Platelet receptor interplay regulates collagen-induced thrombus formation in flowing human blood

The platelet response to exposed subendothelial matrix is fundamental to thrombosis and hemostasis. Uniquely, collagen, the most abundant vessel wall protein, mediates platelet adhesion and activation, localizing and regulating the hemostatic response at sites of injury. Discovering the molecular mechanisms that control platelet-collagen interaction is crucial for understanding the pathogenesis of arteriothrombotic diseases such as stroke and myocardial infarction. Under high shear rate conditions, the glycoprotein (GP) Ib/V/IX complex allows initial platelet rolling over von Willebrand factor (VWF) bound to subendothelial collagen fibers, and subsequently collagen receptors come into contact with their specific binding sequences in the collagen. For the next step, platelet arrest and activation, firm evidence exists of a role for only 2 receptors, integrin α₂β₁ and immunoglobulin superfamily member GPVI, despite the apparent redundancy in collagen receptors (for a review, see Siljander PR-M and Farndale RW¹ ).

According to the 2-site, 2-step model, high-affinity interaction through α₂β₁ stops the platelet, allowing low-affinity binding of GPVI, which generates signaling required for the subsequent thrombus formation. Platelet deposition under flow was found to be dependent on GPIb/V/IX and α₂β₁,^2-4  whereas no platelet deposition occurred on the GPVI-specific substrate collagen-related peptide (CRP), even under low shear rates.⁵  The limited number of studies with human platelets deficient in either GPVI or α₂β₁ support the 2-site, 2-step model. Blood depleted of functional GPVI showed only severe impairment of the collagen-induced “second-wave” aggregation response under flow.⁶  Although thrombus volume was reduced, the initial adhesive layer of platelets was fully preserved. The few patients with defective α₂β₁ displayed bleeding problems, suggesting a central role for the integrin in hemostasis.^7-10  Genetic diversity of α₂β₁ expression also correlates with predisposition to thrombotic events (for a review, see Kunicki TJ¹¹ ).

Recently, several lines of evidence strongly encouraged revision of the step-wise model, promoting the idea that GPVI is the primary collagen receptor. Integrin α₂β₁ was found to change its affinity for collagen when stimulated by noncollagenous agonists such as adenosine diphosphate (ADP) or thrombin, or through GPVI,¹²  implying that high-affinity interaction through α₂β₁ requires the involvement of platelet receptors that induce secretion or inside-out signaling. Moreover, mice with Cre/loxP-mediated loss of α₂β₁ from the platelets displayed undisturbed bleeding times, and their platelets exhibited normal collagen adhesion and signaling.¹³  Similarly, α₂(–/–) mice showed no changes in platelet aggregation and adhesion to fibrillar collagen,¹⁴  whereas Fcγ(–/–) mice, generating a deficiency in the coexpressed GPVI, completely lacked a response to collagen under flow, as did wild-type mouse blood incubated with an anti-GPVI monoclonal antibody (mAb), indicating a key role for GPVI.¹³ 

Following the deposition of platelets, their procoagulant response is essential for thrombus formation. Platelet aggregates are stabilized through fibrin formation, and newly generated thrombin recruits further platelets and activates other cells. At an advanced stage of activation, collagen-adherent platelets undergo a procoagulant transformation that includes the exposure of phosphatidylserine (PS), the secretion of coagulation factors, and the distinctive morphology of blebbing cells and microvesiculation.¹⁵  We have previously shown that GPVI is essential for triggering this response in collagen-adherent platelets, which is preceded by extensive intracellular Ca²⁺-signaling.¹⁶  Supporting this, prothrombinase activity was shown to correlate with GPVI density¹⁷  and genotype.¹⁸ 

This study was undertaken to clarify the role of the human collagen receptors GPVI and α₂β₁ in thrombus formation under flow. Studies with GPVI have been hindered by the lack of antagonists of human GPVI. Recently, 10B12, an anti-GPVI, single-chain, variable domain antibody fragment (scFv) was cloned by phage display from a library of human immunoglobulin variable domains, using its capacity to block the recombinant GPVI-CRP interaction.¹⁹  Because it blocks CRP- and fibrillar collagen-induced platelet activation, 10B12 provides a small and efficient tool for human studies. To inhibit α₂β₁ function, a collagen-derived triple-helical peptide, GFOGER-GPP,²⁰  was used along with well-defined inhibitory mAbs.^21,22  Simultaneous ADP and collagen receptor antagonism were also studied. Under high shear rate conditions, specific parameters of thrombus formation—surface coverage of platelets, aggregate size distribution, and PS exposure—were monitored. Further, the intracellular Ca²⁺ response, which regulates these platelet processes and secretion and thromboxane A₂ (TxA₂) production, was measured from adherent platelets.

