Activation-independent platelet adhesion and aggregation under elevated shear stress

Platelets aggregate at sites of vascular injury, forming thrombi that contribute to arrest bleeding but also occlude atherosclerotic arteries causing cardiac and cerebrovascular diseases.^1,2  Platelet thrombus formation is thought to occur in successive stages. First, individual platelets adhere to altered vascular surfaces and are activated, after which the integrin αIIbβ3 can bind plasma proteins, notably fibrinogen, von Willebrand factor (VWF), and fibronectin; these adhesive substrates immobilized on the membrane surface then recruit additional platelets, resulting in aggregation and thrombus growth.²  Such events take place in flowing blood that generates shear forces. At shear rates exceeding 1000 s^–1 in the human circulation, initial platelet arrest depends on glycoprotein (GP) Ibα binding to immobilized VWF even when extracellular matrices³  or vascular structures⁴  present multiple reactive components. Continued platelet recruitment also becomes dependent on VWF-GP Ibα as growing thrombi narrow the lumen where blood flows, locally increasing the shear rate.⁵  Current knowledge, therefore, is that rapidly forming but short-lived VWF-GP Ibα bonds can keep platelets in contact with a surface or with one another for a limited time, until additional bonds, established mostly through integrin receptors, stabilize adhesion and aggregation.^3,5,6 

A feature distinguishing hemostasis from arterial thrombosis is their occurrence in different hemodynamic environments. A 90% lumen reduction in a coronary artery may cause shear rates of 20 000-40 000 s^–1 at or just upstream of the stenosis,^7-9  values that are 100-fold higher than in the absence of obstruction¹⁰  and 10-fold higher than in microarterioles,¹¹  where platelets participate in hemostasis after trauma. We now have identified a unique mechanism that may be relevant for initiating platelet thrombus formation under extreme hemodynamic conditions, and it involves activation-independent as well as αIIbβ3-independent platelet aggregation at the interface between immobilized and soluble VWF. With shear stress above a threshold value between 350 and 500 dyn/cm,²  soluble VWF multimers bind to platelets as they are tethered to immobilized VWF, thus enhancing platelet recruitment through GP Ibα-mediated aggregation before activation and stable adhesion take place. A dimeric VWF A1 domain fragment also can mediate aggregation, without requiring a threshold shear stress but with progressively less efficiency as shear rates increase. This indicates that soluble VWF function is dynamically regulated by shear stress and depends on multivalent binding to support platelet aggregation at high shear rates. As a consequence of elevated fluid drag, adhering and aggregated platelets are stretched from points of contact with the surface or other platelets, creating elongated structures that can exceed 10 μM in length and support further platelet adhesion and aggregation through VWF binding. These major morphologic changes, which also are activation-independent, appear to stabilize adhesion mediated by VWF-GP Ibα such that it can last minutes rather than seconds. Our findings change the paradigm that platelet aggregation follows adhesion and activation and show that activation-independent aggregation mediated by VWF-GP Ibα bonds can initiate thrombus growth under hemodynamic conditions relevant for arterial occlusion.

