Interaction of von Willebrand factor domains with collagen investigated by single molecule force spectroscopy

Interaction of von Willebrand factor domains with collagen investigated by single molecule force spectroscopy Sandra Posch,1,a) Tobias Obser,2 Gesa König,2 Reinhard Schneppenheim,2 Robert Tampé,3 and Peter Hinterdorfer1,a) 1Institute of Biophysics, Johannes Kepler University, Linz, Austria 2Department of Pediatric Hematology and Oncology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany 3Institute of Biochemistry, Biocenter, Goethe-University Frankfurt, Frankfurt/Main, Germany


I. INTRODUCTION
von Willebrand factor (VWF) is a multimeric plasma glycoprotein critical to primary hemostasis. 1,2It is the essential adhesive protein in mediating platelet adhesion and aggregation during vessel injury.The mature VWF is composed of several domains (2050 amino acids) and interacts with the coagulation factor VIII-an essential blood clotting protein (FVIII), collagen, heparin, and platelet glycoproteins GPIb and GPIIb-IIIa. 3 In early hemostasis, VWF immediately binds to sub-endothelial collagen at the site of injury and acts as a connection between collagen and blood platelets, which induce initial clot formation (Fig. 1).The general understanding is that shear activated elongation of the VWF multimer structure leads to accessibility of the binding sites for collagen I, II, III, and VI in the A1 and A3 domains of VWF. 4 Recent studies show that the VWF can also interact with collagen IV. 5 Collagen III is believed to be essential for the interaction with the A3-domain, and collagen VI is believed to be essential for the interaction with the A1-domain (Fig. 5, left).Even though extensive binding studies using collagen and/or VWF have a) Authors to whom correspondence should be addressed: sandra.posch@jku.at and peter.hinterdorfer@jku.atbeen made, [6][7][8][9][10][11][12][13][14][15] direct experimental investigations on the binding dynamics and energies of the A-domains and clinical relevant A-domain mutations (p.His1786Arg, p.Gln1734His, 16,17 and p.Ser1731Thr 18 ) upon binding to collagen types III and VI are missing.The latter mutations fit to the current von Willebrand disease (VWD) classification model as type 2M:CB, with a qualitative defect in collagen binding but a normal multimer distribution. 19Surprisingly, patients suffering from these VWD type 2M:CB point mutations (p.His1786Arg, p.Gln1734His, and p.Ser1731Thr) exhibit mild or no bleeding symptoms. 20Yet it remains unclear why patients carrying these severe mutations in the VWF A3 collagen binding domain exhibit no or only mild bleeding symptoms.We thus performed VWF/collagen studies on the single-molecule level, utilizing atomic force microscopy (AFM) based single molecule force spectroscopy (SMFS).SMFS offers the potential to directly probe the forces and dynamics [21][22][23] of specific VWF-domain/collagen interactions.

A. Chemicals
3-Aminopropyl-triethoxy silane (APTES, Sigma Aldrich, Vienna, Austria) was distilled at low pressure and stored FIG. 1. Simplified biological system.(a) In the resting state, e.g., under low shear-stress conditions, with no injured vessel wall, resting VWFs and resting blood platelets circulate around in blood.In this state, VWF is incapable of mediating platelet adhesion.(b) After an injury, at higher shear stress conditions, the activated VWF interacts with the exposed collagen in the sub-endothelial matrix via VWF domains A1 and A3 and triggers the adhesion of activated platelets and thus wound closure (red ellipse).These interactions between VWF and collagen are also the area of interest in this study.

C. Plasmid constructs
The cDNA's coding for recombinant human VWF constructs containing the A1A2A3 (aa 1230-1874) or the A1A2 (aa 1230-1672) construct (all domain constructs with 6× Histag) was cloned into the mammalian expression vector pIRES neo2. 29Mutations were inserted by site-directed mutagenesis, employing the QuickChange kit (Stratagene).All primers are available upon request.The vectors were sequenced and used to transform Top10 supercompetent cells (Invitrogen).Plasmid purification was performed using the Endofree Plasmid Maxi Kit (QIAGEN).

D. Cell culture and expression of VWF constructs in 293 cells
HEK-293 cells were cultured in Dulbecco Modified Eagle Medium (DMEM, Invitrogen) with 10% [v/v] fetal bovine serum (Invitrogen) and 1% penicillin/streptavidin at 37 • C and 5% CO 2 .The cells were transfected with the VWF vectors using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.Recombinant expression of VWF variants was performed as previously described. 30

E. Protein purification
The His-tagged VWF domain constructs were purified by employing the His-Pur Ni-NTA resin (Thermo Scientific) according to the manufacturer's instruction for purification of His-tagged proteins using a gravity-flow column.

