10⁻¹M) or no inhibitor, and the rate of substrate cleavage was monitored at A405 nm for 20 min. The percent residual activity of plasmin as a function of the log concentration of inhibitor is shown. B). Effect of Pi on velocity of substrate cleavage by human plasmin. Human plasmin (100 nM, Hematologic Technologies, Inc.) was mixed with or without Pi (50 nM) in the presence of the substrate S2251 (0.1 to 4 mM). The initial rate of substrate cleavage (6 min.) was monitored at A405. Data were plotted and analyzed using Michaelis-Menten kinetics by the Graphpad Prism Program. Means (duplicate observations) are shown and are representative of 4 separate experiments. C). Comparative inhibition of human plasmin by a2AP and P. Human plasmin (200 nM) was mixed with a2AP (12.5 to 200 nM) or Pi (15.625 to 500 nM) or no inhibitor and the cleavage of S2251 was monitored at A405. The means of duplicate observations are shown. D). Fab and chimeric Fab forms of Pi efficiently inhibit plasmin enzymatic activity. Plasmin (25 nM) was mixed with various concentrations (0 – 200 nM) of mouse Fab or chimeric Fab (cFab) and S2251 (0.5 mM) in TBS buffer pH 7.4. Plasmin activity was measured by the release of paranitroanilide product at A405 over time. E) Saturation binding of Pi. Wells of microtiter plate were coated with human plasminogen (2ug/ml, Pg) or nothing, followed by washing and blocking with bovine serum albumin (BSA, control). After washing, various concentrations of purified Pi (shown) were added for 1 h. After washing the amount of antibody bound was detected by a secondary anti-mouse antibody followed by TMB substrate with monitoring at A370 nm. Data shown the means ± SE of triplicate observations, experiments repeated at least twice. The data were analyzed with Graphpad Prism, r=0.95.

Table 1.

Binding of different forms of Pi to human plasmin

Parameter	Antibody form of Pi
	Mouse IgG (n=7)	Chimeric IgG (n=6)	Mouse Fab (n=4)	Chimeric Fab (n=5)
KD (nM)	4.31 ± 0.46	2.72 ± 0.31	2.36 ± 0.06	4.47 ± 1.26
ka (M−1s−1) x 105	3.66 ± 0.24	5.61 ± 0.40	3.06 ± 0.17	2.01 ± 0.09
kd (s-1) x 10−3	1.55 ± 0.21	1.48 ± 0.09	0.72 ± 0.03	0.88 ± 0.22
Data are means ± SE; n= no. of experiments

Comparative binding specificity of the Pi.

By comparison to a control (anti-digoxin) monoclonal antibody, the Pi bound specifically to human plasminogen, but showed no specific binding to mouse plasminogen, human tPA, trypsin, chymotrypsin or bovine serum albumin (control) (Fig. 3A). The specificity of the Pi for inhibiting different animal species plasmins was studied in several different animal plasmas. The Pi suppressed plasmin activity in human plasma generated by urokinase, but the Pi did not inhibit plasmin activity in dog, guinea pig, bovine, cat, gerbil, hamster, pig, rabbit or rat plasma (Fig. 3B). Non-human primate plasmin has the greatest homology with human plasmin. The Pi inhibited plasmin activity in cynomolgus macaque plasma in a dose-related fashion (Fig. 3C). These data show that the Pi specifically inhibited human and non-human primate plasmin and did not inhibit other animal species plasmins or bind to other serine proteases tested.

Fig. 3. Binding and inhibitory specificity of the Pi.

A) Binding of Pi to various serine proteases. The binding of Pi or a control monoclonal antibody (anti-digoxin, Ctl) to human (h) plasminogen (Pg), mouse (m) Pg, tissue plasminogen activator (tPA), trypsin (tryps.), chymotrypsin (chym.) or bovine serum albumin (was assessed in microtiter plates by an ELISA as described in Methods. B) Effect of Pi or no Pi on plasmin activity generated in various animal plasmas after the addition of urokinase. Plasmin activity in human (Hu), dog, guinea pig (GP), bovine (Bov), cat (Cat), gerbil (Ger), hamster (Ham), pig, rabbit (Rab) and rat plasmas was assessed after addition of urokinase (UK) by the cleavage of S2251 substrate as described in Methods. C) Dose-related inhibition of plasmin activity in cynomolgus plasma. Streptokinase (15 U, 10 uL) was added to cynomolgus plasma (10 ul) containing S2251 (10 ul, final 0.5mM), TBS buffer (10 ul, pH 7.4), Pi (0-2 uM, 10 ul). The cleavage of S2251 at 37° C was measured for 20 min. and the percent of activity was determined by comparison to samples without Pi.

