Tapping into the natural PZ-independent anticoagulant function of ZPI to inhibit thrombosis with minimal effect on hemostasis

Abstract

Background

The protein Z-dependent protease inhibitor (ZPI) binds to protein Z (PZ), which enables ZPI to inhibit membrane-bound activated factor X (FXa). ZPI also inhibits activated factor XI (FXIa) independently of PZ.

Methods

To study the PZ-independent ZPI function, we tested the in vitro and in vivo effect of disrupting the ZPI-PZ interaction by mutating ZPI Asp 293 to Ala (D293A).

Results

D293A mutation reduced PZ-dependent FXa inhibition without affecting FXIa inhibition. D293A also diminished FXIIa-induced thrombin generation but reduced tissue factor (TF)-induced thrombin generation only at low TF concentrations. This suggests that D293A selectively inhibits the intrinsic pathway and the thrombin-factor XI (FXI) feedback loop that enhances low-dose TF-initiated coagulation. WT and D293A ZPI both showed selectivity in inhibiting activated partial thromboplastin time (APTT) but not prothrombin time (PT). Increasing PZ in plasma enhances APTT inhibition and enables PT inhibition by WT but not D293A ZPI, further indicating that D293A ZPI selectively inhibits the intrinsic pathway independently of PZ. In mouse models, D293A inhibited FeCl₃-induced occlusive carotid artery thrombosis and venous thrombosis in the inferior vena cava (IVC). Thus, PZ-independent ZPI function plays a major role in ZPI inhibition of occlusive thrombosis, and D293A ZPI is an effective antithrombotic. Importantly, administering D293A ZPI did not affect tail bleeding time and showed improved hemostasis in a saphenous vein hemostasis model as compared with WT ZPI.

Conclusions

The PZ-binding defective variant of ZPI, D293A, selectively inhibits the intrinsic coagulation pathway and is a new anticoagulant with reduced bleeding risk.

Graphical Abstract

Currently available anticoagulants have been proven effective in the prevention and treatment of thrombotic diseases¹. However, these anticoagulants, including direct oral anticoagulants (activated factor X (FXa) and direct thrombin inhibitors), are associated with serious side effects of bleeding². In recent years, the key intrinsic pathway component, factor XI (FXI) or activated factor XI (FXIa), has emerged as a promising antithrombotic target with reduced bleeding risks³. However, bleeding was still reported in patients given FXI or FXIa inhibitors⁴. Additionally, there are recent reports of FXIIa-induced activation of FIX and intrinsic pathway via kallikrein, bypassing FXI^5,6. Here, we investigate whether a natural anticoagulant protein can be modified to effectively inhibit thrombosis with significantly reduced bleeding risk.

The protein Z-dependent protease inhibitor (ZPI) is a natural plasma anticoagulant serpin, which inhibits both FXa and FXIa. One report showed that ZPI may also inhibit activated factor IX (FIXa)⁷, although other studies failed to confirm that finding⁸. ZPI circulates in the plasma mostly in a tight complex (K_D ~1.0–10 nM) with its cofactor protein, Protein Z (PZ)^9,10, with ZPI in modest excess in human and PZ in excess in mice^11,12. PZ is a vitamin K-dependent protein, structurally homologous to factors II, VII, IX, and X, but lacking protease activity due to the absence of serine and histidine residues of the catalytic triad¹³. ZPI-PZ complex formation is required for the rapid inactivation of membrane-associated FXa by ZPI in the presence of Ca²⁺ and phospholipids, while its inhibition of FXIa is PZ-independent⁹. ZPI inhibits both FXa and FXIa by the same suicide-substrate mechanism, with a fast k_a (10⁴-10⁷M⁻¹s⁻¹) and slow k_diss (~0.0002 s⁻¹), typical of serpin-type protein inhibitor interactions with their coagulation proteases⁸. Previous studies have established that the ZPI/PZ complex inhibits membrane-bound FXa by first binding to the membrane through the PZ gamma-carboxy glutamic (Gla) domain to promote the irreversible formation of a stable ZPI-FXa covalent complex that inactivates FXa’s catalytic function. After the formation of the stable ZPI-FXa complex, PZ dissociates from the FXa-bound ZPI to complex with free ZPI, thus maintaining the ZPI/PZ level⁸. X-ray crystal structures of the ZPI/PZ complex have suggested how ZPI and PZ interact^14,15, with the key interacting sites being further verified by mutagenesis studies^10,16.

The ZPI/PZ complex is a physiological anticoagulant¹⁷. Its importance in vivo was demonstrated by the finding that combined deficiencies of ZPI or PZ with factor V Leiden (FVL, an R506Q FV mutant that is resistant to cleavage by activated protein C) result in a severe to mortal thrombosis phenotype in mice as well as in humans^18–22. Interestingly, ZPI deficiency produces a more severe prothrombotic phenotype than PZ deficiency in mice¹⁹, suggesting the PZ-independent anticoagulant function of ZPI is also physiologically relevant. Indeed, ZPI is a potent natural FXIa inhibitor with the fastest k_a among known FXIa inhibitors in human plasma²³. As ZPI inhibits FXIa independently of PZ, we reason that a ZPI mutant that has a reduced affinity for PZ should selectively demonstrate a PZ-independent anticoagulant effect, particularly inhibition of the FXIa-dependent intrinsic coagulation pathway. Previous crystallographic and mutagenesis studies have characterized the interaction between ZPI and its binding partner PZ^14,15. Whereas the interface of ZPI and PZ comprises a broad area, including electrostatic and hydrophobic interactions, a critical interaction was suggested to be between D293 of ZPI and H210 and R298 of PZ (Fig.1A)^10,14,15. A single mutation of Asp 293 to Ala increased the K_D of the ZPI/PZ complex from 1~10 nM to 1~4 μM in physiological buffer and appeared to diminish the PZ-dependent inhibition of FXa at relatively low levels of ZPI and PZ (~30–50 nM)^10,16.

Figure 1. Asp 293 to Ala mutation in ZPI selectively abolished PZ-dependent inhibition of FXa without affecting PZ-independent inhibition of FXIa.

