Force-induced on-rate switching and modulation by mutations in gain-of-function von Willebrand diseases

Significance

Binding of von Willebrand factor (VWF) to platelets is regulated by hydrodynamic forces in the vasculature. VWF can sense force and can bind when the hydrodynamics change due to bleeding. We show that force application switches the A1 domain in VWF to a second state with faster on-rate for its binding partner on platelets, GPIbα. This provides a physiological mechanism for activating VWF binding to platelets at sites of bleeding. Moreover, force increases the effects of gain-of-function mutations found in von Willebrand disease (VWD) and platelet-type VWD by mechanically stabilizing bond formation and strength.

Abstract

Mutations in the ultralong vascular protein von Willebrand factor (VWF) cause the common human bleeding disorder, von Willebrand disease (VWD). The A1 domain in VWF binds to glycoprotein Ibα (GPIbα) on platelets, in a reaction triggered, in part, by alterations in flow during bleeding. Gain-of-function mutations in A1 and GPIbα in VWD suggest conformational regulation. We report that force application switches A1 and/or GPIbα to a second state with faster on-rate, providing a mechanism for activating VWF binding to platelets. Switching occurs near 10 pN, a force that also induces a state of the receptor−ligand complex with slower off-rate. Force greatly increases the effects of VWD mutations, explaining pathophysiology. Conversion of single molecule k_on (s⁻¹) to bulk phase k_on (s⁻¹M⁻¹) and the k_on and k_off values extrapolated to zero force for the low-force pathways show remarkably good agreement with bulk-phase measurements.

Understanding how force affects receptor and ligand binding and unbinding is a long-standing effort in mechanobiology (1–5). Bond dissociation rates typically increase under mechanical stress; however, bond stability can be enhanced through specialized mechanisms induced by force, including catch bonds and switching to a slip bond with a slower off-rate (flex bonds) (6, 7). Bond formation against an applied force has recently been measured (8). Force-regulated switching to a faster on-rate has not yet been reported for any receptor−ligand bond but would have important biological implications for adhesion in environments with high forces such as the circulation.

At sites of vascular injury, hydrodynamic force in the bloodstream acting on von Willebrand factor (VWF) is pivotal in regulating the binding of the VWF A1 domain to GPIbα on platelets and commencing the crosslinking of platelets by VWF to form a platelet plug (9–11). VWF circulates in the form of long, disulfide-bonded concatemers, with tens to hundreds of monomers, which mostly adopt a compact, irregularly coiled conformation during normal hemodynamics (12). At sites of hemorrhage, flow changes from shear to elongational. On transition from low to high shear and from shear to elongational flow, irregularly coiled molecules extend to a thread-like shape, and elongational (tensile) force is exerted throughout their lengths (13–16). Molecular elongation exposes the multiple A1 binding sites in VWF concatamers for multivalent binding to GPIbα (9, 11, 14, 16–18).

In vivo, tensile force transmitted through VWF is applied to the N and C termini of individual domains, and could theoretically change A1 domain conformation before binding to GPIbα. Although this scenario has not yet been observed, single-molecule studies demonstrate two distinct force-dependent dissociation pathways (flex-bond behavior) of the wild-type (WT) A1-GPIbα complex, and thus suggest that two conformational states can be present after formation of the receptor−ligand complex (6).

Mutations in VWF cause von Willebrand disease (VWD), the most common human heritable bleeding disorder (11, 17). In type 2B VWD, gain-of-function mutations localized to the A1 domain enhance binding to GPIbα. These mutations map distal to the GPIbα binding site, near the A1 N and C termini where elongational force is applied to VWF during physiologic activation (19–21). Gain-of-function mutations in GPIbα cause a disease similar to type 2B VWD termed platelet-type VWD (PT-VWD) (22). PT-VWD mutations map to a β-switch region that changes conformation when complexed with A1 to form a β-ribbon structure in GPIbα that adds onto the β-sheet in A1 (20, 21, 23). PT-VWD mutations are thought to favor the conformation that the β-switch assumes when bound to A1, and map adjacent to the major A1-GPIbα interface. Here, using single molecule measurements, we show that the formation of the A1-GPIbα bond is allosterically regulated by force-dependent switching between two distinct association pathways, suggesting two different conformational states before binding. Pathologic gain-of-function mutations retained two-state binding and unbinding and showed faster on-rates together with slower off-rates under force than WT. A1 and GPIbα mutations showed distinct effects on kinetic and mechanical properties.

