Autoinhibitory module underlies species difference in shear activation of von Willebrand factor

Abstract

Background:

Von Willebrand factor (VWF) is a multimeric plasma protein that bridges the gap between vessel injury and platelet capture at high shear rates. Under high shear or tension, VWF can become activated upon the unfolding of its autoinhibitory module (AIM). AIM unfolding exposes the A1 domain, allowing for binding to platelet glycoprotein (GP)Ibα to initiate primary hemostasis. The characteristics of the AIM and its inhibitory properties within mouse VWF are unknown.

Objectives:

To determine and characterize the autoinhibitory properties of mouse VWF.

Methods:

Recombinant mouse VWF A1 fragments containing or lacking the flanking regions around the A1 domain were generated. We tested the ability of these fragments to bind to human or mouse GPIbα and platelets. We compared the unfolding of mouse AIM-A1 to human AIM-A1 by single-molecule force spectroscopy.

Results:

Recombinant mouse AIM-A1 binds with higher affinity to GPIbα than its human counterpart. Recombinant mouse proteins lacking part of the AIM show increased binding to GPIbα. Activated A1 fragments lacking the AIM can effectively agglutinate platelets across the species barrier. Using single-molecule force spectroscopy, we determined that the mouse AIM unfolds under forces similar to the human AIM. Additionally, the human AIM paired with mouse A1 largely recapitulates the behavior of human AIM-A1.

Conclusions:

Our results suggest that the regulation of VWF-GPIbα binding has been specifically tuned to work optimally in different rheological architectures. Differences in the AIM sequence may contribute to the difference in VWF shear response between human and mice.

Introduction

Von Willebrand factor (VWF) is an ultra-large, multimeric, and multi-purpose plasma glycoprotein that initiates primary hemostasis [1]. Circulating in an inactive state, VWF can respond to a vascular injury through immobilization and exposure to high shear stress or elongational flows [2]. This process relieves VWF of its autoinhibited state, exposing the A1 domain that then binds to the VWF receptor on platelets, glycoprotein (GP)Ibα [3].

The organization of the domains that constitute VWF is conserved in mammals, while VWA and VWD domains can be found in vertebrates as old as agnathans [4]. We therefore postulated that the shear dependence of VWF activation is also conserved. Since VWF activity is critical to hemostasis and tightly regulated, it would be naive to expect the biophysical parameters that underlie VWF activation are the same between all mammals given their diversity in physical sizes, vessel diameters, and threat from predation. The comparison of humans and mice is a great juxtaposition of the difference in rheological parameters of mammals. Beating at a faster rate than a human heart, the mouse heart pumps blood through vessels at relatively higher shear rates to the human equivalent [5]. Mouse platelet count in the same volume also greatly outnumbers humans, while also being significantly smaller in size [6]. Regardless of these vast physical differences, mice and humans still achieve a similar hemostatic balance, as a mouse model of VWF deficiency largely replicates the bleeding symptoms observed in humans devoid of VWF (Type 3 VWD) [7, 8].

The generation of a transgenic mouse bearing only the human GPIbα subunit (hTg) was the first test of species specificity between human GPIbα and mouse VWF [9]. These mice can sustain hemostasis, and both wild-type and hTg platelets can bind to human VWF in the presence of botrocetin, a snake venom that has profound procoagulant activity in dozens of species [10–12]. Human VWF activated with ristocetin, a commonly used glycopeptide activator of VWF, showed reduced binding to mouse platelets, but showed normal binding and aggregation with hTg platelets [9]. Despite the orthology between human and mouse GPIbα, mouse platelets do not bind human VWF under flow as well as they do mouse VWF [13]. Further investigation into the species specificity has yielded disparate results [14, 15], but generally conclude that human VWF is not sufficient for hemostasis in mice, yet human GPIbα and mouse VWF can support hemostasis [16, 17].

The direct interaction between GPIbα and VWF, or more specifically between the ligand-binding domain (LBD) of GPIbα and the A1 domain of VWF, has been studied using purified recombinant fragments, with apparent affinities reported to range from nanomolar to micromolar [18–20]. Such heterogeneity in reported binding affinities may likely reflect the variety in the detection method, the source of recombinant proteins, and the chosen termini of the VWF fragment. It is also confounded by the existence of multiple conformational states of the A1 domain. Recent studies suggest that residues flanking the A1 domain modulate the A1 binding activity [21–24]. However, the length of the flanking residues in recombinant A1 proteins in these studies varied greatly, and the activation state of these proteins was often not clearly defined.

