Influence of Fibrillar Collagen Structure on the Mechanisms of Platelet Thrombus Formation Under Flow

Blood was collected through a 19-gauge needle from the antecubital vein of healthy adult donors with normal cell count profiles. The syringes contained the anticoagulant D-phenylalanyl-L-prolyl-L-arginine chloromethyl ketone dihydrochloride (PPACK; 40 μmol/L final concentration) that inhibited thrombin activation. Blood was supplemented with additional PPACK, where necessary, at approximately 2-hour intervals to ensure thrombin inactivation. All donors claimed to have abstained from taking aspirin, or other drugs known to affect platelet function, in the preceding 10 days.

Acid-insoluble fibrillar type I collagen from bovine achilles tendon (Sigma, St Louis, MO) in 0.5 mol/L acetic acid, pH 2.8, at a concentration of 2.5 mg/mL, was prepared as previously described16 and 200 μL was applied evenly over a horizontal glass coverslip (Corning, Inc, Corning, NY; 24 × 50 mm) covering all but the first 10 mm that remained uncoated to facilitate handling. Coated coverslips were then placed in a humid environment at room temperature (22°C to 25°C) for 60 minutes. Excess, unbound collagen was removed by 4 sequential rinses with phosphate-buffered saline (PBS), pH 7.4, and the coverslip was assembled in the flow chamber described below. Uncoated coverslips did not support platelet adhesion under the flow conditions used in these studies, and saturating the surface with bovine serum albumin (coating at 0.1 mg/mL) did not affect initial platelet adhesion or subsequent thrombus formation (unpublished observations).

Acid-soluble type I collagen from human placenta (Sigma) was dissolved in 0.1 mol/L acetic acid to give a stock solution at a final concentration of 2.5 mg/mL. The stock solution was diluted with PBS, pH 7.4, to a final concentration of 200 μg/mL for coating on glass coverslips as described above. Under these conditions, the surface was fully saturated with collagen; increasing the coating concentration up to 2.5 mg/mL had no effect on the extent of platelet adhesion or thrombus formation under all flow conditions examined (unpublished observations). These preparations were used in all flow studies except those experiments using a substrate derived from pepsin treatment of achilles tendon-insoluble collagen. In these studies, a suspension of acid-insoluble type I collagen from bovine achilles tendon (Sigma) at a concentration of 2.5 mg/mL was incubated with pepsin (Sigma) at an enzyme to substrate ratio of 1:1,000 (wt/wt), at 4°C for 18 hours. All insoluble material was sedimented by centrifugation at a relative centrifugal force of 100,000g for 3 hours, and the supernatant was removed. Untreated collagen suspension were also sedimented under the same conditions, and the supernatant was removed for control studies.

Fibrils of human type I collagen were formed by dialyzing acid-soluble type I collagen (from human placenta; Sigma) dissolved in 0.1 mol/L acetic acid at a concentration of 2.5 mg/mL against 4 changes of 4 L of PBS, pH 7.4, for 96 hours at 4°C. The concentration of reconstituted collagen was essentially unchanged from that of the starting material. This procedure has been used to prepare banded collagen fibers from pepsin-solubilized collagen and is based on the observations that purified solutions of collagen under physiological conditions in vitro assemble spontaneously into typical cross-striated fibrils.17 18 The resulting preparation was applied over a glass coverslip and kept in a humid environment for 60 minutes at 22°C to 24°C. Excess, unbound collagen was removed by rinsing the coverslip with PBS, pH 7.4, which was then assembled in the flow chamber.

The monoclonal antibodies used to inhibit the function of platelet receptors α_IIbβ₃, α₂β₁, and GP Ibα were purified from murine ascites using protein A (Sigma) chromatography.19These antibodies have been extensively characterized and their ability to inhibit receptor function has been well-documented. LJ-Ib1 (IgG₁) reacts with the amino-terminal 45-kD domain of GP Ibα containing the vWF binding site20-22 and inhibits fully the ligation of platelet GP Ib with immobilized vWF under a variety of flow conditions.23 LJ-CP8 (IgG₁) is specific for the integrin α_IIbβ₃ (GP IIb-IIIa complex) and blocks the activation-dependent binding of soluble ligands to this receptor24,25 as well as platelet aggregation and thrombus formation.25 MR-5 (IgG₁) binds to the A3 domain of vWF and prevents the binding of vWF to collagen under a variety of experimental conditions.24,25 Antibody R2-7E4 (IgG₁) was obtained from the same fusion as the previously reported antibody R2-8C826 that is specific for the α₂ integrin subunit and prevents the adhesion of Chinese hamster ovary cells transfected with α₂β₁ to type I collagen under stationary conditions. Antibody 12F1 (IgG_2a) binds to platelet α₂β₁, but does not inhibit receptor function.27 All antibodies were used at a final concentration of 100 μg/mL in whole blood. This concentration was shown to produce a maximal specific effect.

