The Interaction of Integrin αIIbβ3 with Fibrin Occurs through Multiple Binding Sites in the αIIb β-Propeller Domain*

Background: During thrombus formation, platelet integrin αIIbβ3 binds fibrin; however, the mechanism of this interaction is unclear. Results: Mutations of discontinuous negatively charged and aromatic residues in the αIIb β-propeller domain impair fibrin clot retraction and cell adhesion. Conclusion: Integrin αIIbβ3 has multiple binding sites for fibrin. Significance: Uncovered recognition specificity of αIIbβ3 for fibrin may be used to select inhibitors of this interaction. The currently available antithrombotic agents target the interaction of platelet integrin αIIbβ3 (GPIIb-IIIa) with fibrinogen during platelet aggregation. Platelets also bind fibrin formed early during thrombus growth. It was proposed that inhibition of platelet-fibrin interactions may be a necessary and important property of αIIbβ3 antagonists; however, the mechanisms by which αIIbβ3 binds fibrin are uncertain. We have previously identified the γ370–381 sequence (P3) in the γC domain of fibrinogen as the fibrin-specific binding site for αIIbβ3 involved in platelet adhesion and platelet-mediated fibrin clot retraction. In the present study, we have demonstrated that P3 can bind to several discontinuous segments within the αIIb β-propeller domain of αIIbβ3 enriched with negatively charged and aromatic residues. By screening peptide libraries spanning the sequence of the αIIb β-propeller, several sequences were identified as candidate contact sites for P3. Synthetic peptides duplicating these segments inhibited platelet adhesion and clot retraction but not platelet aggregation, supporting the role of these regions in fibrin recognition. Mutant αIIbβ3 receptors in which residues identified as critical for P3 binding were substituted for homologous residues in the I-less integrin αMβ2 exhibited reduced cell adhesion and clot retraction. These residues are different from those that are involved in the coordination of the fibrinogen γ404–411 sequence and from auxiliary sites implicated in binding of soluble fibrinogen. These results map the binding of fibrin to multiple sites in the αIIb β-propeller and further indicate that recognition specificity of αIIbβ3 for fibrin differs from that for soluble fibrinogen.

Integrin ␣ IIb ␤ 3 , a major membrane protein expressed on the surface of platelets, plays central roles in normal hemostasis and pathological thrombosis. On stimulated platelets, ␣ IIb ␤ 3 serves as a specific receptor for the plasma protein fibrinogen. Fibrinogen binding to activated ␣ IIb ␤ 3 induces platelet aggregation, the essential cellular event in the formation of the primary hemostatic plug. Furthermore, platelets bind fibrin, a product of the enzymatic transformation of soluble fibrinogen into insoluble fibrin, which is formed early and dominates the entire process of thrombus growth (1)(2)(3). Because the same molecular pathways mediate pathological thrombus formation, the interaction between ␣ IIb ␤ 3 and fibrinogen has been targeted for antithrombotic therapy (for a review, see Ref. 4). It has also been proposed that inhibition of platelet interactions with fibrin may be a necessary and important property of ␣ IIb ␤ 3 antagonists (5). The initial interaction of soluble fibrinogen with ␣ IIb ␤ 3 occurs via the COOH-terminal sequence in the globular ␥C domains of fibrinogen with ␥ 404 GAKQAGDV 411 (␥C peptide) providing critical coordination residues that bind to the interface between the ␣ IIb ␤-propeller domain of the ␣ integrin subunit and the ␤ 3 I domain of the ␤ subunit (6). The ␥C sequence is unique to fibrinogen, and binding of fibrinogen to ␣ IIb ␤ 3 through ␥C is highly specific (7,8). Although four integrin recognition RGD sequences are present in fibrinogen and the RGD peptide inhibits ␣ IIb ␤ 3 adhesive reactions and can bind within the same pocket that is occupied by ␥GAKQAGDV, none of the RGDs in fibrinogen are required for platelet aggregation (9).
Fibrinogen binding to ␣ IIb ␤ 3 is a multistep process: initial reversible contact is followed by irreversible binding such that the bound ligand no longer readily dissociates (10,11). The binding of fibrinogen to the receptor is accompanied by the alteration of fibrinogen conformation and leads to unmasking of cryptic sequences that potentially can serve as new ␣ IIb ␤ 3binding sites (12,13). Also, as the thrombus formation proceeds, the interaction of ␣ IIb ␤ 3 with fibrin engages new contacts that lead to clot retraction. Thus, the overall process of throm-bus formation in vivo involves the interactions of ␣ IIb ␤ 3 with different forms of fibrinogen: soluble fibrinogen and an insoluble fibrin(ogen) matrix. The evidence accumulated so far suggests that these interactions involve differential recognition specificity. In contrast to platelet aggregation, the ␥C sequence is not absolutely required for adhesion to immobilized fibrinogen and fibrin clot retraction (14,15). Furthermore, RGDs do not contribute to ␣ IIb ␤ 3 -mediated clot retraction. Recombinant human fibrinogen in which all RGDs in the A␣ chains were mutated and ␥ 408 AGDV 411 in the ␥C domains were truncated exhibits delayed but otherwise normal clot retraction (16). Also, neither RGD nor ␥C peptides inhibit clot tension development during retraction (17), and some anti-␣ IIb ␤ 3 mAbs inhibit clot retraction but not fibrinogen binding and vice versa (17)(18)(19). In addition, fibrinogen from mice in which the ␥C domain was targeted to delete ␥ 407 QAGDV 411 does not support platelet aggregation but still mediates normal clot retraction (20). Finally, some ␣ IIb ␤ 3 antagonists have different efficacies in inhibiting clot retraction despite the equivalent antiaggregatory potency (21). Taken together, these data indicate that the site(s) involved in the initial binding of fibrinogen to ␣ IIb ␤ 3 during platelet aggregation is different from those that participate in the interaction of platelets with the insoluble fibrin(ogen) matrix during thrombus growth and clot retraction.
