Collagen-bound von Willebrand Factor Has Reduced Affinity for Factor VIII*

von Willebrand factor (vWf) is a multimeric adhesive glycoprotein that serves as a carrier for factor VIII in plasma. Although each vWf subunit displays a high affinity binding site for factor VIII in vitro, in plasma, only 2% of the vWf sites for factor VIII are occupied. We investigated whether interaction of plasma proteins with vWf or adhesion of vWf to collagen may alter the affinity or availability of factor VIII-binding sites on vWf. When vWf was immobilized on agarose-linked monoclonal antibody, factor VIII bound to vWf with high affinity, and neither the affinity nor binding site availability was influenced by the presence of 50% plasma. Therefore, plasma proteins do not alter the affinity or availability of factor VIII-binding sites. In contrast, when vWf was immobilized on agarose-linked collagen, its affinity for factor VIII was reduced 4-fold, with K D increasing from 0.9 to 3.8 nm. However, one factor VIII-binding site remained available on each vWf subunit. A comparable reduction in affinity for factor VIII was observed when vWf was a constituent of the subendothelial cell matrix and when it was bound to purified type VI collagen. In parallel with the decreased affinity for factor VIII, collagen-bound vWf displayed a 6-fold lower affinity for monoclonal antibody W5–6A, with an epitope composed of residues 78–96 within the factor VIII-binding motif of vWf. We conclude that collagen induces a conformational change within the factor VIII-binding motif of vWf that lowers the affinity for factor VIII.

von Willebrand factor (vWf) 1 is a large multimeric glycoprotein that functions both as an adhesive protein and as a carrier for factor VIII (antihemophilic factor) in plasma. vWf is secreted both from the apical surface of cells into plasma and from the basolateral surface of cells, where it binds to type VI collagen and is incorporated into the vascular matrix (1). As a carrier for factor VIII, plasma vWf functions to prevent factor VIII from binding to phosphatidylserine-containing mem-branes, where it may be degraded by anticoagulant proteases such as activated protein C (2)(3)(4). As an adhesive protein, plasma vWf forms the primary link between exposed extravascular collagen at a site of vascular rupture and platelets, which are sequestered from flowing blood by collagen-bound vWf (5,6). vWf secreted from the basolateral surface of endothelial cells also functions as an adhesive protein. When the vascular matrix is exposed to flowing blood, the vWf-bound to type VI collagen supports adhesion of platelets and development of a thrombus (7). Factor VIII is released from vWf following proteolytic activation by thrombin (8) and is then free to bind to phosphatidylserine-containing membranes, where it may assemble with the serine protease, factor IXa, to form an efficient procoagulant enzyme complex (9) or be degraded by activated protein C (4). Only suspended, plasma vWf is known to bind factor VIII. Thus, we were motivated to determine the consequences that may occur when circulating vWf with bound factor VIII binds to collagen at a site of vascular injury.
vWf, which is synthesized in both endothelial cells and megakaryocytes (1,10), consists of disulfide-bonded monomeric subunits of 260,000 Da with multimers ranging from dimers to 80-mers with masses of up to 20 ϫ 10 6 Da (11,12). The primary amino acid sequence of the mature subunit entails homologous domains arranged in the order DЈ-D3-A1-A2-A3-D4-C1-C2 (13)(14)(15). The subunits are linked by C-terminal disulfide bonds prior to transport from the endoplasmic reticulum to the Golgi apparatus (16,17). Within the Golgi apparatus, the dimers are assembled into high molecular mass multimers under the influence of a large propeptide (18,19). Each vWf subunit contains a factor VIII-binding motif with functional residues clustered in the DЈ domain (20,21), whereas the primary collagenbinding motif of vWf is localized in the A3 domain (22). The D domains of vWf are remarkable for a high cysteine content (10), with the vast majority of these residues forming intradomain disulfide bonds. We have recently observed that formation of the intersubunit disulfide bonds between D3 domains of dimers, linking them to form multimers, increases the affinity of vWf subunits for factor VIII 6-fold (23). This suggests that high affinity factor VIII-binding sites reside on the internal subunits of vWf multimers rather than on the two terminating subunits and that the DЈ domain is subject to conformational change despite its eight-intradomain disulfide bond restraints (17).
Factor VIII is a trace plasma glycoprotein with a molecular mass of 280 kDa. It functions as cofactor for factor IXa in the highly efficient factor X-activating enzyme complex (9). Within this complex, factor VIII binds both to a platelet-binding site (24,25) and to the enzyme factor IXa (26). The enzymatic product of this complex, activated factor X, activates prothrombin to thrombin (27). Hemophilia, an X-linked bleeding disorder in which factor VIII or IX levels are decreased in plasma, illustrates the importance of factor VIII for normal hemostasis.
