Effects of Limiting Extension at the αIIb Genu on Ligand Binding to Integrin αIIbβ3*

Structural data of integrin αIIbβ3 have been interpreted as supporting a model in which: 1) the receptor exists primarily in a “bent,” low affinity conformation on unactivated platelets and 2) activation induces an extended, high affinity conformation prior to, or following, ligand binding. Previous studies found that “clasping” the αIIb head domain to the β3 tail decreased fibrinogen binding. To study the role of αIIb extension about the genu, we introduced a disulfide “clamp” between the αIIb thigh and calf-1 domains. Clamped αIIbβ3 had markedly reduced ability to bind the large soluble ligands fibrinogen and PAC-1 when activated with monoclonal antibody (mAb) PT25-2 but not when activated by Mn2+ or by coexpressing the clamped αIIb with a β3 subunit containing the activating mutation N339S. The clamp had little effect on the binding of the snake venom kistrin (Mr 7,500) or αIIbβ3-mediated adhesion to immobilized fibrinogen, but it did diminish the enhanced binding of mAb AP5 in the presence of kistrin. Collectively, our studies support a role for αIIb extension about the genu in the binding of ligands of 340,000 and 900,000 Mr with mAb-induced activation but indicate that it is not an absolute requirement. Our data are consistent with αIIb extension resulting in increased access to the ligand-binding site and/or facilitating the conformational change(s) in β3 that affect the intrinsic affinity of the binding pocket for ligand.

The platelet ␣IIb␤3 receptor plays an important role in both hemostasis and thrombosis (1). Ligand binding to ␣IIb␤3 is controlled by an activation process that affects the conformation of the receptor and ligand binding, in turn, can also affect the conformation of the receptor (2). Several different conformations of ␣IIb␤3 have been identified based on inferences from biochemical analyses (3), studies employing monoclonal antibodies (4 -7) and electron microscopy (8 -10), comparison of the crystal structures of the liganded ␣IIb␤3 headpiece (11) and the unliganded complete ectodomain (12), and analysis of the unliganded and liganded ectodomain of the related ␣V␤3 receptor (13,14). Receptor extension about the regions encompassing the thigh, genu, and calf-1 domains of ␣IIb and the plexin-semaphorin-integrin (PSI), 2 integrin epidermal growth factor-1 (IEGF-1), and IEGF-2 domains of ␤3 or comparable regions of other integrin receptors has been proposed to play an important role in receptor activation (12,(15)(16)(17)(18), but there is uncertainty about whether this conformational change occurs prior to or after ligand binding (19 -21). Thus, "cross-clasping" the ␣IIb headpiece ␤-propeller domain to the ␤3 IEGF-4 domain in the tail region via a newly engineered disulfide bond prevented the binding of fibrinogen induced by activating mAb in concert with the activating divalent cation Mn 2ϩ , and a similar effect was observed with cross-clasped ␣V␤3 (16). In both cases the loss of ligand binding could be rescued by reducing the cross-clasped receptors with dithiothreitol (DTT). In contrast, the binding of a fragment of fibronectin (M r ϭ ϳ50 kDa) to the ectodomain of ␣V␤3 activated by Mn 2ϩ was not associated with receptor extension as judged by electron microscopy (22).
To understand better the role of ␣IIb extension around the ␣IIb genu region in ligand binding, we selectively clamped the ␣IIb thigh domain to the ␣IIb calf-1 domain with a newly engineered disulfide bond created by mutating ␣IIb Arg 597 and Tyr 645 to cysteine residues. We directly confirmed the formation of the disulfide bond by mass spectrometry and tested the interaction of both large and small ligands to clamped ␣IIb␤3, as well as both soluble and immobilized ligands. We also tested the effects of multiple ␣IIb␤3 activation mechanisms, including Mn 2ϩ and coexpression of normal and mutant ␣IIb with a mutant ␤3 subunit that when coexpressed with normal ␣IIb results in spontaneous fibrinogen binding (N339S) (23). Our data support an important role for receptor extension around the genu for the binding of large soluble ligands induced by some, but not all activation mechanisms. In contrast, restricting ␣IIb extension about the genu had only a modest effect on ␣IIb␤3-mediated adhesion to immobilized fibrinogen and no discernable effect on the binding of small soluble ligands. These data provide new insight into the contribution of ␣IIb extension around the genu to high affinity ligand binding.
Molecular Modeling, Target Residue Selection, and Sitedirected Mutagenesis-To "clamp" ␣IIb in a bent position, we engineered a new disulfide bond between the ␣IIb thigh domain and the ␣IIb calf-1 domain. An energy optimized model of the complete extracellular domain of ␣IIb␤3 was built to identify candidate regions of the molecule that might selectively restrict ␣IIb extension around the genu. As previously described (32), the conformation of the ␣IIb chain of the bent, inactive, unliganded ␣IIb␤3 was generated with MODELLER 8v2 (33) using the coordinates of the ␤ propeller from the crystal structure of ␣IIb␤3 (residues 1-453; Protein Data Bank entry 1TY6 (11)) and the coordinates of ␣V for the remainder of the sequence (Protein Data Bank entry 1U8C (34)) Most of the coordinates of the ␤3 chain of the bent, inactive, unliganded ␣IIb␤3 conformation were extracted from the ␣V␤3 complex (Protein Data Bank entry 1U8C), whereas the missing domains, IEGF-1 (435-475) and IEGF-2 (486 -522), were constructed with MODELLER 8v2 using the crystal structures of the ␤2 IEGF-1 (Protein Data Bank entry 1YUK (35)) and the ␣V␤3 IEGF-3 (Protein Data Bank entry 1U8C) domains, respectively, as templates. A model of the extended conformation of ␣IIb␤3 based on changes around the ␣IIb genu and ␤3 IEGF-1 and 2 domains was generated from our energy-minimized model of the complete extracellular unliganded bent conformation of ␣IIb␤3 by a 26°rotation of the ␣IIb 601-960 and ␤3 480 -690 regions around the C␣-N bond of ␣IIb residue 600, followed by a 22°rotation of the ␣IIb 599 -960 and ␤3 480 -690 regions around the C␣-C␣ axis of ␣IIb residues 607-608.
