FXa-α₂-Macroglobulin Complex Neutralizes Direct Oral Anticoagulants Targeting FXa In Vitro and In Vivo

Abstract

Increasing number of patients are treated with direct oral anticoagulants (DOAC). An antidote for dabigatran inhibiting thrombin (idarucizumab) is available but no antidote is yet approved for the factor Xa (FXa) inhibitors (xabans). We hypothesized that a complex between Gla-domainless FXa and α₂-macroglobulin (GDFXa-α₂M) may neutralize the xabans without interfering with normal blood coagulation.

Purified α₂M was incubated with GDFXa to form GDFXa-α₂M. Affinity of apixaban and rivaroxaban for GDFXa-α₂M was only slightly decreased compared to FXa. Efficacy and harmlessness of GDFXa-α₂M were tested in vitro and in vivo. Stoichiometric excess of GDFXa-α₂M neutralized rivaroxaban and apixaban as attested by clot waveform assay and rotational thromboelastometry, whereas GDFXa-α₂M alone had no effect on these assays. Efficacy and pro-thrombotic potential of GDFXa-α₂M were also assessed in vivo. Half-life of GDFXa-α₂M in C57BL6 mice was 4.9 ± 1.1 minutes, but a 0.5 mg/mouse dose resulted in uptake saturation such that 50% persistence was still observed after 170 minutes. Single administration of GDFXa-α₂M significantly decreased the rivaroxaban-induced bleeding time (p < 0.001) and blood loss (p < 0.01). GDFXa-α₂M did not increase D-dimer or thrombin–antithrombin complex formation, suggesting a lack of pro-thrombotic potential.

GDFXa-α₂M is therefore an attractive candidate for xaban neutralization neither pro- nor anticoagulant in vitro as well as in vivo.

Keywords

antidote - apixaban - rivaroxaban - oral anticoagulant

Introduction

Following decades of vitamin K antagonist prescription as unique oral anticoagulant bridging therapy with heparin or its low molecular weight (MW) derivatives, direct oral anticoagulants (DOACs) have been developed[1] targeting thrombin (dabigatran) or coagulation factor Xa (FXa; xabans). As prescriptions expand, experience underlines that bleeding complications also occur with DOAC.[2] [3] [4] [5] This is specially threatening in the absence of available antidote or reversal agent in clinical scenarios such as massive haemorrhage, trauma, stroke requiring thrombolysis or urgent surgery.

Effective specific antidotes are available for vitamin K antagonists, heparin and partially for its low MW derivatives.[6] [7] [8] [9] An antidote for dabigatran had been approved which consists in a humanized monoclonal Fab (idarucizumab) that can be used in case of severe active bleeding and emergency surgery or invasive procedure.[10] [11] [12] Xabans still lack approved antidote. Pro-haemostatic agents (prothrombin complex concentrate activated or not) increase thrombosis risk, thus their use is currently restricted to life-threatening situations.[3] [13] [14] Andexanet-α is the most advanced candidate as antidote to xabans. It is a modified FXa lacking its Gla domain that had been inactivated by a S195A mutation.[15] [16] Andexanet-α efficiently binds FXa inhibitors and does not activate prothrombin nor consume antithrombin albeit still forms a low affinity Michaelis complex with the serpin.[17] [18] ANNEXA-4 clinical trial is currently in progress to confirm its efficacy and safety.[19] Ciraparantag is a small cationic compound antagonizing DOAC as well as heparin and derivatives. Little data are yet available on this molecule which is currently in phase II trial.[20] [21] Besides antidotes, potent by-passing agents are also in the pipeline, such as the I16L FX variant, acting as a zymogen-like acquiring prothrombin activator potential through factor Va binding, or the DOAC resisting 99-loop FX variant.[22] [23] [24] Their key advantage resides in the limited amount needed to correct DOAC-induced bleeding, in contrast to stoichiometric lure or bait antidotes.

We hypothesized that Gla-domainless FXa (GDFXa) sequestered by α₂-macroglobulin (GDFXa-α₂M) would be prevented from interacting with pro- as well as anticoagulant macromolecules (factor Va, prothrombin, antithrombin and tissue factor pathway inhibitor [TFPI]), whereas its active site remaining functional still binds xabans.

Alpha₂M is a broad-spectrum molecular trap inhibitor[25] [26] [27] mainly targeting thrombin, FXa and plasmin in blood.[28] [29] [30] Native α₂M is a homotetrameric glycoprotein (MW 720 kDa; plasma concentration, 3.5 µM in adults). Each sub-unit includes a bait region targeted by numerous proteases and a cysteinyl-glutamyl thiol ester bond. Cleavage induces a major conformational change trapping the protease within a cage-like quaternary structure.[31] [32] [33] Native α₂M and α₂M that had reacted with a protease have markedly different shapes, the latter having a paradoxical enhanced mobility (fast form) in native gel electrophoresis. Cleavage also unmasks the γ-glutamyl groups which react with NH₂ ε-lysyl group of the protease covalently linking the entrapped protease.[34] [35] Steric hindrance prevents macromolecules to interact with the entrapped protease, which nevertheless still cleaves small peptidyl substrates and is neutralized by peptidyl chloromethyl ketone.[29] Thus, it was reasonable to expect that entrapped GDFXa would still bind DOAC inhibiting FXa.

We prepared GDFXa-α₂M complex and assessed its potential regarding xabans neutralization in vitro in platelet-poor plasma (PPP) and whole blood as well as in vivo in a pre-clinical bleeding model.

Materials and Methods

Whole Blood and Platelet-Poor Plasma

Blood was collected by venipuncture (0.105 M buffered trisodium citrate 9/1 v/v) from healthy volunteers who gave their written informed consent (Etablissement Français du Sang, Paris, France; convention C CPSL UNT n°13/EFS/064). Pooled normal PPP was purchased from Cryopep (Montpellier, France).