Our results show that GPVI is crucial to collagen-induced Ca²⁺ responses, aggregate formation, and PS expression but that α₂β₁ also contributes to these processes. Although GPVI is not directly involved in primary platelet adhesion to collagen, the inhibition of GPVI, along with GPIb and α₂β₁, was mandatory to eradicate thrombus formation.

Fibrillar type I collagen (Horm) from equine tendon was obtained from Nycomed (Munich, Germany). Apyrase (grade 7), heparin, and MRS2179, a P2Y₁ antagonist, were from Sigma (St Louis, MO), H-Phe-Pro-Arg chloromethyl ketone (PPACK) was from Calbiochem (La Jolla, CA), and Fluo-3 and calcein acetoxymethyl esters were from Molecular Probes (Leiden, The Netherlands). Annexin V (Apoptest) labeled with Oregon Green 488 (OG488) was obtained from Nexins Research (Hoeven, The Netherlands). AR-C69931MX, an antagonist of the P2Y₁₂ receptor, was from Astra-Zeneca (Charnwood, United Kingdom). Acetylsalicylic acid (ASA) was from Lorex Synthelabo (Maarssen, The Netherlands), and low-molecular–weight heparin (Fragmin) was from Pharmacia N.V. (Puurs, Belgium).

The GFOGER-GPP and CRP peptides were synthesized as described.^20,23  ScFv antibodies 10B12 and 1C3¹⁹  and mAb 12G1 against GPIbα has been previously described^24,25 : 12G1 inhibits platelet deposition to type 1 collagen by 90% at 5 μg/mL, ristocetin-induced platelet aggregation with an IC₅₀ of 1.25 μg/mL, ristocetin-induced VWF binding to GPIb by 35%,²⁴  and 35% of thrombin generation under intermediate shear.²⁵  Fab₂ fragments were generated by standard methodology. The mAb 6F1 against α₂ was a generous gift from Prof B. Coller (Mount Sinai Hospital, New York, NY).²¹  The anti-β₁ mAb 4B4 was from Beckman Coulter (High Wycombe, United Kingdom). Saratin was kindly provided by Dr M. Hoylaerts (KU, Leuven, Belgium).²⁶ 

Blood from 9 healthy volunteers was collected in 40 μM PPACK in 0.1 vol saline, supplemented hourly with 10 μM PPACK. Donors had not taken medication for 2 weeks. Platelet counts were determined with a Coulter counter (Coulter Electronics, Hialeah, FL). Donors were genotyped for receptor polymorphisms: GPVI (a/b), α₂β₁ (α2 807 C/T), and GPIb (5-C/T) using TaqMan at the National Blood Service (Cambridge, United Kingdom). All donors expressed high GPVI levels (7 aa, 2 ab).¹⁸  Five donors were CT for C807T α₂ polymorphism, implying an average density of α₂β₁ on the platelet surface. Three donors were low expressors of α₂β₁ (CC), and one was a high expressor (TT). In this small population, C807T α₂ polymorphism had no apparent effect on any of the platelet responses. Two donors were CT for GPIb C5T.

Static platelet adhesion to CRP and fibrillar collagen was performed exactly as previously described,²⁷  using cation-free conditions to examine the adhesive role of GPVI in the absence of α₂β₁.

Acid citrate dextrose–platelet-rich plasma (ACD-PRP) was incubated with 7 μM Fluo-3 acetoxymethyl ester or calcein for 30 minutes at 37°C. Platelets were centrifuged with 0.1 vol ACD, washed once with HEPES (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid) buffer containing 10 mM HEPES, 136 mM NaCl, 2.7 mM KCl, 2 mM MgCl₂, 0.1% glucose, 0.1% bovine serum albumin (BSA), pH 6.6, centrifuged again with 0.1 vol ACD, and suspended in the original volume of PRP with HEPES buffer, pH 7.45. PPACK-anticoagulated blood was spiked with Fluo-3–labeled platelets at 10% of the original platelet count and with 30 μM PPACK.