Our energy-minimized model of the bent IIb subunit and our model of the extended ␣IIb thigh-genu-calf-1 regions were used to analyze interdomain contacts between residues of the thigh and calf-1 domains in the bent conformation (i.e. distances between C␤ atoms within 10 Å) that would break upon extension. This analysis led to the identification of several candidate interactions, among them the thigh residue Arg 597 and the calf-1 residue Tyr 645 , whose C␤ atoms are 5.7 Å apart in the bent conformation and 13.6 Å in the extended conformation. After these studies were completed, Zhu et al. (12) reported the crystal structure of a complete ectodomain of ␣IIb␤3. In that structure the C␤ atoms of ␣IIb Arg 597 and Tyr 645 are 7.1 Å apart. To clamp ␣IIb about the genu we mutated both of these residues to cysteines with the goal of creating a new disulfide bond (Fig. 1). We separately generated the ␤3 mutation N339S, which has been shown to constitutively activate ␣IIb␤3 and promote fibrinogen binding (23). pEF1/V5-His/␣IIb and pCDNA3.1/Myc-His/␤3 were generous gifts of Drs. Junichi Takagi and Timothy Springer (Harvard Medical School). The ␣IIb double mutant R597C/Y645C and the ␤3 N339S mutant were generated using a QuikChange XL site-directed mutagenesis kit (Stratagene, La Jolla, CA). The mutant cDNA was transformed into XL-10 Gold Ultracompetent Escherichia coli bacteria. cDNA was purified (Maxi kit; Qiagen) and sequenced to confirm mutagenesis.
Stable Cell Line Generation-HEK293 cells were transfected with either normal or mutant cDNA using a cationic lipid transfection reagent according to the manufacturer's instructions (Cellipon 293; America Pharma Source, Gaithersburg, MD). The cells were selected in G418 for 2 weeks. To obtain populations of cells expressing high levels of ␣IIb␤3, the cells were sorted based on their binding of Alexa488-conjugated mAb 10E5 (FACSCalibur; BD Biosciences). The cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 1% penicillin/streptomycin, and 1% nonessential amino acids. Expression of ␣IIb␤3 was assessed by the binding Alexa488 -10E5 mAb on each day of experimentation.
Assessment of Cys 597 -Cyr 645 Disulfide Bond Formation in Clamped ␣IIb␤3 by Mass Spectrometry-Clamped and normal recombinant ␣IIb␤3 were purified from HEK293 cells, and platelet ␣IIb␤3 was purified from washed platelets by solubilizing in 1% Triton X-100 and then immunoprecipitating the lysate with the anti-␣IIb␤3 mAb 10E5 coupled to tosylated magnetic beads (Dynabeads M-280p; Invitrogen). The beads were washed, and the ␣IIb␤3 was eluted with diethylamine at pH 11.0 for 30 min at 22°C, and then the pH of the eluate was immediately reduced to pH 8.0 with HCl. Purified ␣IIb␤3 was prepared for mass spectrometry by overnight digestion with trypsin (2 g/ml; Promega sequencing grade modified trypsin), followed by two overnight digestions with endoproteinase Asp-N (2 g/ml; Roche Applied Science). In some experiments, purified ␣IIb␤3 was treated with iodoacetic acid, DTT, or a combination of DTT and [ 13 C]iodoacetic acid before mass spectrometry. The samples were analyzed by liquid chromatography-tandem mass spectrometry using a nano-reversed phase Immunoprecipitation and Immunoblotting-Studies were performed to assess both the processing of the normal and mutant receptors during biogenesis and the biochemical composition of receptors on the surface of the cells. For the former, the cells were solubilized in 1% Triton X-100, and lysates were treated with SDS. Approximately 550 g of protein was added to each well, and the samples were electrophoresed in 7.0% polyacrylamide gels (SDS-PAGE) under nonreducing or reducing (10% ␤-mercaptoethanol) conditions. After the proteins were transferred to polyvinylidene fluoride membranes, the membranes were incubated with the anti-␣IIb mAb PMI-1. The membranes were washed and incubated with horseradish peroxidase-conjugated goat anti-mouse antibody (Jackson Immunoresearch, West Grove, PA) and visualized by chemiluminescence.