Drugs, Proteins and Reagents

Apixaban and rivaroxaban were kindly provided by Bristol-Myers Squibb/Pfizer (Princeton, New Jersey, United States) and Bayer Healthcare AG (Leverkusen, Germany), respectively. About 4 mg apixaban or rivaroxaban were dissolved in dimethyl sulfoxide (DMSO). Just prior to use, rivaroxaban was rapidly diluted at 1/100 in H₂O and further diluted in 50 mM Tris-HCl pH 7.5 containing 0.15 M NaCl (tris-buffered saline [TBS]) and 1% DMSO. Dilutions were performed directly in TBS containing 1% DMSO for apixaban. Effective (final) concentrations of xabans were measured in PPP by anti-Xa activity on a STA-R and a set of specific calibrators (Stago, Asnières, France). For in vivo studies, pills of rivaroxaban (Xarelto, Bayer) were dissolved in 10 mM HCL (vehicle) and used to force-feed mice at the indicated dose. Enoxaparin (Lovenox) was purchased from Sanofi Aventis (Gentilly, France) and fondaparinux (Arixtra) from Aspen (Marly-le-Roi, France). Dilutions were performed in TBS. Human GDFXa expressed in bacteria was purchased from Cambridge ProteinWorks (Cambridge, UK). Antithrombin was purchased from LFB (Aclotine, Courtaboeuf, France), TFPI from American Diagnostica (Greenwich, Connecticut, United States), recombinant human tissue factor (TF) from Dade Behring (Innovin, Marburg, Germany) and aprotinin from Nordic Group Pharmaceuticals (Paris, France). Phospholipid vesicles were prepared by sonication (2 minutes in pulse mode 0.15/s, 80 W, 4°C) of a 1 mg/mL mixture of L-α-phosphatidylcholine (66%, w/w) with L-α-phosphatidylserine (33%, w/w), both from Avanti Polar Lipids (Alabaster, Alabama, United States) as previously described.[36] Chromogenic substrate N-α-benzyloxycarbonyl-d-Arg-Gly-Arg-pNA (S2765) and the inhibitors D-Phe-Pro-Arg-chloromethyl ketone (PPACK) were purchased from Cryopep, phenylmethylsulfonyl fluoride (PMSF) and methylamine (40% solution) from Sigma Aldrich (Steinheim, Germany). 1,5 Dansyl-Glu-Gly-Arg chloromethyl ketone (DEGRck) was purchased from Merck KGaA (Darmstadt, Germany).

GDFXa-α₂M Preparation and Characterization

Human α₂M was purified according to published protocols.[34] [37] [38] Briefly, a cocktail of inhibitors (1 µM PPACK, 2 mM PMSF, 2 mM ethylenediaminetetraacetic acid (EDTA) and 100 KIU/mL aprotinin) was added to freshly thawed frozen plasma. Plasminogen was removed by chromatography on Lysine-Sepharose 4B and vitamin K-dependent factors by 80 mM barium chloride precipitation and the bulk of fibrinogen by 4% polyethylene glycol (PEG-6000). Supernatant was brought to 12% PEG-6000 and the resulting pellet was re-suspended in 1/10 of the initial PPP volume in 20 mM sodium phosphate, pH 6.0. Alpha₂M was adsorbed onto iminodiacetic acid-Sepharose column (25 × 160 mm, 3 mL/min) saturated by 50 mM zinc acetate then eluted with 0.1 M EDTA pH 8.0. Following concentration by ultrafiltration (Amicon Ultra-15 100K, Merck KGaA), α₂M was further purified by gel filtration on Superose 6B column (16 × 500 mm, 1 mL/min). Purified α₂M was concentrated by ultrafiltration up to 15 mg/mL estimated by immunonephelometry (BN II, Siemens). GDFXa-α₂M was formed by incubating GDFXa (2 µM) with α₂M (8 µM) for 20 minutes at 37°C in TBS containing 5 mM MnCl₂.[29] [30] Preparations were kept at 4°C until use. Complete sequestration of GDFXa by α₂M was controlled by comparing velocity of S2765 hydrolysis in the presence or absence of 2.6 µM antithrombin and 1 IU anti-Xa/mL enoxaparin. Active site concentration of GDFXa-α2M was verified by titration with DEGRck. GDFXa (0.5 µM) free or fully sequestered by α2M were incubated with various DEGRck dilutions prepared immediately before use. Following 30 minutes of incubation at 37°C, residual activities were measured after a 1/20 dilution using S2765. The reported amount of GDFXa-α₂M refers to the concentration of active GDFXa sequestered. Inhibition constants (K_I ) of apixaban and rivaroxaban for GDFXa-α₂M were determined as previously described.[39] Apparent values were corrected for the S2765 competition by taking into account its Michaelis constant for GDFXa-α₂M (60.3 ± 4.4 µM) estimated as previously reported for FXa.[40] For reference purpose, a small amount of purified α₂M was activated into the fast form by incubation with 200 mM methylamine (2 hours at room temperature) followed by extensive dialysis. Native α₂M, GDFXa-α₂M, and methylamine-treated α₂M were analysed by native polyacrylamide gel electrophoresis (Invitrogen NuPAGE 3–8% Tris-Acetate Gel, Thermo Fisher Scientific, France).

Clot Waveform and Rotational Thromboelastometry Assays

The kinetic of fibrin polymerization (clot waveform) was recorded by measuring A₄₀₅ every 8 seconds at 37°C on a Tecan Infinite M200 Pro reader.[41] Coagulation was initiated in 80 µL PPP containing 5 pM TF and 4 µM phospholipid vesicles by adding 20 µL 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid 50 mM pH 7.35 containing 60 mg/mL bovine serum albumin and 100 mM CaCl₂. Data were normalized and the lag time of clot formation was defined as the time needed to reach 15% of the maximum turbidity measured through A₄₀₅. Rotational thromboelastometry was performed on a ROTEM delta (Werfen, Baden-Dättwil, Switzerland). Coagulation was initiated through addition of 300 µL pre-warmed whole blood containing or not containing xaban and/or GDFXa-α₂M to 40 µL triggering solution ensuring final concentrations of 2.5 pM TF, 10 µM phospholipid vesicles and 20 mM CaCl₂. Reported clotting times are as defined by the manufacturer. Whole blood and PPP were spiked (1/20; v/v) with TBS containing 1% DMSO and the xaban as required.

