Specific electrostatic interactions between charged amino acid residues regulate binding of von Willebrand factor to blood platelets

The plasma protein von Willebrand factor (VWF) is essential for hemostasis initiation at sites of vascular injury. The platelet-binding A1 domain of VWF is connected to the VWF N-terminally located D′D3 domain through a relatively unstructured amino acid sequence, called here the N-terminal linker. This region has previously been shown to inhibit the binding of VWF to the platelet surface receptor glycoprotein Ibα (GpIbα). However, the molecular mechanism underlying the inhibitory function of the N-terminal linker has not been elucidated. Here, we show that an aspartate at position 1261 is the most critical residue of the N-terminal linker for inhibiting binding of the VWF A1 domain to GpIbα on platelets in blood flow. Through a combination of molecular dynamics simulations, mutagenesis, and A1–GpIbα binding experiments, we identified a network of salt bridges between Asp1261 and the rest of A1 that lock the N-terminal linker in place such that it reduces binding to GpIbα. Mutations aimed at disrupting any of these salt bridges activated binding unless the mutated residue also formed a salt bridge with GpIbα, in which case the mutations inhibited the binding. These results show that interactions between charged amino acid residues are important both to directly stabilize the A1–GpIbα complex and to indirectly destabilize the complex through the N-terminal linker.

The protein von Willebrand factor (VWF) 4 is essential to initiate hemostasis at sites of vascular injury with high blood flow (1). The A1 domain of VWF binds to glycoprotein Ib␣ (GpIb␣) on the surface of blood platelets (2), whereas the A3 domain binds to collagen in the exposed endothelium. This provides an initial tethering mechanism of platelets to the site of vascular injury. Under normal conditions, circulating VWF does not significantly bind to platelets (3). However, VWF becomes activated either under high shear stress (4) or when it is bound to collagen at the site of exposed endothelium (2). These characteristics motivated our investigation into the structural properties that modulate the binding of the A1 domain to GpIb␣.
The A1 domain of VWF presents a globular fold (Fig. 1a) with a relatively long N-terminal sequence that links it to the neighboring DЈD3 domain and a shorter C-terminal sequence that connects to the A2 domain (Fig. 1b). An important feature of the A1 domain is the presence of a disulfide bond between Cys 1272 and Cys 1458 (5) (Fig. 1a). We refer to the region within the loop defined by this disulfide bond (Fig. 1b) as the "core" A1 domain and the 34 amino acids connecting Cys 1237 at the end of the DЈD3 domain to Cys 1272 at the start of the core A1 domain as the N-terminal linker (Fig. 1b). There is compelling experimental evidence that the N-terminal linker plays an inhibitory function in the binding between A1 and GpIb␣. One of the first studies that provided such evidence showed that a construct with no N-terminal linker bound to GpIb␣ much more strongly in static conditions than both a construct containing half of the N-terminal linker and another construct that contained the entire linker and part of the DЈD3 domain (6). In another study, binding of platelets to recombinant A1 domain immobilized on a surface was tested in dynamic flow conditions (7). Variants of A1 that contained the entire N-terminal linker and part of the DЈD3 domain (Fig. 1b) mediated rolling adhesion in a similar fashion as full-length VWF. In contrast, variants of A1 without the N-terminal linker (starting at position 1271 right next to the disulfide bond) mediated strong stationary adhesion (7).
Unfortunately, it is difficult to infer the molecular mechanism behind the inhibitory function of the N-terminal linker because currently no crystallographic structures of the A1 domain have been reported that include the entire N-terminal sequence. Of the 34 amino acids in the N-terminal linker, a maximum of 13 have been included in the constructs used in the crystallographic studies, and the electron density and coor-dinates were resolved for at most 11 of these amino acids. The structures with these 11 residues resolved were the isolated A1 (5) and the complex of A1, GpIb␣, and botrocetin (8) (shown in Fig. 1, although botrocetin is not displayed). These 11 amino acids are located in proximity of the rest of the A1 domain in both of these structures, but they are not seen to interact with GpIb␣ in the structure of the complex (Fig. 1). This leaves the question open of how these 11 and the remaining 23 amino acids regulate binding of A1 to GpIb␣.
The role of the N-terminal linker in regulating the A1 domain became even more intriguing after the discovery through single molecule force spectroscopy experiments that the lifetime of the bond between A1 with the entire N-terminal linker and GpIb␣ increases with force (9), which is the characteristic of a catch bond. Catch bonds have also been observed for the bacterial adhesin FimH (10), myosin (11), selectins (12), and integrins (13) bound to their respective targets. A later single molecule force spectroscopy study showed that shortening the N-terminal linker to the 11 amino acids closest to the disulfide bond (i.e. starting at position 1261 similarly to the sequence seen in X-ray structures; e.g. Fig. 1) eliminates catchbond behavior and converts the bond between A1 and GpIb␣ to a slip bond (14), for which lifetime decreases with increasing tensile force. However, two studies drew opposite conclusions about why the entire N-terminal linker was required for catch bond behavior and shear-enhanced adhesion. In one study, truncation of 11 amino acids in the linker preferentially weakened binding at high force and high flow (14), whereas in the other study, the same truncation preferentially strengthened binding at low force and low flow (15). To understand the activation of VWF at the molecular level, it is necessary to not only understand how the N-terminal linker regulates A1 in flow or force but also to identify amino acids that are key to its function.
In the present study, we searched systematically for key amino acids in the N-terminal linker that are relevant to the binding function of the A1 domain. The results confirmed that the N-terminal linker has an inhibitory function, and a molecular mechanism behind this observation is presented. First, mutagenesis and flow chamber experiments were used to identify regions and amino acids of the N-terminal linker that activate A1 when deleted or mutated. These experiments showed that deleting or mutating Asp 1261 into a non-negatively charged amino acid dramatically activated the A1 domain at any shear stress. Then molecular dynamics (MD) simulations were run to provide a structural explanation for how Asp 1261 reduces binding of A1 to GpIb␣. The model suggested by the simulations was then validated by further flow chamber experiments. Overall, this study shows how the specific interaction between Asp 1261 in the N-terminal linker and the rest of A1 is crucial for the auto-regulatory function of this protein.