The existence of alternative binding sites in addition to ␥C and RGD that are involved in binding of fibrinogen to ␣ IIb ␤ 3 was first suggested by Parise et al. (22). They found that ␣ IIb ␤ 3 binding to fibrinogen immobilized on agarose was not inhibited by either RGD or ␥ 400 HHLGGAKQAGDV 411 (named the H12 peptide). The subsequent studies have localized two sites in the ␥C domain that may mediate the interaction of ␣ IIb ␤ 3 with insoluble forms of fibrin(ogen). The mutations within the ␥316 -322 sequence of recombinant fibrinogen diminished platelet aggregation and platelet adhesion under flow (23,24). We have previously identified the sequence ␥ 370 ATWKTRWYSMKK 381 (termed P3) as the binding site for ␣ IIb ␤ 3 in adhesion and clot retraction (15,25). We further found that the mechanism by which ␣ IIb ␤ 3 binds P3 is distinct from the ␥C recognition. First, P3-mediated adhesion of platelets to fibrinogen fragments lacking the ␥C residues 406 KQAGDV 411 does not require their prior stimulation, whereas the engagement of ␥C by ␣ IIb ␤ 3 is activation-dependent (15). Second, P3 is fibrin-specific in that it is poorly exposed on the surface of intact soluble fibrinogen but becomes fully available after the transformation of fibrinogen to fibrin or after deposition of fibrinogen on various surfaces, including aggregated platelets (12,26). Third, P3 binding to ␣ IIb ␤ 3 depends on its positively charged residues (25). Because P3 contains no sequences resembling the ␥ 404 GAKQAGDV 411 or RGD motif, it is reasonable to assume that the binding site(s) for P3 in ␣ IIb ␤ 3 is unlike that utilized by RGD or ␥C. Here, we performed the binding analyses to demonstrate that ␣ IIb ␤ 3 contains multiple binding sites for P3. Furthermore, using synthetic peptide libraries and mutational analyses, we have localized these sites in the ␣ IIb ␤-propeller domain of the receptor.
The 9-fluorenylmethoxycarbonyl (Fmoc)-protected and pentafluorophenyl-activated amino acids for synthesis of the peptide libraries were purchased from Bachem (King of Prussia, PA). Pentafluorophenyl-activated Trp was obtained from Novabiochem. mAb AP3 against the ␤3 integrin subunit was isolated from hybridoma cells obtained from ATCC (Manassas, VA). The anti-␣ IIb mAb sc-51654 was from Santa Cruz Biotechnology (Dallas, TX), mAb LM609 was from Millipore (Billerica, MA), and mAb 7E3 was a gift from Dr. B. Coller. The plasmids pcDNA3.1/Neo and pcDNA3.1/Hygro containing the full-length cDNA encoding the human ␣ IIb and ␤ 3 integrin subunits, respectively, were provided by Dr. T. O'Toole. The primers for mutagenesis were obtained from Integrated DNA Technologies, Inc. (Coralville, IA). PfuTurbo DNA polymerase was from Agilent (Santa Clara, CA), and Lipofectamine 2000 was from Invitrogen.
Surface Plasmon Resonance (SPR)-The interaction of ␣ IIb ␤ 3 with P3 was examined using a BIAcore 2000 SPR-based biosensor (Biacore AB, Uppsala, Sweden). Purified ␣ IIb ␤ 3 was coupled to a CM5 sensor chip (Biacore) using the amine coupling kit according to the manufacturer's protocol. The sensor chip was coated to achieve ϳ1500 response units, which correspond to ϳ62 M ␣ IIb ␤ 3 . Different concentrations of P3 in HBSP buffer (10 mM HEPES buffer with pH 7.4, 150 mM NaCl, and 0.005% v/v Surfactant P20) (Biacore) containing 1 mM CaCl 2 and 1 mM MgCl 2 were passed over the flow cell at 10 l/min, and the association between immobilized protein and peptide was detected as the change in the SPR response. All data were corrected for the response obtained using a blank reference flow cell that was activated with 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide/N-hydroxysuccinimide and then blocked with ethanolamine. The chip surface was regenerated with 2 M NaCl and 50 mM NaOH followed by re-equilibration with the binding buffer. Experimental data were analyzed using the BIAevaluation 4.1 program supplied with the instrument. The data for the construction of Scatchard plots were obtained from the equilibrium portions of SPR sensorgrams, and the dissociation equilibrium constant (K d ) was estimated by analysis of the binding curve using the steady-state affinity model provided by the same software.
Synthesis of Cellulose-bound Peptide Libraries and Screening for P3 Binding-The ␣ IIb ␤-propeller (residues 1-451)-derived peptide libraries were prepared by parallel spot synthesis as described (25,29). Peptides were COOH-terminally attached to cellulose via a (␤-Ala) 2 spacer and were acetylated NH 2 -terminally. The cellulose membranes with covalently coupled peptides were incubated for 1 min in methanol and then washed with TBS buffer. After blocking with 1% BSA for 2 h at 22°C, the membranes were incubated with 10 g/ml 125 I-labeled P3 (10 5 cpm/ml) in phosphate-buffered saline (PBS) for 3 h at 22°C. After washing with TBS containing 0.05% Tween 20, the membranes were dried, and P3 binding was detected by autoradiography and analyzed by densitometry as described (25).
Cells and Stable Transfection of Integrin Subunit Constructs-Platelets were collected from fresh aspirin-free human blood in the presence of 2.8 M prostaglandin E 1 and isolated by differential centrifugation followed by gel filtration on Sepharose 2B in divalent cation-free Tyrode's buffer, pH 7.2 containing 0.1% BSA. Human embryonic kidney 293 (HEK293) cells were stably transfected with pcDNA3.1 plasmids with inserted wild-type (WT) ␣ IIb and ␤ 3 or mutant ␣ IIb and WT ␤ 3 using Lipofectamine 2000 reagent (Invitrogen). After 48 h at 37°C in 5% CO 2 , cells were harvested and cultured in medium with 0.5 mg/ml G418 (Invitrogen) and 0.25 mg/ml hygromycin (Invitrogen). After 14 days, surviving cells were collected and sorted. The expression of ␣ IIb ␤ 3 on the surface of the cells was evaluated by FACS analyses using anti-␤ 3 mAb AP3 (10 g/ml) and a FACSCalibur flow cytometer (BD Biosciences). The ␣ IIb ␤ 3expressing HEK293 cells were maintained in DMEM/F-12 (Invitrogen) supplemented with 10% FBS, 2 mM glutamine, 15 mM HEPES, 0.25 mg/ml G418, and 0.1 mg/ml hygromycin.