In control perfusions over collagen, 30% of 7 μM calcein-prelabeled autologous platelets were added. In real-time, Z-stacks of x-y scans were taken during the perfusion (box volume of 171.8 × 171.8 × 50 μm) at a 2-photon excitation wavelength of 800 nm and an emission of 485 to 515 nm (8% of full power) using a Radiance 2000 multiphoton laser scanning fluorescence microscope system (BioRad, Hemel Hempstead, United Kingdom), equipped with a pulsed sub-picosecond Tsunami Ti:sapphire laser (Spectra-Physics, Mountain View, CA). Two-photon images were analyzed with Laserpix software (Bio-Rad).

Whole blood perfusion experiments were performed essentially as described for mouse blood.²⁸  Briefly, glass coverslips were coated with collagen fibers (12.5 μg/cm²) and blocked with HEPES buffer, pH 7.45, containing 1% BSA and 1% glucose. The blood was placed in a syringe and perfused over the coverslip through a transparent 50 μm deep chamber using a pulse-free pump,²⁹  at a shear rate of 1000 s^–1 for 4 minutes. Blood was incubated for 15 minutes before perfusion with various concentrations of scFv 10B12 or 1C3, 10 to 20 μg/mL 6F1, 500 μg/mL GFOGER-GPP, 10 μg/mL 4B4, 20 μg/mL Fab₂ fragment 12G1, ASA (100 μM), MRS2179 (40 μM), and AR-C69931MX (20 μM) or 1 U/mL apyrase, as described. Where indicated, autologous Fluo-3–labeled platelets were added to the blood before antagonists.

Microscopic phase-contrast and fluorescent images from Fluo-3–labeled platelets were recorded in real-time using a Visitech digital imaging system (Sunderland, United Kingdom) equipped with 2 intensified, charge-coupled device (CCD) cameras.³⁰  After perfusion, flow chambers were rinsed at the same flow rate for 4 minutes with HEPES buffer, pH 7.45, supplemented with 1 U/mL heparin and 2 mM CaCl₂. Phase-contrast and fluorescent images were collected with a 40× UV-transparent objective and 15× to 20× image magnification. Exposure of PS was detected after incubation of the slide with 100 μL HEPES/CaCl₂ buffer, pH 7.45, containing OG488-labeled annexin V (1 μg/mL) for 5 minutes. Antibodies and antagonists were also added to the rinsing buffer and the annexin V incubation.

Surface coverage from phase-contrast images was analyzed using Image-Pro (Silver Spring, MD) software version 4.1, and it was analyzed from platelets stained with OG488-annexin V using Quanticell software (Visitech). Distribution of aggregate sizes in phase-contrast images was measured using Leica QWin image analysis software (Leica Imaging Systems, Cambridge, United Kingdom). Changes in Fluo-3 fluorescence from individual platelets were converted to [Ca²⁺]_i as described.³¹  To provide a measure of the proportion of procoagulant cells independent of platelet deposition, the ratio of annexin V–binding surface coverage to phase-contrast surface coverage was calculated and was termed procoagulant index (Pi). Although the procoagulant area was overestimated through fluorescent glare in the optics, Pi provided a means of distinguishing the effects of treatments on procoagulant expression from those on platelet deposition.

Each experimental condition was tested on at least 3 occasions using blood from different donors, and each donor provided blood for at least 10 perfusion experiments. Control conditions were included for each blood sample, together with other permutations of conditions. For each perfusion surface, images from 9 random microscopic fields were collected, and the average percentage area covered by adherent platelets was measured either by phase-contrast or by OG-labeled annexin V fluorescence. Phase-contrast, fluorescent surface coverage (mean ± SE) and Pi were compared among all experimental groups using analysis of variance (ANOVA), with Newman-Keuls or Dunnett posttesting. Data derived from the same donors were compared by paired sample t tests or ANOVA. The effect of ADP blockade under different conditions was tested by 2-way ANOVA.

Image analysis provided estimates (area and roundness of separate features, such as platelets and aggregates) of thrombus size and shape. Features were segmented digitally and measured with minimal operator intervention. From an average of 9 images, using 3 different donors, at least 200 different features were measured for each treatment, the effects of which were determined by χ² analysis. Data are plotted as frequency distributions with mean values per field (± SE). Prism (GraphPad Software, San Diego, CA) was used throughout for multiple comparisons and linear regression.