A modification of the Hele-Shaw perfusion chamber, described in detail elsewhere,28,29 was used to study the interaction of platelets in flowing blood with immobilized collagen under a variety of flow conditions. A flow path height of 254 μm was determined by a silicone rubber gasket placed on a collagen-coated coverslip that formed the lower surface of the chamber. The flow chamber was assembled and filled with PBS, pH 7.4. A syringe pump (Harvard Apparatus Inc, Holliston, MA) was used to draw blood through the flow chamber. Flow rates of 1.94, 0.65, and 0.13 mL/min produced initial wall shear rates of 1,500 s⁻¹, 500 s⁻¹, and 100 s⁻¹, respectively, at the inlet of the flow chamber. Measurements of platelet adhesion and thrombus formation were made at a position proximal to the inlet of the chamber, thereby avoiding pre-exposure of flowing platelets to either the substrate or to preformed platelet thrombi. The fluorescent dye mepacrine (quinacrine dihydrochloride; final concentration, 10 μmol/L) was used to label platelets in whole blood. Fluorescently labeled leukocytes were distinguished from platelets by their relatively large size, nuclear morphology, and sparsity; moreover, permanent leukocyte attachment to collagen was negligible at wall shear rates greater than 500 s⁻¹. At a wall shear rate of 100 s⁻¹, the contribution of adherent leukocytes to the total thrombus volume was typically less than 10% (unpublished observations). Erythrocytes were completely opaque to fluorescence detection and visualization due to fluorescence quenching by hemoglobin. Mepacrine concentrates in the dense granules of platelets and has been shown to have no effect on normal platelet function at the concentration used in these studies.30Platelet secretion after adhesion does not prevent their visualization, and mepacrine does not interfere with platelet adhesion.29The perfusion chamber was positioned on the motorized stage of an Axiovert 135M/LSM 410 inverted epifluorescence/laser scan confocal microscope (Carl Zeiss Inc, Oberkochen, Germany). The platelet adhesion process was visualized in real time and recorded on tape with a videocassette recorder (Magnavox; Philips, Eindhoven, The Netherlands).

The cumulative volume occupied by thrombi in an area of 102,236 μm² was calculated from confocal sections, obtained at 1.0-μm intervals in the z axis while blood was flowing using an excitation laser wavelength of 488 nm and a scanning time of 2 seconds per section. The microscope settings, including contrast, brightness, magnification, and pinhole aperture, were maintained at constant values to facilitate comparisons between different experiments. Confocal sections were analyzed using the Metamorph software package (Universal Imaging Corp, West Chester, PA). A threshold was applied to the image stack to distinguish thrombus sections from the background; this value was then used in all subsequent analyses of confocal sections for a given experiment. The area occupied by all thrombi in a given cross-section was calculated and the volume of a 1.0-μm–thick section was estimated by multiplying the area by the height of the section (1.0 μm). The cumulative volume occupied by all thrombi was then estimated as the sum total of the sectional volumes.

Platelets were defined as moving on the surface when they were displaced by a distance exceeding their diameter.29 A series of images from the recorded experiment was digitized at a sampling rate of 6 frames per second using a computer controlled VCR (Sony 9500; Sony Corp, New York, NY) and a frame grabber (Matrox Image LC; Matrox Electronic Systems Ltd, Dorval, Quebec, Canada). A threshold was applied to the images to distinguish platelets from the background, and the images were then binarized. Time was calculated by referring to the frame number. Image analysis was performed using the Metamorph software package (version 2.76; Universal Imaging Corp). The first 2 consecutive frames in the series were then superimposed (by using the logical AND function) so that the resultant image represented only the overlapping areas of the platelet at 2 different times. The new image was then superimposed to the next frame and the process was continued until the overlapping area was equal to 0. When this occurred, a platelet had moved by a distance greater than its diameter; if this did not occur, a platelet was considered firmly attached during the period of observation. The time for individual platelet motion was computed and the process was repeated until all the platelets in the microscopic field of study moved or for a preselected time interval, whichever occurred first.