Flow chamber experiments with different constructs
To identify key amino acids in the N-terminal linker that regulate binding, we measured binding of GpIb␣ to various truncated A1 domain constructs as follows. The A1 domain constructs were anchored to a surface through a C-terminal biotin tag, and the binding domain of GpIb␣ was anchored to microbeads via a C-terminal biotin tag. The beads were allowed to settle onto the surface, washed at 2 dyn/cm 2 fluidic wall shear stress, and then subjected to stepwise increasing or decreasing flow, whereas their movement was recorded with videomicroscopy. For the purpose of this manuscript, a wild-type construct is named A1_n, where n indicates the position of the N-terminal amino acid (e.g. A1_1238 denotes a A1 construct starting at position 1238 and thus containing the entire N-terminal linker). Constructs bearing a single-point mutation will be identified with A1_n_x, where x is the name of the particular mutation (e.g. D1261Q).
When surfaces were coated with the A1_1238 construct with the full N-terminal linker, the GpIb␣-coated beads moved rapidly across surfaces (Fig. 2a) and detached as shear stress decreased (Fig. 2b). This shear-enhanced rolling adhesion is consistent with that observed previously for platelets of GpIb␣coated beads binding to isolated A1 domain with the full N-terminal linker (15,14) and for full-length VWF (16). In comparison, when the surface was coated with the A1_1261 construct that contains only the last 11 amino acids of the N-terminal linker, beads rolled more slowly at all shear stresses (Fig. 2a). This indicates that the construct with the shorter linker mediated stronger adhesion, consistent with previous studies indicating that the first 23 N-terminal amino acids of the A1 domain that are present in A1_1238 but not in A1_1261 ( Fig. 1) play an inhibitory function (6,15).
Importantly, GpIb␣-coated beads binding to immobilized A1_1261 still demonstrated shear-enhanced adhesion in which GpIb␣ beads detached below a shear threshold (Fig. 2b), demonstrating that the 23 amino acids between 1238 and 1261 were Figure 1. Complex between VWF A1 and GpIb␣ and amino acid sequence of the A1 domain. a, three-dimensional structure of the complex between A1 and GpIb␣ used in this manuscript. The structure was derived from the complex co-crystallized with botrocetin (PDB code 1U0N) (8), because this is the only structure of the complex where 11 amino acids of the A1 N-terminal linker were resolved (other structures had less). The backbone of the A1 N-terminal linker resolved in the X-ray structure (containing 11 amino acids) is highlighted in magenta. The disulfide bond between Cys 1272 and Cys 1458 is shown. Also indicated are the anchoring of the C terminus of GpIb␣ to blood platelets or beads and the attachment of the C terminus of the A1 domain to either a neighboring domain or directly to the surface. b, schematic representation of the A1 domain core with the disulfide bond and N-and C-terminal linkers. The elements in this figure are not to scale.