Immunoprecipitation-Cells (5 ϫ 10 6 ) were labeled with 100 g Immunopure Sulfo-NHS-LC-Biotin (Thermo Scientific, Rochester, NY) in 200 l of PBS for 30 min at 22°C. The cells were solubilized with a lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 1 mM CaCl 2 , 1 mM PMSF, 100 g/ml leupeptin, 10 mM benzamidine) for 30 min at 22°C. The lysates were incubated with 10 g of normal mouse IgG (Sigma) and 50 l of Zysorbin-G (Zymed Laboratories Inc., San Francisco, CA) for 2 h at 4°C. After centrifugation, the supernatant was incubated with anti-␣ IIb mAb sc-51654 (10 g) for 2 h at 4°C. The integrin-mAb complex was captured by incubating with 50 l of protein A-Sepharose (Amersham Biosciences) for 2 h at 4°C. The immunoprecipitated proteins were eluted with SDS-polyacrylamide gel electrophoresis loading buffer and analyzed by Western blotting. The Immobilon-P membranes (Millipore, New Bedford, MA) were incubated with streptavidin conjugated to horseradish peroxidase and developed using enhanced SuperSignal chemiluminescent substrate (Pierce). Fibrin Clot Retraction-Clot retraction assays using isolated platelets were performed as described previously (15). Briefly, the reaction mixture (total volume, 1.0 ml) consisted of 3 ϫ 10 8 platelets in isotonic HEPES buffer (20 mM HEPES, pH 7.3, 137 mM NaCl, 2.7 mM KCl, 1 mM MgCl 2 , 3.3 mM NaH 2 PO 4 containing 35 mg/ml BSA and 1 mg/ml glucose), 0.25 mg/ml fibrinogen, and 1 mM CaCl 2 in glass tubes coated with Sigmacote (Sigma). For clot retraction assays with recombinant Fg ␥407, the volume of the reaction mixture was reduced to 0.25 ml. Clot retraction was initiated by adding of 1 unit/ml thrombin at 22°C. Clot retraction mediated by wild-type HEK293 cells and generated mutant cell lines was performed as described (30). The reaction mixture consisted of 2 ϫ 10 6 cells in 1 ml containing 10 mM tranexamic acid, 0.25 mg/ml fibrinogen, and 2 mM CaCl 2 . Clot retraction was initiated by adding of 1 unit/ml thrombin at 37°C. To block the effect of ␣ v ␤ 3 , cells were first preincubated with mAb LM609 (10 g/ml) for 10 min at 22°C.
Clot retraction triggered by platelets and HEK293 cells was monitored by taking photographs of clots at several time intervals using a digital camera. The images were scanned, and the areas occupied by clots were calculated using NIH ImageJ software. The effect of the ␣ IIb -derived peptides on platelet-mediated clot retraction was evaluated by determining several parameters (lag phase and IC 50 ) obtained from the kinetic curves of retraction as described previously (15). The value of IC 50 is defined as a concentration of the inhibitor that produces 50% of maximal inhibition. The lag phase is defined as the time spanned from the onset of the process until the first visible changes in clot morphology are observed.
Adhesion Assays-The wells of 96-well tissue culture plates (Costar, Cambridge, MA) were coated with the fibrinogen fragment D98 for 3 h at 37°C or overnight at 4°C. The wells were postcoated with 1% BSA inactivated at 75°C for platelet adhesion assays or 1% polyvinyl alcohol for HEK293 cells. Cells were labeled with 10 M calcein AM (Molecular Probes, Inc., Eugene, OR) for 30 min at 37°C. Platelets were washed in isotonic HEPES buffer and resuspended at 1 ϫ 10 8 /ml in the same buffer supplemented with 1% BSA, 1 mM MgCl 2 , and 1 mM CaCl 2 . Calcein-labeled wild-type and ␣ IIb ␤ 3 -expressing HEK293 cells were washed and resuspended in DMEM/F-12 medium at 1 ϫ 10 5 cells/ml. Aliquots (100 l) of cells were added to the wells and incubated at 37°C for 30 min. The nonadherent cells were removed by two washes with PBS, and fluorescence was measured in a fluorescence plate reader (Applied Biosystems, Framingham, MA). The number of adherent cells was determined using the fluorescence of aliquots with a known number of labeled cells.
Platelet Aggregation-Platelet aggregation studies were performed using isolated platelets as described previously (13). 1 ϫ 10 8 /ml platelets were incubated with different concentrations of the ␣ IIb -derived peptides or RGDV peptide for 10 min before the initiation of aggregation. Platelet aggregation with 0.25 mg/ml fibrinogen in the presence of 10 M ADP and 10 M epinephrine was measured in a platelet aggregometer (Chronolog Corp., Haverton, PA) at 37°C with continuous stirring at 1200 rpm. The maximal aggregation, achieved within 5 min after addition of agonists, was determined and expressed as a percentage of aggregation in the absence of peptides. All aggregation assays were conducted within 3 h after venipuncture.

RESULTS
Multiple Binding Sites for the P3 Peptide in ␣ IIb ␤ 3 -To characterize the interaction between ␣ IIb ␤ 3 and the P3 peptide, binding studies using SPR were performed. For these analyses, the isolated ␣ IIb ␤ 3 was coupled to the CM5 sensor surface, and the SPR profiles across a range of P3 concentrations flowed over protein surfaces were examined. Fig. 1A demonstrates the sensorgrams for the binding of P3 in the range of 0 -45 M. The maximal responses achieved in the equilibrium portions of the sensorgrams for each P3 concentration were determined, and binding data were used to construct Scatchard plots and calculate the number of binding sites and the equilibrium dissociation constants (K d ). The binding of P3 to ␣ IIb ␤ 3 was saturable at 45 M (maximal testable concentration) and occurred with K d of 19.4 Ϯ 2 M (Fig. 1B). The stoichiometry of P3 binding to ␣ IIb ␤ 3 obtained from extrapolation of the linear parts of Scatchard plots was found to be 7 Ϯ 0.6:1 (Fig. 1B, inset), indicating that the binding of P3 to ␣ IIb ␤ 3 may occur at several sites.