We recently demonstrated that the exposure of human A1 domain is dependent on the force-resistive nature of the adjoining autoinhibitory module (AIM) [21, 25]. This AIM comprises discontinuous sequences flanking the A1 domain that is delimited by the 1272–1458 disulfide bond (Figure 1A) [26]. While the AIM contains some intrinsically disordered sequence, it contains certain folded structure that likely involves sialylated glycans [25, 27]. It masks a part of the GPIbα-binding site in the A1 domain. The AIM unfolds when sufficient tensile force is applied, thereby relieving the inhibition on A1 [25]. Consistently, the recombinant hAIM-A1 protein, containing residues 1238–1493 of human VWF, like full-length VWF, shows little detectable affinity for the LBD of human GPIbα (hLBD) [25, 26]. In comparison, htAIM-A1 protein, containing residues 1261–1472 of VWF (i.e. the entire A1 domain with only truncated AIM), can spontaneously bind to hLBD and dissociate with a two-phase decay as observed for activated A1 [21, 26]. Furthermore, htAIM-A1, like the active VWF, can activate platelets and induce platelet aggregation [25]. In this paper, we characterized and compared functional and biomechanical differences between human and mouse AIM-A1 proteins, to gain insights on species adaptation of the AIM and the activation mechanism of VWF.

Figure 1. Protection of mouse VWF A1 binding by flanking AIM sequences.

(A) Alignment of human and mouse NAIM (blue) and CAIM (orange) sequences that flank the A1 domain (shown as a circle, delimited by the 1272–1458 disulfide bond) was performed using MUSCLE. Non-identical residues are shown in bold. Alignment of complete AIM-A1 sequences from various organisms can be found in Supplemental Figure 1. (B) Illustration of human AIM-A1 protein and its activation when the tension exceeds the critical threshold (F_crit) and disrupts the AIM. In the inactive form, the A1 domain (gray circle) is masked by the AIM (blue and orange lines). It is not clear if mouse AIM (purple and red lines) confers similar protection to the mouse A1 domain (dark gray circle) and affects its activity. (C) Binding BLI sensorgrams of mouse VWF AIM-A1 (left panel) and tAIM-A1 (dark gray circle with short purple and red lines, right panel), in serial dilution from 4 μM, to mouse GPIbα LBD. Association at each concentration was performed for 100 seconds followed by dissociation into kinetics buffer for an additional 500 seconds. A global fit of heterogenous ligand binding for all concentrations (red traces) is performed. The sum of equilibrium responses from two-state equations for each concentration were fit to a hyperbola and K_D was derived as half-maximal binding (insets). Sensorgrams are reflective of two technical replicates.

Materials and Methods

Materials and reagents

Expi293F cells, DMEM/F12, Freestyle F17, GlutaMAX, penicillin/streptomycin, bovine serum albumin fraction V, hygromycin, and D-biotin were purchased from ThermoFisher Scientific. Ristocetin was purchased from MP Biomedicals and reconstituted to 25 mg/mL stocks in phosphate buffered saline (PBS). Mono/Dibasic sodium phosphate, sodium chloride, and imidazole was purchased from Sigma.

Protein Alignment

Protein sequences were obtained from NCBI VWF Orthologs database (Gene ID: 7450). Individual sequences were extracted to contain the AIM and the A1 domain by selecting all residues after the highly conserved, last cysteine residue of the D3 domain, extending through to the start of the A2 domain, usually marked by the motif GPKRN/K. Selected sequences were imported into DNASTAR MegAlign Pro (v 17.2.1) and aligned with the MUSCLE algorithm [28]. Phylogenetic tree construction was computed using neighbor joining and rooted on Danio rerio (Zebrafish, Supplemental Figures 1, 2). Alignment files are available from the authors upon reasonable request.