Specific electrostatic interactions regulate VWF
not needed for shear dependent adhesion in these studies. In contrast, GpIb␣-coated beads were nearly stationary and did not detach at low flow when bound to the A1_1261, A1_1267, and A1_1272 constructs that contain 10 or fewer amino acids in the N-terminal linker (Fig. 2, a and b). These observations suggest that the aspartate at position 1261 might play a key role in the inhibition of binding between A1 and GpIb␣ and in shearenhanced adhesion because its removal strongly activates binding and eliminates shear enhanced adhesion.
To determine why removal of residue 1261 is so critical, A1_1261 variants with single-point mutations were expressed in which the negatively charged aspartic acid at position 1261 was mutated into the neutral amino acids glutamine (D1261Q) or asparagine (D1261N) or the positively charged amino acid lysine (D1261K). All three variants supported adhesion with significantly lower bead-rolling velocities at low and high shear compared with A1_1261 ( Fig. 2c) and with no detachment of beads at low flow (Fig. 2d). Similar results were observed when the D1261N and D1261K substitutions were introduced into A1_1238 with the longer linker (Fig. 3a). In contrast, when aspartic acid was replaced with glutamic acid to maintain the negative charge (D1261E), rolling velocities remained unchanged compared with wild-type A1 constructs with the same length N-terminal linker (Fig. 3b). Together, these results dem- The error bars represent the standard deviation of values from three independent experiments. An asterisk indicates that the difference between two values is statistically significant, i.e. p Ͻ 0.05. The p values were calculated from a one-tailed Student's t test. For simplicity, the statistical significance is indicated between groups of values. One exception is that at low shear, the velocity of beads rolling over A1_1272 (green line) is not different from the velocity of beads rolling over A1_1261 (black line) in a. This is due to an artifact in one of the experiments where new beads entered the field of view and were still counted. This was an isolated event, which could have been due to an unusually large concentration of injected beads, an inhomogeneous protein concentration on the plate, or a combination of the two.

Specific electrostatic interactions regulate VWF
onstrate that the inhibitory function of the N-terminal linker involves ionic interactions between the residue at position 1261 and the rest of the bound complex.
To understand whether Asp 1261 plays a key inhibitory role in a more physiological context, similar experiments were performed with blood platelets instead of GpIb␣-coated beads. Again, the D1261N and D1261K mutations activated adhesion as indicated by a decrease in platelet rolling velocity (Fig. 3c). To test the importance of Asp 1261 in full-length VWF (flVWF), we expressed recombinant multimeric flVWF with various mutations at position 1261. The two variants with a negative charge at position 1261 (wild-type flVWF and flVWF_D1261E) mediated nearly identical shear-enhanced rolling adhesion of platelets ( Fig. 4, a and b), whereas variants with a neutral charge (D1261Q) or a positive charge (D1261K) at position 1261 mediated strong stationary binding to blood platelets even at low flow ( Fig. 4, a and b). These results indicate the importance of the negative charge at position 1261 for regulating platelet adhesion to VWF.
Finally, we asked whether the change in binding strength caused by elimination of Asp 1261 is likely to be physiologically relevant by comparing these engineered mutations to several mutations that are associated with type 2B von Willebrand disease, a bleeding disorder in which VWF binding to platelets is enhanced, leading to depletion of its largest multimers from the blood (17). Variants of flVWF with neutral (D1261Q) or positive (D1261K) charge at residue 1261 mediated binding with similar level of activation as flVWF variants with the following reported type 2B mutations: H1268D, A1461D, R1341D, and R1306W (Fig. 4, c and d). This indicates that the inhibitory function of Asp 1261 in the A1 domain is of crucial importance to prevent the pathological activation of VWF.