Single frame images were captured from videotapes at various times after the onset of blood flow and a threshold was applied to distinguish both single platelets and platelet thrombi from the background; this value was then used in all subsequent image analyses of surface area coverage. The area occupied by all platelets and thrombi in an image was measured using the Metamorph software package (version 2.76; Universal Imaging Corp).

Aggregation induced by fibrillar type I collagen was studied in a dual channel aggregometer (Chrono-Log Corp, Havertown, PA) using 3 × 10⁸ platelets per milliliter of autologous plasma containing citrate as anticoagulant. Turbidity changes were recorded after the addition of agonist. The pH of the collagen suspension was adjusted to 7.4 before it was added to the platelet suspension.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed using standard procedures.31 

Frozen samples of immobilized collagen were placed in a Balzers freeze fracture machine (BAF-400T; Balzers AG, Liechtenstein) and were shadowed at an angle of 45° with platinum. The resulting replica was supported with a backing of carbon. The replica was subsequently cleaned with bleach, picked up on carbon-coated mesh grids, and examined on a Hitachi HU600 electron microscope (Hitachi Instruments Inc, San Jose, CA). Representative areas of specific interest were documented photographically.

Surface replication images of surface-bound insoluble fibrillar type I collagen (from bovine achilles tendon) and microfibrils derived from pepsin-solubilized type I collagen (from human placenta) showed distinct structural features (Fig 1). A characteristic striated pattern was seen with insoluble collagen due to the regular quarter-staggering of collagen monomers with a 67 nm periodicity,32 yielding regions of alternate maximal electron density regardless of the thickness of the fiber (Fig 1A). In contrast, this banded pattern was noticeably absent from the microfibrils derived from pepsin-solubilized collagen (Fig 1B). Here, collagen monomers polymerize to form thinner microfibrils with a clearly discernible spiraled structure comprising 2 to 3 fibers in a twisted configuration. Gel electrophoresis of pepsin-solubilized type I collagen under denaturing and nonreducing conditions showed the characteristic α1(I) and α2(I) bands (in the predicted ratio of 2:1) as well as high molecular weight multimers that did not enter the gel (Fig 1). The substrates shown in Fig 1 were prepared in an identical manner to those used for evaluating platelet adhesion and thrombus formation under flow described below and are therefore representative of the surfaces presented to flowing blood in these studies. Based on these structural features, the immobilized substrates prepared from insoluble and soluble type I collagen are referred to as banded and nonbanded collagens, respectively, throughout the text. Note also that preparations of nonbanded collagen may contain monomeric collagen in addition to polymerized helical fibrils.

Platelets in flowing whole blood adhered to both banded and nonbanded type I collagens at wall shear rates ranging from 100 s⁻¹ to 1,500 s⁻¹(Fig 2). Banded collagen presented a more thrombogenic surface than nonbanded collagen to flowing control blood at 1,500 s⁻¹, as reflected in the higher total thrombus volumes measured after 3 minutes perfusion, whereas the cumulative thrombus volumes measured with control blood at shear rates of 500 s⁻¹ and 100 s⁻¹ were essentially equivalent for both banded and nonbanded collagens (Fig 2). The concentrations of collagen used to coat the coverslips in these experiments produced maximal platelet adhesion: increasing the coating concentration of nonbanded collagen (up to 2.5 mg/mL) had no significant effect on the extent of platelet adhesion and thrombus development (data not shown). Furthermore, decreasing the coating concentration of banded collagen by up to 10-fold had no measurable effect on thrombus formation at the shear rates indicated.33 Therefore, differences in thrombogenicity seen at 1,500 s⁻¹ with these substrates were not due to differences in the extent of surface saturation with collagen.