A, the crystal structure of the ZPI (Green)-PZ (purple) complex (PDB:3H5C). The contact residues of D293 (red) of ZPI and, H210 (cyan) and R298 (blue) of PZ are shown in the sphere. P1 residue of ZPI (Y387) is shown in a sphere of dark green. B, Inhibition of FXa (0.5 nM) activity by WT (■), D293A (●), or D293A/Y387D (▲) ZPIs in the absence of FVa shown as ZPI-PZ concentration dependence of observed pseudo-first-order rate constants for FXa inhibition. C, Inhibition of prothrombinase (FXa/FVa) activity by WT (■), D293A (●), or D293A/Y387D (▲) ZPIs in the presence of FVa (16 nM) shown as ZPI-PZ concentration dependence of observed pseudo-first-order rate constants for FXa (0.5 nM) inhibition. D, Time-dependent prothrombin activation (1.4μM) by prothrombinase assembled with FVa (16nM), FXa (~0.06nM), lipid (25 μM), and Ca²⁺(5 mM) in the absence (●) or presence of 60 nM WT (○) or D293A (△) ZPI and 60 nM PZ in the reaction buffer (50mM Tris buffer, pH 7.4, 0.1M NaCl, 1mg/mL BSA) at 25°C. Reactions were diluted and quenched at various time points with 10 mM EDTA in the presence of a chromogenic thrombin substrate in the reaction buffer, and thrombin activity was subsequently determined as described in “Methods.” Solid lines are empirical fits of data. E, Observed pseudo-first-order rate constants for WT (■, ▲), D293A (□, Δ) or D293A/Y387D (○, ◇) ZPI-mediated inhibition of FXIa (1nM) (□, ■, ○) and FXa, (0.5nM) (Δ, ▲, ◇), respectively. The assays in B and C were performed in the reaction buffer containing 2.5mM CaCl₂ at 25°C, in the presence of equimolar PZ with ZPI, 25 μM lipids. The assay in E was performed in the reaction buffer containing 2.5 mM CaCl₂ without PZ and lipids. At various time points, the reactions were diluted and quenched with 5 mM EDTA in the presence of an FXa or FXIa substrate, and FXa or FXIa activity was determined as described in “Methods.” Data represents the average of 3 (B to E) independent measurements, presented as mean ±SD. The SD for some points is too small to be visible. Solid lines are linear fits of data (B, C, E) as described in “Methods”.

In this study, we explored the anticoagulant function of the D293A ZPI mutant in vitro and in vivo in comparison with wild type (WT) ZPI. We found that D293A ZPI selectively inhibits the intrinsic coagulation pathway and the ability of the intrinsic pathway to augment thrombin generation induced by low dose tissue factor (TF), one of the mechanisms by which the intrinsic pathway participates in thrombosis. Furthermore, we show that the D293A ZPI mutant is effective in inhibiting FeCl₃-induced arterial and venous thrombosis in vivo. Importantly, D293A ZPI has minimal effect on tail bleeding time. In the saphenous vein hemostasis analysis, which is known to be impacted by FXI deficiency, D293A ZPI, at the concentrations that effectively inhibited thrombosis, showed minimal effect on hemostasis in contrast to the same concentration of WT ZPI, and only showed impact at a much higher concentration, demonstrating safety at the therapeutic concentration, Thus, PZ-independent inhibition of FXIa by D293A ZPI provides a novel and potentially safer approach to developing new antithrombotic.

MATERIALS AND METHODS

Availability of Data

The original data that supports the findings of this study are available from the corresponding author upon reasonable request.

Animals

The Institutional Animal Care Committee of the University of Illinois at Chicago approved all animal studies. C57BL/6J mice were obtained from Jackson Laboratory (Bar Harbor, ME) and were maintained in the university-designated facility.

Proteins

Recombinant human Wild type (WT) and D293A ZPI were expressed in non-mammalian cell systems and purified to homogeneity as previously reported^8,14,24,25. These recombinant ZPIs (E. coli and insect cell) are identical in functions^8,25,26, were used for in vitro and in vivo assays, respectively. Human plasma PZ, prothrombin, FXa, FXIa, and corn trypsin inhibitor (CTI) were purchased from Enzyme Research Laboratories (South Bend, IN). FVa was from Prolytix (Essex Junction, VT). All proteins were judged >95% pure by SDS-PAGE analysis. Molar concentrations of recombinant ZPI and plasma PZ were determined from the absorbance at 280nm using absorption coefficients calculated from the amino acid sequence¹⁰.

Phospholipids

Small, unilamellar phospholipids vesicles (SUV) were prepared by sonication on ice under nitrogen for 60 min from a 7:3 mixture (by weight) of dioleyl phosphatidylcholine and dioleoyl phosphatidylserine (Avanti Polar Lipids) as described previously¹⁷.

Kinetics of WT or D293A inhibition of FXa and FXIa

Assays testing the inhibitory effects of ZPI on FXIa, membrane-associated free FXa and prothrombinase-bound FXa were performed as previously described^8,10,25. In these experiments, purified FXa or FXIa were respectively incubated with various concentrations of ZPI with or without equal molar PZ for increasing lengths of time. For every ZPI concentration, at various time points, the FXa-ZPI/PZ or FXIa-ZPI mixtures were quickly diluted into the solution containing the respective fluorescence substrates for FXa or FXIa to determine the residual FXa or FXIa activity by measuring the initial rate of substrate hydrolysis fluorometrically. The time-dependent inhibition progress curve was generated with the control as zero time point (no ZPI thus full FXa or FXIa activities) and followed by FXa or FXIa residual activities at a series of time points of FXa/FXIa-ZPI incubation as described above. The FXa (or FXIa) activity vs. time progress curve was fitted using Prism software by a single exponential decay function with a nonzero end point to obtain k_obs for each progress curve at each ZPI concentration (k_obs = the fitted rate exponential decay constant)⁸. The progress curve has a nonzero end point due to the presence of small amounts (<5%) of partially degraded proteases that were resistant to inhibition by ZPI. After testing several different ZPI concentrations, a series of k_obs (each for one ZPI concentration) was obtained. The Apparent second-order associate rate constant (k_a) was obtained from the slopes of linear plots of k_obs versus the ZPI concentration according to the equation, k_obs=k_diss+k_a × [ZPI]_o. In this equation, k_diss represents the intrinsic dissociation rate constant for deacylation of the ZPI-factor Xa/FXIa acyl-intermediate complex, whereas [ZPI]_o is the total ZPI concertation. The k_diss was fixed at the experimentally determined value of 2 × 10⁻⁴s⁻¹ for WT and D293A as described^8,10,25. The k_a thus generated for WT or D293A were compared using “Student’s t-tests to obtain p value. The ratio of mutant k_a/WT k_a was calculated to indicate the fold change. Additional details are provided in the supplemental Materials.

Thrombin generation assay

Certified pooled normal human plasma (PNP, from 30 or more screened normal human donors) and pooled C57BL/6 plasma are from George King (Kansas) and Innovative Research (MI), respectively. Five different lots of certified FXI deficient plasma (with FXI activity<1%, George King) with equal volume were pooled to be used as a control. The experiment was conducted as previously described¹⁷. Additional details are provided in the supplemental Materials.