Mice

C57Bl/6JRj male mice (25–30 g) were purchased from Janvier Labs (Le Genest-Saint-Isle, France) and all assays performed at the Animal Platform, CRP2–UMS 3612 CNRS – US25 Inserm-IRD (Université Paris Descartes). Mice were anaesthetized by intra-peritoneal injection of a ketamine (80 mg/kg) and xylazine (16 mg/kg) mixture and were euthanized by cervical dislocation. All animal experiments were approved by the Ethic Committee on Animal Resources of Université Paris Descartes (registration number 201506151109793–V5 APAFiS #2677).

GDFXa-α₂M Half-Life in Mice

Half-life of GDFXa-α₂M was evaluated after a single 100 µL retro-orbital plexus injection of 150 nM GDFXa-α₂M (0.34 mg/kg). Following tail vein transection (see below), 25 µL of blood were collected through challenges at timed intervals and immediately diluted at 1/5 in TBS containing 5 mM EDTA. Residual activity was evaluated in PPP after centrifugation (1,500 × g, 10 minutes, 20°C), by measuring rate of S2765 hydrolysis. Data were normalized with respect to hydrolysis measured in the sample collected 1 minute post-injection. The dependence of the normalized rate of hydrolysis on time was analysed by non-linear regression analysis using a single exponential decay equation to estimate the in vivo half-life of GDFXa-α₂M. Persistence of GDFXa-α₂M in blood was evaluated through 100 µL injection in each retro-orbital plexus of 3.6 µM GDFXa-α₂M (17 mg/kg). Blood was collected in 0.11 M buffered trisodium citrate (9/1; v/v) by cardiac puncture at different time points (1–6 hours post-injection). Residual GDFXa-α₂M activity was evaluated by measuring the rate of S2765 hydrolysis, as above.

Mouse Tail Vein Transection and Bleeding Model

Mouse bleeding model was adapted from published method.[42] Mice were force-fed with 10 mM HCl containing or not 50 mg/kg rivaroxaban. Two hours later, 100 µL GDFXa-α₂M (3.6 µM) or its vehicle were injected in each retro-orbital plexus. Mouse tail was soaked in a mixture of NaCl (0.15 M) and EDTA (2 mM) at 37°C. Mouse was positioned on its right side and tail introduced in a homemade device enabling positioning at precisely its 2.5-mm diameter section. Using a mechanical linear guide, the left lateral tail vein was transected by a 0.5-mm deep incision. Mouse tail was replaced into soaking mixture and initial bleeding time monitored. Fifteen, 30, and 45 minutes post-injury, wound was challenged by gently wiping it twice with a 37°C saline-wetted gauze swab in the distal direction. Following each challenge, mouse tail was placed into a new collection tube containing the soaking mixture and re-bleeding was monitored. Red blood cells in each collection tube were collected (1,500 × g, 10 minutes, 20°C) and lysed in 20 mM Tris pH 7.5. A₄₁₆ was transcribed into microlitre blood loss in reference to a titration curve. Secondary bleeding time and blood loss were defined as the sum of bleeding time and blood loss following the three challenges.

Evaluation of Pro-Coagulant Markers in Mice Receiving GDFXa-α₂M

GDFXa-α₂M (3.6 µM) or its vehicle were injected in each retro-orbital plexus and blood collected 30 minutes later by cardiac puncture, as above. Pro-thrombotic potential of GDFXa-α₂M was evaluated by clot waveform assay as described above. D-dimers and thrombin–antithrombin complexes (TAT) were measured by enzyme-linked immunosorbent assay (ELISA) according to the manufacturer's instructions using Mouse D-Dimers (D2D) ELISA Kit (Cusabio, anticorps-enligne.fr) and TAT complexes Mouse ELISA Kit (Abcam, Paris, France), respectively.

Statistical Analyses

All statistical analyses were computed using the GraphPad Prizm software. Half-life and K_I values were expressed as mean ± standard deviation of three determinations. Coefficients of determination (R ²) were given by linear or non-linear regression analyses. Clotting time, lag time ratio, bleeding time and blood loss values were compared all together by Kruskal–Wallis test followed by Dunn tests performed for pair-wise comparisons. Vehicle versus antidote, D-dimers and TAT values were compared by unpaired two-tailed Mann–Whitney test. Statistical significance was accepted for p-value of < 0.05.

Results

GDFXa-α₂M Preparation and Characterization

Purified human α₂M appeared homogeneous by native polyacrylamide gel electrophoresis analysis ([Fig. 1A]). Following incubation with GDFXa, the main species detectable migrated faster than native α₂M suggesting that in spite of the molar excess of α2M over GDFXa most α₂M had been activated to the fast form. Importantly, complete GDFXa sequestration by α₂M was achieved since hydrolysis of S2765 by GDFXa-α₂M was indistinguishable whether or not a mixture of 2.6 µM antithrombin and 1 IU anti-Xa/mL enoxaparin was added ([Fig. 1B]). On the contrary, rate of S2765 hydrolysis by GDFXa alone was fully inhibited by the antithrombin enoxaparin mixture. The same was true when enoxaparin was substituted with 1.4 mg/mL fondaparinux. We also verified that GDFXa-α₂M resisted TFPI inhibition: TFPI completely inhibited GDFXa but had no effect on GDFXa-α₂M amidolytic activity ([Fig. 1C]). Active site titration confirmed that GDFXa fully sequestered by α₂M remains active albeit with a lower catalytic activity than that of free GDFXa ([Fig. 1D]). Therefore, all reported concentrations of GDFXa-α₂M reflect the active site concentration of GDFXa within the complex and not total amount of α₂M. Adding 1 µM purified α₂M to PPP had no detectable effect on clot waveform assay ([Fig. 2A]). On the contrary, active GDFXa dose-dependently decreased the lag time, thus had pro-coagulant potential ([Fig. 2B]). It was therefore of upmost importance to avoid any traces of free GDFXa. Reproducible and satisfactory results were obtained by incubating α₂M with GDFXa at a 4/1 stoichiometric ratio. The inhibition constants (K_I ) of apixaban and rivaroxaban for GDFXa-α₂M were slightly higher than those we recently reported for FXa[39] in strictly identical conditions (K_I  = 2.41 ± 0.22 nM [R ² = 0.84] vs. 0.74 ± 0.03 nM and 1.29 ± 0.13 nM [R ² = 0.87] vs. 0.47 ± 0.02 nM, respectively; [Fig. 2C]). Above all, the preparation of GDFXa-α₂M was devoid of pro- or anticoagulant activity as attested by clot waveform assay performed in PPP whether or not containing 1.7 µM GDFXa-α₂M.