In marked contrast to banded collagen, platelet adhesion and thrombus development on nonbanded type I collagen required competent α₂β₁ function, because functional inhibition of this integrin completely abolished stable platelet attachment at all shear rates tested (Fig 2), resulting in a transient surface translocation of platelets mediated by platelet GP Ibα and the A1 domain of collagen-bound vWF absorbed from plasma. A nonfunction blocking anti-α₂β₁ monoclonal antibody had no effect on the extent of platelet adhesion and thrombus formation with either banded or nonbanded collagen (data not shown). Blocking platelet α₂β₁ function had no significant effect on the extent of thrombus formation on banded collagen at any shear rate tested (P > .1 for all shear rates; Fig 2). Inhibition of plasma vWF A3 domain binding to collagen, or functional blockade of platelet GP Ibα, completely abolished platelet interaction with both banded and nonbanded collagens at 1,500 s⁻¹, but not at 500 s⁻¹ or 100 s⁻¹ (data not shown), in a manner similar to that previously described.33 

The differential role of platelet α₂β₁ in platelet adhesion to type I collagen fibrils having distinct structural features was confirmed with reconstituted type I collagen, produced from pepsin-solubilized type I collagen (from human placenta) by extensive dialysis of the latter against phosphate buffer, pH 7.4, at 4°C. These conditions have been shown to produce banded collagen fibers from pepsin-solubilized collagen and are based on the observation that purified solutions of collagen under physiological conditions in vitro assemble spontaneously into typical cross-striated fibrils.17 18 Reconstituted type I collagen prepared in this way supported platelet adhesion and thrombus formation in a manner that did not require competent α₂β₁function; blocking platelet α₂β₁ function with a monoclonal antibody resulted in only a slight, albeit significant (P < .05) decrease in the total thrombus volume after 3 minutes of perfusion at 1,500 s⁻¹ compared with controls (Fig 2). Furthermore, at wall shear rates of 100 s⁻¹ and 500 s⁻¹, inhibition of α₂β₁ function had no significant effect on the extent of thrombus formation compared with control blood (P> .1 at both shear rates; Fig 2). These results contrast sharply those seen with nonbanded type I collagen, in which inhibition of α₂β₁ function completely abolished stable platelet attachment and thrombus formation at all shear rates. Thus, reconstituted type I collagen showed distinct functional properties compared with the nonbanded type I collagen from which it was derived, demonstrating that structural characteristics of fibril assembly have a profound effect on the mechanisms of platelet interaction with this substrate under flow.

Further confirmation of the differential role of platelet α₂β₁ in platelet adhesion to different type I collagen preparations was obtained by controlled pepsin-digestion of native, banded type I collagen obtained from achilles tendon. The supernatant obtained after sedimenting pepsin-treated banded type I collagen, when immobilized on glass, supported stable platelet attachment and thrombus formation at shear rates ranging from 100 s⁻¹ to 1,500 s⁻¹ in a manner that required intact α₂β₁ function; blocking α₂β₁ function resulted in a continuous turnover of transiently attached platelets with no further accrual of surface associated platelets, even after prolonged perfusion of blood (data not shown). Thus, pepsin-solubilized type I collagen prepared from either achilles tendon or placenta required competent α₂β₁ function for irreversible platelet attachment and thrombus formation.

The role of platelet α₂β₁ was further studied by measuring the extent of platelet surface coverage as a function of time after the onset of blood flow. Platelets adhered most rapidly to banded collagen in a manner that was essentially independent of α₂β₁ function (Fig 3A). Under these conditions, half maximal surface coverage was achieved after approximately 30 seconds. In contrast, the rate at which the surface became covered with platelets was lower on nonbanded collagen exposed to the same control blood at this shear rate (half maximal surface coverage was achieved after ∼1.5 minutes), and blocking α₂β₁function resulted in the complete inhibition of stable platelet attachment and subsequent lack of surface accumulation of platelets at any time after the onset of blood flow (Fig 3A). Furthermore, reconstituted collagen presented a surface that differed substantially from both the nonbanded collagen from which it was derived and from native, banded collagen. Blocking platelet α₂β₁ function with this substrate resulted in a prolonged lag phase before the accrual of firmly attached platelets (Fig 3A), although after 3 minutes of perfusion, there was significant platelet adhesion and thrombus formation, as shown previously (Fig 2). Although the anti-α₂β₁antibody was clearly effective in blocking α₂β₁ function, as evidenced by its inhibitory effect at early time points after the onset of blood flow over this substrate, prolonged perfusion of blood resulted in platelet adhesion and thrombus formation in a manner that was independent of α₂β₁ function (Fig 3A).