For analysis of surface proteins, the cells (1.5 ϫ 10 6 ) were treated with biotin (1 mg/ml; EZ-Link Sulfo-NHS-LC-Biotin; Thermo Scientific, Rockford, IL) 30 min on ice, and then glycine (5 mM) was added for 10 min to quench the reaction. The cells were then washed and lysed with 1% Triton X-100; then ␣IIb␤3 was immunoprecipitated using mAb 10E5 and protein G-Sepharose beads (GE Healthcare). The samples were treated with SDS, and ϳ550 g of protein was added to each well. The samples were then electrophoresed in 7.0% polyacrylamide gels under nonreducing or reducing (10% ␤-mercaptoethanol) conditions. After the proteins were transferred to polyvinylidene fluoride membranes, the membranes were incubated with a streptavidin-horseradish peroxidase conjugate (Amersham Biosciences) and visualized by chemiluminescence.
To reverse the effect of the disulfide clamp, in some experiments the cells were incubated with the reducing agent DTT for 5 min at 37°C before adding the activator and/or ligand. As previously shown by others, this concentration of DTT can partially activate ␣IIb␤3 (37).
Adhesion to Fibrinogen-Polystyrene 96-well microtiter plates (Nunc, Rochester, NY) were coated with fibrinogen (50 or 10 g/ml; American Diagnostica, Stamford, CT) for 1 h at 22°C and then washed and blocked with HBMT for 1 h. The cells expressing either normal or mutant ␣IIb␤3 were washed and resuspended in HBMT containing 2 mM CaCl 2 and 1 mM MgCl 2 at 1 ϫ 10 6 /ml. Some samples were treated with DTT (5 mM) for 5 min at 37°C and then washed. Negative controls consisted of samples of cells treated with the ␣IIb␤3 antagonist mAb 7E3 (40 g/ml). The cells (50 l/well) were added to wells and allowed to adhere for 1 h at 37°C. The wells were then washed with HBMT, and adhesion was quantified using an endogenous acid phosphatase activity assay (38). To assess any contribution to adhesion mediated by ␣V␤3, initial experiments were conducted in the presence of the ␣V␤3-blocking mAb LM609 (15 g/ml). Because LM609 did not reduce the adhesion, we concluded that ␣V␤3 does not contribute significantly to adhesion under the conditions tested. Receptor expression was assessed prior to adhesion experiments on each of the 5 days of experimentation to ensure comparable surface levels of ␣IIb␤3 (normal ␣IIb␤3 GMFI ϭ 138 Ϯ 53; clamped ␣IIb␤3 ϭ 148 Ϯ 56).
mAb AP5 Binding-The cells were suspended in HBMT containing 2 mM CaCl 2 and 1 mM MgCl 2 at a cell count of 5 ϫ 10 6 /ml. Some cells were incubated with DTT (5 mM) at 37°C for 5 min and were then washed and resuspended to the same count. The cells were incubated with Alexa488-AP5 (5 g/ml) in the presence or absence of kistrin (200 nM) for 1 h at 37°C. The samples were diluted 1:10 with HBMT and analyzed by flow cytometry. Nonspecific binding was determined in the presence of 25-fold excess unlabeled mAb AP5, and the data were normalized for receptor expression on each day of experimentation.

The Clamped Mutant ␣IIb Contains a Cys 597 -Cys 645 Disulfide Bond and Migrates More Rapidly than Normal ␣IIb in SDS-PAGE under Nonreducing, but Not Reducing Conditions-
Sequence analysis of the expected peptide cleavage products of the normal ␣IIb␤3 and clamped ␣IIb␤3 (assuming successful creation of the Cys 597 -Cys 645 bond) led to the prediction of a unique disulfide-linked peptide (␣IIb Asp 589 -Leu 600 disulfidelinked to ␣IIb Asp 636 -Arg 661 ) in the clamped ␣IIb but not in either the normal recombinant ␣IIb or platelet ␣IIb (supplemental Fig. S2). This disulfide-linked peptide was observed in the main spectrum of clamped ␣IIb␤3 at m/z ϭ 1,031.21 corresponding to the MH 4 4ϩ . The identity of the peptide was confirmed by tandem mass spectrometry, in which fragments of both halves of the peptide were observed (data not shown). The peptide ion was not observed in mass spectra for either the recombinant normal ␣IIb␤3 or the ␣IIb␤3 purified from platelets.

␣IIb Extension Variably Impacts Ligand Binding to ␣IIb␤3
To assess the extent of disulfide bond formation by Cys 597 and Cys 645 in clamped ␣IIb, the purified clamped ␣IIb␤3 was treated with iodoacetic acid to label free cysteines before and after reduction with DTT (45 mM). Although a Asp 589 -Leu 600 peptide containing a free sulfhydryl at Cys 597 could be identified in the nonreduced purified ␣IIb, the intensity of this peptide increased by at least 16-fold with reduction. A Asp 636 -Arg 661 peptide containing a free sulfhydryl at Cys 645 could not be detected in the nonreduced sample but was readily detectable after reduction. To further assess the extent of disulfide bond formation, a single sample of purified clamped ␣IIb␤3 was split in half and separately labeled with either iodoacetic acid or the combination of DTT and [ 13 C]iodoacetic acid. The samples were then combined and analyzed by liquid chromatography-tandem mass spectrometry. The ratio of isotope labeling of the Asp 589 -Leu 600 peptide (1:20) indicated that fewer than 5% of the Cys 597 residues contained a free thiol, and the ratio of isotope labeling of the Asp 636 -Arg 661 peptide (1:100) indicated that fewer than 1% of Cys 645 residues contained a free thiol. To assess whether either the Cys 597 or Cys 645 created disulfide bonds with the nearest cysteine residues (Cys 602 and Cys 608 ), which are located adjacent to the genu, we analyzed the ratio of the disulfide-bonded peptide ions produced from this region by enzyme digestion and normalized the value by comparison with another internal standard peptide. The normalized intensities of the peptide ions were virtually identical in clamped ␣IIb, native recombinant ␣IIb, and platelet ␣IIb, indicating that neither Cys 597 nor Cys 645 cross-links in substantial amounts to either of these residues. We thus conclude that the Cys 597 -Cys 645 bond is present in clamped ␣IIb and that ϳ95% or more of the Cys 597 and Cys 645 residues are engaged in disulfide bonds.