Fig. 1 Preparation of Gla-domainless FXa α₂-macroglobulin complex (GDFXa-α₂M). (A) Native polyacrylamide gel electrophoresis (PAGE) (gradient 3–8%): lane 1, GDFXa only (invisible in this gel system); lane 2, purified native α₂M; lane 3, GDFXa-α₂M complex; lane 4, fast form of α₂M (methylamine-treated). In spite of the 4/1 stoichiometric ratio, most if not all native α₂M shifted towards the fast form upon incubation with GDFXa. (B and C) Progress of S2765 hydrolysis by 1 nM GDFXa (closed circles) or GDFXa-α₂M (closed squares). Corresponding open symbols in B were obtained in the presence of 2.6 µM antithrombin and 1 IU anti-Xa/mL enoxaparin; corresponding open symbols in C were obtained in the presence of 145 nM tissue factor pathway inhibitor (TFPI). In both cases, GDFXa was fully inhibited, whereas GDFXa-α₂M was unaffected. (D) Active site titration of GDFXa (open squares) and GDFXa-α₂M (closed squares) by 1,5 Dansyl-Glu-Gly-Arg chloromethyl ketone (DEGRck). GDFXa or GDFXa-α₂M (0.5 µM) were incubated 30 minutes at 37°C with DEGRck at the indicated concentration and the residual catalytic activity was measured after 1/20 dilution using S2765. GDFXa cleaved 1.4-fold faster S2765 than GDFXa-α₂M.

Fig. 2 Characterization of Gla-domainless FXa α₂-macroglobulin complex (GDFXa-α₂M). (A) Clot waveform triggered by adding 5 pM tissue factor (TF) and 4 µM phospholipid vesicles to platelet-poor plasma (PPP) (closed squares), containing 1 µM purified α₂M (open circles) or 1.7 µM GDFXa-α₂M (closed diamonds). Graph represents the A₄₀₅ versus time (mean values ± standard error of the means [SEM] of 6 experiments). Lag time was defined as the time to reach 15% of the maximum. (B) Clot waveform triggered by adding 5 pM TF and 4 µM phospholipid vesicles to PPP alone (squares) or containing increasing amounts of GDFXa: 31.2 nM (circles), 62.5 nM (diamonds), 125 nM (upward triangles) and 250 nM (downward triangles). GDFXa was a potent pro-coagulant. (C) Graph represents the residual rate of 400 µM S2765 hydrolysis by 5 nM GDFXa-α₂M as a function of rivaroxaban (closed squares) or apixaban (closed circles); mean value ± SEM of three determinations. Solid lines were obtained by non-linear regression analysis using the tight-binding inhibition equation.[40] Values of 2.41 ± 0.22 nM (R ² = 0.84) and 1.29 ± 0.13 nM (R ² = 0.87) were obtained for apixaban and rivaroxaban, respectively.

In Vitro Neutralization of Xabans by GDFXa-α₂M

We used clot waveform assay as a coagulation-based assay allowing detection of as little as 5 to 10 ng/mL xaban in PPP. To evaluate xaban neutralization by GDFXa-α₂M, we triggered clot waveform assays in PPP spiked with increasing amount of xabans in the presence or absence of 1.7 µM GDFXa-α₂M. Apixaban or rivaroxaban dose-dependently prolonged the lag time. GDFXa-α₂M fully neutralized xaban anticoagulant effect even at supra-therapeutic levels. Indeed, lag times in the presence of GDFXa-α₂M were comparable to control, irrespective of the xaban concentration ([Fig. 3A]).

Fig. 3 Neutralization of xabans in vitro. (A) Graph represents the lag time of clot waveform triggered by adding 5 pM tissue factor (TF) and 4 µM phospholipid vesicles. Platelet-poor plasma (PPP) was spiked with increasing amounts (as indicated in the abscissa) of rivaroxaban (open squares) or apixaban (open circles). Corresponding closed symbols were obtained in these PPP containing in addition 1.7 µM Gla-domainless FXa α₂-macroglobulin complex (GDFXa-α₂M). Solid lines link the mean value of three determinations (error bars min and max). (B) Graph represents the clotting time (CT in seconds) of rotational thromboelastometry assays triggered by adding 2.5 pM TF and 10 µM phospholipid vesicles to blood spiked with rivaroxaban (squares) or apixaban (circles) and containing or not 1.7 µM GDFXa-α₂M (closed or open symbols, respectively). Xaban concentrations indicated in the abscissa are the actual concentrations measured in PPP by anti-Xa activity. Solid lines were obtained by linear regression analysis. Overall, 1.7 µM GDFXa-α₂M fully neutralized supra-therapeutic amounts of xaban (600 ng/mL) without affecting (in their absence) the lag time in PPP or the clotting time in the whole blood.

ROTEM triggered by low TF is a whole blood coagulation-based assay potentially allowing xaban detection. Apixaban or rivaroxaban dose-dependently prolonged the clotting time. Adding 1.7 µM GDFXa-α₂M fully reversed xaban anticoagulant effect: the clotting times were comparable to those obtained in the absence of xaban, irrespective of the drug concentration of up to 600 ng/mL ([Fig. 3B]).