Because von Willebrand factor protects factor VIII in plasma, factor VIII levels are decreased in patients with reduced or absent circulating vWf and in patients with mutant vWf with impaired factor VIII binding (28).
In plasma, vWf multimers carry only one factor VIII molecule/50 vWf subunits. Binding studies conducted in vitro have yielded conflicting data for the number of available factor VIIIbinding sites on plasma vWf multimers. A recent investigation from this laboratory (23) demonstrated a 1:1 stoichiometry of binding sites/vWf subunit, which is in agreement with the data of Lollar and Parker (29) and Vlot et al. (30). In contrast, other studies reported different factor VIII/vWf ratios, such as 1:4, 1:10, 1:50, and 1:70 (30 -33). Several investigators reported that factor VIII forms a high affinity complex with vWf, with dissociation constants range from 0.2 to 0.94 nM (30,(32)(33)(34)(35). A study by Leyte et al. (33) suggested that the high affinity interaction is restricted to 1-2% of the available sites. We were motivated to determine whether any proteins in plasma reduce the affinity or availability of factor VIII-binding sites on vWf multimers, possibly explaining the discrepancies.
In vitro, plasma vWf does not bind to its platelet receptor, glycoprotein Ib, in the absence of nonphysiologic cofactors such as ristocetin. However, after vWf binds to collagen or is adsorbed to a plastic surface, platelets bind to the immobilized vWf via the platelet glycoprotein Ib-IX receptor (6). The altered affinity of vWf for platelet glycoprotein Ib-IX is presumed due to a collagen-induced conformational change. We hypothesized that the apparent reduction in factor VIII-binding sites on vWf bound to microtiter wells may reflect a conformational change, induced by binding to collagen, that affects the affinity for factor VIII as well as for platelet glycoprotein Ib-IX.
Proteins-vWf was purified from cryoprecipitate and quantified with an enzyme-linked immunosorbent assay as described previously (23). Recombinant vWf lacking the propeptide, therefore present only as C-terminal dimers, was expressed and purified as described previously (19). Recombinant human factor VIII was labeled with fluorescein-5maleimide as described previously (3). Anti-von Willebrand factor monoclonal antibody 16.9.2 was labeled with fluorescein isothiocyanate as described (23).
Preparation of Microspheres-Anti-vWf mAb 13.7.9 was linked to cyanogen bromide-activated Superose 12 microspheres as described previously (23). For calf skin collagen immobilization, cross-linked agarose microspheres 34 m in diameter (Superose 12; 500 l) were equilibrated with 1 M CO 3 Na 2 /HNaCO 3 buffer, pH 11. CNBr (500 g) dissolved in 500 l of acetonitrile was added to the microspheres and incubated for 10 min on an ice bath while swirling. If necessary, 4 N NaOH was added to the mixture to maintain the pH between 10.5 and 11. The microspheres were extensively washed with distilled H 2 O and 100 mM carbonate/bicarbonate buffer, pH 9.2. Subsequently, they were washed with cold 0.1 M sodium citrate, pH 6.5. The slurry of activated microspheres was added to a tube containing 12 ml of calf skin collagen (1-1.5 mg/ml) diluted in 0.1 M sodium citrate and incubated at 4°C for 48 -72 h on an inverting mixer. The microspheres were then washed with cold 0.1 M sodium citrate, pH 6.5. Nonspecific active sites were blocked by incubating the coated microspheres with 10 volumes of 100 mM Tris, pH 8, for 4 h to overnight at 4°C with gentle mixing. Micro-spheres were kept at 4°C until used, but no longer than 10 days. The microspheres were counted using a Coulter ZM counter.
Collagen deposition was confirmed by indirect immunofluorescence microscopy; 10 l of microspheres were incubated with anti-type I collagen mAb (Sigma) in a final volume of 500 l on a inverting mixer. After 15 min of incubation at room temperature, the microspheres were washed with Trizma (Tris base)-buffered saline (TBS; 150 mM NaCl and 20 mM Trizma, pH 7.8), and the fluorescein-labeled anti-mouse antibody (Dako Corp.) was added and allowed to incubate for 15 min. Control microspheres were gelatin-coated microspheres (Sigma) and CNBr-activated, uncoated Superose 12 microspheres. Collagen-positive microspheres had a distinctive, bright fluorescent rim as seen by fluorescence microscopy, whereas this property was not observed either on the gelatin microspheres or on the uncoated ones.
Calf skin collagen was adsorbed to 20-m diameter polystyrene microspheres after extensive washing to remove surfactant. Microspheres (1 ml) were sedimented at 3000 rpm ϫ 1 min using a tabletop microcentrifuge. The pellet was washed with H 2 O (1 ml), and the procedure was repeated six times. The microspheres were resuspended and incubated with 1 ml of 600 g/ml collagen, pH 6.5, in 50 mM Tris and 150 mM NaCl overnight at 4°C while rocking. Unbound collagen was removed by washing once with 10 volumes of TBS. Nonspecific sites were blocked with whole milk following a 1-h incubation at room temperature. Microspheres were counted with a Coulter ZM counter. They were flashfrozen in liquid nitrogen and kept at Ϫ80°C until used.