Construction and expression of VWF fragments

Recombinant human VWF fragments have been described and characterized previously [25]. All mouse VWF sequence numbering is based on NCBI protein accession NP_035838.3. Listed sequences of mouse VWF fragments are listed in supplemental Table 1. Gene sequences encoding mouse VWF fragments were codon optimized for expression in mammalian cells, then synthesized and cloned into pcDNA3.1 Hygro(+) by Genscript (Piscataway, NJ). All sequences in expression plasmids were verified by Sanger sequencing (Genewiz). Complete amino acid sequences of mouse VWF fragments can be found in Supplemental Table 1.

Stable expressing cell lines were established by transfection of adherent Expi293F or Expi293F-BirA with Lipofectamine 3000 followed by selection in DMEM/F12 with 10% FBS and 250 μg/mL hygromycin. Individual colonies were selected, screened for high expression, and adapted to suspension culture in Freestyle F17 media supplemented with 2x GlutaMAX and .1 g/L Pluronic F68 (BASF). Cell lines expressing proteins with BirA for in-vivo biotinylation were cultured in media containing 2 μM D-biotin.

Mouse tAIM-A1 (containing VWF residues 1261–1472) was produced via transient transfection of suspension Expi293F in Freestyle F17 media with Mirus Trans-It Pro according to manufacturer’s directions. Cultures were harvested 6 days after transfection.

Suspension media was clarified by centrifugation at 8000 × g for 30 minutes and loaded to equilibrated (20 mM phosphate buffer, 500 mM NaCl, pH 7.4) Nickel-Sepharose excel resin (Cytiva). The column was washed with the above buffer supplemented with 20 mM imidazole until a flat A₂₈₀ baseline was achieved. Bound proteins were eluted with equilibration buffer plus 500 mM imidazole at pH 7.4. The elution was concentrated to ~2–5 mL and further purified by size exclusion chromatography using a Superdex 75 or 200 pg 16/600 column equilibrated with phosphate buffered saline (PBS), pH 7.4. Proteins eluted at expected retention times given their molecular weights with symmetrical peaks suggestive of a monomeric, monodisperse protein. Proteins were purified to homogeneity as assessed by SDS-PAGE and Coomassie stain (Supplemental Figure 3), with sizes indicative of added glycosylation consistent with the expectation of 8 O-linked glycans on the AIM-A1 protein, or 2 in the tAIM-A1 protein. All proteins were concentrated to ~1 mg/mL using Amicon Ultra 15mL, 3kDa MWCO centrifugal filters (Millipore), aliquoted, and frozen at −80°C. Protein concentrations were calculated based on their extinction coefficients at 280 nm. Protein aliquots were used within a day of thawing and were never subjected to freeze-thaw more than once.

Construction of mouse GPIbα LBD and human GPIbα W230L

Human GPIbα LBD (mature protein residues 1–290) W230L was generated by site-directed mutagenesis of a previously described vector [25] using primers W230L_F and W230L_R (Supplemental Table 2). The gene fragment encoding mouse GPIbα LBD (mLBD, residues 1–290) was generated by overlap extension PCR using the mouse GPIbα cDNA[29] as the template and primers mLBD_F and mLBD_R to append a biotinylation sequence and FLAG-tag at its C-terminus. The DNA fragment was cloned into the same expression vector as for the human GPIbα LBD constructs. Stably expressing cell lines and protein was produced and purified as described above. The purity of purified LBD proteins was verified by SDS-PAGE (Supplemental Figure 3).

Bio-layer Interferometry (BLI)

BLI experiments were performed using an Octet QK^e instrument with data acquired at 0.3 Hz with 1000 rpm shaking during all steps. Samples were diluted in kinetics buffer (ForteBio/Sartorius), comprised of PBS, 0.1% BSA, and 0.02% Tween-20. Equilibrated streptavidin (SA) sensors were loaded with ~10 μg/mL GPIbα-LBD from either species until hitting a set threshold for 2 nm of binding. Sensors were then washed briefly and set to association in various VWF fragments for 100 seconds and dissociate for 500 seconds. Loaded sensors were washed with 1 M NaCl to remove any remaining bound VWF and equilibrated again for the next samples. BLI experiments were repeated at least twice for each binding interaction. Sensorgrams presented are representative samples. Data was fit to a heterogenous ligand binding model using two rate equations. Kinetic parameters can be found in Supplemental Table 3. To obtain equilibrium response (ΣR_eq K_D) values for each binding reaction, Req1 and Req2 were summed for each concentration and fitted to a hyperbola assuming a 1:1 stoichiometric binding ratio.