Blood from healthy human volunteers was drawn from an antecubital vein. For perfusion studies, blood was collected into syringes containing D-phenylalanyl-L-prolyl-L-arginine chloromethyl ketone dihydrochloride (PPACK; final concentration, 93 μM; Bachem Bioscience, King of Prussia, PA) to prevent clotting. Prostaglandin (PG) E₁ (10 μM, Sigma Chemical, St Louis, MO) and the disodium salt of ethylenediamine tetraacetic acid (EDTA, 5 mM; Sigma) were added to the blood when indicated to inhibit platelet activation and block integrin function, respectively. To prepare plasma-free blood cell suspensions, 5 parts of blood containing PPACK were mixed with 1 part of acid-citrate-dextrose (ACD; 60 mM citric acid, 85 mM sodium citrate, and 111 mM dextrose; pH 4.5). The adenosine diphosphate (ADP) scavenger apyrase (grade 7, Sigma; ATPase/ADPase ratio < 2) was added at a final concentration of 1.3 ATPase units/mL. The blood was centrifuged at 2100 g for 13 minutes at room temperature (22°C-25°C). The resultant supernatant plasma was removed from the sedimented cells, which included platelets and leukocytes on top of the erythrocyte cushion, and replaced with an equivalent volume of divalent cation-free Hepes/Tyrode buffer (17 mM Hepes [N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid], 130 mM NaCl, 2.7 mM KCl, 0.4 mM NaH₂PO₄, and 2.8 mM dextrose), pH 6.5, containing a reduced amount of apyrase (0.65 ATPase units/mL). After gentle mixing, the cell suspension was centrifuged again, and the supernatant fluid was removed and replaced by a fresh aliquot of Hepes/Tyrode buffer. This procedure was repeated twice, each time using half the amount of apyrase, and after the final centrifugation the cells were suspended in divalent cation-free Hepes/Tyrode buffer, pH 7.4, containing 50 mg/mL bovine serum albumin (BSA; Calbiochem, La Jolla, CA). The platelet count (180 000-390 000/μL) and hematocrit level (38%-43%) were adjusted to the original values in whole blood. PG E₁ (10 μM) and EDTA were added when indicated. Washed platelets for labeling with different fluorochromes were obtained by centrifuging the last blood cell suspension in Hepes/Tyrode buffer, pH 6.5, at 200 g for 10 minutes. All studies involving human subjects were conducted in accordance with the Declaration of Helsinki and were approved by the Institutional Review Board of The Scripps Research Institute. Informed consent to participate in the studies was obtained from all subjects.

A recombinant VWF fragment comprising residues 445-733 of the mature subunit,¹²  including the carboxyl terminal portion of domain D3, the entire domain A1, and a small amino terminal sequence of domain A2,¹³  was expressed and purified following a procedure described in Supplemental Document S1, available on the Blood website; see the Supplemental Materials link at the top of the online article. Plasma-derived VWF multimers were purified as previously described.¹⁴  The antibody NMC-4 (a generous gift of Dr Akira Yoshioka, Nara Medical College, Nara, Japan) was purified as reported elsewhere.¹⁵  NMC-4 binds to the VWF A1 domain and blocks binding to GP Ibα.

Perfusion experiments were conducted at 37°C by exposing blood to different thrombogenic surfaces, which included fibrillar type 1 collagen from bovine tendon (acid insoluble; Sigma),³  immobilized multimeric VWF⁶  purified from plasma as previously described,¹⁴  extracellular matrix deposited by mouse dermal fibroblasts (Document S1), and platelet thrombi formed by perfusing whole human blood over fibrillar type 1 collagen. The glass coverslips onto which thrombogenic substrates were immobilized were assembled into a parallel plate rectangular flow chamber.³  Blood was perfused through the chamber by aspiration with a syringe pump (Harvard Apparatus, Holliston, MA) at the desired flow rate. The perfusion chamber was mounted on the stage of an inverted microscope (Axiovert 135M; Carl Zeiss, Thornwood, NY) for real-time visualization of platelet interactions with the immobilized substrates and with one another. For epifluorescence microscopy, platelets were rendered fluorescent by the addition of 10 μM mepacrine (quinacrine dihydrochloride; Sigma),⁶  unless otherwise specified. For 2-color experiments, platelets in untreated whole blood were rendered fluorescent with mepacrine, and washed platelets (in Hepes/Tyrode buffer, pH 6.5), whose ability to be activated was blocked with 10 μM PG E₁, were rendered fluorescent by incubation with calcein red-orange, AM (1 mg/mL solution in dimethyl sulfoxide; Cell Trace, Invitrogen, Carlsbad, CA), for 15 minutes in the dark. After labeling, the platelets were sedimented by centrifugation at 600 g for 8 minutes and resuspended in Hepes/Tyrode buffer, pH 7.4. For electron microscopy analysis, platelets and aggregates interacting with immobilized VWF were perfusion-fixed under flow with a buffered 4% paraformaldehyde solution. For scanning electron microscopy (JSM-6300F, JEOL, Munich, Germany) the fixed specimens were dehydrated in ethanol baths of increasing concentration (up to 100%) and then critical-point dried with CO₂ and sputter-coated with platinum. For reflection interference contrast microscopy (RICM), no labeling of the blood cells was required. This technique resolves the cell membrane contact with a surface and indicates the distance between the 2 through interference colors.^16,17  In our studies, interference was caused by the light reflected from the glass surface onto which the substrate was coated and that from the membranes of platelets flowing in close proximity to the coated glass or interacting with it. Since we used a black-and-white video camera, interference colors were obtained on a gray scale, in which zero-order black corresponds to a distance between platelet membrane and coated glass surface of 4 to 12 nm, and white corresponds to a distance of more than 20 to 30 nm.^16,17  Structures that are separated by more than 30 nm are out of focus.