Type VI collagen purified from human placenta (80 g/ml) was dialyzed into 0.5 M Na 2 CO 3 /NaHCO 3 with the pH adjusted to 9.5. 1 ml of CNBr-activated Superose 12 microspheres was added to a 1-ml suspension of type VI collagen, and the reaction was allowed to proceed for 16 h at 4°C with gentle rocking. Unreacted active sites were blocked with 0.1 M ⑀-aminocaproic acid, pH 7. Deposition of type VI collagen was confirmed by staining with mAb VI-I, a secondary biotinylated goat anti-mouse polyclonal antibody, followed by staining with Alexa-conjugated streptavidin with fluorescence detection of microspheres containing and lacking type VI collagen by flow cytometry. For factor VIII binding experiments, nonspecific binding of factor VIII to type VI collagen was reduced by incubating the microspheres with skim milk for 60 min at room temperature, followed by washing with TBS.
Endothelial cell-conditioned microspheres were prepared as follows. Following activation with CNBr as described above, Superose microspheres were incubated with 2% (w/v) gelatin in 50 mM Na 2 HPO 4 , pH 7.8, for 16 h at 4°C. Gelatin-coated microspheres were stored in 20% ethanol at 4°C to minimize the likelihood of microbial contamination. Prior to incubation with human umbilical vein endothelial cells, prepared by standard techniques, microspheres were washed twice with endothelial cell growth medium. Human umbilical vein endothelial cells were grown to confluence on 100-mm gelatin-coated tissue culture dishes. Cells from one dish (ϳ2 ϫ 10 6 cells) were removed by trypsin treatment, sedimented by gentle centrifugation, and resuspended in 5 ml of growth medium. The resuspended cells were combined with a 1-ml suspension of ϳ2.5 ϫ 10 6 gelatin-conjugated Superose 12 microspheres in endothelial cell growth medium and placed on a 100-mm bacteriological Petri dish on a rocker at 37°C. Following 3 days of growth, the microspheres were harvested and washed by centrifugal sedimentation with Hanks' buffered saline. The endothelial cells were removed by treatment with 0.1 M NH 4 OH as described previously (36). Briefly, microspheres were resuspended in 5 ml of 0.1 M NH 4 OH, incubated for 15 min at room temperature, sedimented, and resuspended in 0.1 M NH 4 OH for 15 min. Microspheres were then washed three times with Hanks' buffered saline and finally resuspended in TBS. The conditioned microspheres were stored at 4°C. All experiments were performed within 2 weeks of endothelial cell stripping.
Equilibrium Binding Assay-Binding of fluorescein-labeled factor VIII to vWf immobilized on mAb 13.7.9 or collagen-Superose 12 was measured as described previously (23). The same assay was employed to detect binding of fluorescein-labeled mAb W5-6A to vWf.
Factor VIII binding to collagen-bound vWf on polystyrene microspheres was conducted as following. vWf (6.6 g) was incubated with 3 ϫ 10 6 collagen-coated polystyrene microspheres in a final volume of 720 l in TBS containing 0.01% Tween 80 and BSA (TBST-BSA) for 45 min at room temperature while shaking in a Vortex mixer. Free vWf was separated from collagen-bound vWf on microspheres by sedimentation onto a density gradient. vWf/collagen-coated microspheres were layered onto 5.5 ml of TBST-BSA/Histopaque (1.015 g/liter) over a sublayer of 1 ml of TBST-BSA/Optiprep (1.221 g/liter). The preparation was centrifuged for 30 min at 400 ϫ g before removal of microspheres from the interface between the two density layers. Microspheres were washed with 1 ml of TBST-BSA, and aliquots containing 30,000 microspheres were incubated with increasing concentrations of fluorescein-labeled factor VIII or mAb 16.9.2, as described above. After 15 min, vWf-bound fluorescein-labeled factor VIII or fluorescein-labeled mAb 13.7.9 was measured by flow cytometry as described previously (23).
Kinetic Binding Assays-vWf (117 ng) was immobilized onto 200,000 mAb-or collagen-coated Superose 12 microspheres by incubating for 45 min at room temperature with continuous mixing on a Vortex mixer. The association of fluorescein-labeled factor VIII with vWf was monitored after the addition of either 1 or 10 nM fluorescein-labeled factor VIII to suspended vWf-coated mAb 13.7.9-Superose microspheres or 5 or 10 nM fluorescein-labeled factor VIII to vWf-coated collagen-Superose microspheres. At timed intervals, 50-l aliquots were diluted to 500 l, and bound factor VIII was measured immediately using flow cytometry. The average elapsed time between sample dilution and completion of data acquisition was ϳ20 s.
