Factor VIII binding affects the mechanical unraveling of the A2 domain of von Willebrand factor

Summary

Background.

Proteolytic cleavage of von Willebrand factor (VWF) by ADAMTS13 is crucial for normal hemostasis. Our previous studies demonstrate that binding of coagulation factor VIII (or FVIII) to VWF enhances the proteolytic cleavage of VWF by ADAMTS13 under shear.

Objectives.

Present study aims to determine the mechanism underlying FVIII-mediated enhancing effect on VWF proteolysis by ADAMTS13 under force.

Methods.

Single molecular force spectroscopy, atomic force microscopy, and surface plasmon resonance are all employed.

Results.

Using the single molecule force spectroscopy, we show that an addition of FVIII (~5 nM) to D’D3 or D’D3A1 does not significantly alter force-induced unfolding of these fragments; however, an addition of FVIII at the same concentration to D’D3A1A2 eliminates its long unfolding event at ~40 nm, suggesting that binding of FVIII to D’D3 and/or A2 may result in force-induced conformational changes in A2 domain. Atomic force spectroscopy further demonstrates the direct binding between FVIII and D’D3 (or A2) with an intrinsic 2D off-rate (k₀) of 0.02 ± 0.01 s⁻¹ (or 0.3 ± 0.1 s⁻¹). The direct binding interaction between FVIII and A2 is further confirmed with the surface plasmon resonance assay, with a dissociation constant (K_D) of ~0.2 μM; no binding is detected between FVIII and A1 under the same conditions.

Conclusions.

Our results suggest that binding of FVIII to D’D3 and/or A2 may alter the mechanical property in the central A2 domain. The findings provide novel insight into the molecular mechanism underlying FVIII-dependent regulation of VWF proteolysis by ADAMTS13 under mechanical force.

Introduction

Von Willebrand factor (VWF), a large multimeric plasma protein synthesized and released from endothelial cells, plays an important role in normal hemostasis.[1, 2] Coagulation factor VIII (FVIII) is a key clotting factor that binds to VWF with high affinity through its interaction with the D’D3 domains of VWF.[3, 4] Binding of FVIII to VWF results in prolongation of VWF half-life in circulation,[5] modulates the interaction between its A1 interaction and platelet glycoprotein 1b (GP1b),[6] and accelerates the proteolytic cleavage of VWF by ADAMTS13 under shear in vitro.[7, 8] and in vivo.[9]

Our previous study has also demonstrated that FVIII fails to enhance the cleavage of several VWF variants with type 2N mutations,[10] found in patients with von Willebrand disease. Mutations in the D’D3 domain of VWF result in a diminished binding affinity of type 2N VWF towards FVIII.[11] This suggests that the ability of FVIII to enhance VWF proteolysis by ADAMTS13 under shear is mediated primarily through the direct interactions between FVIII and VWF-D’D3 domains.

The present study sought to determine the effects of FVIII on mechanical unraveling of the central A2 domain of VWF using single-molecule force spectroscopy; additionally, the direct interaction between FVIII and VWF-A2 is determined under various conditions. The results of the study may shed new light on mechanism underlying FVIII-dependent regulation of VWF function under pathophysiological conditions.

Methods

Materials:

Recombinant D’D3, D’D3A1, and D’D3A1A2 fragments were constructed, expressed in E. coli, and affinity-purified to homogeneity using Ni-affinity chromatography as previously described.[12, 13] Recombinant A1 or A2 domain of VWF were expressed and purified from human embryonic kidney (HEK) 293T cells.[14, 15] All recombinant VWF-fragments were tagged in the N-terminus with an Avi-His and the C-terminus with a Spy-tag for purification and pulling experiments. Biotinylation of the D’D3, D’D3A1, and D’D3A1A2 fragments occurred in vivo during protein synthesis in E. coli, while biotinylation of the A1 or A2 fragment was performed in vitro using a biotin-labeling kit (Avidity, Aurora, CO). Recombinant B-domainless FVIII (FVIII-SQ) was expressed and purified from baby hamster kidney (BHK) cells as previously described.[7] SpyCatcher was kindly provided by Dr. Howarth at University of Oxford, London, England.[16]

Optical tweezer experiments:

Carboxyl-polystyrene beads of 2.0-μm diameters (Spherotech, Lake Forest, IL) were covalently coupled with streptavidin (Invitrogen) as previously described.[15, 17] Briefly, streptavidin-coated beads were incubated with a VWF domain fragment. SpyCatcher was coupled to biotin-DNA handle, and then incubated with streptavidin beads. For pulling experiments, one bead with an Avi-His-SpyTag VWF fragment on its surface was fixed by the micropipette, while the other bead coupled with SpyCatcher-DNA handle was trapped and controlled by the optical tweezers. The force measurement was performed at force-ramp mode, where the stretching force increases linearly with the pulling distance. The force-extension data were fitted to the WLC model [18] to determine its mechanical properties.