F. Tip and sample chemistry
The tip functionalization protocol was optimized regarding the reproducibility, stability, and probability to bind statistically one single ligand at the outer tip apex.A desiccator (5 l) was flooded with argon gas to remove air and moisture.Then two small plastic trays were placed inside the desiccator; 30 µl of APTES and 10 µl of triethylamine were separately pipetted into two trays, the clean AFM tips were placed nearby on a clean inert surface, and the desiccator was closed.After 120 min of incubation, APTES and triethylamine were removed, the desiccator was again flooded with argon gas for 5 min, and the tips were left inside for two days in order to cure the APTES coating.
The APTES tips were incubated in 0.5 ml of a 1 mg/ml solution of maleimide-PEG 27 -NHS in chloroform containing 0.5% (v/v) of TEA as catalyst for 2 h.Subsequently, the tips were rinsed in chloroform (3×) and dried in a gentle stream of nitrogen gas.The cantilevers were then placed in a polystyrene Petri dish, and a mixture of 100 µl disulfide-tris-NTA (1 mM in MilliQ water), 2 µl EDTA (100 mM, pH 7.5 in MilliQ water), 5 µl HEPES (1M, pH 7.5), 2 µl TCEP hydrochloride (100 mM in MilliQ water), and 2.5 µl HEPES (1M, pH 9.6) was pipetted on the tips and incubated for 2 h.Afterwards the tips were washed in TBS buffer (3 × 5 min).Again, the cantilevers were placed in a polystyrene Petri dish and pre-loaded with 50 µl NiCl 2 (200 µM in TBS buffer, pH 7.5) for 5 min.Afterwards, 100 µl of the his6-tagged protein was mixed with 4 µl NiCl 2 (5 mM) and again incubated for 2 h.At the end, the tips were washed 3 times for 5 min in TBS. 25 The method to prepare a collagen surface suitable for SMFS is presented in Ref. 31: A small amount of PEG 27 -diamine was pipetted on a Nexterion epoxy slide and then placed in a drying stove for 4 h at 70 • C. Afterwards the slide was rinsed with a warm mixture of chloroform/methanol/acetic acid (v/v/v = 70/30/4) and then rinsed two times with the same mixture at room temperature.Then the slide was dried with nitrogen. 32Next, 0.5 ml of a 1 mg/ml solution of acetal-NHS in chloroform containing 0.5% (v/v) of TEA as catalyst was incubated for 2 h.Subsequently, the slide was rinsed in chloroform (3 × 5 min) and dried in a stream of nitrogen.To obtain the aldehyde function, the slide was immersed for 10 min in citric acid solution [1% (w/v) citric acid dissolved in MilliQ H 2 O], washed 3 times in MilliQ water, and dried under N 2 .The slide was then set on a clean dry Petri dish and a drop of collagen solution (100 µl, 0.15 mg/ml in PBS, pH 7.4) was placed on the slide.This allowed the collagen proteins to covalently bind to the aldehyde functions via their lysine residues.Finally, 2 µl 1M NaCNBH 3 (32 mg NaCNBH 3 , 50 µl 100 mM NaOH in MilliQ H 2 O, 150 µl MilliQ H 2 O) were added to the drop and carefully mixed with the pipette.The solution was allowed to react for 1 h. 10 min before washing the slide, 5 µl ethanolamine (1M, in MilliQ H 2 O, pH 8.0) was added to the drop to passivate unreacted aldehyde groups.The slides were then washed 3 times in PBS buffer.

G. AFM cantilever
For SMFS experiments, silicon nitride MSCT tips (Bruker Corporation, MA, USA, C-cantilever, 0.010 N m 1 nominal spring constant) were used.The actual spring constant was determined according to Hutter and Bechhoefer 33 using the thermal noise method.