For dissociation experiments, 10 nM fluorescein-labeled factor VIII was allowed to equilibrate for 15 min with vWf bound to collagen-Superose or mAb 13.7.9-Superose. The dissociation of factor VIII from bound vWf was measured following the addition of 100 nM unlabeled factor VIII. At timed intervals, 50-l aliquots were taken, and the fluorescence/microsphere was monitored as described above.
Data Analysis-Control experiments were performed to measure fluorescence due to nonspecific binding of fluorescein-labeled factor VIII to microspheres lacking vWf. This background fluorescence, ϳ20% of total fluorescence at 1 nM factor VIII and 50% of total fluorescence at 16 nM factor VIII, was subtracted from the total fluorescence to obtain the specific fluorescence. The figures display specific fluorescence, and Table I reports values derived from specific fluorescence. Equilibrium binding data were analyzed using the equation where F is the measured fluorescence/Superose bead, F 0 is the fluorescence/Superose bead in the absence of fluorescein-labeled factor VIII, F max is the maximum fluorescence with saturating concentrations of factor VIII, K D is the dissociation constant for factor VIII binding to vWf, and VIII f is the molar concentration of fluoresceinlabeled factor VIII. We assumed that the fraction of total factor VIII bound could be neglected (the nominal concentration of factor VIIIbinding sites was 0.1 nM, assuming one binding site/vWf subunit). The variables K D and F max were determined using nonlinear least-squares analysis with FitAll Version 4.0 software (MTR Software, Toronto). The stoichiometry between factor VIII-binding sites and vWf subunits was determined using fluorescein-labeled mAb 13.7.9 as described under "Results." Kinetic association data were fitted according to first-order binding kinetics using nonlinear least-squares software (FitAll Version 4.0), assuming that the concentration of free fluorescein-labeled factor VIII did not change significantly in the course of the reaction (see above). Kinetic dissociation data were fitted according to exponential decay with a non-zero plateau using FitAll Version 4.0.

Plasma Proteins and Factor VIII-vWf
Binding-Purified suspended vWf displays one high affinity binding site for factor VIII on each vWf subunit; yet in plasma, only 1-2% of these sites are actually occupied by factor VIII. We asked whether plasma contains protein(s) that inhibit binding of factor VIII to some fraction of vWf subunits. To address this question, we evaluated binding of factor VIII to immobilized vWf in the presence and absence of plasma. Purified vWf was immobilized on the non-inhibitory mAb 13.7.9 linked to 34-m diameter cross-linked agarose microspheres (Superose). We have recently shown that vWf immobilized in this way displays one high affinity binding site/subunit, equivalent to suspended vWf (23). Varying concentrations of fluorescein-labeled factor VIII were incubated with vWf microspheres and either buffer or a vWf-deficient plasma/buffer mixture at a 1:1 ratio. Fluoresceinlabeled factor VIII bound saturably and with high affinity to immobilized vWf in the presence or absence of plasma (Fig. 1). The equivalent shapes and plateaus of the two binding isotherms indicated that plasma does not contain proteins that substantially inhibit binding of factor VIII to vWf.
Calf Skin Collagen and Factor VIII-vWf Binding-We hypothesized that vWf has decreased affinity for factor VIII when bound to collagen. To test this hypothesis, calf skin collagen was adsorbed to 20-m diameter polystyrene microspheres, and the microspheres were incubated with suspended vWf. Saturable binding of vWf to collagen was detected by flow cytometry with fluorescein-labeled mAb 13.7.9. Half-maximal binding occurred at a vWf concentration of 1 g/ml. We performed kinetic association and dissociation experiments to determine whether collagen-bound vWf would remain stably bound over the time period necessary for factor VIII binding. The results indicated that ϳ40% of the bound material dissociated over 20 min, but that following the initial dissociation, at least 80% of bound vWf remained stably bound for Ͼ2 h. Thus, all equilibrium and kinetic factor VIII binding experiments were initiated at least 20 min after washing free vWf from the microspheres and completed within 2 h.
Increasing concentrations of fluorescein-labeled factor VIII were added to the vWf/collagen-coated microspheres, and binding of factor VIII was measured by flow cytometry. The affinity of collagen-bound vWf for factor VIII (K D ϭ 5 nM) was decreased compared with the affinity of antibody-bound vWf (K D ϭ 0.9 nM) ( Table I). To quantify the number of vWf subunits on the collagen-coated microspheres, binding of fluorescein-labeled anti-vWf mAb 16.9.2 to microspheres was measured (data not shown). Comparison of the maximum fluorescence/ microsphere from vWf-bound mAb 16.9.2 with the maximum fluorescence from vWf-bound factor VIII allowed calculation of the apparent stoichiometry between bound factor VIII and vWf subunits. The results indicated that one vWf subunit bound one factor VIII molecule whether the vWf was bound to mAb 13.7.9 or to collagen (Table I).