PPACK-anticoagulated whole blood was perfused at high shear rate (1000 s^–1) over a surface covered with native collagen type I fibrils. Phase-contrast and fluorescence microscopic images were captured to monitor 2 independent parameters of thrombus formation measured as the surface coverage of deposited platelets and of procoagulant, PS-exposing platelets that stained with OG488-conjugated annexin V. In control perfusions, platelets rapidly adhered to the collagen fibers, forming aggregates that later coalesced (Figure 1A). The deposition of platelets measured by surface coverage was linear in time, indicating regular and continuous scavenging of platelets (Figure 2A). To confirm that the measuring surface area represents a linear progression of thrombus build-up, the thrombus volume was measured by using calcein-labeled platelets from the same donor and real-time scanning of the 2-photon fluorescence signal, which even at more than 50 μm penetration was typically not distorted by the flowing blood. Again, thrombus volume increased linearly in time (Figure 2B). No differences in platelet deposition were observed within the donor population, relating to genetic polymorphisms of GPVI, α₂β₁, or GPIbα. By the 4-minute end point, 17.9% ± 0.8% of the surface was covered by platelets. Most platelets were deposited as interconnected islets originating from individual aggregates. The mean feature size was 176 ± 16 μm², an area corresponding to approximately 40 platelets. Single platelets made up 22% of features, whereas 10% were large aggregates corresponding to 130 platelets or more (Figure 3A). Large numbers of blebbing platelets were observed in direct contact with collagen fibers, and staining with OG488-annexin V revealed 8.3% ± 0.7% coverage of PS-exposing platelets. To provide a measure of procoagulant platelets independent of the total number of adherent cells, the ratio of the 2 surface parameters was expressed as Pi, giving a value of 0.43 ± 0.03. Thrombin generation was absent during and after perfusion because no fibrin or D-dimer was observed. Including Fragmin (0.5 U/mL) as an additional anticoagulant did not change platelet deposition or annexin V binding from PPACK-only controls (data not shown).

Blocking GPVI by 10B12 inhibited CRP- and collagen-induced platelet aggregation in vitro.¹⁹  Static platelet adhesion to CRP was abolished at 5 μg/mL and decreased by 80% ± 4% (n = 4) to collagen fibers at 25 μg/mL. However, under flow, a concentration curve for 10B12 (Figure 4) showed that, up to 300 μg/mL, surface coverage with platelets did not significantly decrease (r² = 0.22; P = .42), though, dramatically, platelet aggregates were absent (Figure 1B-D). Biosensor data indicated that at 300 μg/mL 10B12, occupancy of recombinant GPVI had reached more than 97% (data not shown). The predominance of single platelets with increasing 10B12 concentrations, obvious to the eye, was quantified by image analysis as almost complete elimination of larger aggregates (P < .0001) at 50 μg/mL 10B12, and the mean area was reduced to 37 ± 2 μm² from 176 ± 16 μm² (P < .001) (Figure 3B). In contrast to platelet adhesion, surface coverage of PS-exposing platelets decreased progressively with the 10B12 concentration with an IC₅₀ value of 23 μg/mL (r² = 0.94; P < .01) (Figure 4), and blebbing platelets were no longer visible. At the highest concentration of 10B12 tested, the Pi was reduced to 0.02 ± 0.005 (P < .001), indicating strong blockade of GPVI function.

ScFvs 1C3 and 10B12 recognizes distinct epitopes, and 1C3 is incapable of blocking static platelet adhesion to CRP or collagen-induced platelet aggregation.¹⁹  As expected, 1C3 (100 μg/mL) did not alter either the surface coverage (Figure 1E) or the size and morphology (Figure 3C) of the formed thrombi. However, 1C3 halved the PS expressing surface coverage (Figure 4), suggesting that the 1C3-binding site is relevant for GPVI function, possibly by influencing receptor clustering or dimerization. The Pi obtained with 1C3 was reduced to 0.24 ± 0.07 (P < .05, relative to control), a significantly higher Pi than with 100 μg/mL 10B12 (P < .001). Coinhibition with 10B12 and 1C3 did not change surface coverage from that of 10B12 alone (data not shown).

