Light Chain of Factor VIII Is Sufficient for Accelerating Cleavage of von Willebrand Factor by ADAMTS13 Metalloprotease*

Background: The structural component of FVIII required for regulating VWF proteolysis is not fully understood. Results: A light chain of FVIII appears to be sufficient for accelerating the cleavage of VWF by ADAMTS13 in vitro and in vivo. Conclusion: Proteolytic cleavage of VWF can be regulated by the FVIII light chain. Significance: The findings provide insight into the structure-function relationship of FVIII in maintaining VWF homeostasis. We previously demonstrated that coagulation factor VIII (FVIII) accelerates proteolytic cleavage of von Willebrand factor (VWF) by A disintegrin and metalloprotease with thrombospondin type 1 repeats (ADAMTS13) under fluid shear stress. In this study, the structural elements of FVIII required for the rate-enhancing effect and the biological relevance of this cofactor activity are determined using a murine model. An isolated light chain of human FVIII (hFVIII-LC) increases proteolytic cleavage of VWF by ADAMTS13 under shear in a concentration-dependent manner. The maximal rate-enhancing effect of hFVIII-LC is ∼8-fold, which is comparable with human full-length FVIII and B-domain deleted FVIII (hFVIII-BDD). The heavy chain (hFVIII-HC) and the light chain lacking the acidic (a3) region (hFVIII-LCΔa3) have no effect in accelerating VWF proteolysis by ADAMTS13 under the same conditions. Although recombinant hFVIII-HC and hFVIII-LCΔa3 do not detectably bind immobilized VWF, recombinant hFVIII-LC binds VWF with high affinity (KD, ∼15 nm). Moreover, ultra-large VWF multimers accumulate in the plasma of fVIII−/− mice after hydrodynamic challenge but not in those reconstituted with either hFVIII-BDD or hFVIII-LC. These results suggest that the light chain of FVIII, which is not biologically active for clot formation, is sufficient for accelerating proteolytic cleavage of VWF by ADAMTS13 under fluid shear stress and (patho) physiological conditions. Our findings provide novel insight into the molecular mechanism of how FVIII regulates VWF homeostasis.

We previously demonstrated that coagulation factor VIII (FVIII) accelerates proteolytic cleavage of von Willebrand factor (VWF) by A disintegrin and metalloprotease with thrombospondin type 1 repeats (ADAMTS13) under fluid shear stress. In this study, the structural elements of FVIII required for the rateenhancing effect and the biological relevance of this cofactor activity are determined using a murine model. An isolated light chain of human FVIII (hFVIII-LC) increases proteolytic cleavage of VWF by ADAMTS13 under shear in a concentrationdependent manner. The maximal rate-enhancing effect of hFVIII-LC is ϳ8-fold, which is comparable with human fulllength FVIII and B-domain deleted FVIII (hFVIII-BDD). The heavy chain (hFVIII-HC) and the light chain lacking the acidic (a3) region (hFVIII-LC⌬a3) have no effect in accelerating VWF proteolysis by ADAMTS13 under the same conditions. Although recombinant hFVIII-HC and hFVIII-LC⌬a3 do not detectably bind immobilized VWF, recombinant hFVIII-LC binds VWF with high affinity (K D , ϳ15 nM). Moreover, ultralarge VWF multimers accumulate in the plasma of fVIII ؊/؊ mice after hydrodynamic challenge but not in those reconstituted with either hFVIII-BDD or hFVIII-LC. These results suggest that the light chain of FVIII, which is not biologically active for clot formation, is sufficient for accelerating proteolytic cleavage of VWF by ADAMTS13 under fluid shear stress and (patho) physiological conditions. Our findings provide novel insight into the molecular mechanism of how FVIII regulates VWF homeostasis.