Clotting time

Activated partial thromboplastin time (APTT reagent of Dade Actin-FSL) and Prothrombin time (PT reagent of Dade Innovin) were performed on PNP, and pooled C57BL/6 mouse (Innovative Research) plasma supplemented with inhibitors or control vehicles (<15% in volume) with an Amelung KC1 analyzer according to the manufacturer’s instructions.

Determining ZPI and PZ antigen levels in plasma

ZPI and PZ antigen levels in plasma were determined using ELISA kits (DY811505 of R&D Systems for ZPI; NBP2–60592 of Novus Biologicals for PZ) according to the manufacturer’s instructions.

FeCl₃-induced mouse carotid artery thrombosis model

The carotid artery ferric chloride model performed on C57BL/6 mice anesthetized with 2–3% isoflurane inhalation was conducted using 7.5% FeCl₃ for 3 min to induce thrombosis in 15 min, as previously described²⁷.The C57BL/6 mice were pre-injected with control vehicle(n=11, male 6, female 5), WT(n=9, male 5, female 4), D293A(n=8, male 4, female 4), and D293A/Y387A (n=7, male 4, female 3) ZPIs (7.5mg/kg) via retro-orbital veins. Additional details are provided in the supplemental Materials.

FeCl₃-induced venous thrombosis

The experiment was conducted on C57BL/6 mice anesthetized with ketamine(100mg/kg)/xylazine(5mg/kg) using 5% FeCl₃ for 3 min to induce thrombosis in inferior vena cava (IVC) in 20 min, following previously described procedures²⁸. The C57BL/6 mice were pre-injected with control vehicle(n=9, male 5, female 4), WT (n=8, male 5, female 3, for 7.5mg/kg; n=8, male 4, female 4, for 15mg/kg) and D293A ZPIs (n=8, male 5, female 3, for 7.5mg/kg; n=8, male 4, female 4 for 15mg/kg) via retro-orbital veins. Additional details are provided in the supplemental Materials.

Tail bleeding assay

The tail bleeding assay on C57BL/6 mice was conducted as previously described²⁷. In brief, mice were anesthetized with ketamine(100mg/kg)/xylazine(5mg/kg). After cutting a 0.5 cm-long segment off the distal tip of the tail, the tail bleeding time was observed for up to 15 min. The C57BL/6 mice were pre-injected with control vehicle(n=9, male 6, female 3), WT (n=12, 7.5mg/kg, male 7, female 5), D293A (n=10, male 5, female 5, for 7.5mg/kg; n=8, male 5, female 3, for 15mg/kg), and D293A/Y387A (7.5mg/kg, n=9, male 4, female 5) ZPIs via retro-orbital veins. Additional details are provided in the supplemental Materials.

Saphenous vein hemostasis model

The saphenous vein hemostasis assay on C57BL/6 mice was conducted as described previously^29,30 and in the Supplementary Materials in details. Briefly, a small cut was made on the wall of the surgically exposed saphenous vein, which was observed for hemostatic clot formation. Immediately after bleeding stopped at the cut (which was counted as 1 time of hemostasis), the clot was disrupted by stroking a blunted 30-G needle end in the direction of blood flow to re-initiate a new bleeding episode. The process was then repeatedly performed for the duration of 20 min. The total number of times of hemostasis (reformation of hemostatic clot) was recorded, which reflects how fast hemostasis occurred. The C57BL/6 mice were pre-injected with control vehicle(n=10, male 6, female 4), WT (n=11, male 5, female 6, 7.5mg/kg), D293A (n=10, male 5, female 5, for 7.5mg/kg; n=10, male 4, female 6, for 15mg/kg), and D293A/Y387A (7.5mg/kg, n=8, male 5, female 3) ZPIs via retro-orbital veins.

Statistical analysis

Student’s t-tests were used to compare the differences between two treatments or two treatment time points for vitro assays. D’Agostino-Pearson normality test was performed before the significance tests. In animal models, differences between buffer, WT, and mutant ZPIs were tested by one-way analysis of variance (ANOVA) (with Bonferroni correction for multiple testing). GraphPad Prism version 6.0 for Windows was used to perform data analysis. A value of p < 0.05 was considered statistically significant.

Results

Asp 293 to Ala mutation in ZPI selectively abolishes PZ-dependent inhibition of FXa without affecting PZ-independent inhibition of FXIa

To determine whether D293A mutation could diminish PZ-dependent ZPI inhibition of FXa and prothrombinase at high D293A ZPI and PZ concentrations relevant to pharmacological application, we compared FXa inhibitory effects of WT and D293A ZPI together with equimolar PZ up to 300 nM. As shown in Fig.1B and Table 1, D293A mutation caused a 20-fold decrease in k_a of PZ-dependent inhibition of FXa (not complexed with FVa) as compared with WT ZPI (85.0±5.7 ×10³M⁻¹s⁻¹ (D293A) vs. 1787.0±197.3×10³M⁻¹s^−1, (WT), n=3, p<0.001), although there were residual inhibitory effects, as compared with the D293A/Y387D double mutation, which shows no detectable inhibition. Under the physiological condition when FXa is bound to membrane-associated FVa to form prothrombinase, the k_a of prothrombinase inhibition by D293A mutant ZPI was further reduced to ~50-fold slower than that of WT ZPI (6.8±0.5×10³M⁻¹s⁻¹ vs. 296.4±24.9 ×10³M⁻¹s⁻¹, Fig.1C and Table 1, n=3, p<0.001), which already had a ~6.0-fold decrease in k_a compared with that in the absence of FVa (296.4±24.9 ×10³M⁻¹s⁻¹ vs. 1787.0±197.3×10³M⁻¹s⁻1, Fig. 1B & 1C and Table 1, n=3, p<0.001). Consistently, the D293A mutation diminished the PZ-dependent inhibitory effect on the prothrombinase function of the FVa/FXa complex (Fig. 1D), a key component of the common pathway of blood coagulation. In the absence of PZ and lipids, however, both WT and D293A ZPI similarly showed only residual FXa inhibitory effects (k_a ~1/190 of PZ-dependent inhibition by WT) (9.2±0.4 ×10³M⁻¹s⁻¹ and 9.5±0.4×10³M⁻¹s⁻¹ for WT and D293A FXa inhibition in Fig.1E, respectively, vs 1787.0±197.3×10³M⁻¹s⁻1 of WT FXa inhibition in Fig.1B in the absence of FVa, Table 1, n=3, p<0.001), confirming that the FXa inhibitory effect of ZPI is predominantly PZ-dependent. Importantly, D293A and WT ZPI similarly inhibited FXIa in the absence of PZ and lipids (112.1±5×10³M⁻¹s⁻¹ and 86.2±3.6×10³M⁻¹s⁻¹, respectively, for FXIa inhibition, Fig.1E and Table 1), indicating that the function of ZPI to inhibit FXIa is not negatively affected by the D293A mutation. Thus, D293A ZPI is a selective inhibitor of FXIa with minimal effect on FXa.