To study the dynamics of platelet activation, Ca²⁺ signal generation was measured in real-time under flow from single, Fluo-3-labeled platelets coming in contact with collagen. Control platelets exhibited a strong Ca²⁺ response, sometimes preceded by an initial Ca²⁺ spike (Figure 5A). A submaximal dose (50 μg/mL) of 10B12 decreased the Ca²⁺ signal amplitude (Figure 5B). Although blocking GPVI did not abolish primary adhesion, the increased translocation of platelets across the collagen surface was observed during Ca²⁺ measurements. Increasing the 10B12 concentration to 300 μg/mL reduced the Ca²⁺ signal slightly further (Figure 5C). No difference from controls was observed in the averaged Ca²⁺ response with 1C3 (100 μg/mL), though Ca²⁺-spiking occurred in individual platelets (Figure 5D). Together, these data indicate that blocking of GPVI with 10B12 potently suppressed collagen-dependent platelet activation pathways without eradicating primary adhesion.

High levels of VWF and high α₂β₁ density have been correlated with increased platelet deposition on collagen,^32,33  whereas patients with VWF concomitant with low α₂β₁ expression bleed more than patients with either condition alone.³⁴  The essential nature of GPIb and α₂β₁ in human platelet deposition on collagen under flow has been emphasized,^2,3,32,35  though other studies failed to detect consistent involvement of α₂β₁.⁴  This prompted us to revisit the topic in the context of molecular interplay with GPVI using the following antagonists: 12G1 Fab₂ fragment, recognizing amino acids 1 to 59 of GPIbα, hinders platelet adhesion to collagen by inhibiting VWF-GPIb interaction at high-shear^24,25 ; anti-α₂ mAb 6F1, whose epitope within the I-domain is separate from the MIDAS motif and lies between residues 173 and 259³⁶ ; and the GFOGER-GPP peptide, which inhibits collagen binding through the α₂ MIDAS.³⁷ 

The importance of GPIb in primary adhesion⁴  was reconfirmed under the shear rate conditions used here. Using 12G1 Fab₂ at a maximally effective dose of 40 μg/mL, the overall surface coverage was reduced by 50% ± 14% (Figure 6), which confirms that at (arterial) shear rates of approximately 1000 s^–1, platelet adhesion to collagen is only partially dependent on VWF-GPIb interaction.^38,39  Although in 2 donors surface coverage was strongly inhibited (63% and 67%), 2 gave a weaker reduction in thrombus formation (38% and 45%), but the differential response did not correlate with –5C/T polymorphism. For all donors, blocking of GPIb strongly influenced platelet tethering. GPIbα-blocked platelets moved swiftly over the collagen surface, but, once adherent, they became fully activated. Adding 12G1 Fab₂ reduced the PS exposure of platelets, but this simply reflected the decrease in surface coverage (Figure 6A) without a change in Pi. Like 12G1, saratin, which blocks the binding of VWF to collagen,⁴⁰  caused only a partial reduction of platelet surface coverage (68% ± 14.5%) and annexin V staining (27% ± 0.3%; n = 2).

Inhibition of α₂β₁ integrin by mAb 6F1, at a saturating concentration of 10 to 20 μg/mL, reduced surface coverage in all donors by approximately 30% ± 11% (Figure 6A). The tethering platelets reluctantly anchored to collagen fibers and, on arrest, formed only small, dense aggregates (Figure 1F). Image analysis indicated a smaller mean aggregate size of 117 ± 9 μm² with a shift toward round structures (P < .001). Aggregate sizes showed a kurtotic distribution (P < .0001), with a 50% increase in mid-sized features (Figure 3E). Similarly, GFOGER-GPP significantly reduced surface coverage by approximately 45% ± 13% at a maximally effective dose of 500 μg/mL (Figure 6A). Changes in aggregate morphology closely matched those obtained with 6F1 mAb—that is, aggregates appeared as round structures (Figure 1G), of smaller, average size (Figure 3F).

Inhibition of α₂β₁ with either mAb 6F1 or GFOGER-GPP had a small but significant effect (P < .01) on the exposure of PS (Figure 6A). The Pi was reduced from 0.43 ± 0.03 to 0.30 ± 0.04 (P < .05) and 0.37 ± 0.08 (not significant), respectively. To ensure maximal blocking of α₂β₁, we tested combinations of antagonists binding different epitopes. The most effective of 3 anti-β₁ mAbs tested, 4B4—against the activatory hinge region 207 to 218²² —caused less inhibition than GFOGER or 6F1 (data not shown). Slight further inhibition was seen when 6F1 was combined with 4B4 or GFOGER, but these reductions were not significant (Figure 6B). A role of α₂β₁ in platelet signaling was also supported by the decreased Ca²⁺ response of the collagen-adherent platelets. With both 6F1 (Figure 5E) and GFOGER-GPP (Figure 5F), the average Ca²⁺ response was decreased, and the responses of most single platelets showed transient or spiking increases in intracellular Ca²⁺. These findings reconfirm the contribution of GPIb and α₂β₁ to collagen-induced thrombus formation but underscore their different roles. Blocking GPIb interfered with platelet tethering and initial anchorage and dispersed aggregates with little effect on PS exposure, but α₂β₁ blockade reduced primary adhesion (partially by prolonging firm anchorage) and restricted thrombus size, platelet activation, Ca²⁺ signaling, and PS exposure.

