Essential Domains of ADAMTS13 Metalloprotease Required for Modulation of Arterial Thrombosis

Abstract

Objective

ADAMTS13 inhibits platelet aggregation and arterial thrombosis by cleavage of von Willebrand factor (VWF). However, the structural components of ADAMTS13 required for inhibition of arterial thrombosis are not fully defined.

Methods and Results

Using recombinant proteins and a murine model, we demonstrate that ADAMTS13 variant either truncated after the 8^th TSP1 repeat or spacer domain inhibits ferric chloride-induced arterial thrombosis in Adamts13^−/− mice with similar efficacy to full-length ADAMTS13. The results obtained from monitoring thrombus formation in carotid and mesenteric arteries are highly concordant. Further analyses by site-directed mutagenesis and human monoclonal antibody inhibition assay reveal that the Cys-rich and spacer domains of ADAMTS13, particularly the amino acid residues between Arg⁵⁵⁹ and Glu⁶⁶⁴ in the spacer domain, may be critical for modulation of arterial thrombosis in vivo. Finally, the thrombosis-modulating function of ADAMTS13 and variants/mutants is highly correlated with the VWF-cleavage activity under fluid shear stress.

Conclusion

Our results suggest that the amino-terminus of ADAMTS13, specifically the variable region of the spacer domain, is crucial for modulation of arterial thromboses under (patho) physiological conditions. These findings shed more light on the structure-function relationship of ADAMTS13 in vivo and may be applicable for rational design of protein or gene-based therapy of arterial thromboses.

Introduction

ADAMTS13, a member of ADisintegrin-like And Metalloprotease with ThromboSpondin type 1 repeats (ADAMTS) family ^{1, 2}, is primarily synthesized in the liver and secreted into blood circulation. Plasma ADAMTS13 concentrations in healthy individuals range from 0.5 to 1.0 mg per liter ^{3, 4}. Severe deficiency of plasma ADAMTS13 activity results in thrombotic thrombocytopenic purpura (TTP), a potentially fatal syndrome ⁵. Mild to moderate deficiency of plasma ADAMTS13 activity is associated with increased risk of other arterial thrombotic disorders such as myocardial infarction ⁶ and cerebral ischemic injury ⁷.

Plasma ADAMTS13 is constitutively active in cleaving newly released ultra large (UL) von Willebrand factor (VWF) from stimulated endothelial cells ⁸, and thereby preventing an accumulation of platelets on injured vessel wall. In addition, ADAMTS13 cleaves released and soluble VWF or platelet-bound VWF in circulation under shear stress ^{9, 10}. Studies have shown that the cleavage of endothelial ULVWF occurs very rapidly with ⁸ or without ^{11, 12} fluid shear stress, suggesting that newly released ULVWF on endothelial cell surface may be in its “open” conformation. However, the VWF cleaved from endothelial cells by ADAMTS13 remains ultra large in size by multimer analysis ¹¹. This suggests that further proteolytic processing of these released VWF multimers, likely to occur in microvasculature, may be necessary to reduce the size. In blood, soluble VWF adopts a “closed” conformation ¹³, resistant to cleavage by ADAMTS13. Exposure of the soluble VWF to high shear stress or denaturants may open up the binding and cleavage sites, which are normally buried under the β-sheet in the central A2 domain ¹⁴. Moreover, binding of factor VIII ⁹ or platelets ¹⁵ or FVIII and platelets ¹⁰ to the soluble VWF also increases its cleavage by ADAMTS13 under shear stress. These results suggest a cofactor-dependent mechanism regulating VWF proteolysis by ADAMTS13 under physiologically relevant conditions.

Human ADAMTS13 consists of a metalloprotease domain, a disintegrin domain, the first thrombospondin type 1 repeat (TSP1), a Cys-rich domain, and spacer domains. The more distal C-terminus has seven additional TSP1 repeats and two CUB domains ^{1, 2}. We ^{16, 17} and others ^{18, 19} have shown that N-terminal half of ADAMTS13 appears to be necessary and sufficient for proteolytic cleavage of VWF under various in vitro conditions. But, the role of C-terminal domains of ADAMTS13 in vivo remains controversial. For instance, we reported that a C-terminally truncated ADAMTS13 variant after the spacer domain expressed by an in utero injection of lentiviral vector eliminated plasma ULVWF and inhibited ferric chloride (FeCl₃)-induced arterial occlusion in the carotid artery of Adamts13^−/− mice ²⁰. Banno et al. showed that a naturally occurring murine Adamts13 variant truncated after the 6^th TSP1 repeat (Adamts13^S/S) was less efficacious than full-length Adamts13 inhibiting FeCl₃-induced thrombosis in the mesenteric arteriole ²¹. More recently, de Maeyer et al reported that a recombinant murine Adamts13 variant truncated after the 8^th TSP1 repeat (mT8), infused into Adamts13^−/− mice, was not able to cleave newly released ULVWF/VWF strings on the endothelial cells in the mesenteric arterioles/venules ²².

These discrepant results promote us to systematically investigate the structure-function relationship of ADAMTS13 in vivo using recombinant protein strategy and a murine thrombosis model. In addition, we hope to determine the correlation between thrombus-inhibiting activity and VWF-cleavage activity under more physiologically relevant conditions. Our findings may shed light on the structure-function relationship of ADAMTS13 in vivo. The information obtained may be applicable for rational design of protein or gene-based therapies for TTP and other arterial thrombotic disorders.