Platelet Aggregometry

Human fresh whole blood was obtained from healthy donors via venipuncture into Vacutainer 3.2% sodium citrate tubes (BD). Written informed consent was obtained from participants before their inclusion in studies, and all procedures using donor-derived human blood and platelets were approved by the Institutional Review Board at Emory University. C57BL/6J and VWF^−/− mice [7] were purchased from The Jackson Laboratory. Preparation of mouse platelets used in the study was approved by the IACUC of Emory University. Human and mouse platelet rich plasma and washed platelets were obtained as described [30, 31]. Platelet aggregation studies were performed as previously described [25]. Briefly, platelets were resuspended to 200,000/μL in either modified Tyrode’s buffer or pooled normal plasma from humans (George-King Biomedical) or mice. After 30 seconds, VWF fragments or ristocetin were added to the suspension and aggregation was followed for up to 10 minutes.

Single-molecule force spectroscopy

Single-molecule force measurement was performed largely as described.[25] Briefly, the biotinylated VWF fragment was immobilized on a streptavidin bead held by a fixed micropipette. SpyCatcher003-S49C [32] was produced by expression in SHuffle T7 express cells (NEB) according to standard procedures. Bacterial pellets were lysed with BugBugster (Millipore) and clarified supernatant containing SpyCatcher was purified on Ni-Sepharose 6 FF (Cytiva). The eluate was incubated with 5 mM DTT and 5 mM EDTA for 30 minutes and isolated by size exclusion chromatography in PBS on a Superdex 75 16/600 pg column. SpyCatcher was coupled to a biotin-DNA handle of 802 bp, and then coupled to streptavidin beads of 2.0-μm diameter (Spherotech, Lake Forest, IL). For pulling experiments, the SpyCatcher-DNA handle bead, trapped and controlled by the optical tweezer, was brought to interact with the VWF fragment bead. The force measurement was performed at force-ramp mode with varying pulling speeds (50, 100, 200, 400, and 500 nm/s) in Tris-buffered saline (20 mM Tris, 150 mM NaCl, pH 7.5). The force-extension data were fitted to the worm-like chain (WLC) model [33]. Unfolding extension is defined as the increase in end-to-end distance between the point of unfolding and the point at which the force at unfolding is re-established. To find the most probable extension at various forces, the force-extension data was first binned by force. The unfolding extensions within each binned force were then plotted in a histogram. The most probable extensions were determined by conducting a Gaussian fit to the histograms. Unfolding was also analyzed according to the Bell-Evans model to determine most probable unfolding force of the protein at different loading rates [34–36].

Results

Multiple sequence alignment of mammalian autoinhibitory modules

Multiple sequence alignments were performed for all available mammalian, some avian, and some reptile VWF coding sequences between the last cysteine at the end of the D3 assembly and the start of the A2 domain. We define the A1 domain as all residues inside the 1272–1458 disulfide bond, the N-AIM as residues between D3 and A1, and the C-AIM as residues between A1 and A2 (Figure 1A). Mammalian AIMs appear to largely retain an overall acidic nature, enriched with proline, serine, and threonine (Supplemental Figure 1). Human and mouse AIM-A1 share 84% identical amino acids from 255 residues and are 89% similar. The AIMs specifically share 74% identity and are 83% similar (Figure 1A). There are a number of differences in the AIM itself, such as a charge switch at residue 1269, that could explain why caplacizumab does not bind mouse VWF [37]. In addition, several differences are present in residues 1460–1472 of C-AIM that contains the ristocetin-binding site [21], which may explain why ristocetin is unable to activate full-length mouse VWF. There are no differences with the known interfacial residues between AIM and A1 such as His1268 or Glu1264. We thus speculated that mouse VWF likely retains elements of autoinhibition through its AIM, but given the changes in the primary sequence, it might have different properties compared to human AIM.