In a previous study using native banded type I collagen,33the efficacy of the anti-α₂β₁ monoclonal antibody was confirmed by concurrently blocking platelet α_IIbβ₃ function with a well-characterized anti-α_IIbβ₃ monoclonal antibody.24 25 Because α_IIbβ₃function is required for platelet aggregate and thrombus formation, its selective inhibition allowed direct evaluation of individual platelet surface interactions. In the present study, we have used a similar strategy to demonstrate the inhibitory efficacy of the anti-α₂β₁ antibody using reconstituted type I collagen. Platelets adhered to a surface coated with this substrate in a predominantly stable manner after 3 minutes at a wall shear rate of 1,500 s⁻¹, when α_IIbβ₃function alone was inhibited (Fig 3B); only 30% to 35% of attached platelets were displaced on the surface by a distance greater than their respective diameters over a period of 6 seconds. In contrast, when platelet α₂β₁ and α_IIbβ₃ function were blocked concurrently, a predominantly transient platelet attachment was seen, with approximately 85% of platelets displaced over the same time period (Fig 3B). A control, nonfunction blocking anti-α₂β₁ monoclonal antibody had no effect on the stability of platelet attachment when α_IIbβ₃ function was blocked (data not shown). These differences in the stability of platelet attachment were reflected in the number of surface-attached platelets at 3 minutes after the onset of blood flow, as shown visually in Fig 3B. Thus, when only α_IIbβ₃ function was blocked, the majority of platelets were irreversibly adherent through bonds involving the ligation of α₂β₁ with domains in collagen (after the initial tethering of platelets, mediated by platelet GP Ibα engagement of collagen-bound vWF, derived from plasma, as described below). These results confirm that the anti-α₂β₁ monoclonal antibody used in these studies was of sufficient inhibitory efficacy to prevent ligation of α₂β₁ with domains in reconstituted collagen, despite its inability to prevent stable platelet attachment and thrombus development on this substrate when α_IIbβ₃ function was not concurrently blocked, as shown in Figs 2, 3, and 4.

The lag phase before the onset of stable platelet attachment to reconstituted collagen when α₂β₁ function was inhibited may reflect the time-dependent binding of plasma components to collagen, including vWF or other α_IIbβ₃ ligands such as fibronectin, which can then mediate stable platelet attachment and thrombus development in a manner that does not require the participation of α₂β₁. To test this hypothesis, autologous plasma was perfused over reconstituted collagen at a wall shear rate of 1,500 s⁻¹ for 5 minutes before perfusing with either control whole blood or blood containing the anti-α₂β₁ antibody. When α₂β₁ function was not blocked, the rate of surface coverage was similar, regardless of whether the substrate was previously exposed to plasma under flow or to purified vWF under stationary conditions (Fig 4A). In marked contrast, when α₂β₁ function was inhibited, the initial lag phase was substantially reduced when the substrate was pre-exposed to flowing plasma or precoated with purified vWF under stationary conditions (Fig 4A). In this regard, reconstituted type I collagen differed substantially from the nonbanded collagen from which it was derived, because pre-exposing nonbanded collagen to plasma under flow or to purified vWF did not ameliorate the inhibitory effect of blocking α₂β₁ (Fig 4B). In fact, the extent of surface coverage when nonbanded collagen was precoated with vWF, and α₂β₁ function was blocked, was similar to that seen with vWF coated directly onto a glass surface without collagen (data not shown). When α₂β₁ function was blocked, cumulative thrombus volumes measured after 2.5 minutes of perfusion were significantly higher when reconstituted collagen was pre-exposed to plasma under flow or to purified vWF under stationary conditions (Fig 4A).

Insoluble banded collagen, when added to platelets in plasma with citrate as anticoagulant, caused irreversible platelet aggregation in a dose-dependent manner; maximal platelet aggregation response was seen at an agonist concentration of 20 μg/mL (Fig 5). In contrast, nonbanded collagen was less efficient at inducing platelet aggregation; a 50-fold increase in concentration compared with banded collagen was required for a full platelet response, and even then, there was a prolonged lag phase before the onset of full aggregation (Fig 5). Blocking platelet α₂β₁ function had no effect on the lag phase or extent of platelet aggregation with banded collagen at a concentration of 20 μg/mL, whereas, at suboptimal concentrations of this agonist (2.0 μg/mL), there was a partial inhibition of platelet aggregation; increasing the antibody concentration did not produce further inhibition under these conditions (data not shown), indicating the presence of an activating collagen receptor on platelets other than α₂β₁. In fact, simple collagen-like peptides have been synthesized that are potent platelet agonists whose activity is totally α₂β₁-independent.34 In contrast, blocking platelet α₂β₁ function completely inhibited platelet aggregation induced by nonbanded collagen (Fig 5), indicating that this integrin is essential to mediate platelet activation induced by this agonist. However, reconstituted collagen prepared from nonbanded collagen was a potent platelet agonist, promoting platelet activation and aggregation in a manner that was independent of α₂β₁ function when used at a concentration of 10.0 μg/mL (Fig 5).