Materials and Methods

Preparation of ADAMTS13 (or variants), VWF, and anti-spacer antibody

Plasmids containing a full-length human ADAMTS13 (FL) and various truncated variants (T8, S, T1, and del6a) with a V5-His epitope tagged at their C-termini are shown in Fig. 1A. All plasmids were cloned into pcDNA3.1 V5-His TOPO vector (Invitrogen, Carlsbad, CA). Plasmids (FL, T8, S, and d6a) were stably transfected into human embryonic kidney (HEK293) cells, whereas plasmid T1 was stably transfected into Madin-Darby canine kidney (MDCK) cells using lipofectamine2000 reagent (Invitrogen, Carlsbad, CA). The reason for production of T1 in MDCK is purely technical because the expression level of T1 was higher in MDCK cells than in HEK293 cells. Full-length ADAMTS13 and variants were purified from DMEM/F12 serum-free conditioned media using the methods established previously ²³. Human VWF (hVWF) was purified from pooled normal human plasma using cryoprecipitation, followed by precipitation with polyethylene glycol (PEG) 15,000, and Sephacryl-300 gel filtration as described previously ¹⁰. Murine recombinant VWF (mVWF) was purified from DMEM/F12 serum-free conditioned medium of stably transfected HEK293 cells using Q-fast flow ion exchange and Sephacryl-300 (GE Healthcare, Piscataway, NJ) gel filtration chromatographies. The plasmid encoding full-length murine vwf was a kind gift from Dr. David Motto (University of Iowa, Iowa City, IA). The purity of all proteins was determined by SDS-polyacrylamide gel electrophoresis with Coomassie blue staining. The concentrations of those with greater than 95% purity (hVWF, mVWF, FL, T8, and S) were determined by the optical density (OD) at 280 nm corrected with light scattering at 320 nm (1-cm cuvette) using a Nano-drop2000c spectrophotometer (Thermo Scientific, San Diego, CA). The coefficients at the OD280 (corrected) for hVWF, mVWF, FL, T8, and S were 1.0, 1.0, 0.68, 0.71, and 0.91 mg/ml, respectively. The concentrations of those partially purified proteins (T1 and d6a) were determined by an in-house immunosorbent assay (ELISA) as described below using a purified FL as a standard. Human monoclonal anti-spacer antibody (mAb II-1) derived from B cells from a patient with acquired autoimmune TTP ²⁴ and control antibody against pneumococcus (mAb-c) ²⁵ have been described previously.

Figure 1. Domain compositions and plasma half-lives of recombinant ADAMTS13 and variants.

A. The schematic representation of full-length human ADAMTS13 (FL), variants truncated after the 8^th TSP1 repeat (T8), the spacer domain (S), and the 1^st TSP1 repeat (T1), and a mutant with residues Arg⁶⁵⁹-Glu⁶⁶⁴ in the spacer domain deleted (d6a) is shown. All recombinant human ADAMTS13 proteins contain a V5-His tag at their C-termini. B. Plasma concentrations of human ADAMTS13 and variants in Adamts13^−/− mice at 5 min of post-injection are shown. C. The normalized plasma concentrations of human ADAMTS13 and variants in Adamts13^−/− mice were plotted against time after injection. The half-life (t_1/2) of each recombinant protein was determined from the curve. The data are the mean ± SD of three independent experiments (n=3).

ELISA

A high-binding microtiter plate (Thermo Scientific, Rockford, IL) was coated with 100 µl of monoclonal anti-disintegrin IgG in phosphate-buffered saline (PBS) (40 µg/ml) (Green Mountain Antibodies, Burlington VT). After being blocked with 2.5% bovine serum albumin (BSA) in PBS, 100 µl of diluted samples containing ADAMTS13 or variants in 0.5% BSA in PBS were added and incubated at 25 °C for 2 hours. After three PBS washes, the bound ADAMTS13 and variants were incubated for 1 hour with 100 µl of biotinylated rabbit anti-V5 IgG (0.5 µg/ml) (Novus Biologicals, Littleton, CO), followed by a 30 minutes incubation with streptavidin-peroxidase (1:2,000) (Burlingame, CA). TMB (3,3’, 5,5’-tetramethylbenzidine) substrate (100 µl) (Thermo Scientific, Rockford, IL) was added for color development. After stopping the reaction with 50 µl of sulfuric acid (H₂SO₄), the absorbance (450 nm) was determined on a SpectroMax microtiter plate reader (Molecular Devices, Sunnyvale, CA). A purified recombinant FL at concentrations of 0, 0.025, 0.05, 0.1, 0.2, 0.4 µg/ml in 0.5% BSA in PBS was used for calibration.

Inhibition of ADAMTS13 activity by human monoclonal anti-spacer antibody

Recombinant human ADAMTS13 (0.6 nM) was incubated at 25 °C for 30 minutes with human monoclonal anti-spacer IgG (mAb II-1) (0 to 35 nM) or human monoclonal control antibody (mAb-c) (35 nM). The residual ADAMTS13 activity was determined by the cleavage of a fluorescein-labeled recombinant human VWF73 peptide (rF-vWF73) (2 µM) as described previously ^{26, 27}. Normal human plasma was used for calibration. Relative activity of residual ADAMTS13 (%) was plotted against the concentrations of mAb II-1 or mAb-c used in the reaction.