The mouse AIM provides reduced protection to mouse A1

Various recombinant fragments containing mouse VWF A1 domain have been generated. In SDS gels, the recombinant mouse AIM-A1 protein (mAIM-A1, VWF residues 1238–1493) migrated at a much higher molecular weight than is expected based on the primary sequence. This protein likely contains several O-linked glycosylation sites similar to hAIM-A1, which would explain the apparent molecular weight of ~40 kDa (Supplemental Figure 3). Binding of this protein towards mLBD was moderate, with ~1-μM affinity, as analyzed using bio-layer interferometry (BLI) (Figure 1C). In comparison, a protein containing truncated mouse AIM, mtAIM-A1 (residues 1261–1472), bound to mLBD with a higher affinity at 386 nM (Figure 1C). The difference in mLBD-binding affinity between mAIM-A1 and mtAIM-A1, albeit relatively small when compared to that between human VWF fragments, suggests that the mouse AIM provides some protection for the A1 domain.

Mouse VWF A1 fragments can activate platelets and respond to ristocetin

We previously showed that washed human platelets can be dose-dependently activated by recombinant A1 fragments [25]. Furthermore, htAIM-A1 can induce apparent aggregation of these platelets at low physiological concentrations (60 nM), while micromolar concentrations of hAIM-A1 are necessary for full aggregation [25]. Here we tested the activity of mouse A1 fragments towards mouse platelets. As shown in Figure 2, addition of mAIM-A1 dose-dependently induced apparent aggregation of washed mouse platelets, similar to the responses observed for the human pair. At the physiological circulating VWF level (60 nM), mAIM-A1 was unable to cause platelet aggregation. Addition of recombinant proteins to micromolar concentrations induced a robust aggregation response (Figure 2A). In comparison, mtAIM-A1 is a more potent agonist of platelet activation, showing near complete aggregation at 250 nM (Figure 2B). Furthermore, the aggregation of platelets via AIM-A1 fragments was not dependent on intraplatelet stores of full-length VWF, as VWF^−/− mouse platelets were still aggregated following the addition of mAIM-A1 (Supplemental Figure 4).

Figure 2. Platelet aggregation traces showing the activation of mouse and human platelets by supraphysiological concentrations of mouse AIM-A1 proteins.

Mouse (A,B) or human (C) washed platelets were resuspended in modified Tyrode’s buffer and gently stirred. Reagents were added at 30 seconds following establishment of a stable baseline (arrows). Aggregation is seen as a decrease in the optical density as the platelets agglutinate and clump together. When needed, ristocetin was added to a final concentration of 1.5 mg/mL. Mouse AIM-A1 or tAIM-A1 protein was added to the denoted final concentrations.

Ristocetin is a glycopeptide antibiotic that has been used for decades to activate human VWF in a manner similar to shear-dependent activation [38]. We recently reported evidence that ristocetin induces exposure of the A1 domain for LBD association by binding to the C-AIM of human VWF and disrupting the AIM conformation [21, 25]. However, earlier studies reported that ristocetin was unable to induce aggregation in mouse platelet-rich plasma [16, 38]. Thus to our surprise, 1.5 mg/ml ristocetin was able to induce apparent aggregation of washed mouse platelets in the presence of murine A1 fragments (Figure 2A, B), even though the primary sequence at the presumed ristocetin-binding site is not identical (Human-APEAPPPTLPPDM, Mouse-APEAPAPTQPPQV). This result suggests that either ristocetin exhibits a lower affinity for full-length mouse VWF, or that the disruptive nature of ristocetin on the mouse AIM is less effective in the context of full-length VWF.

As the mouse A1 domain is orthologous to human A1, we tested if the mAIM-A1 could activate human platelets. Indeed, mAIM-A1 dose-dependently induced apparent aggregation of washed human platelets (Figure 2C), indicating that the binding between mAIM and hLBD is capable of inducing activation of human platelets. Based on these results, we conclude that mouse VWF contains similar, although apparently less protective, autoinhibitory properties as human VWF.

Cross reactivity of mouse and human VWF with GPIbα

We next probed the ability of the AIM-A1 fragments to bind the LBD of either species. It should be noted that both AIM-A1 fragments and the LBD are monomeric proteins, and all detectable binding was fitted to global, heterogenous ligand-binding model given a pronounced two-phase decay in the sensorgrams. Similar to the binding behavior of human A1 fragments to hLBD [21, 25, 26], hAIM-A1 showed little binding to mLBD at up to 2 μM, and htAIM-A1 bound to mLBD with a moderate affinity of 2.4 μM, although the slow phase dissociation of htAIM-A1 is less pronounced (Figure 3A, B). In comparison, mAIM-A1 bound to hLBD with an affinity of 515 nM, and the presumably more active mtAIM-A1 shows a lower affinity of around 1 μM although this behavior likely due to a fast off-rate (Figure 3E, F).