Table 1. Second-order association rate constants for WT or D293A ZPI inhibition of FXa and FXIa.

Second-order association rate constants (k_a) determined in Fig.1 were measured in 50mM Tris Buffer pH 7.4, 0.1M NaCl, 1mg/mL BSA, 2.5mM CaCl₂ at 25°C, and in the presence or absence of equimolar PZ, 25 μM lipids, and 16nM FVa as described under “Methods.” The results represented three independent measurements, presented with mean±SD (n=3). The student’s t-test was used to compare the differences between selected two groups. The p values were shown in the results. The normalized values vs. WT (as 1.0) were shown in brackets.

	ka (×103M−1s−1)
	+PZ-FVa: FXa	+PZ+FVa: FXa	-PZ: FXa	-PZ:FXIa
WT	1787.0 ± 197.3 (1.0)	296.4 ± 24.9 (1.0)	9.2± 0.4 (1.0)	86.2 ±3.6 (1.0)
D293A	85.0 ± 5.7 (0.05)	6.8 ± 0.5 (0.02)	9.5± 0.4 (1.0)	112.1±5.0(1.3)
D293A/Y387D	-	-		2.6 ±0.2 (0.03)

Inhibitory effect of WT and D293A ZPI on thrombin generation induced by activated factor XII (FXIIa) in the Calibrated Automated Thrombogram (CAT) assay.

Coagulation can be activated via the FXIIa- and FXIa-dependent intrinsic pathway and the TF-initiated extrinsic pathway, both of which activate FX and the common prothrombinase pathway of thrombin generation. We used the CAT assay to quantify the effects of WT and D293A ZPI on thrombin generation. Human FXIIa was used to initiate the intrinsic pathway of thrombin generation in both human and C57BL/6 mouse plasma, either with a high dose (33 nM for humans, 6.5 nM for mice) or a low dose (6.5 nM for humans, 1.5 nM for mice) FXIIa. In PNP, both WT and D293A ZPI caused a potent dose-dependent inhibiton of thrombin generation induced by FXIIa. In contrast, Inactive ZPI mutant D293A/Y387A had no detectable effect on thrombin generation (Fig.S1). WT ZPI caused a more potent inhibitory effect than D293A on lag time (LT), thrombin peak (TP), and endogenous prothrombin potential (ETP) at lower concentrations of ZPI (0.2–0.7 μM, Fig. 2A: a–d and Fig. S2: A–F) when both WT and D293A only partially inhibited FXIIa-induced thrombin generation. However, when the concentration was increased incrementally from 0.7 to 1 μM, WT and D293A ZPI similarly achieved complete inhibition (>40 min for LT or 0 for TP and ETP) when the assay was induced by either low or high concentrations of FXIIa. Considering that D293A differs from WT only in its deficiency in PZ-dependent FXa inhibition and is similar to WT in inhibiting FXIa, these data suggest that the deficient FXa inhibitory function of D293A manifests in the CAT assay only when the intrinsic pathway is not completely abolished and that similar complete inhibition of thrombin generation at the same high dose likely reflects their similar function in inhibiting the intrinsic pathway. Interestingly, in contrast to the complete inhibition of intrinsic pathway in PNP by high concentrations of WT or D293A ZPIs, FXI deficiency in FXI-deficient plasma (FXI-DP) showed only partial reduction in FXIIa-induced thrombin generation (Fig. 2A: e–f and Fig. S2: A–F). This partial inhibition of FXIIa-induced thrombin generation in FXI-DP may possibly be due to the presence of a low level of FXI-dependent activity in FXI-DP or/and due to the presence of FXI-independent activation of the intrinsic pathway by FXIIa. In this respect, there were reports that FXIIa-induced activation of FIX and thus intrinsic pathway via kallikrein^5,6, bypassing FXI. Whereas these possibilities require further investigation in the future, our data demonstrate that both WT and D293A ZPI can completely inhibit FXIIa-induced activation of intrinsic pathway of thrombin generation. In mouse plasma, dose-dependent inhibitory effects of WT and D293A ZPIs (ZPI concentrations ranged from 0– 6 μM) were also seen on high and low dose FXIIa-induced thrombin generation (Fig. 2B: a–d and Fig. S3), although they were much less potent than in normal human plasma, possibly due to species differences. Also, the inhibitory effect of WT ZPI in mouse plasma was more potent relative to D293A ZPI, possibly due to a higher PZ concentration in mouse plasma and the effect of endogenous PZ concentration in limiting the function of excess ZPI (see below).

Figure 2. The effects of WT and D293A ZPIs on thrombin generation in pooled human or C57BL/6 mouse plasma initiated by various concentrations of FXIIa.

Thrombin generation in pooled normal plasma human (PNP) and factor XI deficient patient plasma (FXI-DP) (A) or pooled C57BL/6 mouse (B) plasma activated by various doses of FXIIa was measured by Calibrated Automated Thrombogram (CAT) at pH 7.4, 37°C, as described under “Methods.” To 60 μL mixture containing 40 μL Plasma, fluorogenic substrate, and lipid was added 20μL of WT or D293A ZPI as indicated. Thrombin generation was initiated by adding 20 μL of pre-warmed CaCl₂ and FXIIa. The final concentration was 420 μM fluorogenic substrate, 25 μM lipids, 20 mM CaCl₂, and indicated concentrations of WT or D293A ZPI and FXIIa. Thrombinoscope^™ software was used to analyze fluorescence data to obtain the lag time (LT), thrombin peak (TP), time to peak (TTP), and endogenous thrombin potential (ETP) (summarized and presented in Fig.S2 and S3, for human and mouse plasma, respectively). The data represented from three independent measurements is the average of triplicate measurements. Student’s t-test was used to compare the differences between selected two groups.

Inhibitory effect of WT and D293A ZPI on thrombin generation induced by tissue factor (TF) in the CAT assay.