Half-life of ADAMTS13 and variants in mice

Adamts13^−/− mice (4–6 weeks old) were anesthetized with intra-peritoneal injection of Nembutal (0.1 mg/g body weight). Recombinant ADAMTS13 or variants diluted with 100 µl of normal saline were injected into mice via a retro orbital sinus plexus. The amount of ADAMTS13 or variants injected was determined based on the body weight and blood volume of mice in order to achieve physiological plasma concentrations (~5–10 nM). We used ~7 ml blood per 100 g body weight in all of our calculations as described ²⁸. Blood samples (50 µl) were collected from a jugular vein at 5, 20, 40, and 70 minutes after protein injection and anticogulated with 3.8% sodium citrate (9:1 blood: anticoagulant). Plasma was obtained after immediate centrifugation at 8,000 rpm for 8 minutes and stored at −80 °C in aliquots. The concentrations of ADAMTS13 and variants in murine plasma after 1:10 dilution with PBS at various time points were determined by the ELISA as described above.

Carotid arterial occlusion assay

Mice at 2–4 months of age were anesthetized with Nembutal. The right carotid artery was exposed by blunt dissection. PBS (100 µl) alone or PBS containing recombinant human ADAMTS13 and variants (final concentration 10 nM) was injected into mice via the retro orbital plexus. The amount of protein injected was determined as described for the half-life assessment. In addition, purified human recombinant ADAMTS13 (100 nM) was incubated with 3.5 µM of mAb II-1 or mAb-c in PBS for 30 minutes (total volume, 100 µl). The ADAMTS13-antibody mixtures or complexes were infused via a jugular vein into Adamts13^−/− mice (~1 ml blood for ~14 g body weight). This resulted in final concentrations of ~10 nM of ADAMTS13 and 0.35 µM of mAb II-1 (or mAb-c) in mouse circulation, respectively. Three minutes after protein infusion, injury was induced by exposing the carotid artery to 10% ferric chloride (FeCl₃) (anhydrous) soaked on a piece of filter paper (1×2 mm) for 2 minutes ²⁰. The filter paper was then removed and the injured field was flushed with PBS. The blood flow was monitored using a Doppler flow probe (Transonic Systems, Ithaca, NY) until cessation or up to 30 minutes if no occlusion was observed. The times to the complete occlusion or cessation of blood flow were recorded. The differences among various means of the times to the complete occlusion were determined by ANOVA one-way analysis of variance using Minitab statistic software (Minitab, State College PA). All assays were performed in a blind fashion (the operator did not know the nature of the protein infused) to avoid bias.

Thrombus formation in mesenteric arteriole

Prior to vessel injury, platelets were isolated from Adamts13^−/− mice and labeled with calceinAM (Invitrogen, Carlsbad, CA) (2.5 µg/ml) for 15 minutes ²⁹. The labeled murine platelets were washed with modified Tyrode buffer (10 mM HEPES, pH 7.2, 137 mM NaCl, 2.68 mM KCl, 0.42 mM NaH₂PO₄, 1.7 mM MgCl₂, 11.9 mM NaHCO₃, and 5 mM glucose) containing 1 µg/ml PGE₁ (Sigma, St. Louis, MO) and then infused via a retro orbital plexus into the anesthetized mice. Five minutes after platelet infusion, recombinant ADAMTS13 proteins diluted into 100 µl of PBS were infused via the same route. The amount of protein infused was the same as that for the carotid arterial occlusion assay described above. The mesentery was gently exteriorized through a midline abdominal incision. The arterioles of 90–120 µm diameters (Suppl. Table 1) were visualized with an inverted fluorescent microscope (Nikon) and a CCD camera system. The arteriole was injured by topical application of 10% FeCl₃ saturated in a filter paper (1×2 mm) for 5 minutes. The filter paper was removed and the injured site was flushed with PBS. The thrombus formation in the mesenteric arteriole was monitored for 30 minutes after injury or until its complete occlusion that lasted for 30 sec. The times to initial thrombus formation (the diameter of the thrombus is >30 µm) and the completion occlusion (defined as the cessation of blood flow for at least 30 seconds) were determined from video clips using the NIS Elements software (Nikon, Melville, NY) ³⁰. The differences among various groups were determined by ANOVA one-way analysis of variance. Again, the experiments were performed in a blind fashion to avoid the operator’s bias.

Kinetic cleavage of VWF by ADAMTS13 and variants/mutants under fluid shear stress

For time-dependent cleavage, purified VWF (150 nM) was incubated with purified recombinant ADAMTS13 proteins (25 nM or 125 nM) in the presence of lyophilized human platelets (150×10³/µl) and FVIII (1 nM) at 25 °C for various times (0–60 minutes) under constant vortex rotation (2,500 rpm) on a MixMate PCR mixer (Eppendorf, Hauppauge, NY). The total volume was 20 µl in all reactions. For kinetic determination, purified VWF at various monomer concentrations (0, 18.75, 37.5, 75, 150, and 300 nM) was incubated with 25 nM of ADAMTS13 or T8, S, and d6a or 125 nM of T1, under physiological concentration of FVIII (1 nM) and lyophilized platelets (150×10³/µl) at 25 °C for 30 minutes with constant mixing at 2,500 rpm. The proteolytic cleavage of multimeric VWF was determined by 1% agarose gel electrophoresis, followed by Western blotting as previously described ⁹. The increase in proteolytic cleavage products and reduction in high molecular weight VWF multimers was quantified by densitometry using the ImageJ software and plotted against time or VWF concentrations. The curve was fitted into the Michaelis-Menten equation to obtain the rate constants K_m and k_cat using SigmaPlot software.