Further support for nearly complete disulfide bond formation by Cys 597 and Cys 645 in clamped ␣IIb came from SDS-PAGE studies (Fig. 2). In studies involving either both internal and surface receptors ( Fig. 2A) or just surface receptors (Fig.   2B), normal ␣IIb migrated at a molecular mass of 140 kDa under nonreducing conditions, whereas the clamped ␣IIb molecules (in complex with either normal ␤3 or ␤3 N339S) migrated more rapidly (molecular mass, 130 kDa). After reducing the disulfide bonds, both the normal and mutant ␣IIb heavy chains migrated identically (molecular mass, 120 kDa). These data are consistent with the newly introduced disulfide bond producing a more compact structure, leading to more rapid migration of the clamped nonreduced ␣IIb. In the study involving both internal and external receptors ( Fig. 2A), the ratio of uncleaved pro-␣IIb (upper bands in right panel) to mature, cleaved ␣IIb (lower bands in right panel) was slightly greater for the clamped mutant, suggesting somewhat less efficient processing of the clamped receptor. Similar results were obtained when the clamped ␣IIb was expressed with ␤3 N339S, although the percentage of uncleaved ␣IIb was greater. In contrast, nearly all of the clamped ␣IIb subunits on the surface of cells expressing either clamped ␣IIb with ␤3 or clamped ␣IIb with ␤3 N339S were cleaved (Fig. 2B, right panel).
Clamped ␣IIb␤3 Has Reduced Ability to Bind Soluble Fibrinogen (M r ϭ 340,000) and the Soluble Ligand-mimetic mAb PAC-1 (M r ϭ 900,000) when Activated with mAb PT25-2, and DTT Reduction Partially Rescues the Defect-The ␣IIb-specific activating mAb PT25-2 bound equally well to normal and clamped ␣IIb␤3 (supplemental Fig. S1). Cells expressing normal ␣IIb␤3 bound only very small amounts of fibrinogen spontaneously but did bind fibrinogen in the presence of the PT25-2 (Figs. 3A and 4A). In sharp contrast, when incubated with PT25-2, cells expressing clamped ␣IIb in complex with normal ␤3 bound less than 9% of the amount of fibrinogen bound to the cells expressing normal ␣IIb␤3 (NNFI ϭ 11 Ϯ 5 for normal ␣IIb␤3 and 1 Ϯ 1 for clamped ␣IIb␤3; n ϭ 4, p Ͻ 0.001). Treating cells expressing normal ␣IIb␤3 with 5 mM DTT, even in the absence of PT25-2, partially enhanced fibrinogen binding. Similarly, treating cells expressing clamped ␣IIb␤3 with DTT led to an increase in fibrinogen binding in the absence of PT25-2. Clamped ␣IIb migrates more rapidly under nonreducing conditions in SDS-PAGE than normal ␣IIb; reducing disulfide bonds with ␤-mercaptoethanol eliminates the difference in migration. A, total cell analysis. Lysates of cells expressing normal and clamped ␣IIb were heated to 100°C in the presence or absence of 10% ␤-mercaptoethanol (␤-ME) before loading on a 7% polyacrylamide gel. After protein transfer to polyvinylidene fluoride membranes, ␣IIb was detected by the anti-␣IIb mAb PMI-1. B, surface receptor analysis. The cells were surface-labeled with biotin and then lysed and immunoprecipitated with a mAb to ␣IIb␤3. The proteins were separated by SDS-PAGE, and then biotin-labeled proteins were identified with avidin-horseradish peroxidase. C, nonreduced SDS-PAGE analysis of normal and clamped ␣IIb before and after treating cells with 5 mM DTT. Note that the increased migration of clamped ␣IIb is largely, but not completely, reversed by 5 mM DTT.
When PT25-2 was added to cells expressing normal ␣IIb␤3 after treatment with DTT, fibrinogen binding increased above the value in the absence of PT25-2, reaching a level similar to that with normal ␣IIb␤3 in the presence of PT25-2. Treating cells expressing clamped ␣IIb␤3 with the combination of DTT and PT25-2 further increased fibrinogen binding, but not to the level of normal ␣IIb␤3. SDS-PAGE analysis provided evidence that the DTT treatment ␣IIb Extension Variably Impacts Ligand Binding to ␣IIb␤3 substantially reduced the engineered disulfide bond with the majority of the clamped ␣IIb showing normalization of the migration pattern (Fig. 2C).