Results

We used receptor and ligand in a single molecule (ReaLiSM) constructs of A1-GPIbα with or without the VWD R1306Q mutation in VWF A1 or the PT-VWD M239V mutation in GPIbα (Fig. 1A). Receptor−ligand unbinding and rebinding gave discrete jumps in tether length in each cycle of stretch and relaxation, respectively (Fig. 1B). Fits to the worm-like chain model (WLC) (24) for ReaLiSM constructs with 26- and 43-residue linkers gave contour lengths of 12.1 ± 0.9 nm and 17.7 ± 0.6 nm (Fig. 1C) in agreement with calculated values of 11.4 nm and 17.9 nm, respectively, based on 3.8 Å per linker residue and N- to C-terminal distances of 1.9 nm (A1), 7 nm (GPIbα), and 7 nm (A1-GPIbα complex) from crystal structures (20, 23). This agreement, together with B to S transitions of the DNA handles at ∼67 pN (24) observed in all of our experiments (Materials and Methods), provided strong support that single A1-GPIbα binding and unbinding events were being measured.

Fig. 1.

The VWF A1 and GP1bα ReaLiSM construct and change in extension upon unbinding and rebinding. (A) Schematic of the ReaLiSM construct and tweezers (6). (B) Successive cycles of stretching (black traces) and relaxation (red traces) with unbinding and rebinding events arrowed, with the R1306Q construct at 40 nm/s. (C) Fits to worm-like chain model showing the persistence length (PL) and contour length (CL). Bars show SD for each force bin.

The distribution of bond dissociation forces for A1/R1306Q-GPIbα/WT and A1/WT-GPIbα/M239V was bimodal (Fig. 2 D−I), as previously reported for WT (6) (Fig. 2 A−C). Thus, all three types of complexes behave as flex bonds, switching from one state at low force to a second state at higher force. However, the mutations shifted the rupture force distributions. The second peak at higher force at a pulling rate of 40 nm/s shifted from WT value of 14.6 pN to 18.0 pN in GPIbα/M239V (Fig. 2 B and E) and to 21.6 pN in A1/R1306Q (Fig. 2H).

Fig. 2.

Force spectroscopic measurements of k_off. (A−I) Unbinding force distributions at different pulling rates in wild-type (A−C), M239V (D−F), and R1306Q (G−I) constructs. Error bars show the variance, estimated assuming a binomial distribution for each histogram bin, with the square of variance = n(1-n/N_c), where n is events in a particular bin and N_c is total events (35). Curves show the predicted rupture force distributions using the constants from J−L for dissociation pathways 1 (left curve, estimated using dark gray histogram bins) and 2 (right curve, estimated using light gray histogram bins). Events in the overlap region were apportioned between the pathways according to iterative fits. Counts are on a linear scale with maximal values of 20 for A, D, and G; 40 for B and C, E and F, and H and I. (J−L) The k_off (F) values (where F is Force) calculated for each bin in A−I are plotted. Lines and constants in each panel are from fits of k_off (F) to k_off⁰ and σ. Error bars show 1 SD.