Inhibitory function of Asp 1261 investigated by simulations and mutagenesis
To investigate the inhibitory function of Asp 1261 at a structural level, MD simulations were run with the complex between A1 and GpIb␣. Two simulations were performed using wild-type A1_1261 and two using the single-point mutant A1_1261_D1261K. The A1-GpIb␣ complex was derived from the X-ray structure crystallized with the snake venom botrocetin (the latter was removed) because this is the only available X-ray structure of the complex where electron density for the N-terminal linker of A1 is visible starting at position 1261 (see "Materials and methods" for details about the setup). However, the structure used here is very similar to the complex between A1 and GpIb␣ crystallized without botrocetin except that in the latter, the N-terminal linker is visible starting only at residue 1269; after superimposition, the C ␣ root mean square deviation (RMSD) of those residues visible in both structures was 1.25 Å.
In all four simulations, the A1-GpIb␣ complex was stable as indicated by the RMSD from the initial structure (supplemental Fig. S1), which is consistent with previous MD studies (18). The individual proteins GpIb␣ and A1 showed even less flexibility, whereas the core A1 domain (defined as residues 1272-1461) was most stable (supplemental Fig. S1). Salt bridge analysis revealed that in wild-type A1 Asp 1261 engaged in salt bridges with some positively charged side chains in core A1, i.e. Arg 1308 ,

Specific electrostatic interactions regulate VWF
Lys 1312 , Arg 1334 , and Arg 1336 (Fig. 5, a and b). However, in the simulations with the single-point mutant D1261K, Lys 1261 did not form salt bridges with any residues located in core A1. Instead, in one run it formed a stable salt bridge with Glu 5 of GpIb␣ (Fig. 5, c and d), whereas in the other run the N-terminal linker was very flexible, as indicated by the root mean square fluctuations (RMSF) of C ␣ atoms (Fig. 6), and consequently Lys 1261 did not significantly engage in salt bridge formation with either GpIb␣ or the A1 core domain.
The differing patterns of salt bridge formation between aspartate and lysine at position 1261 suggested two contrasting hypotheses why A1_1261_D1261K is much more active in binding to GpIb␣ than A1_1261. One explanation could be that Lys 1261 forms a stable salt bridge with Glu 5 in GpIb␣, thus providing an additional interaction between the two proteins, which is not possible with wild-type A1 where an aspartate is present at position 1261. Another plausible explanation is that Asp 1261 forms salt bridges with core A1, thus anchoring the N-terminal linker in such a way that it can interfere with the bond with GpIb␣. Experimental evidence that mutations of Asp 1261 to neutral residues also activate the A1 domain suggests that the disruption of salt bridges with core A1 is likely the major mechanism.
To validate the role of salt bridge interactions in anchoring the N-terminal linker and inhibiting the A1 domain, variants with single-point mutations were expressed where the positively charged residues at positions 1308, 1312, 1334, and 1336 were mutated to aspartate. These amino acids were observed in the simulations to form salt bridges with Asp 1261 (Fig. 5a), so mutation of these residues is expected to activate the A1 domain. In addition, the single-point mutant K1371D was also expressed as a negative control because the residue at this position does not interact with the N-terminal linker but forms a salt bridge with GpIb␣ (Fig. 7, a and b). As predicted, the mutations R1308D and R1336D enhanced the binding activity of the A1 domain at low and high shear, as shown by the decreased rolling velocity (Fig. 8a) and decreased bead detachment (Fig.  8b) compared with the wild-type. The mutations K1312D and R1334D and the negative control K1371D inhibited the adhesive function of the A1 domain (Fig. 8b). All three of these residues formed salt bridges with residues in GpIb␣ in simulations performed with complexes from multiple crystal structures; Arg 1334 and Lys 1371 formed stable salt bridges with GpIb␣ in both the simulations described here (Fig. 7, a and b) and those described in earlier studies (18), whereas Lys 1312 formed salt bridges with Glu 5 or Glu 14 in earlier simulations only (Fig. 7c) (18). Remarkably, the earlier simulations were performed with crystal structures in which no electron density was visible for residue Asp 1261 (PDB codes 1SQ0 and

Specific electrostatic interactions regulate VWF
1M10), suggesting that Lys 1312 is only able to form salt bridges with GpIb␣ when it is not forming a salt bridge with Asp 1261 .
In summary, salt-bridge interactions are key to regulate binding between A1 and GpIb␣. Mutation of positively charged amino acids that form a salt bridge with Asp 1261 but not with GpIb␣ activates binding. However, mutating positively charged residues that interact with GpIb␣ inhibits binding whether or not they also interact with Asp 1261 . This indicates that the destabilizing effect of abrogating an intermolecular interaction is larger than the activation because of removal of the inhibitory effect of the N-terminal linker. The observation that Lys 1312 is not observed to form a salt bridge with both GpIb␣ and Asp 1261 simultaneously also suggests a mechanism by which Asp 1261 and the N-terminal linker may inhibit GpIb␣ binding.