Comparative inhibition of factor V cleavage and fibrinolysis.

In vivo, factor V cleavage by plasmin is a hallmark of unregulated blood plasmin levels. In vitro, plasmin fully cleaved factor V to low molecular weight bands (Fig. 4A) in the presence of a control (anti-digoxin antibody). Factor V cleavage was mildly reduced by therapeutic concentrations of EACA, but it was more effectively diminished by the Pi (Fig. 4A). The Pi significantly suppressed plasmin-mediated fibrinolysis of human fibrin clots in a dose-related fashion (Fig. 4B). EACA also diminished fibrinolysis, but by comparison EACA was approximately 5,000-fold less potent than the Pi (Fig. 4B). When compared to α2-antiplasmin at the same dose, the Pi was 2-3-fold more potent in suppressing the plasmin-mediated dissolution of human fibrin clots induced by increasing doses of tPA (Fig. 4C). The whole IgG, Fab and chimeric Fab forms of the Pi also effectively inhibited clot lysis with similar dose-related potency (Fig. 4D), suggesting that there were no steric effects on fibrinolysis related to the IgG molecule.

Fig. 4. Comparative inhibition of plasmin mediated factor V cleavage and human clot fibrinolysis (dissolution) by Pi, EACA and a2AP.

A) Effect of EACA and Pi on plasmin-mediated cleavage of factor V. Human factor V (2 ug) was mixed with EACA (1 mM), a control monoclonal antibody (CTL, anti-digoxin, 5 ug) or Pi (5 ug) for 1 hr at room temperature, followed by SDS-PAGE on 7.5% gels. The relative migration of molecular weight standards is shown (kDa). Closed arrow heads indicate uncleaved factor V, open arrowheads indicated cleaved factor V. B). Comparative, dose-related effects of EACA and Pi on human clot lysis. Human plasma was clotted with CaCl2 (10 mM) and thrombin (1 U/ml) in the presence of trace amounts of ¹²⁵I fibrinogen for 1 h at 37° C. After washing in 1 ml of TBS pH 7.4, tPA (5 nM) was added in the presence of various amounts of Pi (62.5 to 1000 nM), EACA (0.125 mM to 2mM) or no inhibitor. After 2.5 h the amount of fibrinolysis was determined by measuring the amount of ¹²⁵I-fibrin dissolved in the supernatant. Means of duplicate observations are shown. C) Comparative effects of a2AP vs. Pi on fibrinolysis. Clots were formed as above in the presence of human plasminogen (2 uM) and a2AP (1 uM) or Pi (1 uM) and various amounts of tPA (0-10 nM). Clot dissolution was monitored in duplicate over 2 h in microtiter plate wells at A405 nm. The amount of fibrinolysis was determined by the relative decrease in A405 as described in Methods. D) Comparative effect of different Pi molecular forms on dissolution of human clots. Human plasma clots were mixed with various concentrations of Pi IgG, Pi Fab and chimeric Pi Fab (25nM-400nM) and plasmin was activated by urokinase. The amount of clot lysis was determined as described in Methods.

Suppression of fibrinolytic bleeding in vivo.

The potential utility of the Pi for suppression of fibrinolytic bleeding was examined in vivo in randomized, blinded studies of surgical arterial and venous hemorrhage. Since the Pi specifically binds and inhibits human, but not mouse plasmin, a humanized mouse tail bleeding model was developed in which normal mice were administered human plasminogen and human tPA. In this model, EACA was capable of inhibiting both mouse and human plasmin, though the Pi only interacted with human plasmin. By comparison to saline (control), a clinical dose of EACA (0.51 mmole/kg) did not significantly reduce bleeding time or blood loss (Fig. 5A, B). The Pi (100 nmole/kg) markedly suppressed bleeding time (>80%, p<0.01, Fig. 5A) and blood loss vs. controls (Fig. 5B, >85%, p<0.05) or vs. EACA-treated mice (Fig. 5B, p<0.05).

Fig. 5. Pi suppresses fibrinolytic bleeding in vivo. A) Duration of tail bleeding. B) Blood loss from bleeding.

Anesthetized mice were treated in a randomized, blinded fashion with saline (Control), Pi (100 nmole/kg) or EACA (510 micromole/kg) intravenously. Human plasminogen (100 nmole/kg) and plasminogen activator (80,000 IU/kg) were given over 60 min. Tail bleeding was initiated 40 min. after the infusion. Tails were pre-warmed for 5 mins in 3 mL of 37°C saline and bleeding was monitored as described [23, 43]. N=5-6, mean bleeding time ± SE and mean bleeding volume ± SE are shown. *p<0.05, **p<0.01; ns, not significant. One-way ANOVA, Neuman Keul’s corrections.