All animal care and experimental procedures complied with the Guide for the Care and Use of Laboratory Animals, United States Department of Health and Human Services, and were approved by the Animal Care and Use Committee of The Scripps Research Institute. The procedure to expose mesenteric vessels and visualize thrombus formation at the site of a ferric chloride–induced lesion using fluorescent platelets has been previously described¹⁸  and is reported in detail in Document S1. The microscope used was a Zeiss Axioplan2 equipped with an Achroplan 40 ×/0.8 NA water-immersion objective.

All experiments were recorded on S-VHS videotape using a VCR (SVO-9500MD; Sony, Tokyo, Japan) at the acquisition rate of 30 frames per second. Fluorescence images were acquired with a silicon-intensified (SIT) high-sensitivity camera (Hamamatsu Photonics; Bridge-water, NJ) and RICM images with a CCD (charge-coupled device) camera (DXC-390; Sony). The objectives used (all from Carl Zeiss) were Plan-Neofluar 40 ×/0.75 NA (Figures 1, 2B, 3, 6), Plan-Neofluar 10 ×/0.30 NA (Figure 2A), and Plan-Neofluar Ph3 Antiflex 63 ×/1.25 NA oil immersion (Figure 5). Image analysis was performed offline using the Metamorph software package (Universal Imaging; West Chester, PA).⁶  The supporting movies available online were prepared by digitizing and editing the recorded analog tapes with Adobe Premiere (Adobe Systems, San Jose, CA).

We first evaluated thrombus formation on a model reactive surface composed of collagen type 1 fibrils in a chamber perfused with blood containing 93 μM PPACK as the anticoagulant. At the wall shear rate of 3000 s^–1, as expected, single platelets adhered and accumulated into thrombi that grew at sites of initial arrest (Figure 1; Video S1). At 24 000 s^–1, in contrast, platelets appeared to be linked to one another even before stable adhesion took place, but no interplatelet links were visible by epifluorescence microscopy. The elevated fluid drag could stretch these aggregates and eventually detach them from the growing thrombus (Figure 1; Video S1). Such events were evident in the initial 20 to 40 seconds, before rapid aggregation took place, and delineated a hitherto unrecognized mechanism of platelet cohesion operating in the initial stages of thrombus formation at pathologically elevated shear rates. A peculiar feature of this form of aggregation was that linked platelets were not closely juxtaposed (Figure 1; Video S1).

Because platelet adhesion to collagen depends on bound VWF above a threshold shear rate,³  we evaluated whether immobilized VWF alone could initiate platelet aggregation in rapidly flowing blood. Moreover, since blocking platelet activation has no effect on GP Ibα binding to VWF but impairs integrin function,⁶  we perfused blood containing 10 μM PG E₁, a potent platelet activation inhibitor, to assess the role of different adhesion receptors in the process. EDTA (5 mM) also was added to prevent ligand binding to integrins,^19,20  but the same results were obtained with or without EDTA when PG E₁ was present. Here, therefore, aggregation defines a process of platelet cohesion mediated by physiologic blood components in the absence of activation.

As expected,⁶  individual rolling platelets covered the VWF surface at the shear rate of 3000 s^–1. In contrast, at 20 000 s^–1 rolling platelets formed aggregates within 100 μMto200 μM from the boundary where immobilized VWF was exposed to blood, equivalent to a few seconds from the initial surface contact (Figure 2A; Video S2, part 1). These aggregates grew larger while translocating in the direction of flow and had varying shape and velocity during motion, but were mostly elongated with a stringlike morphology during periods of prolonged arrest that lasted several seconds (Video S2, part 1). We then perfused a plasma-free blood cell suspension to evaluate whether addition of soluble VWF was sufficient to mediate activation-independent platelet aggregation. In agreement with the results in whole blood, at the shear rate of 3000 s^–1 the VWF surface was covered by single rolling platelets in equal number whether soluble VWF was present or not (Figure 2B; Video S2, part 2). At the shear rate of 24 000 s^–1, in contrast, single platelet adhesion was markedly lower than at 3000 s^–1 in the absence of soluble VWF, but activation-independent platelet aggregates formed as in whole blood when VWF was added (Figure 2B; Video S2, part 2). Under extreme shear stress conditions, therefore, enhanced platelet adhesion and aggregation occurred at the interface between immobilized and soluble VWF without requirement for other plasma components. This VWF-mediated platelet aggregation was inhibited by a monoclonal antibody against the VWF A1 domain (Figure 2C) or one against GP Ibα (not shown). Activation-independent platelet aggregates also formed in pulsatile flow characterized by shear rate cycles between 30 and 62 000 s^–1 with a 1-second period, but were smaller when the cycle was between 30 and 57 000 s^–1 and absent when the cycle was between 30 and 37 000 s^–1 (Video S3).