H. Single molecule force spectroscopy (SMFS)
SMFS measurements were performed on a Pico SPM Plus setup (Agilent Technologies, Chandler, AZ, USA) under near physiological conditions.VWF domains [wild type (wt) VWF A1A2A3, wt VWF A1A2, and three mutations p.His1786Arg, p.Gln1734His, and p.Ser1731Thr in the A3 domain of construct A1A2A3] were coupled to AFM tips and collagen III or VI was covalently immobilized on the sample surface, respectively [see Fig. 2(a)].We characterized specific binding of the wt and mutants to both collagen types III and VI.Force distance cycles (FDCs) were acquired at room temperature in TBS buffer, by approaching the tip towards the collagen surface, followed by its retraction.The deflection of the cantilever was continuously monitored in relation to the distance between the tip and the surface.Close to tip-surface contact, the tip-adorned VWF can form a specific molecular bond with collagen on the surface.In this case, an unbinding event is visible in the retraction phase of the force distance cycle.Specific interactions between VWF A-domain constructs and collagen types III and VI were discerned from nonspecific adhesion by a characteristic roughly parabolic force signal arising from the elastic properties of the PEG-linker through which the VWF is covalently coupled to the tip [Fig. 2 The dynamics and the interaction energy landscapes of binding of the VWF A-domains to collagen III and VI was probed by conducting FDCs at different loading rates, whereby the loading rate r is given by the force increase applied to the molecular bond during retraction.At least 1000 force distance cycles (FDCs) were recorded for each loading rate r.We recorded several velocities (50, 100, 200, 400, 600, 800, 1200, 2000, 3000 nm/s).The unbinding force was plotted against the logarithm of the loading rate r and resulted in a linear dependence, according to the single energy barrier binding model. 34he data points (green dots, each of which represents a single unbinding event) were fitted using a maximum likelihood approach as described 35 [Fig.2(c)].The thick blue line represents the best fit for the data.From such fits, the kinetic off rate, k off , and x β , the distance from the free-energy minimum to the dissociation energy barrier, were attained.k off represents a measure of the bond stability, whereby the bond life time τ is defined as the inverse of the kinetic off rate.x β depicts the distance of the energy barrier between the bound and the unbound states and might be related to the length of the binding pocket in a more structural picture.However, the bond life time τ is of main importance for conclusions about the in vivo effects.x β values are used to underpin our findings [Fig.4(a)].The error bars of the k off and x β values in this work show the inaccuracy of the fit.Non-overlapping error bars were used as an indicator for significant differences of the data sets.Control experiments (Fig. S1 of the supplementary material) and raw data (Figs.S2-S10 of the supplementary material) for all the different systems can be found in the supplementary material.