Atomic force microscopy:

The cantilever was coated with FVIII, while the dish substrate glass surface was coated with D’D3, A1 and A2, respectively, using a polyethylene glycol crosslinker.[19] Five different pulling speeds with a reasonable range were performed to get five different force loading rates. The unbinding force- loading rate was determined by fitting the data to the Bell–Evans model for quantitation of dissociation rates.[20]

Surface plasmon resonance:

The affinities of FVIII for various VWF fragments were determined by surface plasmon resonance (SPR) using a Biacore T200 (GE Healthcare, Piscataway, NJ, USA) at 25 °C as described previously.[21, 22] Biotinylated D’D3, D’D3A1A2, A2, and BSA were immobilized to a SA-sensor chip (GE Healthcare) via biotin-streptavidin interaction. FVIII (0–160 nm) diluted in HBS-EP buffer (HEPES 10 mM, NaCl 150 mM, EDTA 3mM, 0.005% Tween-20) was flowed through at 30 μL min⁻¹ for 300 seconds. FVIII and VWF domain complex dissociation was observed for 600 seconds, before the surface was regenerated with 10 mm of glycine at pH 2.0 for 30 seconds, followed by a final 30-second buffer re‐equilibration. Equilibrium dissociation constants (K_D) were determined by fitting all binding curves into a 1:1 binding model (Langmuir isotherm) using BIA evaluation Software (GE Healthcare).

Results

Binding of FVIII to D’D3A1 does not alter force-induced conformational changes.

All recombinant VWF fragments (except for A2 fragment) as shown in Fig. 1A were expressed in E. coli, refolded in vitro, and affinity-purified to homogeneity (Fig. 1B). When pulling with an optical tweezer as illustrated in Fig. 2A, various approach-retract force distance curves were generated. As shown, no unfolding event was detected by pulling the D’D3 up to ~65 pN (Fig. 2B-a). However, an extensional event occurred in ~40% of the tethers when pulling the D’D3A1 in the absence of FVIII, with the most probable extension distance of 12.8 ± 0.2 nm (mean ± SEM) (Fig. 2B-b & Fig. 3A). Addition of FVIII (~5 nM) to D’D3 (not shown) or D’D3A1 did not significantly alter the force-induced extension profile (13.7 ± 0.4 nm) (Fig. 2B-c & Fig. 3A), suggesting that binding of FVIII to the D’D3 does not significantly alter force-induced conformational changes of the D’D3A1 fragment.

Fig. 1. Schematic representation of VWF constructs and purified recombinant VWF fragments.

A. A full-length VWF and various VWF fragments with N-terminal AviTag and C-terminal SpyTag are shown. An arrow above the A2 domain indicates ADAMTS13 cleavage site. B. SDS-polyacrylamide gel with Coomassie blue staining demonstrates various recombinant VWF fragments expressed and purified. Here, I, II, III, and IV on the right side of the gel image indicate A2 (~30 kDa), D’D3 (~50 kDa) D’D3A1 (~86 kDa), and D’D3A2 (~106 kDa), respectively.

Fig. 2. Schematic representation and pulling profiles of various VWF fragments using the optical tweezers.

A. A typical optical tweezer set-up where a streptavidin bead on the left and SpyCatcher held by a DNA handle that is attached to a bead on the right through biotin-streptavidin interaction. B. The force-extension curves generated by pulling at 200 nm/s of the D’D3 in the absence of FVIII (a), the D’D3A1 fragment in the absence (b) or presence (c) of FVIII, and the D’D3A1A2 fragment in the absence (d) or presence (e) of FVIII, as well as the A2 domain in the absence (f) or presence of (g) of FVIII. Arrows indicate the evidence of force- induced conformational changes. Double-headed arrows (inset) indicate end-to-end extension associated with protein unfolding. Note: A2 was not pulled all the way to 65 pN in these experiments in order generate more data at the expected force extension range.