Data with the IgM ligand-mimetic mAb PAC-1 paralleled those with soluble fibrinogen (Figs. 3B and 4B). Thus, in the presence of mAb PT25-2, cells expressing normal ␣IIb␤3 had an NNFI of 17 Ϯ 7, whereas cells expressing clamped ␣IIb␤3 had an NNFI of only 2 Ϯ 1 (n ϭ 3; p Ͻ 0.001). Binding of PAC-1 to clamped ␣IIb␤3 could be partially rescued by DTT treatment. Thus, after DTT treatment, PAC-1 binding to clamped ␣IIb␤3 receptors was ϳ54% of the binding to normal ␣IIb␤3 in the presence of DTT. The further addition of PT25-2 increased PAC-1 binding to both normal and clamped ␣IIb␤3, with the latter reaching a level similar to that of normal ␣IIb␤3 in the presence of PT25-2 alone.
Fibrinogen Binding to Clamped ␣IIb␤3 Is Not Impaired when Clamped ␣IIb Is Expressed with a ␤3 Integrin Subunit Containing the Activating Mutation N339S or when Clamped ␣IIb␤3 Is Activated by Mn 2ϩ -Fibrinogen and PAC-1 bound spontaneously to cells expressing normal ␣IIb in complex with ␤3 N339S (Fig. 4, A and B). The specificity of the binding for ␣IIb␤3 was established by inhibition by eptifibatide and further confirmed in the case of fibrinogen by the lack of effect of mAb LM609 (anti-␣V␤3) (data not shown). Expressing ␤3 N339S with clamped ␣IIb produced similar levels of spontaneous binding of fibrinogen and PAC-1. The addition of PT25-2 had little impact on ligand binding to cells expressing either normal ␣IIb␤3-N339S or clamped ␣IIb␤3-N339S ( Fig. 4B and data not shown), suggesting that these receptors were already maximally activated. Normal and clamped ␣IIb␤3 bound fibrinogen to a similar extent in the presence of Mn 2ϩ , and the specificity of this binding was assessed by its inhibition by eptifibatide (Fig. 4C). were untreated or treated with DTT (5 mM) and incubated with either Alexa488-fibrinogen (200 g/ml) or FITC-PAC-1 (5 g/ml) in the presence or absence of activating antibody PT25-2. Nonspecific binding was determined in the presence of the ␣IIb␤3 inhibitor eptifibatide. Binding was assessed via flow cytometry and expressed as NNFI, in which the geometric mean fluorescence intensity after subtracting nonspecific binding is divided by the relative surface receptor expression as judged by the binding of mAb 10E5. *, p Ͻ 0.001 versus normal ␤3, n ϭ 3; †, p Ͻ 0.001 versus normal ␤3, n ϭ 4. C, cells expressing either clamped or normal ␣IIb␤3 were incubated with fluorescent fibrinogen in the presence of either Ca 2ϩ /Mg 2ϩ or Mn 2ϩ . Nonspecific binding was determined in the presence of the ␣IIb␤3 inhibitor mAb 7E3. Binding was assessed via flow cytometry and expressed as NNFI (calculated as described above) (n ϭ 4).

JOURNAL OF BIOLOGICAL CHEMISTRY 17609
The RGD-containing Snake Venom Disintegrin Kistrin (M r ϭ 7,500) Binds Similarly to Normal and Clamped ␣IIb␤3-Fluorescently labeled kistrin bound to cells expressing clamped ␣IIb in complex with normal ␤3 somewhat less well than cells expressing normal ␣IIb␤3, but the difference was not statistically significant (Fig. 5). DTT treatment did not significantly affect kistrin binding. Expressing normal ␣IIb with ␤3 N339S did not affect kistrin binding, whereas there was an increase in kistrin binding when clamped ␣IIb was expressed with ␤3 N339S.
Clamped and Normal ␣IIb␤3 Adhere to Immobilized Fibrinogen-The number of cells expressing clamped ␣IIb␤3 receptors that adhered to fibrinogen immobilized at high density (coating concentration, 50 g/ml) was 95 Ϯ 30% of the value of cells expressing normal ␣IIb␤3 at similar levels of surface expression (mAb 10E5 GMFI 138 Ϯ 53 and 148 Ϯ 56 for normal and clamped cells, respectively; n ϭ 5; Fig. 6). Adhesion of cells expressing clamped ␣IIb␤3 to fibrinogen immobilized at lower density (coating concentration, 10 g/ml) was somewhat less than adhesion of cells expressing normal ␣IIb␤3, but the difference was not statistically significant (clamped cells 81 Ϯ 38% relative to normal; p ϭ 0.3). Treatment with DTT did not affect adhesion of normal cells to high or low density fibrinogen (92 Ϯ 11 and 93 Ϯ 11% relative to untreated normal cells, respectively). Treating cells expressing the clamped receptor with DTT did increase adhesion to both high and low density fibrinogen in five of five trials, by 41 and 64%, respectively, although there was variability in the degree of increase observed in each experiment (p ϭ 0.06 for high density and p ϭ 0.08 for low density). Adding mAb 7E3 blocked adhesion mediated by both normal and clamped ␣IIb␤3.