Bond lifetimes at each force bin in rupture force histograms were estimated using the Dudko−Hummer−Szabo equation (25). Data at two different pulling rates and linker lengths demonstrated excellent agreement with no adjustable parameters (Fig. 2 J−L). Off-rates for dissociation pathways at low force (k_{1 off}) and high force (k_{2 off}) were each well fit by the Bell model, k_off = k_off⁰ exp (σF/k_BT), where the force across the receptor−ligand bond exponentially increases off-rate. WT results (Fig. 2J) were within error of previous estimates (6). Interestingly, the mutations had significantly different effects on the extrapolated off-rate at zero force, k_off⁰_, and the mechanical stability of the bond, σ, which is equivalent to the distance to the transition state and determines how much force exponentiates k_off. The k_{1 off}⁰ and k_{2 off}⁰ values for wild-type and A1 mutant were comparable, whereas those of the GPIbα mutant were about twofold and fourfold slower, respectively. Conversely, the A1 mutation increased bond strength (decreased σ₁ and σ₂ values) more than the GPIbα mutation (Fig. 2 K and L).

Association kinetics were investigated by observing rebinding forces (Figs. 1B and 3). Interestingly, we saw bimodal rebinding histograms for all three types of A1-GPIbα complexes (Fig. 3). Compared with WT (Fig. 3 A−C), the two peaks were more separated for patient mutations and were shifted to higher force (Fig. 3 D−I). The presence of two peaks in rebinding force demonstrates that before binding, either A1, GPIbα, or both can exist in two different states that differ in binding kinetics. Binding histograms for the PT-VWD mutation showed a shift in both peaks compared with WT; the first pathway shifted from 4.7 pN to 8.8 pN, and the second shifted from 9.3 pN to 15.6 pN at 40 nm/s (Fig. 3 B and E). In the VWD type 2B mutation, the first pathway shifted from 4.7 pN to 6.7 pN, and the second rupture force peak was shifted from 9.3 pN to 12.6 pN at 40 nm/s (Fig. 3H). Bimodal binding force histograms were further observed in the constructs with a shorter linker (Fig. 3 C, F, and I).

Fig. 3.

Force spectroscopic measurements of k_on. (A−I) Binding force distributions at different relaxation rates for wild-type (A−C), M239V (D−F), and R1306Q (G−I) constructs. Errors are variance, estimated as in Fig. 2. Histograms were fit to the Pierse−Dudko equation (bottom equation) (26) using least-squares (SI Materials and Methods) yielding the probability distributions for rebinding pathways 1 (left curve, dark gray bins) and 2 (right curve, light gray bins). Events in the overlap region (histogram bins with dark and light gray) were apportioned between the pathways according to iterative fits. Force relaxation rates (PN/s) were averaged over bins for each pathway and are shown in each panel. Counts are on a linear scale with maximal values of 20 for C and F, 30 for A, B, D, and G, and 40 for E and H–I. (J−L) The k_on(F) values (where F is force) calculated from each bin in A−I are plotted. Error bars show SD. (M) Force-dependent A1-GPIbα RL binding constants, calculated after removal of the effect of the tether (26) (SI Materials and Methods). Data are average RL values from experiments using 43- and 26-residue linkers ±1 SD estimated by propagation of error.

Dissociation and reassociation through low-force and high-force pathways were observed in successive cycles with the same tether (Fig. 1B). We compared the frequency of successive events to that expected based on overall frequency at each pulling rate (Table S1). There was no evidence of hysteresis; e.g., the frequency of unbinding or rebinding at high force was independent of whether the previous event was at high force. This suggests that state switching occurred more rapidly than the half-cycle time of 12 s at 40 nm/s (Fig. 1B) or 24 s at 20 nm/s. In agreement, state-switching rates for bond dissociation were previously estimated to be in the range of 0.13–1.17 s⁻¹ at 10–11 pN (6).