Fig. 3. Pulling profiles of various VWF fragments in the absence or presence of FVIII using optical tweezers.

The histograms of the unfolding extension of pulling D’D3A1 (A), D’D3A1A2 (B), and A2 (C) in the absence (top graph) or presence (bottom graphs) of 5 nM FVIII at 200 nm/s. (D) The relationship between unfolding force (pN) and unfolding extension (nm) of pulling the A2 fragment in the absence (black) or presence (red) of 5 nM or 125 nM FVIII. The data are the fits to the WLC model (dashed lines). Histograms of unfolding extension (inset, as an example) determined the peak extension. Horizontal and vertical error bars are one standard deviation for force and half width of the half bin width for extension, respectively.

Binding of FVIII to D’D3A1A2 and/or A2 domain of VWF results in significant alteration of force-induced conformational changes in the A2 domain. When pulling D’D3A1A2 in the absence (Fig. 2B-d) or in the presence of FVIII (~5 nM) (Fig. 2B-e), significant force-induced conformational changes occurred, indicated by the abrupt force drop and end-to-end extension. There were two unfolding events detected in one approach-retraction cycle in the absence of FVIII (Fig. 3B). These peaks at ~18 nm and ~40 nm likely represent a short extension in A1 domain and a long extension in the A2 domain ¹², respectively. In the presence of FVIII (~5 nM), only one peak at ~20 nm was detected with the most probable unfolding extension at 24.5 ± 0.4 nm (Fig. 3B). These results suggest that binding of FVIII to the D’D3 may alter inter-domain interactions, resulting in force-induced conformational changes in the A2 domain of VWF; alternatively, FVIII may directly interact with and alter the conformation of VWF-A2.

To determine if FVIII binds A2 directly, an isolated VWF-A2 fragment expressed from HEK293 cells, was used for the pulling experiment. When pulling the A2 in the absence of FVIII, an unfolding event occurred (Fig. 2B-f), with the probable extension distance of 31 ± 1 nm (mean ± SEM) (Fig. 3C), consistent with that reported previously ^12,14. However, when FVIII (~5 nM) was added, an unfolding event of the VWF-A2 occurred (Fig. 2B-g), with the most probable unfolding extension reduced to 16.0 ± 0.4 nm (mean ± SEM) (Fig. 3C). Additional illustration and interpretation how force affects pulling profile of VWF-A2 with or without FVIII are shown in Suppl. Fig. 1.

In order to determine the number of amino acids included in the initially folded structure, the most probable unfolding extensions at different unfolding forces were obtained from pulling experiments under varying pulling speeds, and the resultant force-extension curve fitted to the WLC model (Fig. 3D), which yielded a contour length L_C for the initially folded structure. By fitting the data into a worm-like chain (WLC) model, we observed the VWF-A2 unfolding contour length of 27 ± 1 nm (mean ± SEM) in the presence of FVIII (Fig. 3D, red line), which was significantly shorter than that (60 ± 2 nm) in the absence of FVIII (Fig. 3D, black line). The contour length reflects the difference between a completely unfolded structure and a folded structure. These results demonstrate for the first time that FVIII may bind VWF-A2 domain directly, and alter the structure of A2. A potential model to explain how such a change may occur is shown in Suppl. Fig. 1.

In addition, three control experiments were performed to further validate our VWF-A2 unfolding study. First, we found that human serum albumin (5 nM) has no effect on the unfolding force and extension of VWF-A2 (Fig. 3D, black curve), indicating that the decrease of VWF-A2 unfolding contour length is indeed due to FVIII binding. Second, to determine if adding 5 nM of FVIII is sufficient, we added 125 nM of FVIII (i.e., a 25-fold increase). We found that addition of 125 nM FVIII did not affect both unfolding forces and extensions profile. This indicates that the 5 nM of FVIII may be sufficient to saturate single-molecular A2 in the pulling study. Lastly, a control experiment with addition of ADAMTS13 and EDTA further confirms the identity of A2 unfolding signal in the D’D3A1A2 unfolding study (Suppl. Fig. 2). It is known that under EDTA, ADAMTS13 can bind unfolded A2 with high affinity without cleaving A2. This is because the proteolytic activity of ADAMTS13 is inhibited when all the Ca²⁺ is chelated by EDTA.[23] Shown in Suppl. Fig. 2A, in the presence of ADAMTS13 and EDTA, the A2 unfolding signal generated from pulling the A2-alone construct occurred only at the very first pulling trace, and A2 unfolding signal disappeared in all the subsequent pulling cycles. This indicates that the binding of ADAMTS13 to unfolded A2 inhibits A2 refolding. Suppl. Fig. 2B shows after adding ADAMTS13 and EDTA, the frequency of observing A2 unfolding signal drastically decrease. Similarly, after we added ADAMTS13 and EDTA to D’D3A1A2 fragment, A2 unfolding only happened in the first pulling curve of a newly fetched tether (Suppl. Fig. 2C), whereas the short (A1) unfolding remained in all pulling cycles. Comparison of unfolding extension histograms suggests that without ADAMTS13, a bimodal distribution is evident, with the short and long unfolding events stemming presumably from A1 and A2 unfolding, respectively. Under ADAMTS13 and EDTA, the long A2 unfolding peak disappeared (Suppl. Fig. 2D). Overall, these results support the identity of A2 in the long-extension unfolding peak in the D’D3A1A2 experiment.