Blocking the function of platelet integrins α_IIbβ₃ and α₂β₁concurrently with inhibition of the binding of plasma vWF completely abolished platelet interaction, including initial tethering, with nonbanded collagen at wall shear rates ranging from 100 s⁻¹ to 1,500 s⁻¹(Fig 6). Thus, with this substrate, platelet adhesion and thrombus formation under flow require the functional integration of 3 platelet receptors at a wall shear rate of 1,500 s⁻¹, namely GP Ibα, α₂β₁, and α_IIbβ₃. In contrast, α₂β₁ is not required with native banded collagen when coated at high density under the conditions used in the present study (Figs 2 and 3), although blockade of α₂β₁ caused significant inhibition of thrombus formation at low surface densities of this substrate.33 In contrast with the results obtained with nonbanded collagen, reconstituted collagen supported stable platelet attachment at 100 s⁻¹ and 500 s⁻¹when vWF binding to collagen was blocked concurrently with functional inhibition of both α₂β₁ and α_IIbβ₃ (Fig 6), implicating the participation of another collagen receptor(s) other than α₂β₁, possibly GP VI,7,8 GP IV,9-11 or the recently identified 65-kD collagen-binding protein.12 Results similar to those seen for reconstituted collagen were obtained with native, banded collagen (data not shown). Stable platelet attachment under these conditions was completely inhibited when divalent cations were chelated with EDTA (data not shown). Therefore, platelet adhesion to banded (native or reconstituted) collagen type I at shear rates less than 500 s⁻¹, although cation-dependent, does not require vWF or the platelet integrins α₂β₁ or α_IIbβ₃, whereas requisite α₂β₁-collagen pairing leading to stable platelet attachment to nonbanded collagen was always observed at shear rates ranging from 100 s⁻¹ to 1,500 s⁻¹ (Fig 2). Consistent with the present findings are recent binding studies showing that platelets interact with soluble and insoluble collagens through different mechanisms, implying that α₂β₁ and another collagen receptor mediate collagen-platelet interactions with different specificities.35 

Simultaneous inhibition of platelet α_IIbβ₃function and the binding of plasma vWF to collagen resulted in the stable attachment of single platelets at shear rates up to approximately 1,000 s⁻¹ when blood was perfused over nonbanded collagen (Fig 7). The extent of surface coverage was inversely related to the wall shear rate, with virtually no platelet-surface interactions at 1,500 s⁻¹. Under these conditions, platelet attachment was contingent upon ligation of collagen domains with α₂β₁ on platelets; blocking α₂β₁ function under these conditions completely prevented platelet interaction with this substrate (Fig 6). The biomechanical properties of the α₂β₁-collagen bond appear to be similar to those observed for α_IIbβ₃-fibrinogen interactions, in which α_IIbβ₃-mediated platelet adhesion to surface-bound fibrinogen cannot occur above a limit wall shear rate.29 Functional inhibition of α₂β₁ by chelation of divalent cations with EDTA completely abolished platelet interaction under these conditions (data not shown).

The nonbanded collagen structures seen with the substrate derived from pepsin-solubilized type I collagen (Fig 1B) may be a distinctive feature of fibrils generated by enzymatic cleavage of collagen precursors,14 because monomers obtained from pepsin-treated collagen have altered nonhelical extremities.13 Presumably, monomers prepared in this manner lack crucial domains required for the assembly of lateral aggregates present in native fibrillar type I collagen, but can generate spiraled structures by lateral packing of subfibrils. These structural differences had a profound effect on the mechanisms of platelet adhesion and thrombus formation under flow: platelet α₂β₁ function was essential for the permanent arrest of platelets on nonbanded collagen, whereas native banded type I collagen supported platelet adhesion and thrombus growth even when α₂β₁ function was compromised. Studies of patients with a deficiency of platelet α₂subunit or with autoantibodies directed against this protein attest to the involvement of α₂β₁ in collagen-platelet interactions in vivo, because platelet adhesion to collagen and collagen-induced aggregation are both impaired.36-39 However, simple collagen-like peptides consisting essentially of a repeating Gly-Pro-Hyp sequence have been synthesized that are potent platelet agonists whose activity is totally α₂β₁-independent.34 The platelet reactivity of these peptides could be attributed solely to their tertiary (triple helical) and quaternary (polymeric, cross-linked) structures. Thus, platelet-collagen interactions in vivo are likely to involve the integration of at least 2 distinct receptor-collagen interactions acting in concert: one involving α₂β₁ as an adhesion/signaling receptor and the other involving the recognition of collagen structural features by a receptor other than α₂β₁. A likely candidate for the latter function is GP VI which appears to play a major role in collagen-induced platelet activation.8 