Coinhibition of GPVI with 10B12 (100 μg/mL) and of GPIb with 12G1 Fab₂ (40 μg/mL) resulted in almost complete abolition of platelet adhesion, with remaining single platelets covering not more than 2% of the surface; PS-exposing platelets were hardly detected (Figure 7A-B). Similarly, coinhibition of GPVI and α₂β₁ with 6F1 (20 μg/mL) or GFOGER-GPP (500 μg/mL) completely inhibited stable platelet adhesion and all subsequent activation events, eradicating thrombus formation (Figure 7A-B). Calcium responses, measured in platelets adhering in the presence of 6F1 and 10B12, resembled the responses measured in the presence of 10B12 alone (data not shown).

In contrast, combined blocking of GPIb and α₂β₁ had a less dramatic effect on platelet deposition and aggregation because significant numbers of small aggregates could still be observed. Surface coverage after combined blockade of GPIb with α₂β₁ was reduced by approximately 70% (Figure 7A), a greater effect than achieved by blocking either receptor individually (P < .05). The remaining platelets were in an activated state, judged by the relatively high surface coverage of PS-exposing platelets and the apparent increase in Pi to 0.72. In conclusion, our data show that GPIb and α₂β₁ have partially overlapping functions in mediating human platelet adhesion and GPVI-dependent thrombus formation.

ADP regulation of thrombus size has been attributed to the platelet P2Y₁₂ purinergic receptor; P2Y₁ regulates the onset of thrombus formation.⁴¹  To obliterate any effects of ADP, we used apyrase in conjunction with specific P2Y₁₂ and P2Y₁ inhibitors. Full inhibition of ADP- or TxA₂-mediated (inhibited by ASA) effects reduced the surface coverage by 43% (P < .001). The morphology of platelet deposition was similar to that under GPVI blockade: a layer of single platelets remained on the collagen (Figure 1H), and aggregates were eradicated (Figure 3E). However, combined blockade of ADP and TxA₂ actions had no effect on Pi (0.53 ± 0.15 [not significant]) or platelet blebbing, though it slightly reduced the Ca²⁺ response (Figure 5G).

Inhibiting the ADP pathway further decreased surface coverage when combined with GPVI blockade. Irrespective of the 10B12 concentration (50 or 300 μg/mL), ADP receptor antagonism caused an additional reduction in platelet surface coverage (55%; P < .001) (Figure 8A), suggesting that the presence of ADP was not solely derived from GPVI-dependent secretion. Because GPVI inhibition itself abolished aggregate formation, the additional blockade of ADP receptors had no further effect on the mean feature size: 24 ± 2.6 μm² (ADP + GPVI blockage) and 23 ± 1.8 μm² (GPVI blockade alone) (Figure 8B). When applied together with 10B12, ADP receptor blockade reduced the number of PS-exposing platelets by attenuating platelet deposition because the Pi did not significantly decrease.

When the ADP receptors were coinhibited with α₂β₁ (20 μg/mL 6F1), surface coverage decreased by 41% (Figure 8A), and the mean aggregate size decreased from 121 ± 8 to 44 ± 4 μm² (P < .0002) (Figure 8B). Again, this was accompanied by a small decrease in PS-exposing platelets and an unaltered Pi value. These results let us conclude that ADP (from various sources) stimulates GPVI- and α₂β₁-dependent adhesion and contributes to aggregate formation but is insufficient to promote PS exposure. Thus, platelets respond to ADP by inducing one of the thrombus end points, aggregation, in contrast to full activation through GPVI.