Previous studies showed that FXI and, consequently, the intrinsic pathway can be activated by thrombin^31,32. This thrombin-mediated feedback activation loop is crucial in amplifying thrombin generation via the extrinsic-common pathway induced by low initial concentrations of TF, FXa, and thrombin. We analyzed the effects of exogenously added WT and D293A ZPI on human and C57BL/6 mouse plasma thrombin generation activated by various doses of TF (in the presence of 30 μg/mL CTI to prevent FXII activation). At a high dose of TF (6 pM), PNP and FXI-DP had a similar profile of thrombin generation (Fig.3A: a and overlapped dotted and dashed lines in Fig. S4: A–C). However, at lower doses of TF (2 pM and 0.5 pM), FXI-DP exhibited significantly decreased thrombin generation as indicated by prolonged LT, and decreased TP and ETP, consistent with the role of FXI in promoting low-dose TF-initiated coagulation (Fig. 3A: b–c and Fig.S4: D–I PNP vs. FXI-DP). Exogenous WT and D293A ZPIs both had minimal to modest impacts on LT and ETP in both human and mouse plasma at the high TF concentration, although WT ZPI still potently reduced TP (Fig. 3A: d, g & 3B: a, d and Fig. S4 & S5: A–C). At low TF concentrations (0.5–2 pM), WT ZPI more potently inhibited thrombin generation in both human and mouse plasma (Fig.3 A & B: e vs.h & f vs.i and Fig. S4 & S5: D–I) than D293A ZPI and FXI-DP (in human). However, D293A ZPI still dose-dependently inhibited thrombin generation initiated by lower doses of TF, reaching or exceeding FXI-DP at the high concentrations of D293A (Fig.3A: h & i and Fig.S4: D–I). Together, these data suggest that D293A ZPI, while having a reduced inhibitory effect compared with WT due to the loss of the PZ-dependent inhibitory effect on FXa, still effectively inhibits the amplification of low dose TF-induced thrombin generation similar to FXI-DP.

Figure 3. The effects of WT and D293A ZPIs on thrombin generation in pooled human or C57BL/6 mouse plasma initiated by various concentrations of TF.

Thrombin generation in FXI-DP and PNP (A) or pooled C57BL/6 (B) plasma activated by various doses of TF was measured by CAT at pH 7.4, 37°C, as described under “Methods.” To 60 μL mixture containing 40 μL plasma, fluorogenic substrate, CTI, and lipid was added 20μL of TF, WT, or D293A ZPI as indicated. Thrombin generation was initiated by adding 20 μL of pre-warmed CaCl₂. The final concentration was 420 μM fluorogenic substrate, 30 μg/mL CTI, 25 μM lipids, 20 mM CaCl₂, and indicated concentrations of WT or D293A ZPI and TF. The dotted lines in the second and third rows of (A) reproduce the corresponding data of FXI-DP in the first row. Thrombinoscope^™ software was used to analyze fluorescence data to obtain the lag time (LT), thrombin peak (TP), time to peak (TTP), and endogenous thrombin potential (ETP) (summarized and presented in Fig.S4 and S5, for human and mouse plasma, respectively). The data represented from three independent measurements is the average of triplicate measurements. Student’s t-test was used to compare the differences between selected two groups.

The effect of WT and D293A ZPI on APTT and PT in human plasma compared with FXIa inhibitor asundexian and BMS-262084.

To compare the PZ-dependent and independent inhibitory function of D293A and WT ZPI on coagulation triggered through intrinsic and extrinsic pathways, we performed standard APTT and PT assays. The PNP used in the experiments had an APTT and PT of ~30 s and ~10 s, respectively. Adding WT or D293A ZPI to plasma increased APTT dose-dependently and to a similar extent for every concentration employed (0 – 4 μM) (Fig.4A). At 4 μM WT or D293A ZPI added, both APTTs were increased by 2.5-fold as compared with control plasma. In contrast, neither WT nor D293A ZPI affected PT (Fig.4A). Thus, both WT and D293A ZPI potently inhibit the intrinsic/common coagulation pathway as indicated by APTT but not the extrinsic/common coagulation pathway induced by very high concentrations of TF as indicated by PT.

Figure 4. The anticoagulant effects of WT or D293A ZPIs as determined by Activated partial thromboplastin time (APTT) or Prothrombin time (PT).

APTT and PT analyses were performed as described in “methods”. Data represents the average of 3–5 independent measurements, presented as mean±SD. Student’s t-test was used to compare the differences between selected two groups. A, Dose-dependent effects of WT, D293A, D293A/Y387D ZPIs, BMS-262084, and asundexian (0–4 μM) on APTT and PT of PNP. ***p<0.001, WT or D293A or BMS-262084 vs. asundexian of the same concentrations. B, Dose-dependent effects of BMS-262084 and asundexian (0–60 μM) on APTT and PT of PNP. *-**p values were shown as indicated, ***p<0.001, the APTT of BMS-262084 vs. Asundexian of the same concetrations; ^ΔΔΔp<0.001, the PT of BMS-262084 or asundexian vs. no BMS-262084 or asundexian added. C, Dose-dependent effects of WT, D293A, D293A/Y387D ZPIs (0–4 μM) on APTT and PT of pooled C57BL/6 mouse plasma. **p=0.0013; *** p<0.001, the APTT of WT ZPI vs. D293A of the same concentrations. D, Effects of WT or D293A ZPI (2 μM) in combination with varying concentrations of PZ (0–0.9 μM) on APTT of PNP in comparison with the effect of PZ alone (0.2–0.9 μM). ***p<0.001, WT+PZ vs. D293A+PZ. E, Comparative effects of WT ZPI (2 μM) and D293A ZPI (2 μM) in the absence and presence of PZ (1.4 μM) on PT of PNP. PZ (1.4 μM) alone was also tested as a control. ***p<0.001, WT+PZ vs. D293A+PZ. F, Effects of WT ZPI (3 μM) and D293A ZPI (3 μM) with or without PZ (2.6 μM) on PT of C57BL/6 mouse plasma. PZ (2.6 μM) alone was tested as a control. **p=0.0023, WT+PZ vs. D293A+PZ.

Next, we compared the effects of WT and D293A ZPI on APTT and PT with that of asundexian (MedChemExpress, Monmouth Junction, NJ), a small molecule inhibitor of human FXIa currently in clinical trials³³, and BMS-262084 (MedChemExpress), a known potent FXIa inhibitor³⁴. Interestingly, the dose response curve of asundexian’s effect on APTT was not saturated even at 60 μM, and was relatively weaker than BMS-262084 (p < 0.01–0.05, Fig.4B). The dose response curve for asundexian and BMS-262084 on PT, however, had an initial lack of effect (up to 4 μM) followed by a moderate but significant increase in inhibitory effect at higher concentrations (Fig.4B), consistent with previous reports showing that although asundexian and BMS-262084 potently inhibit FXIa, they may also have additional targets^33,35. The inhibitory effects of WT and D293A ZPI were compared with asundexian and BMS-262084 at concentrations that only prolonged APTT but not PT. Both D293A and WT ZPI showed similar inhibitory effects to that of BMS-262084 but were significantly more potent than that achieved by the same concentrations of asundexian (Fig. 4A, p<0.001). These data demonstrate that ZPI (WT or D293A) is likely to be superior in selective inhibition of intrinsic pathway-mediated coagulation as compared with asundexian, but similar to BMS-262084.