Expressions have recently been derived for extracting single-molecule on-rates from distributions of binding forces at different relaxation speeds for a receptor−tether−ligand (RTL) complex (Fig. 3 J−L) (26). The effects of the 43- and 26-residue tethers can be accounted for and removed using the worm-like chain parameters measured in Fig. 1C, revealing true receptor−ligand (RL) binding parameters (26) (See SI Materials and Methods and Figs. S1 and S2). Fitting and converting to RL values yields three distinct parameters: k_on^0RL, the zero-force on-rate corresponding to the intrinsic, unimolecular on-rate measured in s⁻¹ (1); σ_on^RL, the mechanical sensitivity of on-rate to force; and ΔG^RL, the height of the energy barrier to rebinding (Tables S2 and S3). Average RL parameters derived from measurements with the 43- and 26-residue linkers are shown in Fig. 3M. For WT and both mutants, k_{2 on}^0RL was 11- to 17-fold faster than k_{1 on}^0RL, showing a large difference between the two association pathways. Within each pathway, on-rates for WT, A1 R1306Q, and GPIbα M239V were comparable, with differences of less than 1.2-fold. The most dramatic difference between WT and mutant behavior manifested in the exponential σ values, which govern the force dependence of on-rates; σ₁ decreases from 2.5 nm in WT to 1.4 nm and 1.5 nm, and σ₂ decreases from 1.8 nm in WT to 1.0 nm and 1.1 nm for M239V and R1306Q, respectively. Thus, gain of function A1 R1306Q and GPIbα M239V mutations mechanically stabilize bond formation.

Discussion

We have demonstrated that force can switch the states of VWF A1 and/or GPIbα, resulting in two distinct receptor−ligand association pathways. Our finding that force can switch the kinetics of bond formation between A1 and GPIbα is unprecedented for a RL bond. Usually, receptors and ligands have no significant forces exerted on them before binding; it is only after binding that force on cell(s) is applied to the RL bond. Thus, previous theories on how catch or flex bonds work have focused on the RL complex and only considered the effect on bond dissociation (6, 7). Because integrin−ligand and selectin−ligand complexes are more extended in their high- than low-affinity states, one theory posits that by favoring extension, the applied force lowers the energy of the high-affinity relative to the low-affinity state (4).

VWF is an exceptional ligand. The length of VWF concatamers (Fig. 4A) can exceed the diameter of cells and the elongational forces applied to free VWF in the bloodstream range up to 10 pN, similar to the force range studied here (16). When bound to platelets on the vessel wall, the force on VWF would be much greater (14). In the irregularly coiled conformation of VWF concatemers at low flow, A1 may interact with other domains. However, elevated shear and elongational flows found at sites of hemostasis will tend to induce a thread-like, uncoiled conformation of VWF (Fig. 4B) (14). By definition, elongation removes interactions with distal domains in the same or other monomers within a VWF concatemer. Notably, A1 also contains mucin-like N- and C-terminal segments (Fig. 4A) that act as spacers to separate it from neighboring domains (14) after elongation. In elongated VWF, force is applied to the N and C termini of A1 (Fig. 4 B and D) and propagates through A1 similarly to A1 in the unbound state of ReaLiSM (Fig. 4E). In contrast, no significant force would be applied to GPIbα on the surface of a platelet before binding A1. Because of these physiologic considerations, and the proximity of VWD type 2B mutations to the site of force application to A1 (Fig. 4 C and D), it is reasonable to suggest that the two on-rates may correspond to two distinct conformations of A1, but this remains to be formally demonstrated.

Fig. 4.