Results

Cleavage of murine or human VWF by murine or human ADAMTS13

Species difference in proteolytic cleavage of VWF by ADAMTS13 ought to be considered before a proper interpretation of in vivo data could be made. Previous study reported that human recombinant VWF was not efficiently cleaved by murine plasma Adamts13 under urea-denaturing conditions ³¹. However, recombinant human ADAMTS13 infused into Adamts13^−/− mice rapidly cleaved newly released ULVWF on endothelial cells ³⁰, suggesting that human ADAMTS13 does cleave murine VWF under physiological conditions. To further assess whether there was a species-dependent variation in proteolysis of VWF by ADAMTS13, we determined proteolytic cleavage of murine (or human) VWF by murine (or human) ADAMTS13 under two commonly used in vitro assay conditions. As shown, the cleavage of murine VWF by murine Adamts13 and human ADAMTS13 or the cleavage efficacy of human VWF by murine Adamts13 and human ADAMTS13 was quite similar if guanidine was used to denature VWF (Suppl. Fig. 1). However, if urea was used for denaturization, cleavage of murine VWF by human ADAMTS13 was reduced by ~5-fold compared with cleavage of human VWF by human ADAMTS13 (Suppl. Fig. 2). Such difference may be caused by higher urea concentration (2.0 M) used for denaturization of murine VWF than human VWF (1.5 M). Surprisingly, the cleavage of both murine and human VWF by murine Adamts13 was dramatically reduced compared to that by human ADAMTS13 (Suppl. Fig. 2), suggesting that murine Adamts13 is highly sensitive to urea denaturization under these assay conditions.

Half-life of recombinant ADAMTS13 proteins in Adamts13^−/− mice

Full-length ADAMTS13 and variants with different truncations, mutations, or sources of production may exhibit different half-life in vivo ^{20, 32}. To assess the half-life of our recombinant ADAMTS13 proteins used for murine studies, we injected via a retro orbital sinus plexus various recombinant ADAMTS13 proteins into Adamts13^−/− mice (C57BL/6). The amount of injected recombinant proteins was determined by the body weight and blood volume of each mouse. By ELISA, we demonstrated that at 5 minutes after infusion, plasma concentrations of FL, T8, S, T1, and d6a were quite similar (~6–10 nM) (Fig. 1B). The plasma concentrations of all infused ADAMTS13 proteins except for T1 reduced rapidly over time (Fig. 1C). The estimated half-life (t_1/2) for most infused recombinant ADAMTS13 proteins (FL, T8, S, and d6a) derived from HEK293 cells was between ~15 and ~18 minutes, consistent with those reported previously ²⁰. However, the t_1/2 of T1 which was purified from MDCK cells was ~38 minutes, approximately twice as long as those produced in HEK293 cells (Fig. 1C). Therefore, the average plasma concentration of T1 during 30–40 minutes of observation was almost twice as high as that of other variants (Fig. 1C).

Effect of full-length ADAMTS13 and variants on carotid arterial occlusion

Ferric chloride is a widely used chemical for inducing arterial and venous thromboses in mice to study VWF-platelet functions ³³. Tissue damage mediated by iron chemical oxidation predisposes the injured area to platelet adherence and aggregation, followed by coagulation activation and fibrin deposition. Upon topical application of a filter paper soaked with 10% FeCl₃ (anhydrous) on an isolated carotid artery, blood flow was continuously monitored until cessation. Due to the short half-life of infused recombinant ADAMTS13 proteins, we started FeCl₃ injury 3 minutes after protein injection. The entire experiment was completed within 35 minutes after protein infusion. We showed that the time to the complete vessel occlusion (or the cessation of blood flow) in carotid artery in Adamts13^−/− mice (C57BL/6) was 5.4 ± 0.3 minutes (n=14) (mean ± SEM). This was significantly shorter than that in wild-type mice (10.4 ± 0.9 minutes, n=13) with the same genetic background (p<0.0001) (Fig. 2). These results suggest that FeCl₃-induced carotid arterial occlusion assay is sensitive for assessment of ADAMTS13 function in vivo.

Figure 2. Effect of ADAMTS13, variants, and human anti-spacer monoclonal antibodies on FeCl₃-induced carotid arterial occlusion.