Discussion

Despite a need for specific plasmin inhibitors, they have been challenging to generate. The plasmin active site is conserved among other serine proteases and it shares kringle domains with molecules in other pathways. However, the surface-exposed loops of the protease domain have sequences unique to plasmin and these mediate highly specific interactions with substrate modifiers, inhibitors and substrates [19]. A monoclonal antibody Pi directed to these loops showed exquisitely specific, high affinity binding restricted to the protease domain of human plasmin, without detectable binding or activity against other tested serine proteases or non-primate plasmins. The Pi acted as a potent, competitive inhibitor of plasmin-mediated proteolysis. The Pi was ~ 7 logs more potent than EACA for inhibiting small substrate cleavage and ~5000-fold more potent for inhibiting fibrin cleavage. When compared to α2-antiplasmin, the most potent plasmin inhibitor, the Pi was similarly effective for inhibiting plasmin cleavage of small molecular weight substrates. The Pi was 2-3 fold more potent for inhibiting fibrinolytic activity on a molar basis than α2-antiplasmin, which may reflect the bivalency of the Pi antibody. The Pi potently neutralized plasmin is in primate, but not other animal plasmas. In a model of experimental surgical bleeding in mice, the Pi was a more potent than a clinical dose of EACA for stopping fibrinolytic bleeding.

The most widely used fibrinolytic inhibitors are the lysine analogs, EACA and tranexamic acid, which interact with lysine binding sites on plasmin kringles. The lysine analogs prevent plasminogen and tPA from binding to fibrin, thereby inhibiting plasminogen activation and fibrinolysis. Through these mechanisms, the lysine analogs can increase plasmin activity by blocking kringle interactions with α2-antiplasmin [26] and by enhancing plasminogen activation by tPA or uPA in solution [27, 28]. The lysine analogs also inhibit the interactions of plasminogen-plasmin with cellular receptors [29] and block interactions of plasminogen with tissue factor [30]. The biological effects of the interactions of the lysine analogs with other kringle-containing proteins (tPA [31], (pro)thrombin [32], hepatocyte growth factors, uPA, apoprotein (a) of lipoprotein(a) [32]), are not well understood. Lysine analogs cross the placenta and the blood brain barrier; they cause seizures in cardiac surgical patients [10, 13, 14, 16, 33] and increase brain infarction in subarachnoid hemorrhage patients [34]. EACA and tranexamic acid are largely excreted unchanged via the urine; therefore toxicity risk may be significant in patients with kidney disease [15, 22].

BPTI (aprotinin), is an active site inhibitor of plasmin, which is more potent than the lysine analogs for stopping cardiac surgical bleeding [10, 11] and for preventing generation of plasmin activity as measured by formation of plasmin-α2-antiplasmin complexes [12]. However, as its name implies, BPTI is also non-specific inhibitor of other serine proteases including trypsin, thrombin, activated protein C (a natural anticoagulant), kallikrein (which activates the kinin system), neutrophil elastase and other proteases [8]. Through these actions, BPTI has broad anticoagulant, antifibrinolytic, anti-inflammatory and other effects.[9] While more effective than the lysine analogs for stopping cardiac surgical bleeding, BPTI was associated with a significant increase in all-cause 30 day mortality, which resulted in withdrawal from the market [11].

Fibrinolysis affects bleeding in the millions of patients who undergo cardiac surgery, orthopedic surgery, liver transplantation, vascular surgery, thoracic surgery, gynecological surgery, end-stage renal disease, peripartum bleeding, gastrointestinal bleeding, neurosurgery, trauma, traumatic brain injury, intracerebral bleeding and subarachnoid hemorrhage [35-42]. Control of fibrinolysis during bleeding reduces complications, the need for re-operations, transfusions, length of hospitalization and mortality. In the largest study to date, lysine analog therapy reduced death after trauma, though the magnitude of its effect was limited (0.8% reduction in mortality) [7]. As such, for patients with serious, life threatening hemorrhage, the availability of more potent and specific plasmin inhibitors may improve reduce bleeding, morbidity and mortality [18].

Essentials.

• Fibrinolysis contributes to severe bleeding after trauma and surgery

• To reduce morbidity and mortality from bleeding will require more potent and specific agents

• A monoclonal antibody against the plasmin protease domain is a highly specific plasmin inhibitor

• This plasmin inhibitor significantly reduced experimental surgical bleeding in vivo