Blocking von Willebrand factor free thiols inhibits binding to collagen under high and pathological shear stress

Von Willebrand factor (VWF) contains a number of free thiols, the majority of which are located in its C‐domains, and these have been shown to alter VWF function, However, the impact of free thiols on function following acute exposure of VWF to collagen under high and pathological shear stress has not been determined.


| INTRODUC TI ON
Von Willebrand factor (VWF) is a large multimeric plasma glycoprotein essential to normal hemostasis. First, VWF mediates platelet capture to the damaged vessel wall under high shear stress conditions and, second, it is the carrier molecule for factor VIII ( Figure 1). 1,2 Compared with other mammalian proteins, VWF has a uniquely high cysteine content. 3 It was previously understood that all the cysteine residues present in VWF took part in inter-or intra-molecular disulphide bonds. [4][5][6][7] However, several recent studies have demonstrated that each VWF molecule contains a number of unpaired cysteine residues or free thiols. [8][9][10][11] The majority of these free thiols are located in the C-domains of VWF and have been suggested to impact significantly on VWF function. For example, shearing of VWF in a plate and cone viscometer at high shear stress was shown to enhance VWF platelet binding, but reduced the free thiol content. 8 When sheared in the presence of the thiol-blocking agent N-ethylmaleimide (NEM), platelet capture function was lost, suggesting a crucial role for free thiols in mediating VWF function. 8 Furthermore, it was demonstrated that when soluble VWF was perfused over stimulated human umbilical vein endothelial cells, formation of the VWF string network was blocked by the addition of NEM indicating a role for VWF thiols in the association of soluble VWF with endothelial cell anchored VWF. 11 Subsequently, Ganderton et al 9 reported that VWF lateral self-association involved the Cys2431 and Cys2453 residues in the C3 domain, hypothesizing that disulfide exchange between VWF molecules could occur; however, this was not shown in full-length VWF or under shear stress. Moreover, we previously demonstrated that mutation of the predicted unpaired cysteine residues abolished secretion of VWF, suggesting that proper disulphide bond formation within the endoplasmic reticulum is essential for VWF production. 10 It is therefore unclear how and when certain disulfide bonds are broken or for what purpose. Interestingly, it is now known that the C-terminal domains of sequential VWF monomers form a stem structure. 12,13 Formation of the stem is controlled by pH and calcium-dependent interactions between D4 domains, with the D4 domain thought to act as a force sensor enabling the molecule to elongate in response to changes in shear stress. 14,15 Because the majority of the free thiols are present in the adjacent C-domains that form the VWF stem, it is conceivable that they also play a role in mediating stem formation.
Although blocking VWF free thiols impacts VWF function, these previous studies have not addressed their effects on VWF binding to collagen under shear stress where the protein will be exposed to collagen for a few milliseconds or extremely pathological shear rates such as those found in arterial stenosis where wall shear rates can exceed 100 000 s −1 . [16][17][18] In the present study, we have investigated the role of free thiols in VWF using a physiological system in which the key measure is the ability of VWF to mediate platelet capture to collagen under flow. 19 Our data demonstrate that under conditions of high pathological shear stress, the free thiols in the C-domains of VWF play an unexpected role in promoting collagen binding and subsequent platelet capture.

| Purification, labelling and characterization of VWF
Plasma derived VWF was isolated from Voncento (CSL Behring) using gel filtration through a Sephacryl-400 gel filtration column,

Essentials
• Von Willebrand factor contains a number of free-thiols.
• We investigated how blocking these free-thiols impacts on VWF function under high shear stress.
• Blockade of VWF thiols ablated collagen binding at pathological shear rates.
• VWF free thiols are a novel regulator of collagen binding.
F I G U R E 1 Domain organization of VWF. Each full-length VWF monomer comprises a series of repeating domains arranged as depicted. The cysteine knot domain (CK) contains the dimerization site, whereas multimers are formed through disulphide bonding between adjacent D′ and D3 domains and this process requires the propeptide (D1D2 domains). The A1 domain contains the binding site for glycoprotein Ibα, whereas the A2 domain contains the ADAMTS13 cleavage site and the major collagen binding site is located within the A3 domain. The C4 domain contains an RGD sequence that binds to glycoprotein IIbIIIa. The majority of the free thiols that occur in VWF have been previously mapped to the C domains. A monomeric VWF protein spanning the D′ to A3 domains is used in this study and thus lacks the D4 and C domains as previously described. 20 The first eluted fractions containing pu-

| VWF-mediated platelet capture to collagen
Assessment of VWF-mediated platelet capture was performed essentially as described. 21 In brief, Ibidi VI 0.1 flow chambers or custom-made flow channels generated as previously described 24 were coated with 100 µg/mL human type III collagen (Southern