Models for force-dependent activation of VWF and comparison of single-molecule and bulk measurements. (A) Schematic organization of domains in VWF and head-to-head and tail-to-tail linkage of VWF monomers into concatemers. (B) VWF concatemers in low shear are predominantly irregularly coiled; transition to high shear and elongational flow induces a thread-like conformation. High tensile force exerted through concatemers is hypothesized to cause a change in A1 conformation similar to that in the ReaLiSM construct shown in E. (C) Crystal structure of A1 bound to GPIbα (21). Cα spheres show sites of VWD type 2B mutations in A1 (36) (yellow, except R1306 in silver) and PT-VWD mutations in GPIbα (37) (yellow, except M239 in silver). (D) Enlargement of the region near the A1 N and C termini. Hydrogen bonds (black dashes) external to the long-range disulfide (sulfur atoms shown in orange) would be broken by force applied (red arrows) to the N and C termini (small cyan spheres) before the disulfide bond would resist force. Breakage of these hydrogen bonds might trigger conversion of A1 to a high-affinity state (14), as shown schematically in B and E. (E) Model of force-dependent switching between two conformational states consistent with the two pathways for force-dependent unbinding and binding measured here. Red arrows in B, D, and E represent tensile force. (F) Single-molecule encounter complex model. To convert to bulk-phase on-rate units, the effective concentration of one reactant relative to the other is determined as the volume of a sphere of radius equal to the separation between the center of mass of A1 and GPIbα in the bound state (Lower) plus σ_on and σ_off (Upper). (G) Comparison between bulk-phase kinetic and K_D values from Blenner et al. (21) and Miura et al. (29) and our single-molecule (SM) data with k_on converted to bulk-phase units. Values at 0 pN in state 1 are for comparison with bulk-phase values. Values at 15 pN in state 2 enable comparison between wild-type and mutants in a physiologically relevant force range.

It is interesting to convert our intrinsic single molecule k_on^0RL rate estimates in units of s⁻¹ to bulk k_on rates (k_on^sol) in units of M⁻¹s⁻¹ using a model for effective concentration in an encounter complex (Fig. 4F). The encounter complex is formed when two reactants diffuse sufficiently close to one another for the subsequent binding reaction to occur (27, 28). The distance between the two reactants in the encounter complex is used to calculate the concentration at which the intrinsic on-rate in s⁻¹ occurs, and thus to convert to the bulk rate on-rate in M⁻¹s⁻¹. In our model of the encounter complex, we have assumed that σ_on and σ_off correspond to distances to transition states between unbound and bound states, respectively, and added these distances to the distance between the centers of masses of A1 and GPIbα in complex crystal structures (Fig. 4F). The k_{1 on}^sol, k_{1 off}⁰, and 3D dissociation constants (K_D = k_{1 off}⁰/k_{1 on}^sol) calculated from our measurements on WT, VWD type 2B, and PT-VWD ReaLiSM constructs match remarkably well with bulk-phase values from two well-documented reports (21, 29) (Fig. 4G). These agreements provide an important confirmation of the ability of the ReaLiSM construct to measure meaningful force-induced binding kinetics.

Off-rates have been estimated by other studies that report either single molecule or single tether A1-GPIbα measurements. Using thermal fluctuation of beads coated with A1 and antibody bound to GPIbα, a zero-force off-rate of 0.2 s⁻¹ was found (30). Transient tethers of A1-coated beads in shear flow over surfaces coated with platelets were extrapolated to zero force using the Bell model and yielded k_off⁰ of 3 s⁻¹ and σ of 0.03 nm (31). These off-rates differ by two to three orders of magnitude from our k_{1 off} measurement of 0.0047 s⁻¹ and the bulk phase measurement of 0.0036 s⁻¹ (29). The σ value of 0.03 nm (31) also differs greatly from our estimate of 2.5 nm. These discrepancies suggest that bead thermal motion and transient tethers measure different types of events than binding and unbinding of single A1 and GPIbα molecules measured with ReaLiSM or bulk studies.

We found that force greatly increases the effect of VWD mutations. At zero force in state 1, the VWD type 2B R1306Q A1 and PT-VWD M239V GPIbα mutations enhance bond formation and bond lifetime values by less than 1.2-fold (Fig. 4G). These are modest changes considering the resulting disease phenotype but not too dissimilar from bulk measurements that show twofold to fivefold increases in affinity (Fig. 4G) (20, 21, 29). In contrast, large differences in kinetics are observed once force is applied. At 15 pN in state 2, bond formation occurs ∼45-fold faster than WT for both VWD 2B and PT-VWD mutants (Fig. 4G and Table S4). Bond dissociation by the mutants is also slower than WT at 15 pN in state 2, so the VWD 2B and PT-VWD mutations have an effective 260-fold and 230-fold increase in affinity, respectively (Fig. 4G). Therefore, an important concept emerging from these results is that force can accentuate the manifestation of disease phenotypes. In VWD 2B and PT-VWD, enhanced binding of VWF to platelets leads to depletion of VWF, with longer concatemers selectively depleted, and also to depletion of platelets; the final result is bleeding tendency (11, 17).