Discussion
The recruitment of blood platelets to the site of vascular injury during hemostasis is mediated by the binding of the A1 domain of VWF to platelet surface receptors, and it is particularly essential under high shear conditions. Under normal blood flow conditions, the A1 domain is not active, and this has been largely attributed to the inhibitory function of its N-terminal linker (16,15). However, the exact molecular mechanism has not been elucidated by which the N-terminal linker prevents binding of A1 to GpIb␣ under low flow conditions and is involved in high-shear activation of A1 binding to GpIb␣. Using a combination of flow chamber experiments and MD simulations, the present study investigated the role of key amino acids in the N-terminal linker in regulating the adhesive function of the A1 domain. Four main conclusions arise from the results. First, Asp 1261 is the most critical residue for the inhibitory function of the N-terminal linker (Figs. 2 and 3). Second, Asp 1261 is critical for shear-enhanced adhesion (Figs. 2 and 3). Third, Asp 1261 inhibits the A1 domain because it forms salt bridges with positively charged residues in the rest of the protein (Fig. 5). Fourth, interprotein salt bridges between A1 and GpIb␣ are necessary for the stability of the complex (Figs. 7 and 8). Overall, the present study indicates that interactions between charged residues are key to the regulation of the A1 domain either through anchoring of the N-terminal linker to the rest of A1 and inhibiting binding or by stabilizing the complex between A1 and GpIb␣.
A major impediment to understanding the structural basis of the N-terminal linker inhibition has been that residues 1238 -1260 have been considered essential to the inhibitory function of the linker (14, 15) but are not included in any crystal structures of A1. However, we established here that the short N-terminal linker (residues 1261-1270) is sufficient for inhibition and shear-enhanced adhesion (Fig. 2), although residues 1238 -1260 indeed amplify these effects. Thus, we can directly evaluate the structural mechanism of inhibition from the crystal structure of the A1-GpIb␣ complex in which the short linker is fully resolved (PDB code 1U0N) (8). Both crystal structure (8) and our simulations (Fig. 7) clearly demonstrate that the short linker and GpIb␣ can simultaneously bind stably to A1, so the short linker is not a competitive inhibitor of GpIb␣. However, the linker and GpIb␣ bind to overlapping sites on A1 that include Lys 1312 (Fig. 7). This demonstrates that the N-terminal linker acts as a parasteric inhibitor, which is defined as a noncompetitive inhibitor that binds to a site on a receptor that overlaps with the ligand (19). The parasteric mechanism of the N-terminal linker may be reminiscent of toehold nucleic acid strand displacement, in which single-stranded nucleic acid binds to an exposed single-stranded toehold on doublestranded nucleic acid, allowing rapid sequential displacement of the original and otherwise stable complex. That is, we propose that the linker docks to A1 in the A1-GpIb␣ complex and then disrupts the salt bridges between Lys 1312 and GpIb␣, which are essential to binding function (Fig. 8).
This proposed mechanism of parasteric inhibition of GpIb␣ by the N-terminal linker provides a structural explanation for multiple observations about the biophysics of the interaction between VWF and GpIb␣. First, it is known that stretching of VWF in flow activates binding to GpIb␣. This stretching would extend both N-and C-terminal linkers away from A1, which would dislodge Asp 1261 , unmasking the full GpIb␣ binding site on A1. Second, the complex between A1 and GpIb␣ is known to form catch bonds that are longer lived under increased tensile force (9). Because GpIb␣ can bind to core A1 simultaneously as the N-terminal linker, A1 and GpIb␣ can transiently bind to each other even when the N-terminal linker is still docked in place. At that point, tensile force transmitted through the complex would pull away the N-terminal linker, unmasking Lys 1312 to strengthen the A1-GpIb␣ complex. It should not matter whether force is transmitted primarily through the C or N terminus in this model, because MD simulations showed that in both cases, the disulfide bond linking the two terminii of the A1 domain moves away from GpIb␣ (18). Third, the A1-GpIb␣ complex displays two distinct lifetimes (14,20). In the shortlived state, Asp 1261 would be fully docked, forming a salt bridge to Lys 1312 , whereas in the long-lived state, Asp 1261 would be dislodged, and Lys 1312 would form salt bridges with GpIb␣. Fourth, this mechanism explains why no allosteric conformational changes have been observed in the core A1 domain between various crystallographic structures of the A1 domain in isolation (5), in complex with GpIb␣ (8,21) or bearing clinically known gain-of-function mutations (22,23), despite many observations that might otherwise suggest allosteric regulation.
The inhibitory mechanism of the N-terminal linker described here might also explain how some naturally occurring mutations lead to type 2B von Willebrand disease, a bleeding disorder caused by depletion of the hemostatically most active large multimers of VWF caused by their clearance by platelets (17). Type 2B mutations R1308C (24) and R1308P would eliminate a positive charge that helps anchor the N-terminal linker (Fig. 5b). Type 2B mutations R1306W/Q/L, I1309V, S1310F, W1313C, and R1341W/Q/P/L are also located in the vicinity where the N-terminal linker docks, and P1266L/Q, H1268N/D, and C1272G are in the N-terminal linker where it docks to A1. Therefore, these type 2B mutations may also activate platelet binding by destabilizing the linker from the docked position where it inhibits GpIb␣ binding.
In conclusion, the present study highlights that salt bridges are a key component for the binding function of the A1 domain and its regulation. Intermolecular salt bridges stabilize the complex between A1 and GpIb␣, whereas the N-terminal linker of A1 destabilizes the complex, in large part because of salt bridges between Asp 1261 and the core of the A1 domain. Because the N-terminal linker and GpIb␣ can bind simultaneously to the core of the A1 domain yet have overlapping binding sites, they act as noncompetitive parasteric inhibitors of each other. The results presented here provide a structural explanation for many prior observations about the regulation of the complex, including its mechanical activation (9), dual lifetimes (14,20), and the existence of activating type 2B von Willebrand disease mutations (17). The present work also strongly suggests that the N-terminal linker can induce detachment of the A1-GpIb␣ complex and thus could provide a novel mechanism to dissolve thrombi. More generally, design of novel anti-thrombotic drugs that target the VWF-GpIb␣ interaction should take into account the role of charged interactions.