A. Wild-type (WT) or Adamts13 ^−/− mice were injected via retro orbital sinus plexus with 100 µl of PBS alone or PBS containing 10 nM of purified human recombinant ADAMTS13 and variants (i.e. FL, T8, S, T1, and d6a). B. Adamts13^−/− mice injected 100 µl of PBS containing 10 nM of purified human recombinant ADAMTS13 (FL) pretreated with mAb II-1 or mAb-c. ADAMTS13-antibody mixtures (or complexes) were infused via a jugular vein. Three minutes after infusion, an injury to the right carotid artery was performed with topical application of 10% FeCl₃ soaked on a filter paper (1×2 mm) for 2 min. The time to the complete cessation of blood flow was determined by a Doppler ultrasound flow probe. The means and standard errors of the mean (SEM) are shown. The numbers in parenthesis are the total number of mice performed in each group. Statistical analysis was performed using ANOVA one-way analysis of variants with a Tukey correction. The p values < 0.01 are considered statistically highly significant.

We therefore used this assay to determine the function of various C-terminal truncated ADAMTS13 variants using the same protocol. As shown, the time to the complete occlusion of carotid artery after infusion of recombinant FL, T8, and S was 13.4 ± 1.3 minutes (n=16), 11.1 ± 1.4 minutes (n=15), and 20.0 ± 1.9 minutes (n=16), respectively (Fig. 2). These results suggest that the amino terminal half of ADAMTS13 (up to the spacer domain) is sufficient for inhibition of arterial thrombosis. An infusion of T1 and d6a into Adamts13^−/− mice at the same concentrations (~6–10 nM) did not significantly prolong the time to complete occlusion of carotid artery. The time to the complete occlusion after T1 and d6a infusion was 6.4 ± 0.7 minutes (n=14) and 5.9 ± 0.6 minutes (n=11), respectively (Fig. 2). The difference in the mean occlusion time between T1 or d6a-injected and PBS-injected was not statistically significant (p=0.156 or 0.411). These results suggest that the Cys-rich and spacer domains of ADAMTS13, particularly the amino acid residues between Arg⁶⁵⁹ and Glu⁶⁶⁴, are required for modulation of arterial thrombosis in vivo.

To further test this hypothesis, we infused a full-length ADAMTS13 pre-treated with anti-spacer antibody (mAb II-1), which targets specifically at the variable region (Arg⁶⁵⁹-Tyr⁶⁶⁵) of spacer domain ^{24, 34, 35}, into Adamts13^−/− mice. We found that the antibody treated ADAMTS13 had dramatically reduced the ability to attenuate arterial thrombosis in carotid artery induced by FeCl₃. As a control, an infusion of ADAMTS13 pre-treated with a human control monoclonal antibody (mAb-c), into Adamts13^−/− mice significantly attenuated the formation of arterial thrombosis under the same conditions. The difference in the time to complete occlusion between Adamts13^−/− mice receiving ADAMTS13 pre-treated with mAb II-1 (8.2 ± 0.5, mean ± SEM, n=10) and ADAMTS13 pre-treated with mAb-c (19.5 ± 3.5, n=10) was statistically highly significant (p=0.0058) (Fig. 2B). However, the time to the complete occlusion in Adamts13^−/− mice receiving mAb II-1 treated ADAMTS13 (8.2 ± 0.5, n=10) was slightly longer than that in Adamts13^−/− mice (6.6 ± 0.4, n=9) (p=0.037) (Fig. 2B), suggesting low residual activity of ADAMTS13-mAb II-1 complexes under these in vivo conditions.

Prior to infusion into mice, the inhibitory effect of purified mAb II-1 was confirmed by cleavage of rF-VWF73 peptide. As shown, mAb II-1, but not mAb-c inhibited proteolytic activity of recombinant full-length ADAMTS13 in a concentration-dependent manner (Suppl. Fig. 3). In the presence of 35 nM of mAb II-1, proteolytic activity of ADAMTS13 was completely inhibited. The half-maximal inhibitory concentration of mAb II-1 (EC₅₀) was estimated to be ~4.5 nM (Suppl. Fig. 3). In addition, the same mAb II-1 was shown to inhibit proteolytic cleavage of multimeric VWF under denaturing conditions and cell-bound ULVWF on endothelial cells under flow in the previous study ²⁴.