Proteolytic cleavage of ultra-large von Willebrand factor (VWF) 3 on endothelial cells (1,2) and in flowing blood (3,4) by a plasma metalloprotease, ADAMTS13, is crucial for normal hemostasis. ADAMTS13 cleaves VWF at the specific Tyr 1605 -Met 1606 bond in the central A2 domain (5,6). This cleavage is dramatically accelerated by fluid shear stress (4,6,7) or mild denaturization with urea (5,8) or guanidine (6,9), which alters VWF conformation and exposes binding and cleavage sites. Inability to cleave VWF as a result of a severe deficiency of plasma ADAMTS13 leads to an accumulation of ultra-large VWF on endothelial cells or in blood (10). This triggers spontaneous platelet aggregation and disseminated microvascular thrombosis, characteristic of thrombotic thrombocytopenic purpura (11,12). Moreover, mild to moderate deficiency of plasma ADAMTS13 has been shown to be a risk factor for cardiovascular events such as myocardial infarction (13)(14)(15)(16) and ischemic cerebral stroke (17,18).
In addition to fluid shear stress, we and others have shown that cleavage of soluble multimeric VWF by ADAMTS13 under fluid shear stress is dramatically accelerated by coagulation factor VIII (FVIII) (19), platelets (20,21), and glycoprotein 1b␣ (21,22). FVIII and platelets appear to synergistically accelerate cleavage of VWF by ADAMTS13 under these conditions (21). We have previously shown that the FVIII B-domain is not required, but the a3 in the context of two chain B-domainless FVIII, which binds VWF with high affinity, is required for the rate-enhancing effect on proteolytic cleavage of VWF by ADAMTS13 under shear stress (19).
Here, we show that an isolated light chain of FVIII, which is biologically inactive for clot formation, is sufficient for accelerating proteolytic cleavage of VWF by ADAMTS13 in vitro using a fluid shear-based assay and in vivo using fVIII Ϫ/Ϫ mice expressing FVIII variants via a hydrodynamic approach. This rate-enhancing effect by FVIII light chain also depends on its high affinity binding with VWF; a light chain of FVIII lacking the a3 region and a heavy chain of FVIII, which do not bind VWF detectably, exhibit no effect on ADAMTS13-mediated VWF proteolysis under the same conditions. Our findings may shed more light on the structure-function relationship of FVIII in regulation of the VWF-ADAMTS13 axis, which helps in understanding the clinical heterogeneity of patients with severe hemophilia A.
Serum-free condition medium of stable cell lines expressing recombinant FVIII or variants was collected daily for a total of 2.5-10 liters. Recombinant hFVIII-BDD and cFVIII-BDD were purified and quantified as described previously (19,23,26). Recombinant hFVIII-HC without the C-terminal epitope was kindly provided by Dr. Philip Fay at the University of Rochester, New York. Recombinant hFVIII-LC, hFVIII-LC⌬a3, and hFVIII-HC were purified by SP-Sepharose ion exchange chromatography, followed by a nickel-nitrilotriacetic acid affinity column as described previously (19).
The purity and integrity of the purified recombinant FVIII variants were determined by 10% SDS-polyacrylamide gel with Coomassie Blue staining. All purified recombinant FVIII variants except for hFVIII-HC were quantified by absorbance at 280 nm corrected with light scattering at 320 nm. Partially purified hFVIII-HC (V5-His tagged) was quantitated by Western blotting with monoclonal anti-V5 IgG using a positope TM as a reference (Invitrogen) (25).
Preparation of Plasma VWF and Recombinant ADAMTS13-Human VWF was purified from plasma by cryoprecipitation followed by gel filtration on a Sephacryl-300 column (2.5 ϫ 100 cm) (GE Healthcare) as described previously (21). Recombinant human ADAMTS13 (V5-His tagged) was expressed from stably transfected HEK293 cells and purified by Q-fast flow ion exchange followed by a nickel-affinity chromatography as described previously (7,21).