Figure 3. Binding of autoinhibited and activated AIM-A1 proteins to GPIbα LBD across human and mouse species.

AIM-A1 proteins, depicted at the top of each panel following the same format as in Figure 1, were assessed for binding to immobilized LBD by BLI using a dilution series from 2 μM to 63 nM. Association into each protein concentration was performed for 100 seconds followed by dissociation into kinetics buffer for an additional 500 seconds. Global fitting (red traces) was performed as in Figure 1. The sum of equilibrium responses from two-state equations for each concentration were fit to a hyperbola and K_D was derived as half-maximal binding (inset). Sensorgrams are reflective of two to three technical replicates.

The human AIM confers strong protection to the A1 domain, irrespective of species

The AIM largely controls the autoinhibition of A1 in human fragments. As the mouse fragments appear to be more active in their capacity to bind the LBD (Fig. 1, 3), we sought to test if the human AIM confers a stronger protective effect than the mouse AIM. To this end, a chimera AIM-A1 protein, in which the human AIM sequence flanks the mouse A1 (hAIM-mA1, Figure 4A), was generated. In a dose-dependent manner, hAIM-mA1 was able to induce apparent aggregation of platelets from either species and respond to ristocetin (Figure 4B, C).

Figure 4. . Chimeric human AIM-mouse A1 shows similar autoinhibition as observed in hAIM-A1.

(A) The human AIM was fused to the mouse A1 domain to yield a chimeric hAIM-mA1 protein. (B,C) Aggregation traces of mouse (B) and human (C) platelets induced by hAIM-mA1 at noted concentrations. 1.5 mg/ml ristocetin (R, red trace) could induce apparent aggregation in the presence of 60 nM hAIM-mA1. (D, E) Binding BLI sensorgrams of hAIM-mA1, in serial dilution from 2 μM to 63 nM, to mouse (D) and human (E) LBD. In either case, the affinity could not be determined as little binding responses were observed.

We next tested the ability of the chimera to bind to the LBD from either species. The human AIM appears to confer enhanced protection to mouse A1 as the binding response is barely detectable up to 2 μM. Such lack of binding of chimeric construct to hLBD and mLBD is similar to those observed with hAIM-A1 (Figures 4D, E;3B, C).

While it is apparent that various AIM-A1 constructs can bind to platelets and activate them, we wanted to ensure that the lack of binding to the LBD is due to the presence of AIM, and not the inability of A1 to bind to the LBD. To ensure the A1 still retains all the elements sufficient for LBD binding, we generated a gain-of-function hLBD variant that bears a platelet-type (PT) VWD, W230L mutation. This mutation increases the affinity of the LBD to VWF by increasing the binding affinity to the primary binding site on A1 [39, 40]. This mutated PT-LBD can overcome some of the protection provided by the AIM. hAIM-A1, mAIM-A1, and the chimera all show increased binding responses to the PT-LBD compared to the wild-type LBD (Supplemental Figure 5), indicating that the A1 domain in these constructs can still bind to the LBD.

The mouse AIM is weaker than the human AIM by single-molecule force spectroscopy

We previously demonstrated that the human AIM is unfolded by piconewton forces in single-molecule force spectroscopy [25]. We sought to define the unfolding parameters of the mouse AIM by the same optical tweezer setup to exert tension on the protein (Figure 5A). Representative DNA unfolding traces with accompanying protein unfolding events are shown in Figure 5B. Unfolding events in the plot of unfolding force versus unfolding extension are fitted to the worm-like chain model (Figure 5C). Pulling on mouse AIM-A1 showed a single unfolding event, with a contour length, L_C, of 21.8 ± 1.2 nm, which is significantly shorter than the contour length of human AIM at 26.6 ± 0.5 nm (Table 1). As the unfolding force is a function of the loading rate, these two parameters can be used to fit to the Bell-Evans model to yield unfolding force parameters. When compared to human AIM-A1, the mouse AIM-A1 unfolding force is similar to the human unfolding force (Figure 5D). A reduced contour length indicates that the mouse AIM is composed of fewer residues than the human AIM yet unfolds at similar forces. Substitution of the human AIM onto mouse A1 in the chimeric protein does slightly raise the unfolding force but more drastically increases the contour length considerably to 29.0 ± 0.4 nm. The increase in contour length suggests the human AIM may help to obscure the binding site of GPIbα without substantially increasing the contacts between the AIM and A1.