Effect of ADAMTS13 and variants on thrombus formation in mesenteric arteriole

To assess the effect of ADAMTS13 proteins on modulation of thrombosis in smaller arteries with a shear rate of ~1000 to ~1,500s⁻¹, we determined the real-time thrombus formation in mesenteric arterioles. Mesenteric arterioles with diameters between 70–105 µm were chosen for the experiment (suppl. Table 1). After topical application of 10% FeCl₃ for 5 minutes, thrombus formation was monitored using time lapse under an inverted fluorescent microscope for accumulation of fluorescein-labeled murine platelets at the site of injury. Representative images (Fig. 3A) or video clips from Suppl. Videos 2-8 of the real-time thrombus formation (up to 10 minutes) in wild-type mice and Adamts13^−/− mice receiving ADAMTS13 or variants are shown. Moreover, the time to the initial thrombus formation and the complete occlusion of the mesenteric arterioles were determined. The results showed that the times to the initial thrombus formation and the complete occlusion in the mesenteric arteriole of Adamts13^−/− mice were 5.2 ± 0.6 minutes and 9.6 ± 0.8 minutes (mean ± SEM, n=16), respectively. These values were significantly smaller than those in wild-type mice with the same genetic background (8.8 ± 0.6 minutes and 13.8 ± 1.0 minutes, n=12, respectively) (Fig. 3B and Fig. 3C). The results imply that the FeCl₃-induced mesenteric arterial thrombus formation assay is also sensitive for assessing systemic anti-arterial thrombotic function of ADAMTS13 in vivo. Similar to carotid arterial occlusion, an infusion of recombinant FL, T8, and S (~6–10 nM) into Adamts13^−/− mice significantly prolonged the times to both the initial thrombus formation (mean ± SEM) (FL: 9.7 ± 0.9 minutes, n=10; T8: 10.1 ± 1.5 minutes, n=10; S: 10.5 ± 1.9 minutes, n=10) (Figs. 3B and 3C) and the complete occlusion (FL: 15.3 ± 1.4 minutes; T8: 21.9 ± 2.3 minutes; S: 16.2 ± 1.6 minutes) in these mice (Fig. 3B and Fig. 3C). These results further confirmed that the N-terminus of ADAMTS13 is sufficient for inhibition of arterial thrombosis in vivo. Similar to the results obtained from carotid arterial occlusion, an infusion of T1 or d6a into Adamts13^−/− mice at the same concentrations did not prolong both the time to the initial thrombus formation (means ± SEM) (T1: 5.5 ± 0.37 minutes, n=12; d6a: 5.4 ± 0.56 minutes, n=11) (Fig. 3B) and the time to complete occlusion (means ± SEM) (T1: 9.3 ± 0.8 minutes, n=12; d6a: 10.0 ± 1.1 minutes, n=11) (Fig. 3C). The differences in the time to the initial thrombus formation and that to the complete occlusion between T1 or d6a-injected and PBS-injected Adamts13^−/− mice were not statistically significant (all p values were greater than 0.5) (Fig. 3B and Fig. 3C). These results further confirmed the importance of the Cys-rich and spacer domains, particularly the amino acid residues between the Arg⁶⁵⁹ and Glu⁶⁶⁴ in the spacer domain, in systemic anti-arterial thrombosis under (patho) physiological conditions.

Figure 3. Effect of ADAMTS13 and variants on mesenteric arteriolar thrombus formation.

Wild-type and Adamts13^−/− mice (C57BL/6) were injected with 100 µl of PBS alone, or PBS containing FL, T8, S, T1, and d6a. The amount of proteins was determined by body weight and blood volume of the mice. Mesenteric arterioles were injured for 5 min with topical application of 10% FeCl₃ soaked in a filter paper (1×2 mm). Thrombus formation was monitored every 10 sec for 30 min or until complete occlusion under a NIKON inverted fluorescent microscope (with 10× objective). A. The fluorescent signal was generated from the activated and accumulated platelets loaded with calceinAM at the site of injury. The dashed white line indicates the boundary of vessel wall, whereas the red arrows point to the direction of blood flow in the mesenteric arteriole. The times to the initial thrombus formation (defined by the formation of thrombus >30 µm in diameter) (B) and the complete occlusion (C) of the mesenteric arteriole were determined using NIS-Elements software. The data are presented with means ± SEM. The numbers in parenthesis are the total numbers of mice performed in each group. Statistical analysis was performed by ANOVA one-way analysis of the variance with Tukey correction between the control (Adamts13^−/− receiving PBS alone) and various groups. The p values less than 0.01 were considered to be statistically highly significant.

Kinetic cleavage of VWF by ADAMTS13 and variants under fluid shear stress

To correlate thrombosis inhibition function with VWF-cleavage activity under fluid shear stress, we performed kinetic analyses. Addition of FVIII ⁹, platelets ^{10, 15}, or both ¹⁰ increased cleavage of VWF by ADAMTS13 under shear stress. Therefore, we performed all cleavage assays in the presence of FVIII (1 nM) and lyophilized platelets (150 ×10³/µL). As shown, the formation of VWF cleavage product by FL, T8, and S increased as a function of vortexing time (Fig. 4A–C), reaching a plateau after 30 minutes of incubation under constant vortexing at 2,500 rpm on a MixMate PCR mixer. No cleavage products were observed when ADAMTS13 was omitted (not shown) or when 10 mM EDTA was included in the reactions. There appeared to be no difference in terms of the cleavage efficiency among recombinant FL, T8, and S under these conditions (Fig. 4A–C). However, little or no cleavage production was detected when VWF was incubated with 125 nM of T1 or d6a (5 times of FL) (Fig. 4D and 4E).

Figure 4. Time-dependent cleavage of VWF by ADAMTS13 under fluid shear stress.

Purified plasma VWF (150 nM) was incubated for various times (0–60 min) with 25 nM of recombinant FL (A), T8 (B), S (C), 125 nM of T1 (D), and 25 nM d6a (E) in the presence of 1 nM FVIII and 150×10₃/µl lyophilized platelets under constant mixing at the rotation rate of 2,500 rpm. The cleavage products (arrowheads) were determined by 1% agarose gel and Western blotting and quantified by ImageJ software. Uncleaved VWF is marked as U. In each blot, a FL was included for calibration (not shown). The last lane is the negative control (with 10 mM EDTA). T1 (125 nM) was used because no cleavage product was observed with 25 nM in the preliminary experiments. F represents the means ± SEM from three independent experiments.