Proteolytic Cleavage of VWF by ADAMTS13 under Fluid Shear Stress-Purified plasma VWF (37.5 g/ml) was incubated in a PCR tube (Fisher Scientific, Newark, DE) for 10 min with 50 nM recombinant human ADAMTS13 in the absence or in the presence of FVIII variants at various concentrations in 20 mM HEPES, 0.15 M NaCl, 5 mM CaCl 2 , and 1.0 mg/ml BSA, pH 7.5. The reaction mixture (20 l) was subjected to constant vortexing at 2,500 rpm for 10 min at room temperature as described previously (7,19,21). The reaction was quenched by boiling the sample with an equal volume of sample buffer (125 mM Tris, 10% glycerol, 2% SDS, and 0.01% bromphenol blue, pH 6.8) for 5 min. A fraction of sample was fractionated on a 5% Tris-glycine SDS-polyacrylamide gel. Alternatively, samples were denatured at 60°C for 20 min with sample buffer (60 l) (70 mM Tris, pH 6.8, 2.4% SDS, 0.67 M urea, and 4 mM EDTA, 10% glycerol, 0.01% bromphenol blue). Denatured sample (10 l, ϳ90 ng of VWF) was fractionated on a mini-gel containing 1% agarose (Lonza, Rockland, ME). After being transferred onto a nitrocellulose membrane (Bio-Rad), VWF cleavage product or multimers were determined by Western blotting. Membrane was blocked with 1% casein in TBST for 30 min and incubated with rabbit anti-VWF IgG (Dako, Carpinteria, CA) (1:5,000) in 1% casein/TBST, followed by IRDye 800CW-labeled goat anti-rabbit IgG (LI-COR Bioscience, Lincoln, NE) (1:20,000) in the same buffer. When resolved by SDS-PAGE, the cleavage product at 350 K was quantified by densitometry using ImageJ software as described previously (19,21). When resolved by agarose gel electrophoresis, the ratio of the low molecular weight bottom band to the high molecular weight bands was quantified in a similar manner.
Hydrodynamic Injection-The Institutional Animal Care and Use Committee (IACUC) at the University of Pennsylvania and the Children's Hospital of Philadelphia approved the protocol for the mouse study. FVIII-deficient mice (fVIII Ϫ/Ϫ ) in a C57BL6/129 strain with exon-16 deletion were described previously (27). Mice at the age of 6 -8 weeks were injected with 2 ml of saline alone or saline containing 100 g of plasmid DNA (endotoxin-free) via a tail vein within 5 s as described previously (28). 48 h after injection, whole blood (200 l) was collected after tail clip and anti-coagulated with sodium citrate (3.8%). Platelet-poor plasma was obtained after centrifugation at 10,000 rpm for 10 min and stored in small aliquots at Ϫ80°C until use.
Plasma VWF Antigen in fVIII Ϫ/Ϫ Mice before and after Reconstitution with FVIII Variants-Plasma VWF antigen was quantified by an ELISA as described previously (29). A microtiter plate was coated with rabbit anti-vWF IgG (Dako, Carpinteria, CA) (1: 2,000) and blocked with 1% casein in PBS containing 0.05% Tween 20 (TBST). Mouse plasma diluted (1:20 and 1:40) with 0.2% casein in TBST was added and incubated for 1 h. After being washed with PBS, bound VWF was detected by peroxidase-conjugated rabbit anti-VWF IgG (1:2,000) (Dako, Carpinteria, CA). Pooled murine plasma from C57BL6 was used as a reference. A pre-mixed TMB solution was used for color development. The absorbance at 450 nm was determined with a SpectroMax microtiter plate reader (Molecular Device, Sunnyvale, CA).
Statistical Analysis-The difference in means between the control and experimental groups was determined by one-way analysis of variance using Minitab16 software. p values less than 0.05 and 0.01 are considered to be statistically significant and highly significant, respectively.

RESULTS
Biochemical Characterization of Recombinant FVIII Variants-We previously found that full-length FVIII accelerates proteolytic cleavage of VWF by ADAMTS13 under fluid shear stress (19). However, the domain components of FVIII required for the cofactor activity to enhance VWF proteolysis are not fully understood. We therefore prepared various recombinant FVIII variants, including hFVIII-HC, hFVIII-LC, and hFVIII-LC⌬a3, in addition to hFVIII-BDD and cFVIII-BDD (Fig. 1A). All variants except for hFVIII-BDD and cFVIII-BDD contained a V5-His epitope at their C-terminal end to facilitate purifica- hFVIII-HC consists of A1, a1, A2, and a2 domains; hFVIII-LC contains a3, A3, C1, and C2 domains. hFVIII-LC⌬a3 is similar to hFVIII-LC with a deletion of a3 region. B, SDS-polyacrylamide gel and Coomassie Blue staining. hFVIII-BDD was secreted as a heterodimer of 90,000 and 80,000, whereas cFVIII-BDD was secreted as a single chain protein (160,000). hFVIII-HC was partially purified due to limited secretion. The hFVIII-LC was secreted as a single chain of ϳ80,000. C, Western blotting with anti-V5 confirmed the predominant bands (asterisks) revealed on Coomassie Blue-stained gel were in fact the proteins of interest (hFVIII-HC, hFVIII-LC, and hFVIII-LC⌬a3, respectively) with a purified recombinant protein (Positope TM ) as a reference. tion and detection (Fig. 1A). All variants except for hFVIII-HC were purified to homogeneity as demonstrated by SDS-PAGE and Coomassie Blue staining (Fig. 1B). The hFVIII-HC was only partially purified due to low secretion of this variant from stably transfected cells (data not shown). Therefore, a purified preparation of hFVIII-HC (without V5-His) was obtained from Dr. Phillip J. Fay, University of Rochester, School of Medicine and Dentistry, Rochester, NY.