Recent studies involving the adhesion of GP VI-deficient platelets to collagen under flow testify to the importance of GP VI as a signaling receptor required for the activation of platelet α_IIbβ₃.8 Moreover, GP VI-deficient platelets were unresponsive to the collagen-related peptides comprising a repeating Gly-Pro-Hyp motif that mimic the collagen tertiary and quaternary structure.40 GP VI therefore appears to act as a signaling receptor after recognition/coupling with distinct structural components of collagen. In this context, collagen-induced activation of the protein-tyrosine kinase Syk was severely compromised in GP VI-deficient platelets.41 In a recent report, the activation of α_IIbβ₃ induced by platelet adhesion to collagen was found to be mediated by both α₂β₁ and GP VI.42 In this study, platelet adhesion to insoluble fibrillar type I collagen under static conditions was mediated by both α₂β₁and GP VI, whereas platelet adhesion to monomeric type I collagen was exclusively mediated by α₂β₁. This report is entirely consistent with the present study in which we demonstrate that, under physiologically relevant flow conditions, platelet adhesion and subsequent thrombus formation on nonbanded collagen requires competent α₂β₁ function, whereas with banded collagen another receptor other than α₂β₁, possibly GP VI, is involved and can mediate platelet adhesion and thrombus formation even when α₂β₁ function is blocked. However, it should be noted that, although the anti-α₂β₁ antibody blocks adhesion to nonbanded collagen, this does not exclude the role for a second collagen receptor with this substrate, which may act in concert or synergistically with α₂β₁ such that participation of both receptors may be required for stable platelet attachment. Such a mechanism would be analogous to that previously described for the irreversible adhesion of platelets to immobilized vWF that requires the functional integration of 2 receptors, namely GP Ib α and α_IIbβ₃.29,43Furthermore, a conformational change triggered by collagen binding to α₂β₁ may induce an increased affinity of the receptor for its ligand.44 

Reconstituted type I collagen showed distinct functional properties compared with the nonbanded collagen from which it was derived. First, platelet adhesion and thrombus growth under flow did not require competent α₂β₁ function, although when this receptor was blocked, there was a prolonged lag phase before the accrual of firmly attached platelets compared with that seen with native banded collagen (Fig 3A). The delayed response before the onset of α₂β₁-independent platelet attachment proved to be related to the time-dependent binding of plasma components, as evidenced by the significant increase in the total thrombus volume measured after 2.5 minutes of perfusion when reconstituted collagen was pre-exposed to plasma under flow or to purified vWF under stationary conditions (Fig 4A). Thus, the rate of surface coverage by platelets and the development of platelet thrombi on this substrate are processes that are greatly accelerated by the presence of collagen-bound vWF and/or other collagen-bound plasma components when α₂β₁ function is compromised. The present findings indicate that collagen structural characteristics may regulate the extent and/or affinity of the binding of components, including vWF or other α_IIbβ₃ ligands such as fibronectin, present in plasma. In this context, activated α_IIbβ₃ promotes platelet arrest and subsequent thrombus growth on surface-bound vWF by interacting with the Arg-Gly-Asp sequence in the C1 domain near the carboxyl terminus of the vWF subunit.29 The extent to which this process occurs when plasma vWF or other α_IIbβ₃ ligands are immobilized onto collagen may therefore be regulated by the morphology of the collagen fibrils. Despite differences with regard to the role of α₂β₁ in mediating the adhesion of platelets flowing over banded and nonbanded collagens, binding of plasma vWF was requisite at 1,500 s⁻¹, but not at 100 s⁻¹, for both substrates.