Screening of the ␣ IIb Propeller-derived Peptide Libraries for P3 Binding-Previous studies demonstrated that, in addition to a well defined set of amino acid residues in the ␣ IIb ␤-propeller and the ␤ 3 I-like domain involved in ligation of RGD and ␥ 404 GAKQAGDV 411 , numerous residues scattered throughout the ␣ IIb ␤-propeller domain provide contact sites for soluble fibrinogen (31)(32)(33). Therefore, we focused on this region of the receptor. To localize the binding sites for P3 in our initial analyses, we screened the cellulose-bound peptide library representing the complete sequence of the ␣ IIb ␤-propeller. The library consisting of 9-mer peptides with a 3-residue offset spanning the entire ␣ IIb ␤-propeller (residues 1-451) ( Fig. 2A) was synthesized, and the membrane with covalently attached peptides was probed with 125 I-labeled P3 peptide. The results of FIGURE 1. The interaction of ␣ IIb ␤ 3 with the P3 peptide analyzed by SPR. A, representative profiles of the SPR responses for P3 peptide concentrations ranging from 0 to 45 M binding to purified ␣ IIb ␤ 3 coupled onto a CM5 sensor chip. RU, response units. B, saturable binding curve and Scatchard plot (inset) of P3 binding to ␣ IIb ␤ 3 . Req is the response at equilibrium. The abscissa in the Scatchard plot is the ratio of the number of P3 molecules bound per molecule of ␣ IIb ␤ 3 . The ratio of bound to free peptide is given on the ordinate.
To ensure that these types of analyses detect integrin-ligand recognition events reported previously (33), we have independently re-examined the binding of the DD fragment to the ␣ IIb ␤-propeller. This fibrin-derived fragment contains both the H12 and P3 sequences, and thus, its binding to ␣ IIb ␤ 3 may potentially involve both fibrinogen-and fibrin-specific sites. Screening membranes with 125 I-DD revealed that, among 16 binders, 10 corresponded to the segments identified by Kamata et al. (33) as critical for binding of soluble fibrinogen to ␣ IIb ␤ 3 (supplemental Fig. 1S). Furthermore, among 40 individual amino acid residues identified as critical in that study (33), 36 were present in the DD binders (supplemental Fig. 1SB). Several residues, including Tyr 143 , Pro 145 , Asp 163 , Leu 183 , and Thr 207 , mutations of which were identified in Glanzmann thrombasthenia (34), 3 were also found in the DD binders. In addition, in agreement with previous data (33) indicating that mutations in the ␣ IIb 298 -304 region (a predicted binding site for fibrinogen (36)) do not affect fibrinogen binding, no interaction of the DD fragment with this segment was detected. Among other DD-binding regions, five segments (␣ IIb 25-33, ␣ IIb 52-60, ␣ IIb 349 -363, ␣ IIb 379 -393, and ␣ IIb 442-451) were not previously found to be important for fibrinogen binding, and two (␣ IIb 400 -408 and ␣ IIb 415-423) were not analyzed (33). Thus, the screening experiments with the DD fragment largely supported conclusions drawn from previous mutational analyses using soluble fibrinogen (33) and, thus, substantiated the validity of this mapping strategy. However, they also revealed the differences between the binding of the DD fragment and P3 to the ␣ IIb -derived peptide libraries; i.e. only three regions, ␣ IIb 109 -117, ␣ IIb 228 -239, and ␣ IIb 256 -267, bound both ligands.
The ␣ IIb ␤-Propeller-derived Peptides Inhibit Platelet-mediated Clot Retraction and Adhesion but Not Platelet Aggregation-To assess whether the P3-binding peptides identified in the screening experiments above can mirror the effect of P3 in functional analyses, the peptides duplicating the sequences of the P3 binders, including ␣ IIb 64 -78, ␣ IIb 94 -108, ␣ IIb 153-162, ␣ IIb 229 -237, ␣ IIb 241-255, ␣ IIb 361-375, and ␣ IIb 421-435, were synthesized and examined for their ability to inhibit clot retraction and platelet adhesion. The peptide duplicating cluster 6 was not synthesized because only 1 residue FIGURE 2. The binding of P3 to the peptide library spanning the sequence of the ␣ IIb ␤-propeller. A, the amino acid sequence of the of ␣ IIb ␤-propeller (residues 1-451). The ␤-strands from the seven blades are marked and underlined. The numbering of residues is shown with a dot, which marks every 10th residue. B, a peptide library consisting of 9-mer peptides derived from the ␣ IIb ␤-propeller (residues 1-451) was screened for P3 binding. The membrane with covalently attached peptides was incubated with 125 I-labeled P3 and subjected to autoradiography. C, clusters of the overlapping sequences selected based on the highest P3 binding activity are shown. The numbers of peptides correspond to the numbering of spots in B. (Trp 262 ) in its most active part (␣ IIb 259 -271, spots 87-89) is exposed on the surface. Fig. 3 shows the effect of ␣ IIb 241 VGEFDGDLNTTEYVV 255 as an example. The increasing concentrations of peptide progressively blocked clot retraction (Fig. 3, A and B), and at 200 M, fibrin clots did not retract after 4 h. The IC 50 value calculated from the progress curves of retraction was 72 Ϯ 6 M ( Fig. 3C and Table 1). All other peptides, except ␣ IIb 361-375, were capable of inhibiting clot retraction albeit with different efficiencies. The IC 50 value and lag phase determined for the concentration of each peptide allowed the comparison of their potency ( Table 1). The ␣ IIb 241-255 was most active followed by ␣ IIb 94 -108, ␣ IIb 64 -78, ␣ IIb 421-435, ␣ IIb 153-162, and ␣ IIb 229 -237. In additional experiments, we examined the effect of selected ␣ IIb -derived peptides used in combinations. As shown in supplemental Fig.  3S, the mixtures of two or four peptides added to platelets at the concentrations to achieve a final concentration equal to that used with each individual peptide did not produce an additive effect in inhibition of clot retraction. This observation is consistent with a model in which the identified ␣ IIb -derived peptides bind the same P3 site in fibrinogen.