III. RESULTS AND DISCUSSION
Many of the multiple functions of VWF have been related to the multi-domain structure of VWF that contains binding sites for the platelet membrane glycoprotein Ib, collagen, factor VIII, sites for dimerization and multimerization, and cleavage sites for furin and ADAMTS13.In this study, we focused on the interaction between VWF and sub-endothelial collagen, which is an initial and crucial step in wound closure.Collagen types III and VI belong to the most important compounds of tissues and thus provide the perfect interaction partner for our VWF binding studies.
In a first attempt, we used a construct containing all VWF A domains (A1A2A3) as a specific counterpart for collagen binding.We compared the bond life time of single interactions between the VWF A1A2A3 construct and collagen type III or VI and found that these domains bind to collagen type III significantly stronger than to collagen type VI.The bond length, x β , was significantly higher for the VWF A1A2A3/collagen III interaction [Figs.3(a the tables reflect the error of the fit], too.Collagen type III appears to be far more abundant in artery walls and skin than collagen type VI, which makes it the prime binding partner for VWF. Malfunctions in the VWF/collagen interplay must be taken very serious as the VWD is the most common bleeding disorder and may be directly related to such perturbations.In this light, we further characterized specific binding of three different A1A2A3 constructs, each containing a VWD type 2M:CB mutation (p.Ser1731Thr, p.His1786Arg and p.Gln1734His) in VWF domain A3, to both collagen types III and VI.
We initially used the p.Ser1731Thr mutation.In our experimental setting, this construct showed no significant difference in bond life time when compared to the wild type construct on both collagen types III and VI [Fig.3(a), Table II].Our results suggest that the p.Ser1731Thr mutation does not crucially affect the VWF/collagen interactions, which is in good agreement with the former observation that patients carrying this mutation exhibit no or only moderate bleeding symptoms. 16Nevertheless, the x β values appeared to be significantly higher for the mutant systems [Fig.4(b),  II].Serine and threonine are both neutral and polar amino acids.Nevertheless, compared to serine, threonine features one more C-C single bond and thus is approximately 1.5 Å bigger in size.This slightly bigger mutant might cause sterical hindrance in the binding pocket and cause the higher bond length (xβ).
For comparison, we used the A3-domain mutations p.His1786Arg and p.Gln1734His.In contrast to the previously studied p.Ser1731Thr construct, both A1A2A3 peptides formed a slightly more stable bond with collagen type III when compared to wt A1A2A3 [Fig.3(b) and Table III].As described earlier, 36 arginine and histidine residues within VWF are involved in its interaction with collagen by formation of hydrogen bonds to collagen residues.Mutation of a histidine to an arginine residue introduces a slightly more basic residue in the protein that might increase the life time of the bond.In the case of mutant p.Gln1734His, a neutral residue is exchanged with histidine, thereby introducing an additional basic residue that could provide an additional hydrogen bond.x β was, as before, higher for the mutant system [Fig.4(b) and Table III].This effect probably results from additional  hydrogen bonds and indicates that the mutant forms longer bond lengths.
We then took the same mutants (p.His1786Arg and p.Gln1734His) but switched to collagen type VI as a receptor surface (Table IV).In this configuration, the bond life time of the mutants showed a drastic increase, compared to the wt system [Fig.3(b), Table III].Both the VWF A1 and A3 domains are capable of binding collagen.VWF domain A1 was reported to be essential for the interaction with collagen type VI, 37 and VWF domain A3 was suggested to be the main binding domain for collagen type III 38 (see also Figs. 1 and 5).However, with a defect VWF A3 domain, the interaction between VWF and collagen was suspected to be strongly hampered.Nevertheless, patients who carry these mutations showed no bleeding abnormality or just weak symptoms.Along this line, it was suggested by Bonnefoy et al. 39 and Flood et al. 17 that domain A1 might compensate for the VWF A3 domain mutation p.His1786Arg under shear flow conditions.We now provide experimental evidence for this assumption for single VWF domain constructs.Our results show that A1A2A3 constructs with the severe p.His1786Arg and p.Gln1734His mutations in domain A3 significantly increase the binding to collagen.This finding leads to the assumption that VWF domain A1 increased its binding strength to compensate for the defect A3 domain collagen binding.We hypothesize that the mutant VWF A3 lost the capability to bind to collagen.This missing interaction could be compensated by additional hydrogen bonds between VWF and collagen, but mainly by a much stronger binding between VWF domain A1 and collagen.Constructs where the VWF domain A3 was missing showed the same result [Fig.3(c)].In vivo this suggests that a stable VWF binding to collagen can be established, even if there is a severe mutation in one of the collagen binding domains.Due to this self-regulatory process, the patient will not suffer of any bleeding symptoms.The p.Ser1731Thr mutation has no effect on wound healing and blood clotting (Fig. 5).Thus, our findings confirm this hypothesis and additionally suggest that the mutations could even stabilize the interaction between VWF and collagen without shear.
In summary, we investigated the interaction strength between three mutations in the A3-domain of the VWF A1-A2-A3 construct (p.His1786Arg, p.Gln1734His, and p.Ser1731Thr) and collagen type III or VI and compared it to the wt A1-A2-A3 construct.The interactions between the p.Ser1731Thr mutant and collagen III showed no effect, while the bond life time slightly increased for the p.His1786Arg and p.Gln1734His mutants, most likely due to additional hydrogen bonds that can be formed with the latter mutants (Fig. 5, gray box, studies on collagen III surfaces).
With collagen type VI, the p.Ser1731Thr mutation had no effect as well.p.His1786Arg and p.Gln1734His were found for a dramatic increase in bond stability in contrast to the wt system.These results suggest that VWF domain A1 can compensate for severe mutations in VWF domain A3 when it comes to collagen binding (Fig. 5, blue box, studies on collagen VI surfaces).
Our data allow us to derive a fairly detailed molecular picture on the interplay between collagen and VWF domains and strengthen the hypothesis that defective interactions between one VWF and collagen can be compensated by alternative binding strategies to preserve functionality.These findings might provide an explanation for the mild or even inconspicuous symptoms in patients carrying the here described VWF A3 mutations.

SUPPLEMENTARY MATERIAL
See supplementary material for control experiments (Fig. S1) and raw data (Figs.S2-S10) for all the different systems.

4 Posch
FIG. 2. Force spectroscopy of VWF A-domain/collagen interactions.(a) The AFM tip was functionalized with VWF A-domains.Collagen was covalently bound to glass slides.(b) Example of a force-distance cycle showing a specific VWF A-domain-collagen unbinding event.[Inset in (b)] In the blocking experiment, no specific unbinding events occurred.(c) Loading rate dependence plot.A scatter plot of the rupture force as a function of the loading rate of all measured binding events is shown.The blue line represents the most probable fit of the data cloud.k off and x β can be obtained by fitting the data.
FIG. 3. Comparison of the bond life time τ of VWF A1A2A3 wt and mutant (p.His1786Arg, p.Gln1734His, and p.Ser1731Thr) binding to collagen III and VI.

FIG. 5 .
FIG. 5. Graphical conclusion.Experiments using collagen type III are shown in the gray box and the measurements using collagen type VI in the blue box.

TABLE I .
Bond life time and x β values for the interaction of wt VWF A1A2A3 with collagen type III or VI.

TABLE II .
Bond life time and x β values for wt and mutant (p.Ser1731Thr) VWF/collagen type III or VI interactions.

TABLE III .
Bond life time and x β values for wt and mutant (p.Gln1734His and p.His1786Arg) VWF/collagen type III interactions.