We were concerned that the decreased affinity of collagenbound vWf for factor VIII may not be due solely to an effect of collagen on vWf, but may also reflect an effect of the polystyrene matrix to which collagen was adsorbed. Therefore, we prepared cyanogen bromide-activated Superose microspheres and allowed them to couple to calf skin collagen, which contains type I and III collagens. Deposition of collagen on the Superose microspheres was confirmed by indirect immunofluorescence FIG. 1. Effect of plasma on factor VIII binding to von Willebrand factor. vWf was immobilized on mAb 13.7.9 coupled to Superose microspheres as described under "Experimental Procedures." Increasing amounts of fluorescein-labeled factor VIII were incubated with the immobilized vWf for 15 min in buffer (q) or 50% plasma (f) prior to evaluation by flow cytometry. Plasma and the buffer control each contained 10 mM sodium citrate and 12.5 mM CaCl 2 with 1 unit/ml heparin, 5 mM benzamidine, and 5 mM ⑀-aminocaproic acid to inhibit plasma proteases. The results displayed are from a single experiment representative of two such experiments. microscopy using an anti-type I collagen mAb. Collagen attachment was also evident by inspection as the Superose microspheres sedimented to a less dense pellet than control microspheres and had a fuzzy-appearing margin when observed by phase-contrast microscopy. vWf bound to collagen-Superose microspheres was detected by fluorescein-labeled mAb 16.9.2. Binding of fluorescein-labeled factor VIII to vWf on collagen-Superose microspheres was evaluated in comparison with vWf on mAb 13.7.9-Superose microspheres (Fig. 2). The results confirmed that the affinity of collagen-bound vWf for factor VIII was decreased ϳ4-fold compared with that of mAb-immobilized vWf (Fig. 2 and Table I). The stoichiometry of factor VIII to vWf subunits was 1:1 on collagen-Superose and mAb 13.7.9-Superose microspheres.
Subendothelial Matrix and Type VI Collagen-The experiments described above provide a model of plasma vWf that adheres to collagen at a site of vascular injury. In vivo vWf is also secreted from the basolateral surface of endothelial cells and is then found associated with type VI collagen fibrils in the subendothelial matrix. We wished to know if vWf deposited in the subendothelial matrix also bound factor VIII with reduced affinity. Accordingly, we cultured endothelial cells on gelatincoated Superose microspheres. After 72 h, the endothelial cells were removed, and microspheres were stained with fluorescein-labeled mAb 16.9.2. The results indicated that vWf had been deposited. Factor VIII bound saturably to vWf, with a K D of 4.1 nM (Fig. 3A and Table I), comparable to vWf bound to calf skin collagen. There were 3.5-fold more factor VIII-binding sites accessible than mAb 16.9.2 epitopes, indicating that access to the mAb 16.9.2 epitope is hindered by matrix proteins to a greater degree than access to factor VIII-binding sites. These results indicate that vWf in the subendothelial matrix has decreased affinity for factor VIII.
Smooth muscle cells efficiently produce type VI collagen, which is incorporated into the vascular matrix, whereas endothelial cells are inefficient producers of type VI collagen (37). Therefore, it is likely that the results in Fig. 3A do not reflect the affinity of factor VIII for vWf bound to type VI collagen. To explicitly examine the factor VIII-binding properties of vWf bound only to type VI collagen, we linked purified human type VI collagen to Superose microspheres. vWf was incubated with these microspheres, and deposition of vWf was confirmed with fluorescein-labeled mAb 16.9.2. Factor VIII bound saturably to type VI collagen-bound vWf, with an average K D of 8.3 nM (Fig.  3B and Table I). This dissociation constant indicates that vWf has reduced affinity for factor VIII when it is bound to type VI collagen, comparable to the reduced affinity when bound to calf skin collagen. In contrast to calf skin collagen-bound vWf, type VI collagen-bound vWf displayed only one factor VIII-binding site for every three epitopes of mAb 16.9.2. Thus, binding of vWf to type VI collagen differs from binding to calf skin collagen in that some factor VIII-binding sites are made inaccessible by the interaction between vWf and collagen.