| Direct binding of VWF to collagen under shear
Flow channels coated with collagen were perfused with 10 µg/ mL VWF in 20 mmol/L Tris-HCl, pH 7.4 at designated shear rates.
Following perfusion, the channels were washed with 20 mmol/L Tris-

| Atomic force microscopy
For the atomic force microscopy (AFM) studies, human type I collagen at 1 mg/mL (Sigma-Aldrich) was diluted to a final concentration of 0.1 mg/mL in phosphate buffered saline and was covalently attached to an AFM cantilever (MLCT: Bruker Nano) via an established crosslinking protocol using a heterobifunctional PEG linker. 25 Con-VWF or NEM-VWF was also attached to the silanized glass coverslips (Nanocs) using the same protocol. The AFM force measurements were performed on an apparatus designed for operation in the force spectroscopy mode. [26][27][28] Using a piezoelectric translator, the collagen-functionalized cantilever was lowered onto the glass coverslip with the attached VWF sample until binding between the collagen and the VWF sample occurred.
The binding interaction was then determined from the deflection of the cantilever via a position-sensitive two-segment photodiode.
To calibrate the cantilever (320 µm long × 22 µm wide triangle), the spring constant at the tip was characterized via thermally induced fluctuations. The spring constants (13 ± 3 pN/nm) of the calibrated cantilevers agreed with the values specified by the manufacturer. Except for the adhesion specificity test, the contact time and indentation force between the cantilever and the sample were minimized to obtain measurements of the unitary VWFcollagen unbinding force. An adhesion frequency of ~30% in the force measurements ensured that there was a >83% probability that the adhesion event was mediated by a single VWF-collagen bond. AFM measurements were collected at cantilever retraction speeds ranging from 0.4 to 10 μm/s to achieve the desired loading rate (~400-10 000 pN/s). All measurements were conducted at 25℃ in TBS buffer.

| Molecular dynamics
The starting structure of the VWF C4 domain (PDB code 6FWN) was energy minimized and solvated in Tip3p waters, using a dodecahedron box of dimensions 12.9 nm × 6.9 nm × 6.9 nm. 29 The charge of the system was kept neutral by addition of four sodium ions. All simulations were performed in the NPT ensemble, with a timestep of 2 fs. Temperature was coupled using the V-rescale method 30 (300 K, with coupling constant of 0.1 ps), and pressure used the Berendsen algorithm (reference pressure of 1 bar, coupling constant of 1 ps). 31 We used periodic boundary conditions and the LINCS algorithm for constraints. 32 Electrostatic interactions were accounted for by the Particle-mesh Ewald method. 33 All simulations used the force field AMBER99SB-ILDN. 34 Initially every form of the VWF C4 domain was pulled by applying opposite forces of 1000 kJ/ mol (along the x-axis direction) on the N-and C-termini. We then extracted equally spaced structures along this pulling pathway, with a spacing of 0.1 nm. Each of these structures was simulated for 5 ns using a biasing umbrella potential used to later calculate the potential of mean force (PMF). 35 The final energy profiles of the PMF calculation were reconstructed using the weighted histogram analysis method, discarding the first nanosecond of each simulation as an equilibration phase. Finally, the thiol exchange was modelled by replacing one of the disulphide forming residues with the free cysteine, and applying opposite forces to the termini (1000 kJ/mol along the x-axis direction) until the maximum extension of the domain was reached. Simulations and analyses were performed with the GROMACS molecular simulations package. 36

| Statistical analysis
Statistical analysis was performed using Prism7 software, using either a standard one-way ANOVA with multiple comparisons or Student t test.

| Blockade of free thiols reduces VWF-mediated platelet capture in a shear-dependent manner
To investigate the role of free thiols in VWF-mediated platelet capture, VWF was treated with NEM to block unpaired cysteines.