Construction of expression plasmids for VWF A1 domain wildtype and mutants
The DNA fragments containing A1_1261 or A1_1238 were first obtained by PCR amplification using the full-length human VWF cDNA plasmid as a template and corresponding primers that included a NheI site at the 5Ј end and a XhoI site at the 3Ј end of the A1 domain. All inserted A1 domain constructs ended at position 1472. The amplified fragments were then cloned by TA cloning into the vector pCR2.1-TOPO following the protocols provided by the manufacturer (Invitrogen). The single-site mutations D1261K, D1261N, and D1261Q in A1_1261 and A1_1238, respectively, and the deletion mutations in A1_1261 to generate the constructs A1_1261, A1_1267, and A1_1272 were introduced into the corresponding wild-type A1 fragments according to the protocol of the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The desired DNA sequences of the wild-type and various mutant A1 fragments were confirmed by DNA sequencing. By taking advantage of the NheI and XhoI restriction sites on the vector, these A1 DNA fragments were further subcloned into the pBIG expression vector in frame with four repeats of GGGGS linker between the C terminus of the A1 domain and a Biotin tag. The resulting sequence of the expressed protein was therefore MPSSVSWGILLLAGLCCLVPVSLA2ERAS (VWF insert) LEGRLNDIFEAQKIEWH, with the arrow indicating the location at which the signal peptide is cleaved, so that the secreted protein starts just after this.

Expression of various recombinant VWF A1 domain proteins
The recombinant expression constructs were transiently transfected into CHO Tet-On cells with Lipofectamine 2000 (Invitrogen) for protein expression. The expression was induced by doxycycline (2 g/ml) in the presence of biotin (50 M) in DMEM culture medium (Invitrogen) with 10% tetracycline-free fetal bovine serum (Atlanta Biologicals, Lawrenceville, GA) for 72 h before harvest. Culture medium containing secreted recombinant A1 protein was collected and concentrated five times using an Amicon Ultra-15 centrifugal filters (EMD Millipore, Billerica, MA). Free biotin was removed from the samples using a desalting PD-10 column (GE Healthcare).