Fibrin-specific Binding Sites in the ␣
Previous studies demonstrated that, although mutation of each of the RGD sequences in the A␣ chains of recombinant fibrinogens did not affect clot retraction, deletion of the ␥C sequence 408 -411 resulted in the slight delay of retraction when compared with normal fibrinogen (16). However, after the delay, retraction rates and the final size of clots for both mutant Fg ␥407 and normal fibrinogen were similar, consistent with a mechanism in which ␥C may contribute to an initial phase of clot retraction but not to a subsequent step. We compared the potency of two representative ␣ IIb ␤-propeller-derived peptides, 94 -108 and 241-255, to inhibit retraction of clots formed by mutant ␥407 and normal fibrinogens. In agreement with published data (16), the retraction rates for both fibrinogens after 20 min were similar (Fig. 4, A and B). The inhibitory effect of peptides on Fg ␥407-mediated clot retraction was stronger than that on retraction of clots generated from normal fibrinogen ( Fig. 4 and supplemental Fig. 2S; shown for ␣ IIb 241-255), suggesting that peptides mainly block the ␥C-independent step of clot retraction.
The effect of ␣ IIb -derived peptides on platelet adhesion was tested using the immobilized D98 fragment. This fibrinogen fragment lacks the COOH-terminal ␥404 -411 sequence, and therefore, ␣ IIb ␤ 3 -dependent platelet adhesion is mediated solely by P3 (15). The effect of peptide ␣ IIb 94 -108 is shown as an example (Fig. 5A). The peptides inhibited adhesion of resting platelets in a dose-dependent manner and at 250 M produced ϳ40 -60% inhibition ( Fig. 5B and Table 1). The inhibition was specific as ␣ IIb -derived peptide ␣ IIb 10 -20 had no activity. As anticipated, the inhibitory effect of peptides on platelet adhesion to intact fibrinogen was less potent apparently due to the presence of strong ␥404 -411-binding sites (Table 1). In agreement with the clot retraction data, the equimolar mixtures of peptides did not produce additional inhibition of platelet adhesion to D98 (supplemental Fig. 4S).
To examine whether the ␣ IIb -derived peptides were able to inhibit binding of soluble fibrinogen, we tested their effect on platelet aggregation. Isolated platelets were preincubated with different concentrations of synthetic peptides or RGDV (positive control) for 10 min before the initiation of aggregation. Among selected peptides, ␣ IIb 64 -78 inhibited platelet aggregation by ϳ30% at 200 M (supplemental Fig. 5S). No inhibition Fibrin-specific Binding Sites in the ␣ IIb ␤-Propeller of aggregation by other peptides was detected. As expected, RGDV blocked aggregation in a dose-dependent manner and completely inhibited it at 100 M. These observations lend further support to the idea that the sites responsible for binding of soluble fibrinogen are different from those involved in the interaction of ␣ IIb ␤ 3 with insoluble forms of fibrin(ogen).
Localization of Critical Amino Acid Residues in the P3-binding Clusters-With the above data indicating that the ␣ IIb -derived peptides are able to inhibit platelet-mediated clot retraction, we sought to identify critical residues for P3 binding. Additional peptide libraries in which each residue in the identified clusters was mutated to Ala were synthesized and examined for P3 binding (supplemental Fig. 6S Table 2). The finding that the majority of critical residues in the ␣ IIb ␤-propeller are negatively charged is consistent with the high positive charge of P3. In addition, hydrophobic residues (Leu and Phe) and aromatic residues (Trp and Tyr) were found to contribute to binding. The role of these residues is also highlighted by the finding that not all negatively charged peptides in the ␣ IIb ␤-propeller scan bound P3 (e.g. 43 LGP-SQEETG 51 , 115 VLEKTEEAE 123 , and 295 VTDVNGDGR 303 ).
The analyses of the three-dimensional structure of the ␣ IIb ␤-propeller (Protein Data Bank code 2VDO (6)) indicated that not all residues in the P3 binders are exposed on the surface (supplemental Fig. 6S). The following residues that are exposed on the surface of the ␣ IIb ␤-propeller and when mutated to Ala exhibited significant loss of P3 binding were selected as initial candidates for subsequent analyses: HEK293 Cells Expressing Mutant ␣ IIb ␤ 3 Receptors Support Reduced Cell Adhesion and Clot Retraction-To determine whether the residues identified as critical in the experiments above are involved in the ␣ IIb ␤ 3 function in adhesion and clot retraction, we replaced them with corresponding residues of the I-less ␣ M ␤-propeller of integrin ␣ M ␤ 2 . The rationale for this strategy is based upon the observations that, although both the ␣ IIb and I-less ␣ M ␤-propeller domains have relatively high homology (30% identical residues and 48% conservative substitutions) and both ␣ IIb ␤ 3 and ␣ M ␤ 2 bind fibrinogen, it is the ␣ M I domain inserted between the second and third repeats of the ␣ M propeller that is responsible for the binding of fibrinogen by

Effect of the ␣ IIb ␤-propeller-derived peptides on platelet-mediated fibrin clot retraction and platelet adhesion to fibrinogen or its D98 fragment
The potency of each peptide in adhesion and clot retraction assays was determined as described under "Experimental Procedures."   ␣ M ␤ 2 . The deletion of the ␣ M I domain generates the I-less integrin, which supports ϳ10 -15% adhesion to fibrinogen compared with wild-type ␣ M ␤ 2 (37). Because of the considerable homology between the two propellers, substitutions of individual residues in ␣ IIb propeller with corresponding ␣ M residues are not expected to alter the conformation of the mutant receptor. The sequence alignment of two domains (supplemental Fig. 7S) revealed that many residues identified as candidates for P3 binding, especially those that are negatively charged, are not conserved between the two domains. For example, the 69 LFDLRDE 75 sequence (cluster 1; critical residues are underlined) in the ␣ IIb ␤-propeller is replaced with RLQVPVE in the ␣ M ␤-propeller. Thus, Leu 69 , Asp 70 , and Asp 74 were substituted with homologous residues in ␣ M . Glu 75 , which is identical in both propellers, remained unchanged. In another example, Trp and Asp in ␣ IIb 100 WSD 102 were replaced with Ser and Pro present in the homologous ␣ M SPP sequence. Residues Asp 224 , Glu 229 , and Phe 231 in cluster 4 were excluded from the analyses because Asp 224 and Phe 231 coordinate ␥Lys 406 in soluble fibrinogen (6); therefore, it is unlikely that residues in this segment bind P3. In addition, in view of high homology between the ␣ IIb residues in cluster 7 and those in ␣ M , they were left unchanged. The residues that have been mutated are boxed in supplemental Fig. 7S. Wild-type and mutant ␣ IIb subunits were transfected into the HEK293 cells together with wild-type ␤ 3 subunit, and stable cell lines carrying mutant integrins were established. Heterodimer association of mutants was evaluated by immunoprecipitation of detergent-lysed surface-labeled cells, and mutant cell lines were sorted to select the expressors with similar levels of integrins (supplemental Fig. 8S).