Our finding of force-induced switching to a faster on-rate extends the concept of flex bonds from bond dissociation to bond association. For WT, VWD-2B, and PT-VWD, switching to a state with faster on-rate resulted in a 40- to 100-fold increase in bond association kinetics at 15 pN. Force-induced switching may therefore dramatically enhance bond formation under flow. The hypothesis that a similar conformational change underlies second states of both bond dissociation and association will be an important subject for future structural studies. Switching to each of these second states occurs at ∼10 pN; furthermore, the GPIbα cytoplasmic domain remains bound to filamin at forces in this region (32) and, together with the covalently linked GPIbβ subunits (33), helps prevent uprooting from the cell.

At sites of hemostasis and stenosis, alterations in flow are predicted to elongate VWF and increase tensile force exerted throughout its length (14). A1 thus becomes better exposed for binding to platelets; furthermore, our results provide a mechanism for switching A1 to a second state with faster on-rate for GPIbα. In agreement, shear thresholds have been observed, above which, VWF agglutinates platelets in flow and VWF adsorbed to a vessel wall mediates binding and rolling of platelets (9, 10, 34). Thus, the forces unleashed in hemorrhage can trigger binding of VWF to platelets and formation of a hemostatic plug. Moreover, the second state of the A1-GPIbα complex and its greater mechanical strength than state 1 enable resistance to the unusually high forces that must be overcome by a biological RL bond for hemostatic plug formation and final closure of a bleeding vessel.

Materials and Methods

Proteins were expressed and purified as previously described (6). The ReaLiSM construct consists of human VWF A1 domain (Asp-1261 to Pro-1466 with prepro-VWF numbering) and human platelet GPIbα (His-1 to Arg-290) connected by linkers of 26 residues, GTGENLYFQGGSSSSTTGWTGGHVGT; or 43 residues, GTGENLYFQGHHHHHH(GSSSS)₃GTTGWRGGHVGT. Our current 43-residue linker lacks the Pro residues present in our previous 43-residue linker (6). Protein was made with or without mutations M239V in GPIbα or R1306Q in A1.

DNA handles (802 bp), protein−DNA coupling, anti-digoxygenin Fab, and streptavidin beads were as previously described (6).

We performed force rip experiments (constant trap velocity) by stretching and relaxing the tether between force values of 2 pN and 15–30 pN at pulling rates of either 20 nm/s or 40 nm/s. The unbinding distance was measured between two points on the force trap position curves, from the point just before the dissociation to the point when the force returned to the same level after dissociation. The extension between these two points arises solely from stretching the flexible polypeptide tether. Force loading rates (pN/s) before each rip event were estimated from the curve by measuring the slope of the force vs time data; this value was then averaged over all events in a given histogram bin of unbinding events at a given pulling speed.

Rebinding was observed in force rip experiments as force was lowered during the relaxation phase of the cycle. We define the rebinding force as the highest force at which rebinding was observed in one relaxation cycle. Most cycles only showed one unbinding and one rebinding event; however, hopping between bound and unbound states in one cycle was not uncommon for the R1306Q construct, as shown in Fig. 1B. Binding events were binned as a function of force.

Equations, fitting methods, and estimations required to convert RTL values to RL k_on, σ, and ΔG values are described in SI Materials and Methods. Briefly, off-rate fitting was performed as described (6). On-rate fitting was performed by minimizing the sum of squared errors between the normalized binding histogram data and the probability of binding function described in ref. 26 using the fminsearch tool in MATLAB. Error bars and errors shown in figures and tables show SD estimated by propagation of error (35).