Collagen and Factor VIII-vWf Binding Kinetics-To determine whether the decreased affinity of factor VIII for collagenbound vWf was due to a decreased association rate or to an increased dissociation rate, we performed kinetic binding experiments. vWf was immobilized on mAb 13.7.9-or collagen-Superose microspheres. Fluorescent factor VIII was added to mAb 13.7.9-immobilized or collagen-immobilized vWf. At timed intervals, 50-l aliquots were removed from the reaction mixture, and bound factor VIII was measured by flow cytometry (Fig. 4A). The averaged association rates were 0.07 nM Ϫ1 min Ϫ1 for mAb-immobilized vWf and 0.05 nM Ϫ1 min Ϫ1 for collagenimmobilized vWf (Table I), indicating that the rate of association of factor VIII with vWf is not significantly affected after vWf binds to collagen. The dissociation of factor VIII from collagen-bound and mAb 13.7.9-bound vWf was monitored after the addition of 100 nM unlabeled factor VIII to the fluores-FIG. 2. Decreased affinity of collagen-bound vWf for factor VIII. vWf was allowed to adhere to collagen-Superose and mAb 13.7.9-Superose microspheres as described under "Experimental Procedures." Increasing concentrations of fluorescein-labeled factor VIII were incubated for 45 min with vWf bound to mAb 13.7.9 (q) or collagen (E). The quantity of bound factor VIII/bead was measured by flow cytometry. The fitted curves correspond to K D values of 4.3 and 1.3 nM for collagenbound and mAb 13.7.9-bound vWf, respectively.
The ratio of factor VIII-binding sites to mAb 16.9.2 epitopes was 3.5 for vWf in the subendothelial matrix, suggesting that access to the mAb 16.9.2 epitope was impaired by matrix proteins. The ratio of factor VIII-binding sites to mAb W5-6A-binding sites remained 1:1. cent factor VIII-vWf equilibrium complex (Fig. 4B). The dissociation of factor VIII from mAb 13.7.9-immobilized vWf was relatively slow, comparable to the previously reported value (34). By comparison, the dissociation from collagen-bound vWf was 7-fold faster. The averaged separation constants (k off ) were 0.07 and 0.77 min Ϫ1 for mAb 13.7.9-immobilized and collagenimmobilized vWf, respectively (Fig. 4B). The precision of experiments in which the dissociation of factor VIII from collagenbound vWf was measured is limited because the 30-s increments at which bound factor VIII was measured allowed most of the dissociation to occur within only the first two readings. However, the data are sufficient to indicate that the lower affinity of factor VIII for collagen-bound vWf relates primarily to a more rapid dissociation rate.
Collagen and Affinity of vWf for a Monoclonal Antibody-The decreased affinity of collagen-bound vWf for factor VIII is most readily explained by a collagen-induced conformational change in vWf. If the conformation of the factor VIII-binding motif is altered, then the affinity of vWf for other molecules that bind this motif may be changed. We asked whether the affinity of mAb W5-6A, which competes with factor VIII for its binding site on vWf (21), may also be sensitive to conformational changes that occur within the vWf site. Binding of fluoresceinlabeled mAb W5-6A to vWf was measured on collagen-Superose microspheres versus mAb 13.7.9-Superose microspheres (Fig. 5). The results indicated that the affinity of mAb W5-6A was decreased ϳ6-fold when vWf was bound to collagen. This reduction in affinity parallels the reduced affinity of factor VIII for collagen-bound vWf (Fig. 2), suggesting that the same change in vWf alters binding affinity for both proteins. We also measured the affinity of mAb W5-6A for vWf deposited on Superose microspheres by endothelial cells and for vWf bound to purified type VI collagen (data not shown). In both cases, the affinity was reduced at least 6-fold, with a K D of ϳ35 nM, comparable to the affinity for vWf bound to calf skin collagen.
Since binding of mAb W5-6A is apparently sensitive to the collagen-induced change in the factor VIII-binding motif of vWf, we asked whether it may also be sensitive to another conformational change. We have recently observed that di- FIG. 3. Decreased affinity of subendothelial matrix vWf and type VI collagen-bound vWf for factor VIII. A, vWf and other matrix proteins were deposited by human umbilical vein endothelial cells on Superose microspheres as described under "Experimental Procedures." Increasing concentrations of fluorescein-labeled factor VIII were incubated 45 min with endothelial cell-conditioned Superose microspheres, and the quantity of bound factor VIII/bead was measured by flow cytometry. The fitted curve corresponds to a K D of 5.1 nM for this representative experiment. B, varying concentrations of fluoresceinlabeled factor VIII were incubated with vWf bound to purified type VI collagen coupled to Superose microspheres. After 15 min, bound factor VIII was evaluated by flow cytometry. The fitted curve corresponds to a K D of 6.4 nM for this representative experiment.

FIG. 4. Accelerated dissociation of factor VIII from collagenbound vWf.