Binding affinity between FVIII to VWF-A2.

Using atomic force microscopy and surface plasmon resonance, we were able to further demonstrate the direct binding interactions between FVIII and VWF-A2 and assess the binding affinity. As shown in Fig. 4A, there was very low background adhesion (<4%) when one of the binding partners was absent in our assay system. The adhesion frequency increased to ≥30% when FVIII was brought into contact with either D’D3 or A2. Again, no adhesion activity was detected between FVIII and A1 under the same conditions. Upon fitting the data to the Bell–Evans model ¹⁷, we obtained an intrinsic off-rate (k₀) of 0.02 ± 0.01 s⁻¹ (mean ± SEM) between FVIII and D’D3 with an energy barrier position parameter (Δx^‡) of 7.3 ± 0.3 Å (Fig. 4B, red line) and an intrinsic off-rate of 0.3 ± 0.1 s⁻¹ (mean ± SEM) between FVIII and A2 with an energy barrier position of 3.3 ± 0.2 Å (Fig. 4B, black line). Surface plasmon resonance demonstrated the direct binding interaction between FVIII and A2 with an apparent dissociation constant (K_D) of ~0.2 μM, 67-fold lower affinity than that between FVIII and D’D3 (K_D, 2.9 nM) (Fig. 5). The AFM and SPR detect two different types of affinities. In AFM, both ligand and receptor are immobilized on surfaces (i.e., AFM tip and dish substrate), and protein-protein interactions occur on the interface of the two surfaces. Therefore, what AFM detects is a 2-dimentional (2D) affinity. In contrast, in SPR, the interactions are taken place in solution (3D) with at least one interacting molecular species in the fluid phase. The 2D reaction kinetics are oftentimes different from 3D kinetics. For the same interaction, the dissociate rate detected from a 2D assay is typically faster than that of a 3D assay.[24, 25] Consistently, in this study, the k_off values detected by AFM for FVIII-D’D3 and FVIII-A2 interactions are 0.02 ± 0.01 s⁻¹ and 0.3 ± 0.1 s⁻¹, respectively, which are approximately 10-fold faster than the dissociate rates detected by SPR (0.0017 s⁻¹ for FVIII-D’D3 and 0.042 s⁻¹ for FVIII-A2 interactions). Nonetheless, both methods are consistent that the dissociation of FVIII-D’D3 complex is approximately 20 times slower than that of the FVIII-A2 complex. These results suggest that FVIII binding to the D’D3 domain may be required for the A2 conformational changes in the D’D3A1A2 fragment.

Fig. 4. Binding of FVIII to VWF-A2 fragment.

A. The adhesion frequency (%) between FVIII and various VWF fragments using atomic force microscopy. BSA was used as a negative control. B. The relationship between the most probable unbinding force determined from force histograms (inset, as an example) (pN) and loading rate (pN/s) for binding of FVIII to either D’D3 (red) or A2 (black). The error bars are the half bin width. The solid lines are the linear fits of the data to the Bell-Evans model.

Fig. 5. Surface plasmon resonance detects binding of FVIII to various VWF fragments.

The kinetic binding of FVIII (0–160 nM) to D’D3 (A), D’D3A1A2 (B), and A2 (C), respectively, immobilized on a biosensor chip via the biotin-streptavidin interaction. The representative sensorgrams are shown. The equilibrium dissociation constants (K_D) (s) were determined using the BIA evaluation software (1:1 Langmuir model).