Because activation-independent platelet aggregation occurred above a threshold shear rate, we evaluated formation and stability of rolling aggregates moving across a perfusion chamber with flow velocity (shear rate) varying bidirectionally as a function of distance from the inlet and position relative to the centerline of the flow path²¹  (Figure 3; Video S4, part 1). We assumed that blood is a Newtonian fluid at the shear rates relevant for these studies, to which this relationship applies:

with viscosity = 0.04 P. The Reynolds number was compatible with laminar flow under all conditions tested. At a position with shear rate of 5000 s^–1 (shear stress = 200 dyn/cm²), only single platelets rolled on the VWF surface regardless of conditions upstream. At 12 000 s^–1 (shear stress = 480 dyn/cm²) there were mostly single platelets downstream of lower shear rates, but small rolling aggregates appeared downstream of higher shear rates. This observation indicates that the shear stress resulting from such a flow condition is not sufficient for the formation of activation-independent platelet aggregates but is compatible with their persistence if they formed at higher shear rates. At 20 000 s^–1 (shear stress = 800 dyn/cm²) large platelet aggregates were present regardless of the upstream shear rate, but their shape and mode of interaction with the surface differed depending on upstream flow conditions. When upstream shear rates were higher, stringlike aggregates exhibited prolonged periods of stationary adhesion; in contrast, when upstream shear rates were lower, rolling aggregates with ellipsoidal shape were predominant. Adherent aggregates were prevalent at 26 000 s^–1 (shear stress = 1040 dyn/cm²) regardless of upstream flow conditions, and they exhibited arrest times on the surface that could exceed 1 minute (Figure 3; Video S4, part 1). Thus, there was a transition from rolling to adherent aggregates at shear stress levels above those required to initiate VWF-mediated platelet aggregation. The reversibility of activation-independent aggregates appeared to depend on whether they were rolling onto or attached to immobilized VWF. The progressive disappearance of rolling aggregates upon decreasing the shear stress to which they were exposed could be visualized in real time (Video S4, part 2) and explains the absence of aggregates in areas exposed to lower shear rates downstream of higher shear rates (Figure 3). In contrast, elongated aggregates that established a firm attachment to the surface when exposed to the highest shear rates tested remained adherent even when the shear rate decreased. These aggregates retracted in the direction opposite to flow when exposed to a lesser hydrodynamic drag, with a reduction in length but no substantial loss of interplatelet cohesion as shown by the return to an extended shape as the shear stress increased (Video S4, part 3). It appears, therefore, that interplatelet adhesive interactions mediated by VWF-GP Ibα bonds are enhanced by increasing shear stress, and this positive modulation may become less reversible as the bonds are exposed to higher stress and, possibly, more interactions are established at the interface of soluble and immobilized VWF (stretched, chainlike aggregates).

The links between platelets aggregated through VWF-GP Ibα bonds in the absence of activation had distinctive properties. Rolling aggregates could wrap around surface adherent platelets or more slowly moving aggregates (Video S5, part 1), with subsequent fusion into a single entity often followed by detachment from the surface or separation into different parts under the effect of tensile stress (Video S5, part 2). Hydrodynamic drag stretched surface adherent aggregates to the point that interplatelet distances reached 10 μM to 20 μM, until rupture occurred, giving origin to rolling aggregates (Video S5, part 3). When this happened, the recoil of the platelets remaining on the surface highlighted the elasticity of the long interplatelet links (Video S5, part 4). Thus, the transition from stationary to rolling aggregates and the fusion of smaller into larger or separation of larger into smaller aggregates were dynamic events, dependent on the properties of interplatelet links formed in the presence of soluble VWF and modulated by hydrodynamic forces in the absence of platelet activation.