Figure 5. Single-molecule force spectroscopy reveals unfolding parameters of the mouse AIM.

(A) Schematic of single-molecule optical tweezer apparatus in this study. An N-terminal biotinylated tag and C-terminal SpyTag003 were used to conjugate the AIM-A1 protein to a streptavidin bead and SpyCatcher DNA handle, respectively. (B) Representative force-extension traces of mouse AIM-A1 (red) and chimera hAIM-mA1 (blue). The extension event in each trace is marked by an arrowhead. (C) Plots of unfolding force versus unfolding extension for noted AIM-A1 fragments and fits to the worm-like chain model. The human AIM-A1 unfolding data has been published previously and is overlaid here as a black dashed line for comparison. Force data are presented as mean values ± standard deviation, and extension data are presented as the peak of the Gaussian fit ± the full width at half maximum (FWHM) of Gaussian fit divided by the square root of counts. (D) Plots of unfolding force versus loading rate for noted AIM-A1 fragments and fits to the Bell-Evans model. Unfolding force data are presented as the center of the tallest bin of the histogram ± one-half of the bin width. Previous data from hAIM-A1 are shown in black dashed line with 99% confidence interval predictions of the upper and lower limits based on the error shown as fine dashed lines.

Table 1.

Single-molecule force spectroscopy parameters associated with unfolding events of various constructs and additives

Construct	Contour length LC (nm)a	Persistence length Lp (nm)a	Unstressed unfolding rate ${\mathit{k}}_{\mathit{u}}^{\mathbf{0}}$ (s−1)b	Barrier position γu(nm)b
Mouse AIM	21.8 ± 1.2	0.21 ± 0.02	0.074 ± 0.035	0.75 ± 0.13
Chimera AIM – one unfolding	29.0 ± 0.4	0.25 ± 0.01	0.027 ± 0.056	0.99 ± 0.24
Chimera AIM – two unfolding – lower force	28.0 ± 1.0	0.20 ± 0.01

^aContour length and persistence length are fitted worm-like chain model parameters. Uncertainties are the standard error of the fits.

^bUnstressed unfolding rate and barrier position are fitted Bell-Evans model parameters. Uncertainties are the standard error of the fits.

Discussion

Our investigation into the binding between both mouse and human GPIbα and VWF demonstrates the presence of A1-specific autoinhibition in mouse VWF, albeit with less protection compared to human VWF. Like its human counterpart, mouse AIM-A1 could dose dependently activate platelets from both humans and mice, although at supraphysiological concentrations. The removal of part of the AIM in the tAIM-A1 construct created a more potent agonist for platelet activation, although mouse AIM-A1 itself is more active than human AIM-A1. A more active VWF fragment may be more promiscuous in binding to non-optimized binding partners, as observed in mouse VWF binding to human GPIbα. Furthermore, the contour length of the mouse AIM is considerably shorter than the human AIM, suggesting that fewer residues flanking the A1 domain are folded and constitute the mouse AIM. As the unfolding forces of the human AIM and mouse AIM are not significantly different from each other, the mouse AIM likely shields less of the A1 domain from GPIbα, which could explain the higher binding activity of wild-type mouse AIM-A1 than human AIM-A1.