Overall, GDFXa-α₂M neutralized supra-therapeutic amounts of apixaban and rivaroxaban in blood and PPP, whereas it was without effect in the absence of xabans.

Half-Life and Persistence of GDFXa-α₂M in Mice

The reported half-life of FXa-α₂M in mice is 2 minutes[43] and that of subtilisin-α₂M in rats is 6 minutes.[44] In accord with these data, we estimated a half-life of 4.9 ± 1.1 minutes (R ² = 0.90; [Fig. 4A]) after injection of 10 µg/mice GDFXa-α₂M. However, uptake of protease-α₂M is saturable.[45] [46] [47] Thus, we evaluated the actual persistence of GDFXa-α₂M following injection at the high dose needed for stoichiometric xaban neutralization. When 0.5 mg/mice GDFXa-α₂M was injected, 50% of the catalytic activity of GDFXa-α₂M was still detectable 170 minutes following injection ([Fig. 4B]). Consequently, we designed the in vivo experiments taking into account the actual persistence of GDFXa-α₂M in mice.

Fig. 4 Half-life and persistence of Gla-domainless FXa α₂-macroglobulin complex (GDFXa-α₂M) in mice. (A) Kinetic of the normalized residual rate of S2765 hydrolysis when 100 µL GDFXa-α₂M (150 nM) was injected in mice. Data points are mean value of three determinations (error bars min and max). Solid line was obtained by nonlinear regression analysis using a single exponential decay equation. Estimated half-life of GDFXa-α₂M in mice was 4.9 ± 1.1 minutes (R ² = 0.90). (B) Kinetic of the normalized residual rate of S2765 hydrolysis when 100 µL GDFXa-α₂M (3.6 µM) were injected in each retro-orbital plexus in mice. Data points are mean value of three determinations; solid line links the mean values (error bars min and max). Probably due to saturation of the uptake mechanisms, 50% of GDFXa-α₂M catalytic activity persisted 170 minutes after injection.

Neutralization of Xabans by GDFXa-α₂M in Mice

GDFXa-α₂M efficacy for xaban neutralization was evaluated in vivo using a mouse bleeding model. Mice were force-fed with rivaroxaban or placebo and 2 hours later, vehicle or GDFXa-α₂M (100 µL, 3.6 µM) were injected in each retro-orbital plexus. Primary endpoints were bleeding time and blood loss following lateral tail vein transection. Initial bleeding time and blood loss were comparable whether or not mice were force-fed with rivaroxaban (p > 0.05). However, following challenges, rivaroxaban increased secondary bleeding time (p < 0.001) as well as blood loss (p < 0.01), whereas both were comparable to placebo following GDFXa-α₂M injection (p > 0.05). Overall, GDFXa-α₂M significantly decreased rivaroxaban induced bleeding time (from 4.92 ± 1.0 to 2.84 ± 0.4 minutes; p < 0.001; [Fig. 5A]) and blood loss (from 351 ± 45 to 180 ± 87 µL; p < 0.01; [Fig. 5B]). In the absence of rivaroxaban, GDFXa-α₂M had no effect on bleeding time (2.90 ± 0.28 vs. 2.95 ± 0.51 minutes; p > 0.05) and blood loss (158 ± 99 vs. 163 ± 73 µL; p > 0.05). We concluded that GDFXa-α₂M effectively neutralized rivaroxaban anticoagulant effect in this in vivo model and that GDFXa-α₂M alone had no adverse effect on mouse haemostasis.

Fig. 5 Neutralization of rivaroxaban in mice. (A) Sum of secondary bleeding time (in minutes) following challenges at time 15, 30 and 45 minutes past initial tail vein transection. (B) Corresponding sum of secondary blood loss (in µL). Mice were force-fed with 10 mM HCl (open symbols) or rivaroxaban (closed symbols). Two hours later, 100 µL vehicle alone (circles) or containing 3.6 µM Gla-domainless FXa α₂-macroglobulin complex (GDFXa-α₂M) (squares) were injected in each retro-orbital plexus of the mice. Rivaroxaban significantly increased the bleeding time (p <0.001) and blood loss (p <0.01). This rivaroxaban-induced bleeding was significantly reduced by GDFXa-α₂M both for the bleeding time (p <0.001) and blood loss (p <0.01) becoming non-significantly different from the control or GDFXa-α₂M alone groups (p > 0.05). Drug/vehicle versus drug/antidote pair-wise comparisons estimated by Dunn tests (**p < 0.01; ***p < 0.001).

Evaluation of Pro-Coagulant Markers in Mice Receiving GDFXa-α₂M

The above data suggested that GDFXa-α₂M was devoid of in vivo adverse effect. To exclude the potential pro-thrombotic effect of GDFXa-α₂M in mice, we performed clot waveform assay in samples collected 30 minutes post-injection of vehicle or GDFXa-α₂M. Whether or not mice received GDFXa-α₂M, the lag time of clot waveform did not differ between groups (p = 0.18; [Fig. 6A]). Moreover, D-dimer and TAT levels were comparable (p = 0.50 and p = 0.13; [Fig. 6B] and [C]). Overall results suggested that GDFXa-α₂M was devoid of pro- as well as anticoagulant properties in vivo.

Fig. 6 Lack of Gla-domainless FXa α₂-macroglobulin complex (GDFXa-α₂M) pro-thrombotic potential in mice. Pro-thrombotic potential of GDFXa-α₂M was evaluated in platelet-poor plasma (PPP) from mice injected with vehicle alone (open circles) or containing 3.6 µM GDFXa-α₂M (closed circles). (A) Lag time (LT in seconds) of clot waveform. (B) Amount of D-dimers and (C) of thrombin–antithrombin (TAT) measured by enzyme-linked immunosorbent assay (ELISA). Determined by unpaired two-tailed Mann-Whitney test (vehicle vs. GDFXa-α₂M) no significant difference was observed between lag times (p = 0.18), D-dimers (p = 0.50) or TAT (p = 0.13) suggesting a lack of pro-thrombotic potential in mice for GDFXa-α₂M.