The ability of these interconnecting structures to fuse with other platelets and their elasticity suggested that they could comprise cell membrane. Evaluation of perfusion-fixed activation-independent aggregates by electron microscopy (Figure 4) demonstrated the presence of membrane protrusions connecting platelets to one another and to the surface. The connecting membrane links were longer in what appeared to be elongated (stringlike) surface-adherent aggregates (Figure 4A,B) than in ellipsoid rolling aggregates (Figure 4C). The effects of perfusion-fixation were monitored by microscopic observation; thus, we had positive evidence that the morphology of rolling aggregates was preserved after fixation. All aggregates seen during perfusion that were not in a stringlike configuration were rolling on the surface.

To elucidate in real time the nature and dynamic properties of the links that support activation-independent platelet aggregation, we used RICM, a technique that visualizes areas of a cell membrane in close contact with a surface,^16,17  and followed the process from the arrest of single platelets onto immobilized VWF. From a discrete point of tight adhesion, over a period of several seconds, the platelet body was stretched by the fluid drag into a structure that exceeded 10 μM in length but was less than 1 μM thick at the narrowest point (Figure 5A; Video S6, part 1). During stretching, most of the platelet body slid forward, and only a limited upstream area of the membrane provided firm attachment to immobilized VWF. Additional platelets then adhered to the ones that had become elongated on the surface. Some that were stretched by the fluid drag contributed to the increasing length of a stringlike formation, while many retained their discoid morphology (Figure 5B; Video S6, part 2). In the end, stretched platelets joined to one another formed continuous structures that reached lengths of 100 μM to 200 μM and remained attached to immobilized VWF for minutes, while hundreds of discoid platelets adhered to them and to one another with arrest times of variable duration until eventually detaching as rolling aggregates (Figure 5C; Video S6, part 3). In the absence of soluble VWF, in contrast, only single platelets interacted with immobilized VWF, and thin tethers occasionally appeared between points of tight adhesion and the moving cell body,^22,23  but no interplatelet links developed, confirming the essential role of soluble VWF in the process (Video S6, part 4).

When activation was not inhibited, platelets within a stringlike aggregate that were firmly adherent to VWF were spread, while those at the downstream end, which often detached into rolling aggregates, were not (Figure 5D; Video S7). This observation is in agreement with the concept that VWF-mediated platelet aggregation can precede activation and stable surface adhesion.

Activation-independent platelet aggregation occurs on different substrates. To confirm the data obtained with immobilized collagen (Figure 1; Video S1) and demonstrate that activation-independent platelet aggregation mediated by soluble VWF could occur on several relevant surfaces, we perfused blood at varying wall shear rates over the extracellular matrix deposited by mouse skin fibroblast (Video S8) or platelet thrombi formed by a previous perfusion of whole blood over fibrillar type 1 collagen (Video S9). In the latter experiment, PG E₁–treated, washed platelets labeled with a red fluorochrome were added into a washed erythrocyte suspension supplemented with purified VWF (15 μg/mL) and perfused over aggregated platelets, visualized with a green fluorochrome. In either case, activation-independent rolling platelet aggregates and more firmly adherent elongated aggregates formed at progressively increasing shear rates, in a manner similar to that observed on surface-immobilized purified VWF. Activation-independent rolling aggregates were visible on the surface of platelet thrombi at a shear rate of 10 000 s^–1 (Video S9), thus lower than on purified VWF. This may be due to the fact that platelet activation results in the release of VWF stored in α-granules, which contains larger and possibly more active multimers than those circulating in blood.

To explain why activation-independent platelet aggregation mediated by soluble VWF occurs only above a threshold shear rate, we compared the adhesive properties of multimeric plasma-derived VWF and recombinant isolated VWF A1 domain. The latter was expressed as a dimer (dVWFA1) with interchain disulfide bonds in a portion of domain D3 preceding A1, thus reflecting the assembly of VWF multimers in this region.^24-26  When added to a washed blood cell suspension perfused in a variable shear rate flow chamber (Figure 3), dVWFA1 mediated activation-independent platelet aggregation at the lower shear rates tested, well below the threshold required for function of multimeric VWF (Figure 6; Video S10). In contrast, above this shear rate threshold, multimeric VWF mediated the formation of aggregates that could become elongated and firmly adherent under the effect of increasing hemodynamic force, while dVWFA1 could no longer support platelet aggregation (Figure 6; Video S10). The same results were observed when the concentration of dVWFA1 in solution was increased by 10-fold (50 μg/mL as opposed to 5 μg/mL; Video S10), a clear indication that the functional difference between multimeric VWF and isolated A1 domain fragment was the consequence of distinct molecular architecture rather than concentration of active sites.