AP5 Binds Less Well to Clamped ␣IIb␤3 than Normal ␣IIb␤3 in the Presence of Kistrin, and Treatment with DTT Rescues the Clamped Mutant-AP5 binding to cells expressing normal ␣IIb␤3 and clamped ␣IIb␤3 was similarly low, reflecting ϳ10% of maximal binding. Incubating cells expressing normal ␣IIb␤3 with kistrin increased AP5 binding ϳ6-fold (Fig. 7) but pro-duced only a 2-3-fold increase in binding to cells expressing the clamped ␣IIb␤3 (p Ͻ 0.001; n ϭ 10). Treatment with DTT in the absence of kistrin increased the binding of AP5 to both normal and clamped ␣IIb␤3, with both reaching similar levels. Adding kistrin to DTT-treated cells of both types further increased AP5 binding. AP5 binding to ␣IIb␤3 N339S was higher than AP5 binding to normal ␣IIb␤3 (p Ͻ 0.001; n ϭ 5) and clamped ␣IIb␤3 (p Ͻ 0.001; n ϭ 5); it showed a trend to higher binding when compared with clamped ␣IIb␤3 N339S (p ϭ 0.06; n ϭ 5). Adding kistrin increased the binding of AP5 to ␣IIb␤3 N339S, reaching levels that were higher than those for normal ␣IIb␤3 (p Ͻ 0.001), clamped ␣IIb␤3 (p Ͻ 0.001), or clamped ␣IIb␤3 N339S (p Ͻ 0.001) in the presence of kistrin. After DTT treatment, the binding of AP5 to ␣IIb␤3 N339S was similar to its binding in the presence of kistrin, whereas DTT treatment of clamped ␣IIb␤3 significantly increased AP5 binding relative to its binding to the receptor in the presence of kistrin (p ϭ 0.004). Adding kistrin in addition to DTT further increased binding to all of the receptors but to only a modest extent.

DISCUSSION
Our studies were designed to elucidate the role of ␣IIb extension about the genu in ␣IIb␤3 ligand binding. Previous studies clasping the ␣IIb ␤-propeller domain to the ␤3 IEGF-4 domain suggested a role for extension of ␣IIb, ␤3, or both in the binding of fibrinogen to ␣IIb␤3 when cells were activated by a combination of Mn 2ϩ and an activating mAb (16). We have extended these studies by analyzing the effect of an ␣IIb clamp confined to the region immediately adjacent to the genu and assessing the binding of both high and low M r soluble ligands, as well as the ability of the mutant receptor to support adhesion to FIGURE 6. Cells expressing normal and clamped ␣IIb␤3 bind to immobilized fibrinogen; treatment with DTT increases adhesion of cells expressing clamped ␣IIb␤3. When added to wells coated with fibrinogen at either high density (50 g/ml) or low density (10 g/ml), cells expressing either normal or clamped ␣IIb␤3 adhered to the surface. Treatment with DTT (5 mM) did not affect adhesion of cells expressing normal ␣IIb␤3 but did increase adhesion of clamped ␣IIb␤3-expressing cells (n ϭ 5). FIGURE 7. Kistrin induces binding of AP5 to clamped ␣IIb␤3 to a lesser extent than to normal ␣IIb␤3; the defect is rescued by incubation with DTT. The fluorescently labeled ligand-induced binding site mAb AP5 was incubated with either untreated cells or cells pretreated with kistrin (200 nM) and/or DTT (5 mM) for 1 h at 37°C. Antibody binding was assessed via flow cytometry and expressed as NNFI as indicated in the legend to Fig. 4, but using excess unlabeled AP5 to assess nonspecific binding. The values were further normalized for the ϳ40% difference in the intensity of fluorescent labeling (F/P ratio) of the two AP5 preparations used in this study. The data for normal ␣IIb␤3 and clamped ␣IIb␤3 are from 10 experiments, and the data for ␣IIb␤3 N339S and clamped ␣IIb␤3 are from five experiments.
immobilized fibrinogen and bind a ligand-induced binding site antibody whose epitope is on the ␤3 PSI domain.
The ␣IIb R597C-Y645C clamp was designed based on the predicted proximity of these residues in the ␣IIb molecular model we constructed using the crystal structure of the bent ␣V␤3 receptor ectodomains as a template. After these studies were completed, the close proximity of the residues was directly confirmed in the crystal structure of the ectodomain of ␣IIb␤3 (12). Support for the successful creation of the new disulfide bond came from analyzing both mass spectrometric data and the electrophorectic mobility of ␣IIb in SDS-PAGE under both nonreducing and reducing conditions.
We found that clamped ␣IIb␤3 has a dramatically reduced ability to bind the soluble ligands fibrinogen (M r ϭ 340,000) and IgM mAb PAC-1 (M r ϭ 900,000) induced by the activating mAb PT25-2, which binds to ␣IIb. The defect could be largely rescued by adding the reducing agent DTT. These studies have the inherit limitation that they depend on the ability of the mAb to bind equally to the normal and clamped receptor. We excluded the possibility that the reduction in ligand binding was due to the clamp altering the binding of PT25-2 by directly measuring PT25-2 binding to cells expressing normal or clamped ␣IIb␤3 and observing equivalent binding. DTT was able to partially activate the receptor itself and partially rescued the ability of the clamped receptor to increase ligand binding in response to PT25-2. SDS-PAGE analysis indicated that DTT treatment reduced the engineered disulfide bond in most, but not all, of the clamped ␣IIb␤3 molecules, providing an explanation for the partial rescue of ligand binding. We cannot, however, exclude the possibility that substituting cysteines for arginine at position 597 and/or tyrosine at position 645 affects ligand binding.