Effects of WT and D293A on APTT and PT in mouse plasma and the role of PZ in regulating the pharmacological effect of ZPI.

We also tested the effect of exogenously added WT and D293A ZPIs on APTT in pooled C57BL/6 mouse plasma. WT or D293A ZPI increased the APTT of mouse plasma dose-dependently. Different from human plasma APTT, however, WT ZPI exhibited a stronger inhibitory effect than D293A ZPI at all ZPI concentrations (Fig.4C), suggesting the relative importance of PZ-dependent inhibitory effects of ZPI on mouse APTT. This is consistent with the results of thrombin generation assays above (FXIIa triggered CAT, Fig. 2B and Fig.S3). C57BL/6 mouse plasma has a higher concentration of endogenous PZ (~240 nM) than the PNP we used (~ 37 nM, similar to previous reports^9,36). As the PZ-dependent inhibitory function of ZPI on FXa is abolished in D293A ZPI, we hypothesized that the more potent effect of WT vs. D293A ZPI on APTT in mice may be due to the presence of higher PZ concentrations in mouse plasma. Indeed, adding exogenous PZ (0.2–0.9 μM) together with exogenous ZPI to human plasma, but not adding PZ alone, significantly enhanced the effect of exogenous WT ZPI in prolonging APTT without affecting the inhibitory effect of D293A ZPI (Fig. 4D). These data suggest that the enhanced effects of exogenously added WT PZ in prolonging APTT relative to the effect of D293A ZPI is PZ-dependent and is limited by the relatively low PZ concentrations in plasma. This supports the notion that PZ-dependent ZPI function is diminished in D293A ZPI and that the anticoagulant effect of D293A is mostly PZ-independent. These results also indicate that due to the limiting effect of endogenous PZ on the WT ZPI-mediated inhibition of the FXa-dependent common pathway, the WT ZPI-mediated inhibitory effect on clotting time (especially in human plasma) mostly resulted from the PZ-independent inhibitory effect of ZPI on the intrinsic pathway, as displayed by the effect of D293A. Interestingly, exogenous WT or D293A ZPI (0–4 μM), in the absence of exogenous PZ, minimally affected PT in C57BL/6 mouse plasma (Fig.4C), which is similar to human plasma (Fig.4A). Adding PZ along with ZPI selectively induced prolongation of the PT only in WT ZPI-treated human and mouse plasma, but not D293A-treated human and mouse plasma (Fig. 4E & F). Thus, the inability of even WT ZPI to prolong PT is not only due to the large amount of TF in the PT assay (that likely overwhelms the inhibitory effects of ZPI by activating FXa via the extrinsic pathway) but also due to the limiting effect of low endogenous PZ concentrations on the PZ-dependent function of exogenous WT ZPI to inhibit Xa. These results are consistent with the results obtained from the CAT assays, indicating the minimally changed LTs of WT and D293A in high dose TF triggered CAT assays (Fig.3A: d, g & 3B: a, d; and Fig. S4 & S5: A), although thrombin generation is still more significantly inhibited by WT than D293A ZPI (TPs in the same Fig.3A: d,g, Fig. 3B: a,d and Fig. S4 & S5: B).

Comparison of antithrombotic effects of WT and D293A ZPIs using the mouse carotid artery thrombosis model

Clinical anticoagulation therapies usually have an APTT target of 1.5–2.5 folds of normal³⁷, equivalent to that achieved by 1~2 μM of ZPIs. The pilot studies showed that retro-orbital injection of 7.5 mg/kg of ZPIs gave the plasma concentrations of ~2.0 μM in C57BL/6 mice at 3–4 min, and the concentration went below 1.0 μM after ~20 min. Thus, 7.5 mg/ml i.v. injection was selected as the concentration to be used in our in vivo studies.

To compare the ability of WT and D293A ZPI to inhibit arterial thrombosis in vivo, the FeCl₃-induced carotid artery thrombosis model was used. After induction of injury to the carotid artery with 7.5% FeCl₃ for 3 minutes, C57BL/6 mice formed stable occlusive thrombi with a median time of 330 s. The occlusion time was significantly delayed (~525 s) in mice injected with a one-time bolus of WT ZPI (7.5 mg/kg retro-orbitally, p<0.001, WT vs. buffer, Fig.5). Interestingly, despite the absence of FXa inhibitory function, the same dose of D293A ZPI had a potent inhibitory effect (~480 s, p=0.0043, D293A vs. buffer, Fig.5), only slightly lower than that of WT ZPI, though no significant difference was detected. In contrast, the loss of function mutant ZPI, D293A/Y387A, at the same dose, did not affect the occlusion time.

Figure 5. WT and D293A ZPI inhibit FeCl₃-induced mouse carotid artery thrombosis.

Carotid artery thrombosis was induced using 7.5% FeCl₃ in mice retro-orbitally injected with the same volume of control buffer (5 mM sodium phosphate pH 6.5 containing 0.16M NaCl, 1mg/ml mouse albumin, n=11, male 6, female 5), WT (n=9, male 5, female 4), D293A (n=8, male 4, female 4), and D293A/Y387D (n=7, male 4, female 3) ZPIs (7.5mg/Kg). Time to stable occlusion was then monitored as described in the “Methods”. p values were shown as indicated. D’Agostino-Pearson normality test was performed before one-way ANOVA testing with Bonferroni correction for multiple testing. The median of each is shown.

Effect of D293A on venous thrombosis

We used the FeCl₃-induced IVC injury model to evaluate the effect of D293A on venous thrombosis in vivo. In this experiment, D293A and WT administered at the same dose that inhibited carotid artery thrombosis (7.5 mg/kg) decreased venous thrombus size significantly (p < 0.001, Fig.6, 7.5mg/kg D293A or WT vs. buffer). A double dose of D293A and WT (15 mg/kg) further decreased thrombus size (D293A 7.5mg/kg vs. 15mg/kg, p < 0.001; WT 7.5mg/kg vs.15mg/kg, p=0.0017, Fig.6). At the same concentration, the inhibitory effect of WT ZPI showed no significant differences compared with D293A, although showing a trend of enhanced inhibition (p=0.082, Fig.6). Thus, D293A ZPI and WT ZPIs inhibit both arterial and venous thrombosis in vivo.

Figure 6. Comparison of effects of ZPIs inhibiting inferior vena cava (IVC) thrombosis induced by 5% FeCl₃.