A, representative experiments comparing the association of fluorescein-labeled factor VIII with vWf on collagen-Superose (E) versus mAb 13.7.9-Superose (q). 10 nM fluorescein-labeled factor VIII was added to suspended vWf-coated microspheres. At timed intervals, 50-l aliquots were removed, and the quantity of bound factor VIII/ microsphere was measured immediately by flow cytometry. The fitted curves correspond to k on values of 0.084 and 0.044 nM Ϫ1 min Ϫ1 for collagen-Superose and mAb 13.7.9-Superose, respectively. B, representative experiments comparing the dissociation of factor VIII from vWf on collagen-Superose (E) versus mAb 13.7.9-Superose (q) after the addition of excess unlabeled factor VIII. After 10 nM fluorescein-labeled factor VIII was allowed to equilibrate with immobilized vWf for 15 min, 100 nM unlabeled factor VIII was added, and the residual bound vWf was measured at timed intervals. The fitted curves displayed correspond to k off values of 1.3 and 0.1 min Ϫ1 for collagen-Superose and mAb 13.7.9-Superose, respectively. meric vWf, lacking N-terminal intersubunit disulfide bonds, has a 6-fold decreased affinity for factor VIII (23). We have hypothesized that the gain in affinity exhibited by dimeric vWf upon formation of the N-terminal intersubunit disulfide bond reflects a conformational change in the factor VIII-binding motif. We therefore compared binding of fluorescein-labeled mAb W5-6A to plasma vWf versus recombinant dimeric vWf lacking N-terminal intersubunit disulfide bonds (Fig. 6). Both forms of vWf were immobilized on mAb 13.7.9-Superose microspheres and incubated with fluorescein-labeled mAb W5-6A. The results indicated that mAb W5-6A recognized plasma vWf and dimeric vWf with affinities that were within 2-fold of each other. We interpret these results as indicating that mAb W5-6A is not sensitive to the conformational change in vWf that results from formation of N-terminal intersubunit disulfide bonds. These data suggest that the N-terminal portion of vWf subunits is susceptible to two conformational changes. One change is induced by formation of intersubunit disulfide bonds, increasing affinity for factor VIII; and the other is induced by collagen binding, decreasing affinity for factor VIII. DISCUSSION The results in this report indicate that plasma proteins do not alter the affinity of factor VIII for vWf or the stoichiometry with which factor VIII binds to vWf subunits. Therefore, the ϳ1:50 ratio between factor VIII and vWf subunits in circulating blood (31,38) cannot be explained by plasma proteins that modify the availability of vWf subunits for factor VIII binding or the affinity of those sites for factor VIII. However, our results indicate that when vWf adheres to collagen, the affinity of vWf subunits for factor VIII is reduced. The decreased affinity of vWf is due to an enhanced off-rate of factor VIII, which is increased ϳ10-fold. The dissociation half-time is 1 min for collagen-bound vWf versus 10 min for suspended vWf. The more rapid release suggests the possibility that, when vWf binds to collagen at the site of vascular injury, factor VIII released by this mechanism may contribute to initiation of the coagulation cascade. It also suggests that collagen-bound vWf, which is a ubiquitous constituent of the subendothelial vascular matrix, has a reduced capacity to compete for factor VIII compared with plasma vWf.
We have found that the number of factor VIII-binding sites/ vWf subunit is 1:1. These findings agree with those of Lollar and Parker (29), who studied factor VIII-vWf association by velocity sedimentation; with those of Vlot et al. (30), who studied VIII/vWf stoichiometry either on colloidal gold or using gel-filtration chromatography on Sepharose; and with those of Saenko and Scandella (35), who measured factor VIII binding to suspended vWf in a competition assay. In contrast, other studies have concluded that the number of factor VIII-binding sites/vWf subunit is Ͻ1:1. From binding competition experiments, Nesheim et al. (32) inferred that the factor VIII/vWf stoichiometry is ϳ1:10 at saturation. In studies in which vWf was adsorbed to polystyrene microtiter wells, the factor VIII/ vWf subunit stoichiometries were 1:40 (39), 1:50 (30), and 1:70 (33). The detected factor VIII binding on microtiter plates is greatly diminished if the rate of dissociation of factor VIII from vWf is increased so that bound factor VIII is removed in the wash steps. Therefore, it remains possible that interaction of vWf with microtiter plate plastic causes a reduced affinity of most vWf subunits for factor VIII, analogous to the effect of collagen. Binding to the subunits with reduced affinity may not have been detected because of the experimental design. In our preliminary experiments in which vWf was immobilized on collagen adsorbed to polystyrene microspheres, we detected reduced affinity of vWf subunits comparable to the reduced affinity on collagen-Superose microspheres, suggesting that a polystyrene matrix does not necessarily inhibit the affinity of vWf subunits for factor VIII to a greater degree than collagen alone.