In sharp contrast to the findings with fibrinogen and PAC-1, the binding of the lower M r snake venom kistrin, which contains an Arg-Gly-Asp (RGD) sequence and does not require exogenous receptor activation, was unaffected by the clamp, indicating receptor extension about the genu is not required for its binding. The smaller M r of kistrin compared with fibrinogen and PAC-1 may account for the difference in its ability to bind to ␣IIb␤3 because it is established that short peptides containing the RGD sequence can bind to unactivated platelets (39).
Because activating ␣IIb␤3 with a mAb may not simulate activation by physiologic agonists (10), especially because it introduces a large molecule that may cause steric hindrance, we also activated the receptor by coexpression with the activating ␤3 subunit mutant N339S and by incubation with Mn 2ϩ . Cells expressing normal ␣IIb with the ␤3 mutant demonstrated high level spontaneous binding of both fibrinogen and PAC-1, and adding an activating mAb had little or no effect. Fibrinogen and PAC-1 both also showed high level binding to cells expressing the clamped ␣IIb in combination with the ␤3 activating mutant, and adding an activating mAb also did not affect the results. Additional structural studies are required to assess the mechanisms by which N339S activates the receptor, but it is strategically located in the region of ␤3 encompassing the ␤6-␣7 loop and ␣7 helix; the latter have been implicated in controlling ligand affinity because they occupy different locations in the unliganded and liganded ␤3 crystal structures (11)(12)(13)(14). Because size-dependent access appears to play a role in ␣IIb␤3 ligand binding (40), it is possible that the ␤3 N339S mutation alters the conformation of the receptor so as to allow greater access to the ligand-binding pocket; alternatively or additionally, it is possible that ␤3 N339S directly affects the affinity for the ligands either by reorienting Asp 251 (12) or by facilitating the movement of the ␣1-␤1 loop toward the MIDAS so as to bring two of the loop's backbone N atoms close enough to interact with the ligand carboxyl oxygen that does not coordinate the MIDAS (12). Our studies of AP5 binding to the normal and mutant ␣IIb␤3 receptors provided some data that bear on these issues. Thus, in the absence of ligand, AP5 bound equally well to normal ␣IIb␤3 and clamped ␣IIb␤3; ␣IIb␤3 N339S, however, bound more AP5 than the other receptors, suggesting that the ␤3 N339S mutation induces conformational changes in the ␤3 ␤A (I-like) domain that are propagated to the PSI domain.
Mn 2ϩ activation was also able to overcome the defect in binding soluble fibrinogen induced by clamping ␣IIb. The mechanism by which Mn 2ϩ activates integrin receptors is unknown, but studies swapping different regions of ␣IIb and ␣V implicated the calf-2 region as controlling the process (41). Ye et al. (10) reported that purified ␣IIb␤3 reconstituted into liposomes did not undergo a major change in height after activation by Mn 2ϩ despite increased adhesion of the liposomes to immobilized fibrinogen. They interpreted their data as inconsistent with the need for integrin extension for ligand binding, but their study had a number of limitations, including lack of quantitation of the percentage of receptors that mediated ligand binding, the identification of a small subpopulation of receptors that did assume a more upright conformation in the presence of Mn 2ϩ , and the absence of cytoskeletal and contractile elements in their system that may play an important role in generating the conformational changes in ␣IIb␤3 (12).
Regardless of the precise mechanisms, the data from the ␤3 N339S mutant and Mn 2ϩ activation indicate that it is possible under certain circumstances for high M r ligands to bind to ␣IIb␤3 receptors that have limited mobility about the ␣IIb genu.
If, as suggested by our data, ligand size is an important determinant of its ability to bind to integrin receptors that have not undergone extension about the ␣-subunit genu, this may provide an explanation for the finding by Adair et al. (22) that an ϳ50-kDa fragment of fibronectin, which is between kistrin and fibrinogen in M r , can bind to ␣V␤3 without evidence of headpiece extension. However, because Mn 2ϩ was used to induce the binding of the fibronectin fragment, and we found that Mn 2ϩ could rescue the ability of the clamped ␣IIb␤3 receptor to bind soluble fibrinogen, it is not possible to separate the potential contributions of ligand M r and Mn 2ϩ activation.
In contrast to our finding that Mn 2ϩ activation could rescue soluble ligand binding to clamped ␣IIb␤3, Takagi et al. (16) found reduced soluble fibrinogen binding to cross-clasped ␣IIb␤3 even in the presence of both Mn 2ϩ and PT25-2. Thus, the alteration induced by cross-clasping differs from that induced by our clamping procedure. Collectively our data and those of Takagi et al. could be explained by postulating that PT25-2 (and perhaps AP5) induces activation via an initial effect on ␣IIb extension, with subsequent effects on the ␤3 subunit. The effect of the ␤3 N339S mutant and Mn 2ϩ may thus be downstream of the effect of PT25-2, but the Mn 2ϩ effect may require mobility between regions of ␣IIb and ␤3 that are restricted by the cross-clasp introduced by Takagi et al.