The present study shows that, in humans, collagen-induced thrombus formation under high shear rate requires the orchestrated interplay of several platelet receptors. Using antibodies and peptides allowed clarification of the complementary and overlapping roles of the GPIb/V/IX complex, α₂β₁ integrin, GPVI, and the platelet purinergic receptors in the processes of platelet adhesion, aggregation, and procoagulant activity. Our results, though formally in line with the original 2-step adhesion-activation model, also revealed its oversimplification. Although the main role of GPIb is to provide initial contact with collagen-bound VWF, it stabilizes the anchorage of thrombi on collagen together with α₂β₁. GPVI is involved in GPIb- and α₂β₁-mediated adhesion through activation, and it is the main signaling receptor during thrombus formation. Conversely, α₂β₁ assists GPVI in signaling. Complete abrogation of adhesion and subsequent activation thus requires simultaneous inhibition of GPVI together with either GPIb or with α₂β₁.

We find that GPVI has a crucial role in regulating human thrombus growth. Inhibiting GPVI resulted in the complete loss of platelet aggregates and the full inhibition of collagen-induced Ca²⁺ mobilization and PS exposure. In contrast, the adhesion of single, nonaggregated platelets was not significantly diminished unless ADP was also antagonized, probably eliminating the contribution of α₂β₁. These results corroborate findings from earlier flow studies with GPVI-deficient platelets in which the surface coverage equaled that obtained with α_IIbβ₃ antagonists⁶ . A recent study using the blood of patients with GPVI deficiency also showed the obliteration of collagen-induced thrombus formation though a layer of single platelets was still visible in the PDF-format micrographs,⁴²  supporting our conclusions. In contrast, these results clearly deviate from those obtained in the first flow study using GPVI-deficient mice,¹³  which showed the total abolition of platelet-collagen adhesion. In mice, GPVI function was severely impaired when more than 90% of receptors are blocked. At the highest level of the scFv 10B12 used here (300 μg/mL), calculated receptor occupancy was 97%, supporting the idea that unaltered platelet adhesion was not caused by the inadequate inhibition of GPVI. Although it is possible that antibody dissociation or displacement by ligand may allow a proportion of GPVI to remain functional during multireceptor interaction under flow, 10B12 caused the near-complete, concentration-dependent abolition of platelet activation processes such as PS exposure, Ca²⁺ response, and aggregate formation. While this paper was in preparation, a true GPVI knockout model was generated in mice in which thrombus formation was similar to that observed in our study.⁴³  The reasons for the discrepancies regarding the different mouse data remain to be elucidated.

Both GPIb and α₂β₁ were implicated in the primary adhesive step at high shear rate, as others suggested after using recombinant receptors in liposomes.⁴⁴  Inhibiting GPIb or α₂β₁ on its own, but not GPVI, dramatically reduced the number of collagen-adherent platelets, though those remaining became activated. The interplay of GPVI with both receptors is implicit in the near-complete suppression of adhesion using the combined blockade of GPIb and GPVI or α₂β₁ and GPVI. This indicates that all 3 receptors are required for optimal adhesion. As a first step, GPIb tethers the platelet, allowing other receptors to interact, which perhaps further regulates GPIb function, as recently reported.^42,45  Adhesion through α₂β₁ follows closely, with GPVI and α₂β₁ acting interdependently. Initial α₂β₁ binding to high-affinity collagen sequences, not requiring activation, supports the weak GPVI-collagen interaction, which, in turn, mediates the switch of the integrin to the high affinity required for stable adhesion. In line with this, blocking GPVI with 10B12 caused a more dramatic (but not full) reduction in surface coverage on collagen type III containing only low-affinity xxxGER sequences, but not on collagen type I containing the high-affinity GFOGER sequence used in this study (P.A.S., unpublished data, 2002). Collagen-dependent differences in GPVI inhibition were also observed elsewhere.⁶  The failure of α2β1 blockade to cause further reduction in Ca²⁺ response and PS exposure in the presence of 10B12 implies that α₂β₁ signaling occurs downstream from GPVI. This conclusion was reinforced by almost complete loss of binding to adherent platelets, in the presence of 300 μg/mL 10B12, of an antibody that recognizes only the fully activated state of α₂-integrin, IAC-1 (7F6).⁴⁶  However, the signaling end points of this study (Ca²⁺, PS exposure, and aggregate size) were measured to identify the contribution of specific receptors rather than to study the detail of the signaling processes, which will be subject to further study.