Expression of recombinant full-length VWF
The cDNA containing full-length VWF, wild type, or mutants, in the pBIG vector (25) was transiently transfected into HEK293 Tet-On cells (Clontech) with Lipofectamine 2000 (Invitrogen). Proteins were expressed for 48 h in serum-free FreeStyle 293 medium (Life Technologies) in the presence of 2 g/ml of doxycycline and 50 M biotin. The culture medium containing secreted VWF was concentrated, and the free biotin was removed as described above.

Preparation of surfaces with VWF constructs
VWF constructs were anchored to surfaces via the biotin tag as follows. 35-mm tissue culture polystyrene dishes were incubated with a 100-l droplet of 200 g/ml biotin-BSA for 2 h at 37°C, washed three times with 0.2% BSA-PBS (w/v), incubated with 100 g/ml streptavidin for 30 min at 37°C, washed three times with 0.2% BSA-PBS (w/v), and incubated overnight at 4°C in 0.2% PBS-BSA. The dishes were then incubated for 30 min at 37°C with a solution containing the biotinylated VWF A1 or full-length VWF prepared as described above, washed three times with 0.2% BSA-PBS (w/v), and blocked with 0.2% BSA-PBS (w/v). The amount of biotinylated VWF used in this surface functionalization was determined to be sufficient to saturate the available biotin-binding sites by using biotin-4-fluorescin to test for the presence of free biotin-binding sites using the method of Kada et al. (26). Briefly, flat bottomed 96-well dishes were functionalized with biotin-BSA and streptavidin as described above and shown to quench up to 25 nM of biotin-4fluorescein in 100 l. This indicates the availability of binding sites for 2 pmol biotin on the streptavidin-functionalized surfaces. After incubation with biotinylated VWF, the surfaces displayed essentially the same (unquenched) fluorescence as control dishes with no streptavidin, indicating that all biotinbinding sites were now saturated. Given an expected functionalized surface area of 32 mm 2 in the wells, 2 pmol translates to a site density of two biotin-binding sites per 50 nm 2 area, consistent with the size of streptavidin tetramers and the expectation that two biotin-binding sites will be free because they face upward away from the biotin-BSA layer on the surface. A 3-fold excess of this amount of biotinylated VWF was used to functionalize the 35-mm dishes for insurance in case the droplet used in the flow chamber had a slightly larger surface area than in the dishes.

Preparation of washed platelets
The platelets were isolated from the blood of healthy donors, which had been drawn into acid citrate dextrose tubes. The platelets were separated by differential centrifugation in the presence of apyrase and PGE-1 (to inhibit platelet activation) and resuspended in Hepes Tyrodes buffer containing 0.1% BSA (27).

Preparation of GPIb␣-coated microspheres
The N-terminal domain of GPIb␣ (amino acids 1-300 of the extracellular domain of GPIb␣ with a C-terminal biotin tag for oriented anchoring) was made using a previously described expression vector (28) and expressed in CHO cells. After desalting and concentrating via ammonium sulfate precipitation, the GC300 was bound to 3-m-diameter streptavidin-coated Dynabeads (Invitrogen) with gentle mixing for 30 min at room temperature. The beads were then washed and resuspended in PBS with 0.1% BSA (w/v) for use in the flow chamber.

Platelet and microsphere adhesion in flow
Platelet adhesion studies were performed in a Glycotech TM parallel plate flow chamber as previously described (29), using functionalized 35-mm dishes prepared as described above as the lower surface. A 300-l bolus of washed platelets or GPIb␣coated microspheres was introduced into a flow chamber and allowed to settle for 30 s, before PBS-BSA buffer was pushed through the chamber at the indicated shear stress and platelets or microspheres observed with a 10ϫ objective and CCD camera. To calculate the velocity of platelets or microspheres at each shear stress, time-lapse videos were taken with a 10ϫ objec-

Specific electrostatic interactions regulate VWF
tive using MetaMorph or MicroManager software, at 1 frame/s, and the platelets or microspheres in the videos tracked using SVCell RS (DRVision Technologies, Bellevue, WA). The tracks were processed using MATLAB scripts to calculate the velocity of each platelet or microsphere at each time point. If the velocity in any frame was equal to or greater than the hydrodynamic velocity (particle radius times the shear rate or the estimated velocity of fluid at the midpoint of a platelet or microsphere touching the surface), then the platelet or microsphere was assumed to not be bound at that time point. The average velocity of all bound platelets or microspheres was calculated at each time point, and this value was averaged over all time points at a given shear stress to calculate the translocation velocity at a given shear stress.