Discussion

Our goal was designing a neutralizing agent for xabans in vitro and in vivo. Our specification was that agent should neutralize xabans without otherwise affecting haemostasis. GDFXa-α₂M was an attractive candidate because the active site of FXa within the complex would be preserved and bind xabans, whereas steric hindrance would prevent interaction with macromolecules (prothrombin, factor Va, antithrombin or TFPI). We herein documented that 1.7 µM GDFXa-α₂M neutralized supra-therapeutic amounts of apixaban or rivaroxaban while devoid by itself of pro- or anticoagulant properties.

Mouse tail vein transection is a pre-clinical model sensitive to pharmacological intervention in haemophilia A mice.[42] We evaluated GDFXa-α₂M efficacy as xabans neutralizing agent using this model. We confirmed as previously reported[15] that oral administration of rivaroxaban produced variable initial blood loss. In contrast, secondary bleeding time and blood loss were significantly increased in mice treated with rivaroxaban alone, whereas in the groups having received in addition GDFXa-α₂M, values were comparable to those of the control group. Presumably rivaroxaban binding to GDFXa-α₂M neutralized its anticoagulant effect. That bleeding time and blood loss were comparable whether or not mice received GDFXa-α₂M suggested harmlessness in addition to effectiveness. In accord with this hypothesis, neither D-dimer nor TAT levels increased following GDFXa-α₂M injection.

Ideally, an antidote should neutralize and eliminate its target in vivo rather than just sequestering it. Cleavage of the bait region in α₂M exposes previously concealed recognition sites for receptors on fibroblasts, macrophages and hepatocytes that uptake α₂M-protease from the circulation.[45] [46] [47] [48] [49] Clearance is indistinguishable whether or not the active site of the α₂M-linked protease is inhibited by low MW inhibitors.[48] Assuming that such depletion occurs with GDFXa-α₂M, it could literally deplete xaban from blood, bypassing the natural clearance mechanisms. On the other hand, we evidenced that high amount of GDFXa-α₂M had a relatively long persistence in blood, presumably due to saturation of the uptake mechanisms. GDFXa-α₂M would rapidly neutralize xabans but eliminate them only slowly. Accordingly, GDFXa-α₂M may constitute a fairly long-lasting antidote not requiring continuous perfusion or multiple injections. It is worth mentioning that α₂M and FXa are native blood components thus not immunogenic and that 1.7 µM GDFXa-α₂M neutralizing supra-therapeutic levels of xabans is below the normal plasma concentration of α₂M (3.5 µM in adults; 5.6 µM in newborn). Affinity of apixaban and rivaroxaban for GDFXa-α₂M was just slightly less than that for FXa, supporting the hypothesis of stoichiometric interaction with the xabans.

Thrombin generation assay is affected by xabans and thus would allow evaluating their neutralization.[50] However, GDFXa-α₂M ruined the thrombin generation assay because of its ability to cleave FluCa, precluding measurement of thrombin generation in its presence. Amount of GDFXa-α₂M added to neutralize xabans (1.7 µM) actually exceeded the thrombin potential (1.3 µM).

The prospect of using GDFXa-α₂M as an antidote to xaban has limitations. First, it is difficult to consider α₂M purified from human plasma for therapeutic use and large-scale production of recombinant α₂M may turn out difficult. Second, our GDFXa-α₂M preparation was a mixture of native α₂M, and activated α₂M having sequestered or not GDFXa. Adding micromolar quantities of native α₂M to PPP had no detectable effect on blood coagulation. Blood readily contains large amounts of α₂M which does not prevent xaban effects. We showed that active GDFXa is highly pro-coagulant and verified that GDFXa-α₂M preparation did not contain traces of GDFXa. We also verified that GDFXa-α₂M preparation itself was devoid of adverse effect in PPP (neither pro- nor anticoagulant). It can be concluded that sequestered GDFXa neutralized the xabans. We unexpectedly observed that partial neutralization of dabigatran as well as of heparin or its derivative occurs with activated forms of α₂M such as methylamine-treated α₂M in addition to GDFXa-α₂M. Specifically, the catalytic activity of sequestered GDFXa is not affected by dabigatran, yet GDFXa-α₂M partially neutralizes the anticoagulant effect of dabigatran. Mechanism likely involves secondary binding sites latent in native α₂M. During blood coagulation, FXa-α₂M and thrombin-α₂M are formed. FXa-α₂M formation with its 170 nM FXa potential is minor with respect to therapeutic xaban concentration. Again xabans are efficient coagulation inhibitors in spite of blood α₂M content.

Overall, GDFXa-α₂M represented an in vitro and in vivo attractive neutralizing agent of xabans, neither pro- nor anticoagulant.

Conflict of Interest

I.G.-T. received honoraria for participating in expert meetings on apixaban (Bristol-Myers Squibb/Pfizer). P.G. received honoraria for participating in expert meetings on enoxaparin (Sanofi) and together with I.G.-T. and V.S. on rivaroxaban (Bayer Healthcare AG). The other authors declare no conflict of interest.

Acknowledgements

The authors thank the generous donators of the Conny-Maeva charitable foundation. They also thank N. Neveux (AP-HP Hôpital Cochin, Paris, France) for the specific dosage of α₂M and K. Kamaleswaran together with F. Martin for excellent technical skills.

Authors' Contributions

B.L.B. conceived the study and together with G.J. designed and performed research, analysed data and wrote the manuscript; I.G.-T., V.S., S.G. and P.G. critically discussed the data, revised the manuscript and gave final approval.