Light Chain of FVIII Accelerates Cleavage of VWF by ADAMTS13 under Fluid Shear Stress-To determine whether the isolated light chain of FVIII, a biologically inactive FVIII variant for clot formation, accelerates VWF proteolysis by ADAMTS13, we incubated a fixed concentration of VWF (150 nM) with ADAMTS13 (50 nM) in the presence of various concentrations of recombinant hFVIII-LC (0, 0.5, 1.0, 2.5, 5.0, and 10 nM) for 10 min under constant vortex at the rotation rate of 2,500 rpm. The proteolytic cleavage of VWF was determined by SDS-PAGE and Western blotting as described under "Experimental Procedures." Recombinant hFVIII-LC increased the formation of the proteolytic cleavage product (ϳ350,000, a dimer of C-terminal fragments) in a concentration-dependent manner (Fig. 2, A and B). The maximal rate-enhancing effect was ϳ8-fold (Fig. 2B). Similar fold of rate-enhancing effect by hFVIII-LC without a V5-His epitope on the cleavage of VWF by ADAMTS13 under the same conditions was observed (data not shown). The concentration of hFVIII-LC achieving 50% of the maximal enhancing effect (C 50 ) was ϳ1.0 nM (Fig. 2B), quite similar to that of full-length FVIII and hFVIII-BDD that we previously reported (19). Addition of EDTA (15 mM) into the reaction completely inhibited proteolytic cleavage of VWF by ADAMTS13 even in the presence of 10 nM hFVIII-LC (Fig. 2, A,  2nd lane, and B). These results demonstrate for the first time that the isolated light chain of FVIII is sufficient for accelerating proteolytic cleavage of VWF by ADAMTS13 under fluid shear stress.

Light Chain of FVIIII Lacking the a3 or the Heavy Chain Has No Effect on Cleavage of VWF by ADAMTS13 under Fluid Shear
Stress-To determine whether the light chain of FVIII or the heavy chain, both lacking the high affinity binding site for VWF, was able to enhance VWF proteolysis by ADAMTS13 under the same conditions, a fixed concentration of human VWF (150 nM) and human ADAMTS13 (50 nM) was incubated with various concentrations of recombinant hFVIII-LC⌬a3 or hFVIII-HC (0, 0.5, 1.0, 2.5, 5.0, and 10 nM) for 10 min under constant vortexing at a rotation rate of 2,500 rpm. The proteolytic cleavage of VWF was determined by SDS-polyacrylamide gel and Western blotting. As shown, recombinant hFVIII-LC⌬a3 (Fig. 2, C and D) or hFVIII-HC (Fig. 2, E and F) at any given concentration (up to 10 nM) did not exhibit a rate-enhancing effect on cleavage of VWF by ADAMTS13 under the same conditions. Recombinant hFVIII-BDD (20 nM) used as a positive control dramatically increased the formation of the proteolytic cleavage product (Fig. 2, 1st lane). These results suggest that the acidic a3 region in the light chain of FVIII is required for accelerating VWF proteolysis by ADAMTS13 under shear stress.