Initial conformations for the MD simulations
To generate the initial conformations, the crystallographic structure with PDB code 1U0N was used, which contains the triplex between the A1 domain, GpIb␣, and the snake venom botrocetin (8). The reason why this X-ray structure was chosen versus the one consisting of only A1 and GpIb␣ (PDB code 1SQ0) is that in the former the A1 domain contains a longer N-terminal linker, and in particular its N-terminal residue at position 1261 is of particular relevance for this manuscript. Botrocetin was deleted from the PDB file, and the resulting structure of the A1-GpIb␣ complex was then subjected to 200 steps of steepest descent minimization in vacuo using the program CHARMM (30) to relieve any possible strain in the system. The conformation with the single-point mutation D1261K was created per analogy from the wild-type using internal coordinates for the lysine side chain contained in the PARAM22 parameter set (31). The different simulations systems are summarized in Table 1.

Simulation setup
The MD simulations were performed with the program NAMD (32) using the CHARMM all-hydrogen force field (PARAM22) (31) with the CMAP extension (33,34). The proteins were inserted into a cubic water box with side length of 110 Å, resulting in a system with a total of ϳ127,000 atoms. Chloride and sodium ions were added to neutralize the system and approximate a salt concentration of 150 mM. The water molecules overlapping with the protein or the ions were removed if the distance between the water oxygen and any atom of the protein or any ion was smaller than 3.1 Å. To avoid finite size effects, periodic boundary conditions were applied. After solvation, the system underwent 500 steps of minimization, whereas the coordinates of the heavy atoms of the protein were held fixed and subsequent 500 steps with no restraints. Each simulation was started with different initial random velocities to ensure that different trajectories were sampled whenever starting with the same initial state. Electrostatic interactions were calculated within a cutoff of 10 Å, whereas long-range electrostatic effects were taken into account by the particle mesh Ewald summation method (35). van der Waals interactions were treated with the use of a switch function starting at 8 Å and turning off at 10 Å. The dynamics were integrated with a time step of 2 fs. The covalent bonds involving hydrogens were rigidly constrained by means of the SHAKE algorithm with a tolerance of 10 Ϫ8 . Snapshots were saved every 10 ps for trajectory analysis.
Before production runs, harmonic constraints were applied to the positions of all heavy atoms of the protein to equilibrate the system at 300 K during a time length of 0.2 ns. For those systems containing single-point mutations, harmonic constraints were kept on all heavy atoms except those of the mutated amino acid and of amino acids for which any side chain atom was within 4 Å of any side chain atom of the mutated residue (these were Lys 1261 , Ile 1261 , Ser 1263 , Lys 1312 , and Glu 5 in GpIb␣), and equilibration was continued for another 2 ns. After this equilibration phase, the harmonic constraints were released. The systems were simulated for in total 50 ns, and the first 10 ns of unconstrained simulation time were also considered part of the equilibration and were thus not used for the analysis. During both the equilibration and production phases, the temperature was kept constant at 300 K by using the Langevin thermostat (36) with a damping coefficient of 1 ps Ϫ1 , whereas the pressure was held constant at 1 atm by applying a pressure piston (37). The C ␣ RMSD calculated along the trajectories fluctuated ϳ2 Å for the entire complex in all simulations except in complex D1261K_1, in which the N-terminal linker was flexible (supplemental Fig. S1). This indicates that the simulations have likely converged and sampled conformational space around the bound state.

Determination of salt bridge formation
The trajectories were screened for the presence of salt bridges between Asp 1261 or Lys 1261 , respectively, and any other charged residue of A1 or GpIb␣. An interaction was defined as salt bridge if the atoms N of Lys or C of Arg were closer than 4 or 5 Å, respectively, from either the C ␥ of Asp or C ␦ of Glu. All histidines were assumed neutral. Only salt bridges present in at least 10% of the simulation frames were considered for the analysis.