These findings demonstrate that A1 domain binding to GP Ibα is not intrinsically modulated by shear stress and are compatible with the concept that hydrodynamic force releases A1 domain from functional shielding within VWF multimers, possibly through changes of VWF polymer shape from coiled to extended.^27,28  Moreover, the multivalent binding potential of large multimers appears to be essential to mediate the adhesive functions of VWF in solution under high shear stress.

To evaluate whether VWF-mediated platelet aggregation occurs within the vasculature of a living animal, we performed experiments in the mesenteric circulation of anesthetized mice. PG E₁–treated platelets labeled with a red fluorochrome and noninhibited platelets labeled with a green fluorochrome were injected concurrently and their accumulation monitored at a site of injury. Noninhibited platelets formed stable and compact thrombi, while PG E₁–treated platelets formed loose aggregates that became stretched by flow forces while rolling over the stable thrombi (Video S11). The latter resembled the VWF-mediated aggregates observed during in vitro perfusion studies. Similar events were observed both in injured arterioles and venules with initial wall shear rates in the order of 1300 and 700 s^–1, respectively, thus below the threshold that we found was required in ex vivo perfusion experiments for the induction of VWF-mediated aggregates independently of activation. These aggregates, nonetheless, appeared on the surface of thrombi, thus at sites of vascular lumen restriction where shear stress was higher than at the wall. Moreover, the thrombi were formed by activated platelets that release VWF multimers larger than those in circulating blood, which may be more active and contribute to lowering the shear stress threshold required for the formation of activation-independent aggregates.

We have identified a mechanism of activation-independent platelet aggregation mediated by VWF binding to GP Ibα that occurs exclusively at the interface of immobilized and soluble VWF, is modulated by shear stress, and may thus contribute to thrombotic arterial occlusion. As delineated here, the process is distinct from previously identified mechanisms of platelet adhesion and aggregation under flow conditions. It is known that shear stress in excess of 60 dyn/cm² can induce VWF binding to GP Ibα,²⁹  resulting in platelet aggregation, but only if activation can occur and αIIbβ3 is fully functional.^30-32  In contrast, VWF-mediated platelet aggregation as described here is independent of activation and αIIb-β3 function. The previously described “beads-on-a-string,” that is, platelets bound to ultralarge VWF multimers on endothelial cell membranes,³³  appear somewhat similar to the elongated aggregates shown here under extremely elevated shear stress but, in fact, form through a distinct process. These “strings” are newly released and extremely long VWF molecules with enhanced affinity for GP Ibα,³⁴  to which single platelets adhere at shear rates between 250 and 5000 s^–1 and in the absence of plasma,³³  thus of soluble VWF, 2 conditions incompatible with activation-independent aggregation. In the aggregates shown here under high shear stress, in contrast, the interplatelet connections are formed by stretched membrane segments rendered adhesive by plasma VWF multimers functionally modulated by tensile stress, which constitutes a previously unrecognized mechanism of platelet aggregation.