Binding of FVIII Variants to Immobilized VWF-To assess the binding affinity between FVIII variants and VWF, increasing concentrations of recombinant cFVIII-BDD, hFVIII-BDD, hFVIII-LC, hFVIII-HC, and hFVIII-LC⌬a3 (0, 1.56, 3.1, 6.25, 12.5, 25, 50, and 75 nM) were incubated with human VWF immobilized on a microtiter plate (1.0 g/well). The bound FVIII variants were determined by anti-FVIII IgG or anti-V5 IgG (if V5-His tagged) as described under "Experimental Procedures." We showed that cFVIII-BDD (Fig. 3A), hFVIII-BDD (Fig. 3B), and hFVIII-LC (Fig. 3C) bound human VWF in a concentration-dependent manner. The dissociation constants (K D ) for cFVIII-BDD, hFVIII-BDD, and hFVIII-LC binding to immobilized human VWF were ϳ1.1, ϳ1.3, and ϳ15.6 nM, respectively (Table 1). These results indicate that the heavy chain may contribute to the overall binding affinity of FVIII to immobilized VWF. Moreover, the K D values for cFVIII-BDD, hFVIII-BDD, and hFVIII-LC to bind immobilized murine VWF was 1.9, 1.7, and 26 nM, respectively. These results suggest that there is little species difference between VWF and FVIII binding. As predicted, no detectable binding was observed between hFVIII-LC⌬a3 (or hFVIII-HC) and immobilized human VWF and murine VWF under the same conditions (data not shown). These results indicate that the light chain, particularly the a3 region in the light chain of FVIII, contains the major binding site for VWF.
Plasma VWF Multimer Distribution in fVIII Ϫ/Ϫ Mice Expressing FVIII Variants-Although full-length FVIII, B-domainless FVIII variant, and the isolated light chain of FVIII are able to accelerate proteolytic cleavage of VWF by ADAMTS13 under mechanically induced shear stress in vitro, the physiological relevance of such an enhancing effect has never not been determined in vivo. Using a hydrodynamic approach, we were able to reconstitute plasma FVIII in fVIII Ϫ/Ϫ mice with various FVIII variants, including cFVIII-BDD, hFVIII-BDD, hFVIII-LC, and hFVIII-LC⌬a3 at levels between 0.7 and 2.8 g/ml measured 48 h after injection (Table 2). Plasma VWF antigen and multimer distribution in fVIII Ϫ/Ϫ mice were determined after 48 h of post-reconstitution. We showed that although plasma VWF antigen did not change dramatically in different , and hFVIII-LC (C) at various concentrations were incubated with immobilized human (•) and murine (E) VWF (10 g/ml) on a microtiter plate. The bound FVIII variants were detected by peroxidase-conjugated monoclonal anti-FVIII IgG (ESH-8HR) as described under "Experimental Procedures." The specific binding was obtained after subtracting from the total the nonspecific binding in the wells without VWF immobilized but with same concentrations of FVIII variants added. The K D value(s) was obtained by fitting the binding data (n ϭ 3) into a nonlinear binding equation using the SigmaPlot software.

DISCUSSION
In this study, we demonstrate both in vitro and in vivo that an isolated light chain of FVIII, which is biologically inactive for enhancing clot formation, appears to be sufficient for accelerating proteolytic cleavage of VWF by ADAMTS13. The maximal rate enhancing effect with 5 nM of hFVIII-LC was ϳ8-fold, comparable with that of hFVIII-BDD (Fig. 2, A and B). The C 50 is estimated to be ϳ1.0 nM (Fig. 2, A and B), which is within the physiological ranges of FVIII in human plasma. In contrast, a light chain of FVIII lacking the acidic a3 region (hFVIII-LC⌬a3) and hFVIII-HC, both of which do not bind VWF, exhibit no enhancing effect under the same conditions (Fig. 2, C-F). These results indicate that high affinity binding of FVIII to VWF through the acidic a3 region may be critical for accelerating VWF proteolysis.
However, the rate-enhancing effect of FVIII on VWF proteolysis does not appear to be in a linear relationship with its VWF binding affinity. For instance, although hFVIII-LC binds human VWF (K D ϭ 15 nM) and murine VWF (K D ϭ 26 nM) ϳ10 times less than binding of hFVIII-BDD and full-length FVIII to human VWF (K D ϭ 1.3 nM) and murine VWF (K D ϭ 1.9 nM) ( Table 1), the rate-enhancing activity is quite similar to that of hFVIII-BDD (Fig. 2) and full-length FVIII (19).