Comparison thrombus weights of mice treated with buffer (5mM sodium phosphate pH 6.5 containing 0.16M NaCl, 1mg/ml mouse albumin, n=9, male 5, female 4), 7.5 mg/kg WT (n=8, male 5, female 3) and D293A ZPI (n=8, male 5, female 3) and 15 mg/kg WT (n=8, male 4, female 4) and D293A ZPI (n=8, male 4 female 4). Buffer or ZPIs were injected retro-orbitally right before FeCl₃ application. Thrombus was dissected from IVC and weighted as described under “Methods.” p values were shown as indicated. D’Agostino-Pearson normality test was performed before one-way ANOVA testing with Bonferroni correction for multiple testing. The median of each is shown.

Effect of WT and D293A ZPI on mouse tail bleeding time and vena saphenous hemostasis model

We used mouse tail bleeding time analysis and a saphenous vein hemostasis model to determine the effect of WT and D293A ZPI on hemostasis. WT ZPI, when injected at the same dose (7.5 mg/kg) that potently inhibited thrombosis, increased tail bleeding time (p=0.028, WT vs. buffer, Fig. 7A) and decreased hemostasis incidence of injured saphenous vein (p < 0.001, WT vs. buffer or D293A/Y387D, Fig.7B). In contrast, D293A at the same dose (7.5 mg/kg) that potently inhibited thrombosis did not increase the tail bleeding time and did not decrease hemostasis incidence in saphenous vein bleeding model (Fig.7B). As a control, the functionally dead D293A/Y387D double mutant of ZPI also had no effect on tail bleeding time and saphenous vein bleeding. To determine whether even higher concentrations of D293A could affect hemostasis, we doubled the D293A dose to 15mg/kg. Again, D293A did not increase the tail bleeding time (Fig.7A) but moderately decreased saphenous vein hemostasis incidence (p=0.0081 or 0.0054, D293A 15mg/kg vs. buffer or D293A/Y387D, Fig.7B).

Figure. 7. Comparison of effects of ZPIs on tail bleeding time and on hemostasis in saphenous vein model.

A, Tail bleeding times of mice treated with control buffer (5mM sodium phosphate pH 6.5, containing 0.16M NaCl, 1mg/mL mouse albumin, n=9, male 6, female 3), 7.5mg/kg WT (n=12, male 7, female 5), 7.5mg/kg (n=10, male 5, female 5) and 15mg/kg (n=8, male 5, female 3) D293A, and 7.5mg/kg D293A/Y387D (n=9, male 4, female 5) ZPIs. B, Effects on hemostasis (indicated by the number of times of formation/reformation of hemostatic clot is achieved in 20 min) in saphenous vein model treated with buffer (n=10, male 6, female 4), 7.5mg/kg WT (n=11, male 5, female 6), 7.5mg/kg (n=10, male 5, female 5) and 15mg/kg D293A(n=10, male 4, female 6), and 7.5mg/kg D293A/Y387D ZPIs (n=8, male 5, female 3). See “Methods” for details. p values were shown as indicated. D’Agostino-Pearson normality test was performed before one-way ANOVA testing with Bonferroni correction for multiple testing. The median of each group is shown.

Discussion

In this study, we demonstrate that (1) by a single amino acid mutation (D293A) that abolishes PZ-binding, ZPI is converted into a selective intrinsic pathway inhibitor with minimal effect on the common pathway; (2) D293A ZPI shows strong selectivity in inhibiting APTT with minimal effect on PT. At concentrations that are selective in inhibiting APTT, both D293A ZPI and WT ZPI show similar inhibitory effect as the potent FXIa inhibitor BMS-262084, and were more potent than asundexian, a small molecule FXIa inhibitor currently in clinical trials³³, suggesting ZPI is likely to be a better selective inhibitor of the intrinsic pathway than asundexian. Due to the species difference between mouse and human FXI/FXIa, asundexian, and BMS-262084 are not effective in mice and thus cannot be compared with ZPIs in mice in vivo. (3) Importantly, despite the loss of FXa inhibitory function, at pharmacological concentrations, D293A potently inhibits both arterial and venous thrombosis induced by FeCl₃ in vivo in mice. (4) In contrast to WT ZPI, which exacerbates bleeding, D293A ZPI did not affect bleeding at a therapeutically effective concentration, and only showed moderate effect on hemostasis at a much higher concentration. Thus, D293A ZPI is potentially a new and more selective intrinsic pathway anticoagulant that had minimal effect on bleeding.

Whereas WT and D293A ZPI both potently inhibit the intrinsic pathway induced coagulation, D293A is more selective with minimal effects on the FXa-dependent common pathway (Fig.1). WT ZPI also inhibits the common pathway in a PZ-dependent fashion and thus showed more potent inhibitory effects on thrombin generation assays initiated through the extrinsic pathway. However, at very high concentrations of TF, particularly in PT assays, the large amount of FXa generated via the extrinsic pathway apparently overwhelms the inhibitory effect of WT ZPI. Furthermore, the PZ-dependent inhibition of the FXa-mediated common pathway by exogenously added WT ZPI is limited by the relatively low levels of endogenous PZ. The high TF concentration and limited FXa inhibitory effects combined result in no inhibition of PT by WT ZPI, similar to D293A. However, this does not in any way suggest that WT ZPI lacks the ability to inhibit the TF induced coagulation process, as WT ZPI is proven to be significantly superior to D293A in inhibiting TF induced CAT assays in the presence of endogenous PZ (Fig.3).

We found that D293A ZPI potently inhibits both arterial and venous thrombosis but minimally affects mouse tail bleeding and saphenous vein hemostasis at a therapeutically effective concentration as compared with WT ZPI. Only at a higher concentration, D293A showed a selective inhibitory effect on the efficiency of saphenous vein hemostasis (but not tail bleeding time), consistent with the previously shown effect of FXI-knockout mouse²⁹. This is in contrast to WT ZPI, which, even at a much lower concentration affected both tail bleeding time and saphenous vein hemostasis. Clearly, D293A is a safer anti-coagulant with reduced bleeding risk. It may seem surprising that pharmacological doses of D293A are similarly effective as WT ZPI in inhibiting thrombosis in the mouse carotid artery and IVC thrombosis models. However, the data from vitro thrombin generation and APTT assays demonstrate that the pharmacological concentration of exogenously added WT or D293A ZPI similarly inhibited the intrinsic pathway in human plasma, and D293A ZPI is still effective in inhibiting coagulation in mouse plasma, despite the loss of PZ-dependent inhibition of FXa and the common pathway rendering it less potent than WT ZPI in inhibiting thrombin generation in vitro (Figs. 2–4). Furthermore, the inhibitory effect of WT ZPI on the common pathway is limited by the plasma’s relatively low endogenous PZ concentrations, as shown in the APTT and PT assays. Thus, the effectiveness of both pharmacological doses of WT and D293A ZPIs in PZ-independent FXIa inhibition combined with the limitation of the PZ-dependent FXa-inhibitory function of WT ZPI by low endogenous PZ concentrations can explain why exogenous D293A ZPI is similarly effective in inhibiting in vivo thrombosis. These data thus suggest that the antithrombotic effect of pharmacological doses of ZPI is primarily mediated by its PZ-independent FXIa inhibition, at least in the models we used. It is important to note that FeCl₃-induced thrombosis has its limitations. While this model reflects the common process of thrombosis and importance of intrinsic pathway³⁸ in conditions associated with limited TF activation, FeCl₃ is not a natural cause of thrombosis and unlikely the same as many types of natural causes, particularly thrombosis-induced by massive TF activation. Further investigation into the effects of WT and D293A ZPIs in other thrombosis models is warranted in future development of ZPI and ZPI mutants as novel anti-coagulant agents.