Our results indicate that hydrodynamic force exerts multiple effects during activation-independent platelet aggregation. A first shear stress threshold regulates the induction of platelet-platelet contacts, leading to the formation of “rolling” aggregates, in which interplatelet cohesion predominates over surface adhesion. The sequence of events in the process appears to involve single platelet tethering to immobilized VWF, binding of soluble VWF to GP Ibα on the membrane of adherent platelets,²⁹  and subsequent cohesion of more platelets into aggregates with no dependence on activation. The presence of soluble VWF is essential for aggregation, demonstrating that this is the molecule promoting interplatelet contacts. Such a concept also is supported by the recent observation that activation-independent platelet aggregates formed under flow conditions are reduced in size by the action of ADAMTS13, the enzyme that specifically cleaves VWF multimers.³⁵  A threshold shear stress is required for induction of aggregation through VWF multimers but not isolated dimeric A1 domain. Thus, the effect of shear stress apparently is not on the formation or stability of VWFA1-GP Ibα bonds but on the exposure of active A1 domain within VWF multimers. This mechanism may involve changes in the shape of VWF molecules as they bind to the platelet membrane, from coiled to extended as visualized for surface-bound VWF.²⁸  The results with isolated A1 domain show decreasing adhesive function with increasing shear stress, as demonstrated by the diminishing size of rolling aggregates after a peak is reached at an optimal shear rate (2500 s^–1). The fact that rolling aggregates mediated by isolated VWF A1 domain became progressively larger between 0 and 2500 s^–1 likely reflects effects of flow forces on platelet transport to the surface rather than on the properties of VWFA1-GP Ibα bonds. It appears, therefore, that VWF A1 domain and GP Ibα form selectinlike adhesive bonds,^36,37  at least at the shear rates most relevant for function, rather than “catch” bonds that become more stable as they are subjected to increasing tensile stress.³⁸  The enhanced adhesive properties exhibited by soluble VWF multimers at the higher shear rates tested may thus be controlled by available A1 domain density rather than direct modulation of bond properties. In this regard, an extended VWF multimer represents an efficient structure for the localization of a high number of functional A1 domains clustered in close proximity on the platelet membrane, as each multimer bound to GP Ibα contains multiple A1 domains and can mediate multivalent interactions. This is not the case with isolated dimeric A1 domain. On the other hand, molecular recoiling causing decreased A1 domain availability may explain the reversibility of activation-independent VWF-mediated platelet aggregation with decreasing shear stress.

A higher shear stress threshold than required for induction of VWF-mediated aggregation regulates the formation of activation-independent aggregates that remain stationary on the immobilized VWF surface for prolonged periods of time. This additional effect of shear stress appears to depend on major morphologic changes of platelets. By elongating and, thus, reducing their cross-sectional area perpendicular to the direction of flow, stretched platelets are subjected to lesser hydrodynamic drag than discoid platelets. This adaptation decreases the tensile stress on adhesive bonds at points of surface contact, prolonging their lifetime. Stretched platelets also may provide an extended membrane surface for the high-density binding of VWF multimers uncoiled by shear forces,²⁹  possibly reinforced by VWF self-association.^39,40  In fact, availability of soluble VWF is a strict requirement for stretched platelets to form and act as the elastic links that support interplatelet cohesion within activation-independent aggregates anchored to the surface. In this regard, stretched platelets are distinct from single platelets with thin tethers transiently adherent to immobilized VWF,^22,23  which have adhesion times of less than a second to a few seconds as opposed to minutes (Video S6, part 4). Stretching of platelets requires a point of firm attachment to the surface, which may exist only when soluble VWF associates with immobilized VWF to increase the local density of VWF A1 domain. The concept that soluble VWF can enhance the functional activity of immobilized VWF expressed through the A1 domain is in agreement with previous findings.³⁹  Platelet stretching and the resulting amplification of adhesive properties, therefore, are expressions of a specific synergistic function of soluble and immobilized VWF and not passive adaptations to hydrodynamic force.

In conclusion, we have outlined a mechanism mediated by soluble VWF that supports the attachment of large clusters of platelets to immobilized VWF without requiring activation and αIIbβ3 participation. Physiologic surfaces relevant to thrombus formation can support this VWF function, including extracellular matrix and the membrane of activated platelets, and the process appears to occur in the vessels of live animals. The lack of activation requirement may be crucial for the rapid initial positioning of a large number of platelets on vascular lesions exposed to extremely high shear stress. This could favor thrombus development under stringent flow conditions through subsequent activation-dependent mechanisms that remain essential to reinforce and stabilize platelet adhesion and aggregation. It is interesting to note that the genetically induced deficiency of VWF in mice does not inhibit the initial deposition of platelets on the surface of injured arterioles but prevents the definitive occlusion of the lumen after the diameter has been greatly reduced by the growing thrombus; thus, when shear stress reaches the highest values.⁴¹  The VWF-dependent process described here may be a key determinant of platelet accumulation in stenotic arteries leading to acute thrombotic occlusion.

We thank Brian Savage for helpful discussion and insights, and James R. Roberts and Richard A. McClintock for technical assistance.