The biological relevance of FVIII and its variants on regulating VWF proteolysis by ADAMTS13 was further assessed using   . Plasma VWF multimer distribution in fVIII ؊/؊ mice expressing various FVIII variants. A, plasma multimer distribution determined by 1% agarose gel electrophoresis and Western blotting. Representative images of plasma VWF multimers are shown in fVIII Ϫ/Ϫ mice before (unchallenged) and after hydrodynamic injection of saline or saline containing a plasmid encoding cFVIII-BDD, hFVIII-BDD, hFVIII-LC⌬a3, hFVIII-LC, and hFVIII-HCϩLC. B, ratios of high to low molecular weight VWF multimers, which were determined by densitometry using ImageJ software, are shown for each group of mice before (unchallenged) and after hydrodynamic injection of saline or saline containing various FVIII variants. Statistical analysis was performed using one-way analysis of variance with Tuley correction. p values Ͻ 0.05 and 0.01 are considered to be statistically significant and highly significant as compared with the saline control.
a murine model. At the steady state, plasma levels of VWF antigen in fVIII Ϫ/Ϫ mice are increased by ϳ2-fold when compared with wild-type mice in the same genetic background C57BL6/ 129 (data not shown), whereas the ratio of high to low molecular weight VWF multimers was not statistically different between the two groups of mice. Similar increases in plasma levels of VWF antigen and ristocetin-cofactor activity were observed in patients with severe hemophilia A compared with healthy controls (32). These findings suggest that FVIII may be involved in regulating VWF homeostasis under physiological conditions. However, the interpretation of these data may be complicated by the fact that plasma VWF antigen levels differ significantly among various strains of mice or human individuals.
Although hydrodynamic challenge is considered to be a nonphysiological transfection method, it activates endothelial cells and triggers the release of ultra-large VWF from endothelial cells, resulting in an accumulation of ultra-large VWF multimers in plasma of the fVIII Ϫ/Ϫ mice receiving saline alone. However, these ultra-large VWF multimers were not observed in the same mice receiving plasmids encoding cFVIII-BDD, hFVIII-BDD, and hFVIII-LC or hFVIII-LCϩHC ( Fig. 4 and Table 2). These results demonstrate for the first time that reconstitution of functional and nonfunctional FVIII variants restores the distribution of plasma VWF multimers in severe hemophilia A mice.
Consistent with the in vitro data, reconstitution in the fVIII Ϫ/Ϫ mice with a plasmid encoding hFVIII-LC⌬a3 ( Fig. 4 and Table 2) has no effect on plasma VWF multimer distribution. However, the in vivo effect of hFVIII-HC alone remains to be determined because plasma levels of the expressed hFVIII-HC were low, only 1/10 of the plasma levels of other constructs after the hydrodynamic injection (data not shown). The low expression of the heavy chain is due to inefficient secretion of this chain when it is expressed alone (24,26). Interestingly, co-expression of hFVIII-LC dramatically improves the secretion of the heavy chain, thereby synergistically enhancing proteolytic cleavage of VWF by ADAMTS13 in vivo. The efficacy of the co-expressed hFVIII-LC and hFVIII-HC appears to be similar to that of hFVIII-BDD but better than that of hFVIII-LC alone ( Table 2), suggesting that the heavy chain is able to stabilize the light chain to improve the light chain function.
The implication of these findings is not clear. Patients with severe hemophilia A (with FVIII activity Ͻ1%) are heterogeneous in their clinical presentations (31). Further investigation of plasma VWF multimer distribution in correlation with the genetic basis that results in severe FVIII deficiency may shed new light on how FVIII-dependent proteolysis of VWF may play a role in the modification of the clinical phenotype in patients with severe hemophilia A.
In conclusion, our findings provide novel insight into the structure-function relationship of FVIII in the regulation of ADAMTS13-mediated proteolysis in vitro under fluid shear stress and in vivo under (patho) physiological conditions. This may help explain the heterogeneity of plasma VWF multimer distribution and clinical phenotype in patients with hemophilia A.