The present model of human thrombus formation deviates in certain respects from that derived from murine data. Although murine thrombus formation seems to rely more on GPVI, it was recently shown that the interplay of α₂β₁ with GPVI is required for full GPVI-induced platelet activation and procoagulant activity, using β₁-deficient mouse platelets.²⁸  In addition, in another α₂(–/–) mouse flow model, platelet-collagen interaction was severely impaired.⁴⁷  Thus, the main deviation between human and mouse concerns the role of primary receptors in platelet adhesion to collagen under flow. There may be several nonexclusive reasons for discrepancy. First, human α₂β₁ may be competent to bind GFOGER sequences in type I collagen but not those of lower affinity, such as GASGER,⁴⁸  before GPVI involvement. Second, human α₂β₁ may become activated independently of GPVI, perhaps caused by GPIb activity. It has been proposed that GPIb uses the same Fcγ-mediated signaling pathway as GPVI,⁴⁹  and it reportedly induces rapid Ca²⁺ signaling under flow.⁵⁰  Similarly, the GPVI-independent presence of ADP in the blood may activate α₂β₁ at an early stage. Third, the threshold for α₂β₁ activation of human platelets may be low, that is, small signals from the slight percentage of GPVI not blocked by 10B12 may still be sufficient to trigger α₂β₁ for adhesion. Fourth, the human integrin-collagen contact—though dependent on GPIb and GPVI—may be more prone to autocrine signaling than that in mouse platelets. Finally, the considerable differences in human and murine GPVI structures may influence their contribution to the platelet-collagen interaction. Murine GPVI shares only 64% homology with the human receptor, and its cytoplasmic domain, half the length of the human 51-amino acid tail, lacks signaling motifs.^51,52  A human-to-mouse mutation, K59E, was shown to decrease GPVI binding to CRP, indicating important differences in the function of human and murine receptors.¹⁹ 

After the primary platelet-collagen contact, platelets form aggregates or become procoagulant; both processes contribute to full-blown thrombus formation. We defined 2 conditions in which the formation of thrombi was inhibited: blocking GPVI and blocking ADP- and TxA₂-mediated events. Inhibiting GPVI suppressed aggregate formation and signaling processes, detected as the greatly reduced Ca²⁺ and procoagulant response. However, ADP/TxA₂ antagonism reduced Ca²⁺ signaling only slightly, though no aggregates were formed, with no effect on Pi. GPVI regulates aggregate formation through autocrine ADP/TxA₂ secretion, in part through synergism with G_i/q-coupled receptors.⁵³  We found that ADP contributed to aggregation and primary adhesion but that its influence was not eliminated by GPVI inhibition. This implies that ADP was derived from sources other than GPVI-stimulated platelets.

Platelet aggregation was mostly accompanied by high Ca²⁺ signaling and PS exposure, even when blocking shear-dependent adhesion through GPIb. The platelets that escaped inhibition were activated, as demonstrated by an unchanged Pi. A notable exception occurred with the blocking of α₂β function. Ca²⁺ signaling, PS exposure, and aggregate formation were all affected. This implies that α₂β₁ has an activatory effect on these processes, in synergy with GPVI. Most, if not all, integrins exhibit inside-out and outside-in signaling,⁵⁴  suggesting that outside-in signaling from activated α₂β₁ can participate in platelet activation. We have previously shown that integrin-mediated adhesion to fibrinogen or collagen increases the responsiveness of platelets toward GPVI agonists.⁵⁵  Similar cross-talk between GPVI and α₂β₁ also likely controls platelet activation in the present experiments. Previously, others have linked α₂β₁ activation with α_IIbβ₃ up-regulation^56,57  and established cross-talk between α₂β₁ and α_IIbβ₃.⁵⁸ 

Together, the results presented here advance the idea that human thrombus formation occurs through the concerted action of several receptors: the interplay of GPIb/V/IX, integrin α₂β₁ and GPVI enables platelet deposition, and GPVI in cross-talk with α₂β₁ mediates subsequent thrombus formation, additionally supported by ADP. Yet these receptors may not be the only determinants of human platelet collagen interaction because other collagen receptors^35,57,59,60  may participate in the early stages of platelet contact with collagen, and new receptors are being discovered.^61,62  Ultimately, these other receptors may participate in the interaction and may be shown to allow further fine-tuning of the receptor cross-talk identified in this study.

We thank Dr Angela Rankin for kindly genotyping the donor blood and Dr Marc van Zandvoort for assisting with multiphoton microscopy.