Moreover, the VWF-cleavage product formation increased as a function of increasing concentrations of VWF substrate (Fig. 5A and 5B). The proteolytic cleavage efficiencies (the ratios of k_cat/K_m), measured after 5 minutes of vortexing (the linear portion of the curve), for recombinant FL, T8, and S to cleave VWF were 1.0×10⁹, 0.9×10⁹, and 0.7×10⁹ M⁻¹s⁻¹, respectively (Table 1). The differences in kinetic parameters among FL, T8, and S were not statistically significant (p>0.05). When assessed after 30 minutes of vortexing, the ratios of k_cat/K_m for FL (0.3×10⁹ M⁻¹s⁻¹), T8 (0.4×10⁹ M⁻¹s⁻¹), and S (0.3×10⁹ M⁻¹s⁻¹) to cleave VWF remained the same. However, removal of the Cys-rich and spacer domains (T1) almost completely abolished proteolytic activity toward VWF under the same conditions (Fig. 4D and Fig. 5A–D). No cleavage product was detected even with addition of 125 nM of T1 after 60 minutes of vortexing (Fig. 4D, lane 5). Furthermore, d6a exhibited dramatically reduced proteolytic cleavage of VWF under the same conditions (Fig. 4E and Fig. 5). No cleavage products were detectable after 5 minutes of vortexing with both T1 and d6a; thus, no kinetic parameters were obtained (Table 2). A small amount of cleavage product was detected after 30 minutes of vortexing with d6a (Fig. 4D and Fig. 5), but not with T1 (Fig. 4E and Fig. 5). These results suggest that the Cys-rich and spacer domains, particularly the spacer domain, are critical for substrate recognition. Together, our data demonstrate that thrombus-inhibition function of human ADAMTS13 and variants in mice is highly correlated to the VWF-cleavage activity under fluid shear stress.

Figure 5. Kinetic analysis of VWF cleavage by ADAMTS13 and variants under fluid shear stress.

Purified plasma VWF at various concentrations (0–300 nM) was incubated without (−) or with (+) 25 nM of recombinant FL, T8, S, T1 and d6a at 25 °C for 5 min (A and B) or 30 min (C and D) under vortexing rotation rate of 2,500 rpm in the presence of 1 nM FVIII and 150×10³/µl lyophilized platelets. The proteolytic cleavage of VWF was determined by 1% agarose gel and Western blot as described in Materials and Methods. A and C show representative images of the Western blot for the formation of cleavage products (arrowheads) after 5 min and 30 min of vortexing, respectively. Notice that although the VWF concentration in the reaction and the amount loaded into the gel for Western blotting in both A and B were the same, the contrast in A was higher than that in C for better visualization of the cleavage product. ImageJ was used to determine the intensity of cleavage product. The specific signal intensity (after subtracted by that with EDTA) normalized to that of FL cleavage in each blot (not shown) was plotted against the concentrations of VWF substrate (means± SEM) (n=3) (B and D). The quantitative data were fitted into a Michaelis-Menten equation to derive the kinetic constants, K_m and k_cat, for each ADAMTS13 variant. The molar concentration of VWF was assessed based on the monomer mass of 250 kDa.

Table 1.

Kinetic Parameters of Proteolytic Cleavage of VWF by ADAMTS13 and Variants Under Fluid Shear Stresses

	kcat(s−1)	Km(×10−9 M)	kcat/Km(×109) M−1s−1	N
FL	55.9 ± 14.8*	58.4 ± 14.1	1.0 ± 1.1	4
T8	44.1 ± 51.1	60.0 ± 55.5	0.9 ± 0.8	4
S	51.1 ± 37.6	73.5 ± 66.2	0.7 ± 0.6	4
T1	1.2 ± 0.7	ND	ND	4
d6a	2.7 ± 0.5	ND	ND	3

N, number of independent experiments; ND, not determined.

^*Means ± standard deviation; Constructs FL, T8, S, and T1 represent full-length ADAMTS13, variants truncated after the 8th TSP1 repeat, the spacer domain, and the first TSP1 repeat, respectively, d6a indicates the full-length ADAMTS13 lacking amino acid residues Arg⁶⁵⁹-Glu⁶⁶⁴.