• References

• 1 Weitz JI, Harenberg J. New developments in anticoagulants: past, present and future. Thromb Haemost 2017; 117 (07) 1283-1288
• 2 Greinacher A, Thiele T, Selleng K. Reversal of anticoagulants: an overview of current developments. Thromb Haemost 2015; 113 (05) 931-942
• 3 Siegal DM, Garcia DA, Crowther MA. How I treat target-specific oral anticoagulant-associated bleeding. Blood 2014; 123 (08) 1152-1158
• 4 Reiffel JA, Weitz JI, Reilly P. , et al; Cardiac Safety Research Consortium presenters and participants. NOAC monitoring, reversal agents, and post-approval safety and effectiveness evaluation: a cardiac safety research consortium think tank. Am Heart J 2016; 177: 74-86
• 5 Raval AN, Cigarroa JE, Chung MK. , et al; American Heart Association Clinical Pharmacology Subcommittee of the Acute Cardiac Care and General Cardiology Committee of the Council on Clinical Cardiology; Council on Cardiovascular Disease in the Young; and Council on Quality of Care and Outcomes Research. Management of patients on non-vitamin K antagonist oral anticoagulants in the acute care and periprocedural setting: a scientific statement from the American Heart Association. Circulation 2017; 135 (10) e604-e633
• 6 Crowther M, Crowther MA. Antidotes for novel oral anticoagulants: current status and future potential. Arterioscler Thromb Vasc Biol 2015; 35 (08) 1736-1745
• 7 Wolzt M, Weltermann A, Nieszpaur-Los M. , et al. Studies on the neutralizing effects of protamine on unfractionated and low molecular weight heparin (Fragmin) at the site of activation of the coagulation system in man. Thromb Haemost 1995; 73 (03) 439-443
• 8 Ageno W, Büller HR, Falanga A. , et al. Managing reversal of direct oral anticoagulants in emergency situations. Anticoagulation Education Task Force White Paper. Thromb Haemost 2016; 116 (06) 1003-1010
• 9 Goldstein JN, Refaai MA, Milling Jr TJ. , et al. Four-factor prothrombin complex concentrate versus plasma for rapid vitamin K antagonist reversal in patients needing urgent surgical or invasive interventions: a phase 3b, open-label, non-inferiority, randomised trial. Lancet 2015; 385 (9982): 2077-2087
• 10 Schiele F, van Ryn J, Canada K. , et al. A specific antidote for dabigatran: functional and structural characterization. Blood 2013; 121 (18) 3554-3562
• 11 Glund S, Stangier J, Schmohl M. , et al. Safety, tolerability, and efficacy of idarucizumab for the reversal of the anticoagulant effect of dabigatran in healthy male volunteers: a randomised, placebo-controlled, double-blind phase 1 trial. Lancet 2015; 386 (9994): 680-690
• 12 Pollack Jr CV, Reilly PA, van Ryn J. , et al. Idarucizumab for dabigatran reversal - full cohort analysis. N Engl J Med 2017; 377 (05) 431-441
• 13 Majeed A, Ågren A, Holmström M. , et al. Management of rivaroxaban- or apixaban-associated major bleeding with prothrombin complex concentrates: a cohort study. Blood 2017; 130 (15) 1706-1712
• 14 Martin AC, Le Bonniec B, Fischer AM. , et al. Evaluation of recombinant activated factor VII, prothrombin complex concentrate, and fibrinogen concentrate to reverse apixaban in a rabbit model of bleeding and thrombosis. Int J Cardiol 2013; 168 (04) 4228-4233
• 15 Lu G, DeGuzman FR, Hollenbach SJ. , et al. A specific antidote for reversal of anticoagulation by direct and indirect inhibitors of coagulation factor Xa. Nat Med 2013; 19 (04) 446-451
• 16 Siegal DM, Curnutte JT, Connolly SJ. , et al. Andexanet alfa for the reversal of factor Xa inhibitor activity. N Engl J Med 2015; 373 (25) 2413-2424
• 17 Stone SR, Le Bonniec BF. Inhibitory mechanism of serpins. Identification of steps involving the active-site serine residue of the protease. J Mol Biol 1997; 265 (03) 344-362
• 18 Johnson DJ, Li W, Adams TE, Huntington JA. Antithrombin-S195A factor Xa-heparin structure reveals the allosteric mechanism of antithrombin activation. EMBO J 2006; 25 (09) 2029-2037
• 19 Connolly SJ, Milling Jr TJ, Eikelboom JW. , et al; ANNEXA-4 Investigators. Andexanet alfa for acute major bleeding associated with factor Xa inhibitors. N Engl J Med 2016; 375 (12) 1131-1141
• 20 Ansell JE, Bakhru SH, Laulicht BE. , et al. Use of PER977 to reverse the anticoagulant effect of edoxaban. N Engl J Med 2014; 371 (22) 2141-2142
• 21 Ansell JE, Bakhru SH, Laulicht BE. , et al. Single-dose ciraparantag safely and completely reverses anticoagulant effects of edoxaban. Thromb Haemost 2017; 117 (02) 238-245
• 22 Thalji NK, Ivanciu L, Davidson R, Gimotty PA, Krishnaswamy S, Camire RM. A rapid pro-hemostatic approach to overcome direct oral anticoagulants. Nat Med 2016; 22 (08) 924-932
• 23 Parsons-Rich D, Hua F, Li G, Kantaridis C, Pittman DD, Arkin S. Phase 1 dose-escalating study to evaluate the safety, pharmacokinetics, and pharmacodynamics of a recombinant factor Xa variant (FXa^I16L ). J Thromb Haemost 2017; 15 (05) 931-937
• 24 Verhoef D, Visscher KM, Vosmeer CR. , et al. Engineered factor Xa variants retain procoagulant activity independent of direct factor Xa inhibitors. Nat Commun 2017; 8 (01) 528
• 25 Sottrup-Jensen L. Alpha-macroglobulins: structure, shape, and mechanism of proteinase complex formation. J Biol Chem 1989; 264 (20) 11539-11542