Discussion

In this study, we demonstrate by a single molecule technique that FVIII binding to the D’D3 domain and/or the A2 domain of VWF results in dramatic conformational changes under mechanical force in the central A2 domain. Our findings for the first time may provide mechanistic insight into how FVIII binding to VWF accelerates its force-induced proteolysis by ADAMTS13.

Crystal structure of FVIII-D’D3 complexes demonstrates that FVIII binds not only to the N-terminal TIL’ region of D’ domain but also extends across one side of D3 domain.[4] Our results show that FVIII can also bind an isolated VWF-A2 domain with an affinity comparable to that of binding to the D’D3 domain under force, but ~67-fold less than that between FVIII and D’D3 without fluidic shear. These findings suggest that FVIII may cover more surface of VWF than we have originally thought, particularly under mechanical shear force. It remains to be determined whether binding of FVIII to either D’D3 or A2 or both domains is necessary for the cofactor activity of FVIII enhancing proteolytic cleavage of VWF by ADAMTS13 under arterial shear.

Negative-stain electron microscopy and hydrogen-deuterium exchange demonstrate that the C1 domain of FVIII contains the major binding site for binding VWF-D’D3 domain,[26] although the acidic (a3) region, the A3 domain, and the C2 domains of FVIII may also be involved in the FVIII-VWF interaction.[26] Consistent with these findings, binding of an isolated light chain of FVIII to VWF appears to be sufficient for enhancing the proteolytic cleavage of multimeric VWF by ADAMTS13 under physiological shear conditions [9]. Interestingly, binding of platelets or soluble glycoprotein 1b to the A1 domain of VWF and FVIII to the D’D3 domain of VWF also synergistically enhances the proteolytic cleavage of VWF by ADAMTS13 under shear.[8] These results underscore the importance of altering domain-domain interactions for modulation of force-induced proteolysis by ADAMTS13 metalloprotease.

Using single molecule spectroscopy, AFM, and SPR, we are able to detect a moderate-to-high-affinity binding interaction between FVIII and A2 domain. Interestingly, binding of FVIII to A2 domain also results in the force-induced conformational changes of VWF-A2. It is, however, not clear whether such conformational changes affect the accessibility of the scissor bond (Tyr¹⁶⁰⁵-Met¹⁶⁰⁶), which is buried under the β-sheet and further stabilized by calcium ion.[14, 27]. FVIII apparently fails to enhance the proteolysis of a type 2N VWF by ADAMTS13 under shear.[10] The mutations in the D’D3 domains of the type 2N VWF variants (e.g. H817Q, R854Q+T789A, and R782W) exhibit severely impaired ability to bind FVIII.[28, 29]

The findings of proteolysis of type 2N VWF suggest that binding between FVIII and the omains may be a prerequisite for the enhancement of VWF proteolysis. This seems to be contradictory to our current finding that FVIII can directly bind isolated A2 domain, inducing its conformational changes. Perhaps FVIII is not capable of binding A2 within the context of a multimeric VWF due to interdomain interactions and constraints, which may either hide the binding site for FVIII or create a steric hindrance inhibition for FVIII binding (Fig. 6). Binding of FVIII to the D’D3 domains may either break VWF’s interdomain interactions or bring the FVIII close to A2. Such a speculative working model is yet to be determined.

Fig. 6. Working model how FVIII binding may affect VWF conformation.

A and B show force-induced conformational changes of VWF fragment (i.e. D’D3A1A2A3) in the absence and presence of recombinant FVIII, respectively.

In conclusion, we demonstrate that direct binding of FVIII to the A2 domain of VWF results in conformational change in the A2 domain under mechanic force. These findings appear to shed new light on the molecular mechanism underlying FVIII cofactor-dependent regulation of VWF proteolysis by ADAMTS13 under shear.

Essentials:

• Binding of coagulation factor (FVIII) to von Willebrand factor (VWF) accelerates the proteolytic cleavage of VWF by ADAMTS13 under shear.

• However, the mechanism underlying such a rate-enhancing activity of FVIII in ADAMTS13-mediated cleavage of VWF is not known.

• Single molecule force spectroscopy, atomic force microscopy, and surface plasma resonance determine the interactions between FVIII and various recombinant VWF fragments.

• The results demonstrate that binding of FVIII to the D’D3 and/or A2 of VWF alters the force-induced conformational changes in the A2 domain.