Although platelet α₂β₁ has generally been considered to be constitutively active,45 recent studies suggest that α₂β₁ on nonactivated platelets does not bind soluble type III collagen, whereas activation of platelets transforms it to a state with higher affinity for soluble type III collagen.35 It can be inferred from the present study that soluble type I collagen binds to α₂β₁ on resting platelets, because it induces platelet aggregation in an α₂β₁-dependent manner (Fig 5). Furthermore, when soluble type I collagen was immobilized on a surface, blocking platelet activation with prostaglandin E₁(PGE₁) did not prevent α₂β₁-dependent platelet attachment under flow (data not shown). It therefore appears that α₂β₁ on nonactivated platelets can recognize and ligate domains in soluble type I collagen when the latter is immobilized on a surface. We have also demonstrated the inhibitory efficacy of the anti-α₂β₁ antibody under conditions in which platelets are known to be activated (eg, under high shear flow when α_IIbβ₃ function is blocked to prevent platelet aggregation; Fig 3B). Thus, the anti-α₂β₁ antibody used in these studies blocks α₂β₁ function on both resting and activated platelets.

Because there are multiple potential collagen receptors on platelets, their relative contribution in adhesion/signaling induced by collagen is difficult to discern. As shown in the present study, multiple collagen receptors can participate in adhesion to banded (native or reconstituted) type I collagen under flow, whereas α₂β₁ is absolutely required for the initial attachment of platelets to nonbanded collagen fibrils. Furthermore, conditions can be obtained (by concurrent blocking of GP Ibα and α_IIbβ₃) where competent α₂β₁ function is requisite for platelet attachment to this substrate under flow. Such conditions could therefore be exploited to compare distinct signaling pathways associated with α₂β₁-dependent and α₂β₁-independent adhesion to immobilized collagen under flow.

Although at least 7 genetically distinct collagens (types I, III, IV, V, VI, VIII, and XIII) have been identified as constituents of the vessel wall,46 their relative contribution to the thrombogenicity of the extracellular matrix after vascular injury by tissue trauma or by inflammatory and degenerative processes has not been clearly defined. Structural differences in platelet binding to these different collagen types under flow have been reported.2,47 For example, pepsin-digested collagen type VI-coated surfaces were found to support platelet adhesion only at relatively low wall shear rates (100 s⁻¹), although these surfaces were found to be more reactive than collagen type I surfaces under the same low shear conditions.47 Such studies imply that the relevance of different collagen types in hemostasis and thrombosis may depend on their relative distribution in the vascular extracelluar matrix as well as the prevailing flow conditions in different regions of the vasculature. However, more recent studies have demonstrated that intact tetrameric collagen type VI supports platelet adhesion and aggregation at high shear rates in a manner similar to that seen for collagen type I,48suggesting a stringent requirement for an intact conformational macromolecular structure. Whereas collagen types III, IV, and VI may be more relevant in normal blood vessels, collagen type I may be important in fibrous atherosclerotic plaques that contain a higher proportion of collagen type I compared with normal arterial vessel walls49 and in deep vascular injuries where it is abundant in the media and adventitia of blood vessel.50 Physiologic degradation of collagen fibrils by interstitial enzymes such as matrix metalloproteinases is known to occur at extracellular spaces; their assembly to form spiral, nonbanded collagen may represent a process that regulates platelet responses to vascular injury at sites of vascular wound repair during remodeling of the vasculature.

In conclusion, our results show that type I collagens with different structural characteristics elicit platelet adhesion and aggregation through distinct mechanisms as evidenced by the differential role of platelet α₂β₁ under various flow conditions. We also demonstrate that a collagen receptor(s) other than α₂β₁ can selectively engage domains in banded, but not nonbanded type I collagen when α₂β₁ function is blocked and that the state of collagen fibril assembly may regulate the extent and affinity of the binding of plasma vWF as well as other plasma components, crucial for a full thrombotic response under high shear flow.

The authors thank James R. Roberts and Benjamin Gutierrez for preparing monoclonal antibodies; Rolf Habermann, Judith Dent, and Dr Enrique Saldivar for help with videomicroscopy, computation, and image analysis; Dr Malcolm Wood and George Klier (deceased) for assistance with surface replication analysis; and Dr Virgil Woods for providing the nonfunction blocking anti-α₂β₁monoclonal antibody, 12F1.