Discussion

ADAMTS13 limits platelet thrombogenesis presumably through proteolytic cleavage of newly released ULVWF anchored on endothelial cells ³⁶, VWF in circulating blood, and/or VWF bound with platelets at the site of thrombus formation ³⁷. The structural components of ADAMTS13 for modulation of arterial thrombosis are systemically investigated using recombinant ADAMTS13 proteins and murine models. We demonstrate that an infusion via retro orbital sinus plexus of recombinant FL, T8, and S (~6–10 nM) into Adamts13^−/− mice (Fig. 1) are able to inhibit FeCl₃-induced thrombus formation in both carotid artery and mesenteric arteriole (Fig. 2 and Fig. 3). Our data are consistent with those previously reported using a different model system ^{20, 21} with one exception 22. For instance, Adamts13^−/− mice expressing S via an in utero injection of lentiviral vector exhibited prolonged FeCl₃-induced carotid arterial occlusion time ²⁰. However, the relative efficacy of this variant to full-length ADAMTS13 was not assessed because of different plasma expression levels (0.5–0.7 U/ml vs. 0.07–0.1 U/ml)²⁰. Moreover, the effect from ectopic expression of the ADAMTS13 variant such as in leukocytes and platelets could not be excluded. Other study in congenic mice showed that Adamts13 variant truncated after the 6^th TSP1 repeat (Adamts13^S/S) was nearly as efficacious as wild-type Adamts13 (Adamts13^L/L) for inhibition of initial thrombus formation in mesenteric arterioles ²¹. However, Adamts13^S/S was found to be slightly less efficacious than Adamts13^L/L in blocking the later phase of thrombus formation and collagen-induced platelet aggregation under shear stress greater than 5000 s^{−1 21}. However, our results do not reveal significant difference regarding the effect on both early and later phases of thrombus formation among FL, T8, and S using either carotid arterial occlusion assay (Fig. 2) or mesenteric arteriolar thrombus formation assay (Fig. 3). Notably, our results are not fully in agreement with those by de Maeyer et al., in which an infusion of concentrated conditioned medium containing recombinant murine Adamts13 truncated after the 8^th TSP1 repeats (mT8) did not cleave the newly released ULVWF strings at all ²². The reason for such discrepancy is yet to be determined. One possibility is that cleavage of ULVWF string requires the distal C-terminal domain, but inhibition of thrombus growth does not. Species-dependent variation may also play a role, although human ADAMTS13 cleaves murine VWF as efficiently as does murine Adamts13 under guanidine-denaturing conditions (Suppl. Fig. 1). Under urea-denaturization conditions, murine Adamts13 cleaves both murine and human VWF much less efficiently than human ADAMTS13 does (Suppl. Fig. 2), suggesting an inactivation of murine Adamts13 in the presence of 1.5 M urea. Together, the data from ours and others suggest that the N-terminal half of ADAMTS13 (up to the spacer domain) is sufficient for substrate recognition and anti-arterial thrombosis under (patho) physiological conditions.

Further analyses have demonstrated that the ADAMTS13 variant lacking the Cys-rich and spacer domains (T1) or bearing a small deletion of 6 amino acid residues within the spacer domain (d6a), which does not cleaved VWF in vitro ³⁸, also exhibits dramatically reduced efficacy in inhibiting thrombus formation in both carotid artery and mesenteric arteriole after FeCl₃ injury (Fig. 2 and Fig. 3). Moreover, human monoclonal antibody against the variable region of β9–β10 in the spacer domain (mAb II-1) significantly inhibits proteolytic cleavage of VWF and FeCl₃-induced arterial thrombosis. These results suggest the critical role of the Cys-rich and spacer domains, particularly the amino acid residues between Arg⁶⁵⁹ and Glu⁶⁶⁴ for systemic anti-thrombotic function in vivo. The importance of the Cys-rich and spacer domains, particularly the variable region of β9–β10 in the spacer domain, in substrate recognition and anti-arterial thrombotic function is highlighted by the fact that this region contains the core antigenic epitopes for IgG autoantibodies against ADAMTS13 in patients with autoimmune TTP ^{39, 40}. A substitution of a single or a cluster of residues in this region reduced substrate recognition and cleavage of VWF, as well as the binding of these autoantibodies against ADAMTS13 ^{34, 35}.

Our data also demonstrate the direct correlation between anti-arterial thrombotic activity of ADAMTS13 (or variants) (Fig. 2 and Fig. 3) and its VWF-cleavage activity under fluid shear stress (Fig. 4, Fig. 5, and Table 1), suggesting the critical role of ADAMTS13-mediated VWF proteolysis in modulation of arterial thrombosis. However, it remains to be determined whether the VWF-cleavage independent activity, such as disulfide bond reducing activity of ADAMTS13 ⁴¹, plays a role in modulating arterial thrombosis. Recent studies suggest that both VWF and ADAMTS13 contains several surface exposed free thiols ⁴¹. The interaction between VWF and ADAMTS13 via disulfide bond formation under shear stress may prevent shear induced lateral association of VWF multimers to form bundles and strings ^{42, 43}, thereby modulating adhesion function of VWF. Further investigation of VWF-reducing activity of ADAMTS13 may shed light on the biological function of ADAMTS13 in vivo.

With growing interest in the therapeutic potential of recombinant ADAMTS13 protein ⁴⁴ or gene transfer ^{20, 45} in TTP and many other diseases, our findings may provide molecular basis for rational design of therapeutic targets. For instance, recombinant C-terminal truncated ADAMTS13 variants may be utilized instead of fulllength ADAMTS13 to prevent TTP or inhibit pathological thromboses such as myocardial infarction ⁶ and ischemic cerebral infarction ^{7, 32}. Mouse model demonstrates that deletion of Adamts13 aggregates reperfusion injury in the brain, whereas infusion of recombinant ADAMTS13 into mice reduces infarct size and improves functional outcome without producing cerebral hemorrhage ^{32, 46}. The truncated ADAMTS13 variants are easier to express in large quantity ^{16, 17, 23}, more resistant to proteolysis, and potentially less immunogenic than full-length ADAMTS13. In addition, supplementation of ADAMTS13 and variants may also be beneficial to patients with severe preeclampsia ⁴⁷, systemic inflammation ⁴⁸ or severe sepsis ⁴⁹, and multi-organ failure ⁵⁰, in which plasma ADAMTS13 activity was found to be severely reduced.