Peptide
The effect of point mutations on the ␣ IIb ␤ 3 function was explored using adhesion and clot retraction assays. Fig. 6A shows adhesion of wild-type ␣ IIb ␤ 3 -expressing HEK293 cells and W100S/D102P mutant as an example to the increasing concentrations of D98. To exclude the potential effect of ␣ v ␤ 3 on cell adhesion to D98 (38), cells were preincubated with anti-␣ v ␤ 3 mAb LM609. Adhesion of wild-type and mutants cells reached a maximal level at 10 g/ml D98, and the effect of mutations was expressed as a percentage of maximal adhesion attained with wild-type expressing cells. As shown in Fig. 6B, adhesion of cells expressing W100S/D102P, E157A/D159S/ W162F, W235L, E243L/D247A/E252S, and D248T/D429M mutants was reduced by ϳ40 -70% of WT cells. Adhesion of cells expressing the L69R/D71Q/D74V triple mutant was not impaired (not shown), and that of cells expressing the quadru-

Fibrin-specific Binding Sites in the ␣ IIb ␤-Propeller
ple L69R/F70L/D71Q/D74V mutant was reduced by ϳ20%. A mutant in which L69R, D71Q, D74V, W100S, D102P, W235L, E243L, E252S, D428T, and D429M (10-residue mutant) were simultaneously mutated supported ϳ20% adhesion (Fig. 6B). Further evidence for the role of selected residues in ␣ IIb ␤ 3 function was obtained in clot retraction experiments. These analyses were performed in the presence of mAb LM609 to block ␣ v ␤ 3 , which is known to support clot retraction of HEK293 cells transfected with the ␤ 3 integrin subunit (30). While cells expressing wild-type ␣ IIb ␤ 3 supported clot retraction, retraction of cells expressing the 10-residue mutant was significantly delayed. Compared with cells bearing wild-type integrins that began to retract clots after ϳ38 min, the lag phase for mutant cells was prolonged to ϳ100 min (Fig. 7). Furthermore, the final clot size retracted by mutant cells after 4 h was ϳ70% compared with 45% retracted by cells expressing wildtype ␣ IIb ␤ 3 . Clot retraction of cells expressing E157A/D159S/ W162F, E243L/D247A/E252S, and D428T/D429M mutants was also delayed (Table 3); however, final clot size was not significantly different from that retracted by cells expressing wildtype receptor. No significant change in clot retraction by the L69R/F70L/D71Q/D74V mutant cell line was detected, suggesting that the loss of one site may be compensated by another site(s). The activity of the W100S/D102P cell line was not tested. These data indicate that mutations of selected negatively charged and aromatic residues in ␣ IIb ␤ 3 impair the interaction of the receptor with fibrin required for the development of contractile force during clot retraction.

DISCUSSION
In this study, we characterized the interaction of integrin ␣ IIb ␤ 3 with the fibrin-specific peptide P3 (␥ 370 -ATWKTRWYSMKK 381 ) and identified residues critical for P3 binding in the ␣ IIb ␤-propeller domain of the receptor. The analyses of the binding data obtained by SPR indicate that P3 can bind to multiple sites in ␣ IIb ␤ 3 . In agreement with this finding, P3 bound to various peptides in the peptide library spanning the sequence of the ␣ IIb ␤-propeller. The peptides duplicating the P3-binding sequences inhibit clot retraction and platelet adhesion but not platelet aggregation. A common feature of the peptides is their enrichment with negatively charged and aromatic residues. Indeed, substitutions of Leu 69 , Phe 70 , Asp 71 , Asp 74 , Trp 100 , Asp 102 , Glu 157 , Asp 159 , Trp 162 , Trp 235 , Glu 243 , Asp 247 , Glu 252 , Asp 428 , and Asp 429 in the ␣ IIb ␤-propeller blocked adhesion and clot retraction mediated by HEK293 cells expressing mutant receptors. These amino acid residues potentially represent the sites through which ␣ IIb ␤ 3 contacts fibrin fibers during clot retraction.