Consistent with previous observations, VWF could incorporate
MPB demonstrating the presence of free thiols, and MPB binding was lost following NEM treatment ( Figure S1). 8,37 Additionally, NEM labelling did not alter the ability of VWF to interact with collagen, GPIbα, or GPIIbIIIa under static conditions and did not alter its multimeric profile ( Figure S1). Subsequently, NEM-labelled VWF was added to plasma free blood and perfused over collagen at a range of shear rates for 5 minutes. After 5 minutes' perfusion at 500 s −1 and 1000 s −1 , there was no observable difference in the ability of NEM-VWF to mediate platelet capture compared to control VWF at all-time points. When the shear rate was increased to 1500 s −1 , NEM-VWF became visibly less able to mediate platelet capture although this failed to reach statistical significance. At 3000 s −1 , there was a significant reduction in the ability of NEM-VWF to capture F I G U R E 2 Blocking VWF free thiols with NEM reduces VWF-mediated platelets capture to collagen in a shear-dependent manner. Ibidi VI 0.1 flow slides coated with 100 µg/mL human type III collagen were perfused with plasma free blood supplemented with control or NEMtreated VWF at indicated shear rates. Platelets were rendered fluorescent with DiOC6 and platelet surface coverage was monitored in realtime. A, Platelet surface coverage after 5 min perfusion at 500 s -1 to 5000 s -1 for control VWF (black bars) and NEM-VWF (gray bars). Data are mean ± SD (n = 4). B, Real-time images captured after 5 min perfusion at 1500, 3000, and 5000 s -1 . C-E, Time course of platelet surface coverage between 0 and 300 s at 1500 s -1 (C), 3000 s -1 (D), and 5000 s -1 (E) for control VWF (solid lines) and NEM-VWF (dashed lines). Data are mean ± SD (n = 4). **P < .005, ***P < .0005, ****P < .00005   Figure S2).

| Blockade of VWF free thiols reduces platelet capture at stenotic sites
Because blocking VWF free thiols with NEM reduced VWF mediated platelet capture in a shear dependent manner, we hypothesized that at sites of altered shear stress such as stenoses where shear rates can exceed physiological levels, platelet capture would be similarly affected. Microfluidic slides designed to mimic a constricted vessel were coated with collagen and perfused with plasma free blood supplemented with either con-or NEM-VWF and platelet capture recorded at different sections of the channel corresponding to increasing shear rates ( Figure 3A). In keeping with the previous results, at regions of physiological shear rates (~1500 s −1 ) a similar extent of platelet surface coverage was observed. However, as the shear rate increased to ~4000 s −1 , there was a marked reduction in platelet capture. At the constricted region of the channel where the wall shear rate was highest (~25 000 s −1 ), con-VWF mediated the formation of dense platelet rich thrombi; however, no platelet capture was observed with NEM-VWF ( Figure 3B).

| Free thiols in the C-domains of VWF are required for collagen binding under high shear stress
Because NEM-VWF was able to effectively mediate platelet capture to collagen at 500 to 1500 s −1 and when coated to collagen surfaces under static conditions; the effect of blocking VWF free thiols was unlikely to be attributable to altered platelet-VWF binding. To further confirm this, flow slides were coated with VWF-D′A3 protein treated with NEM and then perfused with washed red blood cells and platelets. NEM treatment of the D′A3 fragment did not alter its ability to capture platelets ( Figure 5A). Moreover, when the D′A3 was bound to collagen under static conditions, NEM treatment again did not alter platelet capture ( Figure 5B) Fluor 488 sulfodichlorophenol ester did not alter multimeric content ( Figure S3A).

F I G U R E 4
Blockade of VWF free thiols has minimal effect on collagen binding under static conditions. Ibidi VI 0.1 flow slides coated with 100 µg/mL human type III collagen were incubated with 10 µg/mL control or NEM-VWF under static conditions for 60 min at room temperature or perfused with 10 µg/mL control or NEM-VWF diluted in 20 mmol/L Tris, pH 7.4 for 5 min at 3000 or 5000 s −1 . Channels were then subsequently perfused with plasma free blood at the stated shear rates and capture of DiOC6-labelled platelets recorded after 5 min perfusion (A). Platelet surface coverage after 5 min was determined using ImageJ (coat static, black bars) (coat-flow, gray bars) (B). Data are mean ± SD (n = 4)