Previous studies on the species barrier related to VWF-platelet interaction have noted that while the combination of mouse VWF and human GPIbα can maintain hemostasis and prevent bleeding in hTg mice [41], the opposite combination of human VWF and mouse GPIbα is insufficient for hemostasis, as demonstrated by ex vivo perfusion studies on collagen,[42] hydrodynamic injection [15, 16], and genetic replacement [16]. Similarly, while human platelets appear to readily agglutinate with plasma of pig, cow, sheep, and goat [43], human VWF is a poor binder to pig platelets [44]. However, the molecular mechanism underlying such apparent one-way species barrier has not been clear. Investigations in this area have produced disparate results and incomplete explanations. We note that most of these studies have focused on the species difference in the A1 domain or the A1/LBD binding interface. Our results, particularly that of the hAIM-mA1 chimera, suggest that the difference in VWF activity between species may also result from the sequence difference in the AIM. Since the AIM is a key element in VWF autoinhibition and its response to elevated shear flow, the difference between human and mouse AIM could help explain many previously reported VWF-platelet interactions, or lack thereof. For instance, the weak binding of human tAIM-A1 to mouse GPIbα coupled with a fast off rate could explain the inability of human VWF to initiate adhesion of mouse platelets and thus the bleeding tendency in mice expressing human VWF [15, 16]. In addition, a larger AIM in human VWF offers stronger autoinhibition that the mouse AIM, which is congruent with the lower activity of human VWF [15, 16, 42]. Sequence differences result in the mouse AIM having more positive charge and possibly different O-linked glycosylation, further increasing positive charge. Given the importance of O-glycans [27, 45, 46], negatively charged amino acids on both VWF autoinhibition [47] and A1-LBD guidance [48], an increase in positive charge may change the function of the AIM and facilitate the A1-LBD interaction.

It was reported in a previous study that mutation of residue 1326 in mouse VWF from arginine to histidine could allow for human platelet binding [42]. It was further reported that this mutation inhibited mouse VWF binding to mouse platelets [14]. However, this mutation is not necessary for mouse VWF to bind human GPIbα and maintain hemostasis in hTg mice and another similar strain [9, 16]. Navarrete et al. were unable to demonstrate this mutation to the mouse residue within human VWF corrects hemostasis [15]. Replacement of key binding interfaces within human VWF (residues 1326–1333 and 1370–1385 in the A1 domain) to mouse specific sequences was unable to correct bleeding in mice, suggesting that this mutant human VWF may not be properly activated in the mouse, possibly due to the strong protection provided by the human AIM [15]. Additional evidence for a weaker mouse AIM can be found in a recent investigation into ancestral reconstruction of VWF [49]. Compared to human VWF, full-length variant An88 derived from a lineage of rodents shows increased GPIbα-dependent activity, whereas An101 from a lineage of primates shows decreased activity, which the authors attributed to sequence differences within the AIM, not A1 [49].

The binding of VWF to platelets must occur under high shear conditions, and under force, the human LBD-A1 interaction is strengthened, i.e. a catch-bond is formed [50]. This catch-bond behavior is essential to slow platelets down during elongational flows that occur during vessel injury as a result of vasoconstriction. Elimination of the catch-bond behavior by mutations in either GPIbα or VWF results in a bleeding diathesis [51, 52]. The GPIbα-A1 binding interface occurs over two discontinuous sections. The primary binding site on A1 is always exposed, whereas the secondary binding site is obscured by the AIM [21, 53]. When both binding sites are available, the reinforced catch-bond can form more readily. The disappearance of the slow phase decay in activated tAIM-A1 constructs to the opposite species LBD suggests that the conformational change associated with catch-bond behavior could be impaired. From these results, it is clear that the LBD can sample the primary binding site of A1, regardless of species, as there is an instantaneous equilibrium binding response, but the reinforcement of this interaction by the secondary binding site of A1 is altered given the loss of the slow phase decay (Figure 3B, F). Specific residues within the leucine rich repeats of GPIbα have previously been identified as important mediators of the catch bond and may not be conserved between species [54].

The function of a balanced hemostatic system is essential to the survival of mice and humans alike. While there are unique features between human and mouse platelets, as well as their VWF multimers, the main features of shear activation of VWF and VWF binding to platelets are common in both species and all mammals. The exposure of the A1 domain is carefully controlled to avoid platelet accumulation in the absence of injury. The presence of the AIM makes it possible to partly separate the machinery regulating VWF activity from residues that directly mediate the A1-LBD interaction. Conceivably, the AIM could evolve independently of the A1 domain, thus enabling VWF in different organisms to adapt to different shear stresses and at the same time maintain its platelet-binding activity.

Essentials.

• Human and mouse VWF exhibit differential activities towards platelets.

• The autoinhibitory module (AIM) flanking mouse VWF A1 domain modulates its activity

• The AIM is less protective in mouse VWF compared to human VWF

• Mouse AIM that is smaller than human counterpart but unfolds under similar force