The amino acid residues identified as critical for P3 binding in the ␣ IIb ␤-propeller are largely different from those that coordinate the fibrinogen recognition peptide ␥ 404 -GAKQAGDV 411 and situated at the interface between the ␣ IIb ␤-propeller and ␤ 3 I domains (6). One notable exception is ␣ IIb Asp 224 and ␣ IIb Phe 231 , mutation of which in peptides constituting the substitutional peptide library (supplemental Fig. 6S) modestly reduced P3 binding. However, the fact that synthetic peptide 229 EYFDGYWGY 237 , which contains Phe 231 , does not inhibit platelet aggregation, although it efficiently blocks clot retraction and platelet adhesion, suggests that its activity depends on other residues. Indeed, mutations of Glu 229 and Trp 235 strongly reduced P3 binding and decreased adhesion of HEK293 cells expressing ␣ IIb ␤ 3 carrying the W235L mutation. The P3-binding residues are also distinct from 40 discontinuous residues in the ␣ IIb ␤-propeller identified by Takada and co-workers (32,33) as critical for binding of soluble fibrinogen. Those residues have been mapped to the loops in repeats 2-4 and at the boundary between repeats 4 and 5 of the ␣ IIb ␤-propeller. The crystal structure of ␣ IIb ␤ 3 in complex with the ␥C peptide (6) has subsequently revealed that among these residues not only are those that coordinate ␥C but many residues Fibrin-specific Binding Sites in the ␣ IIb ␤-Propeller that form the ␣ IIb cap subdomain, the region of the ␣ IIb ␤-propeller where epitopes for several function-blocking antibodies were identified (33,39). In the three-dimensional structure of the ␣ IIb ␤-propeller, the P3-binding residues Leu 69 , Phe 70 , Asp 71 , Asp 74 , Trp 100 , and Asp 102 surround Insert 1 of the cap subdomain; Glu 157 , Asp 159 , and Trp 162 are present within Insert 3 of the cap subdomain; and Trp 235 is adjacent to Insert 4 of the cap (Fig. 8). Other residues, including Glu 243 , Glu 252 , Asp 428 , and Asp 429 , are found within or in the vicinity of the Ca 2ϩ -binding sites located in blades W4 and W6 of the ␤-propeller.
Identification of distinct binding sites for the ␥C peptide, whole fibrinogen, and the P3 peptide is consistent with a mechanism where ␣ IIb ␤ 3 exhibits differential recognition specificity for soluble fibrinogen and the insoluble fibrin matrix in platelet aggregation and clot retraction. The binding of soluble fibrinogen to ␣ IIb ␤ 3 during platelet aggregation was described as a two-step process with potential engagement of different amino acid residues at each step. Accordingly, the binding of soluble fibrinogen to agonist-activated platelets results in the formation of platelet aggregates that can dissociate under certain conditions (10,40). The underlying mechanism for this phenomenon appears to be a reversible binding of fibrinogen to platelets (40). That initial binding of fibrinogen to ␣ IIb ␤ 3 is mediated by ␥C 408 AGDV 411 has been documented in numerous studies that showed that recombinant fibrinogen lacking this sequence does not support platelet aggregation (9,14) and was recently confirmed with isolated receptor (41). Reversible platelet aggregation is followed by an irreversible step that was proposed to result from the progressive stabilization of the complex between ␣ IIb ␤ 3 and fibrinogen (10,11). Indeed, biophysical studies revealed the two-step binding mechanism of fibrinogen binding to isolated ␣ IIb ␤ 3 in which fast weak binding is followed by slow strong complex formation (42,43). Although conclusive data are not available and other mechanisms may account for the irreversible step, it is reasonable to assume that stabilization of the complex between ␣ IIb ␤ 3 and fibrinogen involves the amino acid residues in the ␣ IIb ␤-propeller cap domain (33,39). These residues together with those that coordinate the ␥C sequences may constitute the complete integrin-fibrinogen binding interface formed during platelet aggregation.
In contrast to platelet aggregation, the interaction of ␣ IIb ␤ 3 with fibrin is strongly associated with the development of platelet contractile activity. A requirement for clot retraction is the spatially even distribution of platelets within fibrin because preaggregated platelets do not retract clots (44). On the basis of electron microscopic studies of clots retracting under isometric tension, it has been proposed that close apposition of platelet pseudopods to long fibrin fibers is the major mechanism for the transmission of contractile force (45). As pseudopods crawl and pull on bound fibrin, the fibers become stretched and aligned in the direction of tension. The contacts between the platelet surface and fibrin strands in retracting clots were observed across a ϳ15-nm space established by the structures that initially were called "stubs" and presumably represent integrins (46,47). In contrast, the interplatelet bridges in aggregating platelets span the space of ϳ50 nm (48,49). Another distinction between platelet-fibrin and platelet-fibrinogen interactions is their sensitivity to EDTA: i.e. although EDTA-treated and then washed platelets do not aggregate in response to agonists, they support normal clot retraction (50). From the comparison of these char-   (16,20). The proposal that P3 serves as the binding site for ␣ IIb ␤ 3 in fibrin was put forward based upon the ability of mAb 2G5 directed against P3 and synthetic peptides duplicating the P3 sequence to inhibit platelet-mediated clot retraction (15). It has also been shown that natural and recombinant fibrinogens with mutations in the P3 sequence exhibit delayed clot retraction (25). The direct interaction between P3 and ␣ IIb ␤ 3 was confirmed by affinity chromatography using a P3-agarose affinity matrix (15). Further evidence that ␣ IIb ␤ 3 utilizes distinct recognition specificity toward fibrin comes from the fact that P3 is poorly exposed in soluble fibrinogen and becomes available after its conversion to fibrin (12). Because during clot retraction platelet-fibrin clumps stretch and align fibrin in the direction of tensile force, it is tempting to speculate that P3 may become fully exposed under tension. It should be noted that even though the ␥C sequence may be exposed on the surface of fibrin fibers it is unlikely that ␣ IIb ␤ 3 is capable of utilizing it. In fibrin, ␥Lys 406 , one of the amino acid residues that coordinate several residues in ␣ IIb (6), is cross-linked by Factor XIIIa to ␥Gln 399/398 on the neighboring molecule and may not be available for ␣ IIb ␤ 3 binding.