| Blocking VWF free thiols reduces the strength and life-time of the vwf-collagen bond
To directly detect alterations in the VWF-collagen binding interaction resulting from loss of free thiols, we used AFM to probe the collagen-VWF bond. The specificity of the VWF-collagen interaction was first validated by the significant decrease in adhesion frequencies when either collagen or VWF was absent and furthermore by significantly decreased adhesion to the W1745C collagen binding mutation ( Figure 6A). We have previously demonstrated that this von Willebrand disease causing mutation ablates binding to collagen. 39 Figure 6B shows typical pulling traces without (upper) or with (lower) the rupture force of the collagen-VWF bond. The unbinding force of the collagen-VWF bond is derived from the force jump that accompanies the unbinding event. The unbinding forces were sorted and plotted as a dynamic force spectrum ( Figure 6C); a plot of most probable unbinding force ( Figure S4) as a function of the loading rate, and is a quantitation of the mechanical property of a ligand-receptor bond. 40 The unbinding force of con-VWFcollagen bond increased linearly with the logarithm of the loading rate, ranging from 37 to 95 pN over a loading rate of 400 to 6400 pN/s, respectively ( Figure 6C). Compared with the con-VWF-collagen interaction, NEM-VWF showed significantly reduced unbinding forces under the six loading rates tested, indicating binding defects in NEM-VWF ( Figure 6C). Fitting the Bell-Evans model to these data revealed that the con-VWF-collagen bond has a dissociate rate (ko) F I G U R E 6 AFM measurement of VWF-collagen interaction. A, The adhesion frequency of the AFM measurements for different interacting pairs. Contact force, contact time, and retraction speed for all the interacting AFM tips and surfaces were set at 200 pN, 0.15 s, and 3.7 mm/s, respectively. The error bar is standard deviation with n = 5 (AFM cantilever) in each case. *Indicates P < .001 compared with any control groups. The P value was calculated by the Student t test. B, The upper panel shows two sample AFM pulling traces of the VWFcollagen interaction. The first (upper) trace had no interaction and the second (lower) trace shows the rupture force of a single conVWFcollagen complex. Fu is the unbinding force. ks is the system spring constant and was derived from the slope of the force-displacement trace. The lower panel illustrates the four stages of stretching and rupturing a single ligand-receptor complex using the AFM. C, The dynamic force spectra (ie, the plot of most probable unbinding force, Fu, as a function of loading rate, rF) of the VWF-collagen interactions. Unbinding forces at different loading rates were plotted as histograms ( Figure S3). The peak of each histogram was plotted against the loading rate; uncertainty in force is shown as half bin-width. Some error bars are within the symbol. Equations to extract dissociate rate k 0 are shown.  Figure 6C). Figure 6D presents the force-dependent lifetime of the VWF-collagen interactions. Compared with con-VWF, NEM-VWF yielded an over 2-fold shorter lifetime at no force, and the lifetime of the NEM-VWF-collagen bond decreased more rapidly than con-VWF with increasing force. At 100 pN, the lifetime of NEM-VWFcollagen bond becomes 9-fold shorter than the conVWF-collagen bond. Because higher shear rate is associated with greater forces acting on the VWF-collagen bond, the AFM data can explain the effect of NEM treatment on VWF-collagen binding.

| Free thiols in the C-domains of VWF may affect domain flexibility
Our data suggest that under shear stress the bond between NEM-VWF and collagen either does not form or cannot respond normally to shear stress. Because this appears not to involve the formation of covalent interactions, we explored the possibility that free thiols increase the flexibility of the VWF C-domains. Using the recently described structure of the VWF C4 domain we performed molecular simulation on either the intact domain, or with the C2499-C2533 bond reduced, the C2528-C2570 bond reduced or the C2549-C2571 bond reduced, by applying a force to pull the domain from either terminus. The reduced cysteines have been previously demonstrated to be unpaired in some VWF molecules. We performed a PMF calculation to gain a quantitative insight into the differences in the energy required to stretch the domain in each case ( Figure 7B). In the base state, the termini are 3.8 nm apart and during extension, more extension force is required as the distance between them becomes larger.
Interestingly, the maximum extension reached with the nonreduced C4 domain is 7.7 nm, ( Figure 7C); however, when the C2499-C2533 bond is reduced, less force is required to separate the termini and the domain can be further stretched to 15.5 nm ( Figure 7D). These  work is required to investigate this. Following blockade with NEM, the initial free thiols are not present and therefore spontaneous disulphide rearrangement cannot occur following collagen binding. As such, the C-terminus of VWF is less flexible and is unable to withstand high shear forces causing the VWF-collagen bond to rupture. Under lower shear rates, the forces generated on the collagen bound molecule are not significant enough to break the bond and free thiol exchange is not required. Further work is now required to fully understand this process.

| D ISCUSS I ON
Nonetheless, although the exact mechanism requires defining, the effect of blocking VWF free thiols has the most significant effect at very high and, importantly, pathological shear rates and thus represents a way to modulate VWF function specifically under pathological conditions without affecting the function of VWF in normal hemostasis. In conclusion, our data demonstrate a previously unrecognized role for VWF free thiols in mediating collagen binding at high shear rates and targeting VWF thiols may be a novel and safe antithrombotic strategy.

T.A.J. McKinnon was supported by a British Heart Foundation Basic
Science Intermediate Fellowship Grant (FS/11/3/28632). We are grateful for support from Imperial College BRC.

CO N FLI C T O F I NTE R E S T
The authors state they have no conflict of interest. and Thomas A.J. McKinnon designed the study, performed experiments, analyzed data, and wrote the paper.