The results in this report indicate that the affinity of vWf subunits for factor VIII is reduced 4 -6-fold when vWf is bound to collagen. The reduced affinity results from an enhanced dissociation rate with an essentially unchanged association rate. By comparing the effects of collagen adsorbed to solid polystyrene microspheres versus collagen covalently coupled to FIG. 5. Anti-vWf mAb W5-6A, with an epitope in the factor VIII-binding site, has reduced affinity for collagen-bound vWf. Collagen-Superose-immobilized vWf (E) and mAb 13.7.9-Superose-immobilized vWf (q) were prepared as described under "Experimental Procedures." Increasing concentrations of fluorescein-labeled mAb W5-6A were incubated with vWf for 15 min, and binding was monitored by flow cytometry. The dissociation constants corresponding to the fitted curve were 27 nM for collagen-bound vWf and 4.7 nM for mAb 13.7.9-bound vWf, respectively. The fluorescence data for the two experiments shown are normalized to the asymptotic saturation point of the fitted curves for visual comparison. Data are representative of three experiments for collagen-bound vWf and six experiments for mAb 13.7.9-bound vWf. porous agarose microspheres, we have confirmed that the decreased affinity is related to an effect of collagen rather than an effect of the matrix to which the collagen was attached. In these experiments, we have measured an average affinity of vWf subunits for factor VIII. Our results do not indicate what fraction of vWf subunits actually bind to collagen versus reside as constituents of vWf multimers tethered to the collagen surface. Therefore, it is possible that vWf subunits that bind to collagen have a lower affinity for factor VIII and that non-interacting subunits retain higher affinity. Alternatively, all vWf subunits may have affinity reduced to the same degree if a conformational change induced by collagen propagates through intersubunit disulfide bonds from collagen-binding subunits to nonbinding subunits.
We believe that our results may be best explained by a conformational change in the factor VIII-binding DЈ domain of vWf. We argue that the alternative explanation, steric hindrance of the factor VIII-binding site by collagen, is less likely, first because the collagen-binding motif in the A3 domain of vWf is 922 amino acids distant from the DЈ domain in linear sequence, and second because steric hindrance would probably ablate factor VIII-binding sites rather than reduce affinity 4 -6-fold. The probability of a collagen-induced conformational change is further increased by the effect of collagen on the affinity of vWf subunits for mAb W5-6A, which recognizes part of the factor VIII-binding motif in the DЈ domain of vWf (21). The affinity of vWf for mAb W5-6A was reduced ϳ6-fold, suggesting that the epitope of mAb W5-6A, including amino acid residues 78 -96 within the DЈ domain of vWf, is a local region within the factor VIII-binding motif that is subject to conformational change.
We have recently observed that the terminal subunits of vWf multimers, which lack N-terminal intersubunit disulfide bonds, have 6-fold reduced affinity for factor VIII (23). The similarity of this reduction in affinity to the effect produced by collagen led us to ask whether collagen induces a relaxed conformation of the factor VIII-binding motif equivalent to nondisulfide-linked vWf N termini. To address this question, we asked whether the affinity of mAb W5-6A is altered to the same degree on vWf dimers lacking N-terminal intersubunit disulfide bonds as on collagen-bound vWf. Since the affinity of mAb W5-6A is not reduced on dimeric vWf, but is reduced 6-fold on collagen-bound vWf, we speculate that the factor VIII-binding motif, located primarily in the DЈ domain of vWf subunits, has three conformations. In this scheme, the conformation is relaxed, with a low affinity for factor VIII, prior to formation of intersubunit disulfide bonds. Following formation of N-terminal intersubunit disulfide bonds (multimer formation), the conformation of the DЈ domain is strained into a high affinity factor VIII-binding conformation. This high affinity conformation is altered when vWf binds to collagen. Our results do not elucidate the mechanism through which binding of vWf to collagen via the A3 domain may transmit a conformational change to the DЈ domain, nor do they indicate whether a conformational change may be propagated from one collagenbinding vWf subunit to another that may not bind vWf.
The affinity of vWf for platelet glycoprotein Ib-IX is increased when vWf is exposed to shear stress (40,41) or binds to polycations such as ristocetin and polymerizing fibrin (42) or collagen (6). The most likely explanation for the increased affinity upon collagen binding is a conformational change in the platelet glycoprotein Ib-IX-binding A1 domain (43). The relationship of this conformational change to the affinity of vWf subunits for factor VIII is currently unknown. We are attracted to the possibility that a single conformational change increases the affinity for platelet glycoprotein Ib-IX and decreases the affinity for factor VIII.
Because vWf lines the entire vascular tree, our results have implications for the possible partitioning of factor VIII between plasma vWf and subendothelial matrix vWf. Our results indicate that factor VIII should preferentially partition to plasma vWf (as it apparently does) because of higher affinity for suspended vWf subunits. However, they leave open the possibility of subendothelial matrix vWf serving as a lower affinity reservoir for factor VIII, which rapidly equilibrates with the plasma factor VIII compartment. In support of this possibility, our experiments with endothelial cell-deposited vWf indicate an affinity for factor VIII equivalent to the affinity of collagenbound vWf for factor VIII. However, the ability of factor VIII to traverse the endothelial cell barrier between plasma and the subendothelial matrix is a critical unknown factor that may limit the capacity of matrix vWf to serve as a factor VIII reservoir.