ZPI is a natural FXIa inhibitor in plasma. ZPI is highly specific. Except for FXa and FXIa, ZPI does not inhibit thrombin, meizothrombin, FVIIa, FIXa, APC, tissue-type plasminogen activator(t-PA), urokinase-type plasminogen activator (u-PA), plasmin, trypsin, leukocyte elastase, chymotrypsin, or cathepsin G in presence or absence of PZ, lipids, and calcium based on protease screening³⁹. Although an earlier report showed that ZPI can inhibit FIXa independent of PZ⁷, a later study led by Olson and Broze et al. failed to confirm that finding⁸. Unique serpin structures, including P1 and exosites around the reactive center loop, determine such high specificity⁴⁰. The k_a for the inhibition of FXIa by WT ZPI is more than two orders of magnitude greater than that of antithrombin and similar to that of C1 inhibitor (C1-INH) in the presence of glycosaminoglycans^23,41. WT and D293A ZPI similarly have a k_a ~10⁵M⁻¹s⁻¹ and a k_diss ~0.0002 s^{−1 8,9}. Because the dissociation of ZPI-FXIa complex produces active FXIa and inactive ZPI, an apparent K_D=k_diss/k_a of ~2nM can be calculated for FXIa. This affinity is similar to AB023 (monoclonal antibody) (K_D of ~1.4nM for FXI)⁴², asundexian (small molecular compound) (K_i of ~1.0nM for FXIa)³³, and BMS-262084 (small molecular compound, K_i ~2.8nM)⁴³. We demonstrated that D293A is more potent in inhibiting APTT than asundexian and is similar to BMS-262084 at the same concentrations that do not inhibit PT (Fig.4A). It is noteworthy that the affinity of asundexian for FXIa significantly decreases when moving from physiological buffer to human plasma (IC50 or K_i increases from ~1nM to ~140nM)³³. Overall, D293A ZPI may have unique advantages compared with existing FXI or FXIa inhibitors in development: i) it is a natural FXIa regulator in the plasma with a single mutation, unlikely to have unexpected adverse effects often associated with artificial inhibitors, although it could still induce alloantibodies; ii) it is likely to have an inherent ZPI characteristic of a fast onset of action; iii) As ZPI also showed an anti-inflammatory function^44,45, which is additional benefit for thrombo-inflammatory conditions; iv) it has been reported that FXIIa may activate FIX via kallikrein independent of FXI^5,6. Whereas the effect of ZPI on kallikrein has not been evaluated in the previous studies³⁹, we show that both WT and D293 ZPIs at a relatively high concentration completely inhibits intrinsic pathway of thrombin generation. As prekallikrein and FXI share high homology in structure and sequence, especially in active sites⁴⁶, it is thus logical to speculate that ZPI may possibly inhibit the FXIa-independent activation of intrinsic pathways (e.g. kallikrein), and thus may possibly be advantageous over specific inhibitors of FXIa in inhibiting the intrinsic pathway of coagulation.

The relative short half-life (~20min) of the non-mammalian-originated recombinant ZPIs (~50 KD on SDS-PAGE) used in this study is in contrast to a previous study showing a much longer half-life of mammalian-originated or plasma ZPI (~72KD on SDS-PAGE) in mouse circulation¹², suggesting the lack of glycosylation of recombinant human protein being prepared from non-mammalian cell systems affects the pharmacological kinetics of ZPI in mice. Thus, the fully glycosylated recombinant ZPI prepared from mammalian cells or similar modifications should greatly improve pharmacokinetics in vivo and greatly improve its potential for therapeutic use in the future.

ZPI and PZ from humans and mice share 77% and 66% identify in sequence, respectively. Both human and mouse ZPIs have a P1 tyrosine at the reactive center loop. The contact residues of ZPI for PZ are also identical in human and mouse ZPIs¹⁰. Therefore, human ZPI can directly bind to mouse PZ and vice versa, as indicated by the human and mouse plasma assays. However, recombinant human ZPI is still more effective in inhibiting coagulation in human plasma than mouse plasma, indicating that the recombinant human ZPI would have better inhibitory effect in humans than mice. In addition, in contrast to human plasma, which has ZPI and PZ levels 30–50nM with ZPI in excess, mouse (C57BL/6) plasma has a relatively higher PZ level as compared with ZPI level (~240nM for PZ and ~70nM for ZPI)¹². Thus, our results obtained in mice show that D293A ZPI potently inhibits thrombosis with minimal effects on hemostasis are likely to be reflected in humans in which the PZ-dependent function of ZPI is less potent than in mice, as demonstrated in this study. This notion will be tested in future human trials.

Highlights.

• ZPI mutant (D293A) shows a severely weakened PZ affinity and diminished FXa inhibitory function.

• ZPI mutant (D293A) is a selectively effective inhibitor of FXIa and the intrinsic coagulation pathway.

• D293A ZPI inhibits arterial and venous thrombosis in vivo without exacerbating bleeding.

Nonstandard Abbreviations and Acronyms

ZPI

protein Z-dependent protease inhibitor

PZ

protein Z

FIX

factor IX

FX

factor X

FXa

activated factor X

FXI

factor XI

FXIa

activated factor XI

FXII

factor XII

FXIIa

activated FXII

FV

factor V

FVa

activated FV

FVL

factor V Lenden

CAT

Calibrated Automated Thrombogram

APTT

activated partial thromboplastin time

PT

prothrombin time

IVC

inferior vena cava

TF

tissue factor

PNP

pooled normal human plasma

FXI-DP

factor XI deficient plasma

PK

prekallikrein

Ka

kallikrein

DOAC

direct oral anticoagulant

VTE

venous thromboembolism

APC

activated protein C

BSA

bovine serum albumin

PL

phospholipids

WT

wild type