The identification of several P3-binding segments in the ␣ IIb ␤-propeller suggests a model in which a single receptor may form multiple low affinity contacts with a fibrin polymer. The P3-␣ IIb ␤ 3 interactions are largely electrostatic; i.e. positively charged P3 binds negatively charged amino acid residues in the ␣ IIb ␤-propeller, although aromatic residues are also involved. The relatively limited requirements for P3 recognition by the ␣ IIb ␤-propeller domain imply that other sequences in fibrin-(ogen) that contain analogous combinations of amino acid residues or have similar physicochemical properties could function as the ␣ IIb ␤ 3 -binding sites in fibrin. For example, the second homologous domain in the D region of fibrinogen, ␤C, contains a sequence highly homologous to P3 (␤ 438 MNWKGSWYSMRK 449 ). In contrast to platelet aggregation where each ␣ IIb ␤ 3 binds a single fibrinogen molecule, during clot retraction, the receptor makes contacts with already preformed fibrin. In fibrin polymer, the formation of fibers through the lateral association of protofibrils may cluster the P3 and ␤438 -449 sequences in the interacting ␥C and ␤C domains. Although the lateral association of protofibrils that gives rise to mature fibrin has not been defined at the level of atomic resolution, the packing of human and chicken fibrinogens in crystals revealed several interactions that have been proposed to be good candidates for those that occur in fiber formation, including the ␥C-␥C and ␤C-␤C interfaces (51,52). It is not impossible that each ␣ IIb ␤ 3 molecule may establish multiple contacts with P3 and/or the ␤C sequences brought together in fibrin. However, an alternative possibility is that P3 and P3-like sequences may indiscriminately engage one of the negatively charged clusters in ␣ IIb ␤-propeller.
Among seven P3-binding sites, three have been identified in the segments that span the Ca 2ϩ -binding sites in blades W4, W6, and W7. Within these peptides, for example in ␣ IIb 241-255 that spans Ca 2ϩ -binding site 1 in W4 and includes flanking residues, mutations of several negatively charged residues, including Glu 243 , Asp 247 , and Glu 252 , resulted in the loss of P3 binding ( Table 2 and supplemental Fig. 6S). Among these residues, Glu 243 and Asp 247 provide metal-coordinating side-chain oxygen atoms, whereas Glu 252 is outside of the 9-residue Ca 2ϩcoordinating segment (Ref. 53 and references therein). Likewise, although mutations of both Asp 428 and Asp 429 in site 4 reduced P3 binding, Asp 429 is not involved in metal coordination. Mutations of these residues introduced into the whole receptors reduced cell adhesion and clot retraction ( Fig. 6 and Table 3). At first glance, this finding might seem to indicate that the loss of Ca 2ϩ resulted in the alteration of the receptor con-  Bank code 2VDO). The ␣ IIb subunit is shown in gray, and ␤ 3 is shown in tan. Amino acid residues identified as critical for the binding of fibrin-specific peptide P3 in the ␣ IIb ␤-propeller are shown in magenta (selected residues are labeled). Two views (A and B) are rotated relative to each other by 180°about the vertical axis.
Fibrin-specific Binding Sites in the ␣ IIb ␤-Propeller formation, providing support for previous reports that Ca 2ϩbinding sites are essential for ␣ IIb folding and ␣ IIb ␤ 3 heterodimer formation (53). However, receptors carrying triple and double mutations in the Ca 2ϩ -binding sites were assembled normally and expressed on the cell surface at levels comparable with that of WT integrin, suggesting that even though the local conformation may be distorted it does not hamper the biogenesis of integrin. Furthermore, clot retraction is only marginally sensitive to Ca 2ϩ (50). 4 Although the dependence on the ␣ IIb negatively charged residues is compatible with the overall P3 cationic nature, other residues in the vicinity of or within the Ca 2ϩ -binding sites, including Tyr 253 , Tyr 371 , and Gly 423 , may contribute to binding (supplemental Fig. 6S). Thus, although coordination of calcium by oxygen atoms from side chains of Asp and Glu reduces the negative surface electrostatic potential of this region, other negatively charged and aromatic residues in the vicinity may be involved in P3 binding. One puzzling observation is that no binding of either P3 or the DD fragment was detected to peptides duplicating or overlapping Ca 2ϩbinding site 2 in W5. The lack of DD binding to this region is consistent with previous studies (33) showing that swapping the Ca 2ϩ -binding loop W5:1-2 did not affect the binding of soluble fibrinogen; however, the reason for the absence of P3 binding is not clear. Like others, the Ca 2ϩ -binding site in W5 contains negatively charged residues forming the consensus motif. The only difference between the residues that constitute this Ca 2ϩ -binding site is the absence of aromatic residues present in sites 1, 3, and 4 that may impart additional specificity to P3 recognition. Previous studies showed that the ␤-propeller Ca 2ϩ -binding sites are involved in ␣ IIb ␤ 3 -fibrinogen interactions and in binding ␣ 4 ␤ 1 to several ligands (54,55). Clarifying whether the Ca 2ϩ -binding sites also contribute to fibrin recognition and the role of Ca 2ϩ in this process will require further efforts.
It has long been proposed that inhibition of platelet interactions with fibrin may be a necessary and important property of ␣ IIb ␤ 3 antagonists (5). Recent studies using intravital confocal microscopy as well as traditional histological methods have demonstrated the presence of fibrin in early thrombi (1-3), suggesting that ␣ IIb ␤ 3 may interact with fibrin not only in retracting clots that contain large masses of fibrin but also during platelet aggregation. The role of ␣ IIb ␤ 3 -fibrin interactions in platelet aggregation is consistent with the finding of the unique GPIb-thrombin pathway that does not depend on the binding of fibrinogen to platelets but instead requires fibrin (56). The relatively simple complementarity between the cationic P3 site and negatively charged residues in the ␣ IIb ␤-propeller may explain a well known ability of various positively charged compounds, including natural polyamines, to interact with platelets and modulate their responses (57)(58)(59). Likewise, negatively charged compounds may affect the ␣ IIb ␤ 3 -fibrin interactions. Consistent with this proposal, polyphosphates released from activated platelets (35) inhibit clot retraction. 5 Further studies may help to define the reagents that specifically target plateletfibrin bonds in thrombus formation.