• 26 Kolodziej SJ, Wagenknecht T, Strickland DK, Stoops JK. The three-dimensional structure of the human α 2-macroglobulin dimer reveals its structural organization in the tetrameric native and chymotrypsin α 2-macroglobulin complexes. J Biol Chem 2002; 277 (31) 28031-28037
• 27 Marrero A, Duquerroy S, Trapani S. , et al. The crystal structure of human α2-macroglobulin reveals a unique molecular cage. Angew Chem Int Ed Engl 2012; 51 (14) 3340-3344
• 28 Sottrup-Jensen L, Sand O, Kristensen L, Fey GH. The alpha-macroglobulin bait region. Sequence diversity and localization of cleavage sites for proteinases in five mammalian alpha-macroglobulins. J Biol Chem 1989; 264 (27) 15781-15789
• 29 Meijers JCM, Tijburg PNM, Bouma BN. Inhibition of human blood coagulation factor Xa by α 2-macroglobulin. Biochemistry 1987; 26 (18) 5932-5937
• 30 Heeb MJ, Gruber A, Griffin JH. Identification of divalent metal ion-dependent inhibition of activated protein C by α 2-macroglobulin and α 2-antiplasmin in blood and comparisons to inhibition of factor Xa, thrombin, and plasmin. J Biol Chem 1991; 266 (26) 17606-17612
• 31 Gettins PG, Boel E, Crews BC. Thiol ester role in correct folding and conformation of human alpha 2-macroglobulin. Properties of recombinant C949S variant. FEBS Lett 1994; 339 (03) 276-280
• 32 Qazi U, Gettins PG, Stoops JK. On the structural changes of native human alpha2-macroglobulin upon proteinase entrapment. Three-dimensional structure of the half-transformed molecule. J Biol Chem 1998; 273 (15) 8987-8993
• 33 Qazi U, Kolodziej SJ, Gettins PG, Stoops JK. The structure of the C949S mutant human alpha(2)-macroglobulin demonstrates the critical role of the internal thiol esters in its proteinase-entrapping structural transformation. J Struct Biol 2000; 131 (01) 19-26
• 34 Wyatt AR, Kumita JR, Farrawell NE, Dobson CM, Wilson MR. Alpha-2-macroglobulin is acutely sensitive to freezing and lyophilization: implications for structural and functional studies. PLoS One 2015; 10 (06) e0130036
• 35 Sottrup-Jensen L, Hansen HF, Pedersen HS, Kristensen L. Localization of epsilon-lysyl-gamma-glutamyl cross-links in five human alpha 2-macroglobulin-proteinase complexes. Nature of the high molecular weight cross-linked products. J Biol Chem 1990; 265 (29) 17727-17737
• 36 Le Bonniec BF, Guinto ER, Esmon CT. The role of calcium ions in factor X activation by thrombin E192Q. J Biol Chem 1992; 267 (10) 6970-6976
• 37 Arnold JN, Wallis R, Willis AC. , et al. Interaction of mannan binding lectin with alpha2 macroglobulin via exposed oligomannose glycans: a conserved feature of the thiol ester protein family?. J Biol Chem 2006; 281 (11) 6955-6963
• 38 French K, Yerbury JJ, Wilson MR. Protease activation of α2-macroglobulin modulates a chaperone-like action with broad specificity. Biochemistry 2008; 47 (04) 1176-1185
• 39 Jourdi G, Siguret V, Martin AC. , et al. Association rate constants rationalise the pharmacodynamics of apixaban and rivaroxaban. Thromb Haemost 2015; 114 (01) 78-86
• 40 Bianchini EP, Pike RN, Le Bonniec BF. The elusive role of the potential factor X cation-binding exosite-1 in substrate and inhibitor interactions. J Biol Chem 2004; 279 (05) 3671-3679
• 41 Hantgan RR, Hermans J. Assembly of fibrin. A light scattering study. J Biol Chem 1979; 254 (22) 11272-11281
• 42 Johansen PB, Tranholm M, Haaning J, Knudsen T. Development of a tail vein transection bleeding model in fully anaesthetized haemophilia A mice - characterization of two novel FVIII molecules. Haemophilia 2016; 22 (04) 625-631
• 43 Fuchs HE, Pizzo SV. Regulation of factor Xa in vitro in human and mouse plasma and in vivo in mouse. Role of the endothelium and plasma proteinase inhibitors. J Clin Invest 1983; 72 (06) 2041-2049
• 44 Bergsma J, Vije J, Duursma AM, Schutter WG, Bouma JM, Gruber M. The alpha-macroglobulins from rat plasma: structure, plasma clearance and endocytosis of complexes with subtilisin. Biomed Biochim Acta 1986; 45 (11-12): 1549-1556
• 45 Debanne MT, Bell R, Dolovich J. Uptake of proteinase-alpha-macroglobulin complexes by macrophages. Biochim Biophys Acta 1975; 411 (02) 295-304
• 46 Imber MJ, Pizzo SV. Clearance and binding of two electrophoretic “fast” forms of human alpha 2-macroglobulin. J Biol Chem 1981; 256 (15) 8134-8139
• 47 Gliemann J, Larsen TR, Sottrup-Jensen L. Cell association and degradation of alpha 2-macroglobulin-trypsin complexes in hepatocytes and adipocytes. Biochim Biophys Acta 1983; 756 (02) 230-237
• 48 Eddeland A, Ohlsson K. The elimination in dogs of trypsin-alpha-macroglobulin complexes inactivated by the Kazal or the Kunitz inhibitor. Hoppe Seylers Z Physiol Chem 1978; 359 (03) 379-384
• 49 Narita M, Rudolph AE, Miletich JP, Schwartz AL. The low-density lipoprotein receptor-related protein (LRP) mediates clearance of coagulation factor Xa in vivo. Blood 1998; 91 (02) 555-560
• 50 Herrmann R, Thom J, Wood A, Phillips M, Muhammad S, Baker R. Thrombin generation using the calibrated automated thrombinoscope to assess reversibility of dabigatran and rivaroxaban. Thromb Haemost 2014; 111 (05) 989-995