Fibrinogen is unusual among ␣IIb␤3 ligands in that the binding of soluble fibrinogen requires receptor activation, whereas adhesion to immobilized fibrinogen does not (42)(43)(44)(45)(46). One possible explanation for this behavior is that activation-induced ␣IIb␤3 extension is required for the binding of soluble fibrinogen, but not for mediating adhesion to immobilized fibrinogen. To test this hypothesis, we also studied the effect of clamping ␣IIb about the genu on ␣IIb␤3-mediated adhesion to fibrinogen. Clamping ␣IIb␤3 produced little or no reduction in adhesion to fibrinogen immobilized at high density and only a slight reduction in adhesion to fibrinogen immobilized at low density. Thus, ␣IIb extension about the genu had much less of an effect on ␣IIb␤3 interactions with immobilized fibrinogen than on binding of soluble fibrinogen. A number of mechanisms have been proposed to account for the difference in requirements for ␣IIb␤3 binding of soluble fibrinogen versus mediating adhesion to immobilized fibrinogen, including conformational changes in fibrinogen induced by immobilization (47) and the potential for multivalent interactions between receptors and ligand molecules stemming from the higher density of fibrinogen achieved with immobilization (43). We noted a greater impact of clamping ␣IIb at the lower fibrinogen coating density, suggesting a correlation with ligand density, but these data need to be interpreted in light of the known impact of fibrinogen coating density on the conformational changes induced by immobilizing the ligand (43,47). The recent study by Zhu et al. (12) provides an additional hypothesis, namely adhesion-mediated engagement of the cytoplasmic region of ␤3 with the cytoskeleton, leading to activation of the receptor via contractile traction forces initiating receptor extension and ␤3 swing-out. Thus, even an initial low affinity interaction with immobilized fibrinogen could induce higher affinity by this traction activation mechanism. If this is a contributing factor, our data indicate that higher affinity can be achieved even without extension at the genu. Most recently, Ye et al. found that binding of the talin head domain to the cytoplasmic domain of ␤3 was sufficient to induce both ␣IIb␤3 extension and PAC-1 binding (10), but this does not exclude an additional contribution to enhanced ␣IIb␤3 affinity via a traction mechanism (12).
Binding of ligands to ␣IIb␤3 is known to induce conformational changes in the receptor (4 -7), resulting in outside-in signaling that leads to profound cellular changes, most dramatically cytoskeletal reorganization and cell spreading (2). Although receptor extension has been postulated to affect ligand binding, little is known about its potential contribution to postligand binding events. Therefore, we assessed the effect of clamping ␣IIb on the ability of the snake venom kistrin (which we demonstrated binds similarly to both normal and clamped ␣IIb␤3) to expose the AP5 epitope, which is located in the PSI domain of ␤3 (amino acids 1-6) and becomes more available with ligand binding (5). As expected, AP5 binding to normal ␣IIb␤3 was low in the absence of kistrin and increased ϳ6-fold in the presence of kistrin. AP5 binding to clamped ␣IIb␤3 was similarly low in the absence of kistrin, but the addition of kistrin produced less enhancement than was observed with normal ␣IIb␤3. This was not due to an irreversible alteration in the clamped molecule or reduced expression of surface receptors because treatment with DTT completely rescued the response. Clamping the ␣IIb␤3 N339S mutant also decreased kistrin-induced AP5 binding. In the crystal structure of the complete ectodomain of ␣IIb␤3, the AP5 epitope is masked by IEGF-2 in the bent conformation, but it is predicted to be accessible with headpiece extension (12). Thus, our data provide evidence that AP5 epitope exposure secondary to ligand binding requires extension at the genu and raises the possibility that the latter also contributes to outside-in signaling. Moreover, because Zhu et al. reported that the hybrid and PSI domains bury a total 822 Å 2 of solvent-accessible surface area in the bent receptor compared with a model of the extended receptor, it is also possible that headpiece extension initiated by inside-out signaling contributes to enhanced ligand affinity by reducing the energy barrier for the swing-out motion of the hybrid and PSI domains.
It is of interest to compare our data on the clamped receptor with observations on the platelets of a patient with Glanzmann thrombasthenia who had intact ␣IIb␤3 ectodomains but whose receptors were insensitive to inside-out signaling as a result of a mutation in the ␤3 cytoplasmic domain (26,36). The patient's receptors were able to support platelet adhesion to immobilized fibrinogen but not soluble fibrinogen binding in response to agonist stimulation. The similarities between the patient data and the data using clamped ␣IIb␤3 are consistent with a model in which inside-out signaling acts, at least in part, by inducing extension at the ␣IIb genu.
In conclusion, our data support a role for ␣IIb extension about the genu in the binding of high M r , but not lower M r ligands under some, but not all conditions of activation. Additional data are required to define which pathways of activation are initiated by physiologic and pathologic stimuli. Although there is support for ␣IIb extension controlling binding of high M r ligands via an effect on access to the ligand-binding site (40), it remains possible that ␣IIb extension affects the activation responsiveness and/or affinity of the ligand-binding pocket in addition to, or instead of, affecting size-dependent access. One possibility is that the loss of the headpiece-tailpiece interactions that stabilize the ␤3 hybrid and PSI domains in the bent structure frees the ␤3 subunit to undergo the swing-out motion and the movement of the ␣1-␤1 loop toward the MIDAS. These data also provide a potential explanation for the difference in activation dependence of soluble fibrinogen binding versus adhesion to immobilized fibrinogen. Finally, our data raise the novel possibility that ␣IIb extension also contributes to postligand binding outside-in signaling by facilitating conformational change in ␤3.