Probing Chemical and Conformational Differences in the Resting and Active Conformers of Platelet Integrin αIIbβ3 *

Integrin αIIbβ3is the fibrinogen receptor that mediates platelet adhesion and aggregation. The ligand binding function of αIIbβ3 is “activated” on the platelet surface by physiologic stimuli. Two forms of αIIbβ3 can be purified from platelet lysates. These forms are facsimiles of the resting (Activation State-1 or AS-1) and the active (Activation State-2 or AS-2) conformations of the integrin found on the platelet surface. Here, the differences between purified AS-1 and AS-2 were examined to gain insight into the mechanism of activation. Four major findings are put forth. 1) The association rate (k 1) between fibrinogen and the integrin is a key difference between AS-1 and AS-2. 2) Although the divalent ion Mn2+ enhances the ligand binding function of AS-1, this ion is unable to convert AS-1 to AS-2. Therefore, its effect on integrin is unrelated to activation. 3) Peptide mass fingerprints indicate that the chemical structure of AS-1 and AS-2 are virtually identical, calling into question the idea that post-translational modifications are necessary for activation. 4) The two forms of αIIbβ3 have significant conformational differences at three positions. These include the junction of the heavy and light chain of αIIb, the divalent ion binding sites on αIIb, and at a disulfide-bonded knot linking the amino terminus of β3 to the cysteine-rich domain. These observations indicate that integrin is activated by a series of specific conformational rearrangements in the ectodomain that increase the rate of ligand association.

The integrin ligand binding event is a dynamic process that is key to cellular function. Most integrins are displayed in at least two distinct conformations on the cell. These conformations have varying affinities for ligand. The transition between one affinity state and another is typically referred to as "activation," and such transitions are brought about by "inside-out" signaling (for reviews, see Refs. [1][2][3][4][5]. The activation of integrins is crucial to events like morphogenesis, cell migration and invasion, and platelet aggregation (6 -8).
Platelet integrin ␣ IIb ␤ 3 has served as a paradigm of many aspects of integrin structure and function. This receptor is a particularly good model of an integrin whose activation is of immense physiologic significance (9). Integrin ␣ IIb ␤ 3 is pivotal to the control of bleeding because it mediates platelet aggregation. Integrin ␣ IIb ␤ 3 is activated when platelets are stimulated by agonists like ADP or thrombin. Activation endows ␣ IIb ␤ 3 with the ability to bind soluble fibrinogen (Fg) 1 present in the plasma. Because Fg is a dimer, its binding leads to platelet aggregation and ultimately halts the loss of blood. Because the concentrations of both Fg and platelets are so high in the blood, ␣ IIb ␤ 3 must be maintained in a resting, or inactive state, to ensure proper blood flow. Therefore, an understanding of how ␣ IIb ␤ 3 is activated is of great importance.
The precise mechanism by which the ligand binding affinity of ␣ IIb ␤ 3 , and other integrins, is modulated by cellular stimulation is still not completely understood. Several pathways to activation have been put forth. Many of these have focused on the role of the integrin cytoplasmic domains in activation (10). One hypothesis suggests that changes in the conformation of the cytoplasmic tails can release a conformational constraint or open an "integrin hinge" (11). This hinge could potentially be opened by proteolysis of the integrin cytoplasmic domains (12)(13)(14). Alternatively, the hinge could be released by phosphorylation of the integrin. The cytoplasmic tail of ␤ 3 contains several potential phosphorylation sites (15,16). One of these phosphorylation sites, an NPXY motif within the tail of ␤ 3 , is essential for integrin activation (17,18). Therefore, phosphorylation must also be considered as a potential route to activation. Another pathway to activation may involve proteolysis within the ectodomain of the integrin. In fact, there is evidence that ␣ IIb ␤ 3 can be activated in this manner (19). Still other work indicates that divalent ions, which bind to the integrin and control ligand affinity, may activate the integrin (20 -22). The physical association of integrins with regulatory proteins (23)(24)(25) could also enact conformational changes that lead to activation. Intracellular signaling pathways like those controlled by Ha-and R-Ras (26,27) have also been implicated in the activation of integrins.
Despite this progress, there is still no unifying hypothesis on exactly what biochemical changes occur within an integrin during activation. We reasoned that one way to gain insight into this issue was to compare the differences in the structure and function of the resting and active integrin. We have used Surface Plasmon Resonance (SPR) to compare the kinetic behavior of the active and resting form of ␣ IIb ␤ 3 and mass spectrometry to probe the structural differences between the two conformers. The study reveals that the active and resting integrin are virtually identical in chemical structure but reveals important conformational distinctions. These observations lay the fundamental groundwork required to properly interpret other studies aimed at deciphering how inside-out signaling alters the ligand binding function of integrins.

MATERIALS AND METHODS
Purification of ␣ IIb ␤ 3 -The two forms of integrin ␣ IIb ␤ 3 , AS-1 and AS-2, were purified according to previously published procedures (28 -30) with minor modifications. Briefly, washed, outdated platelets were lysed in a 20 mM Tris buffer, pH 7.4, containing 150 mM NaCl, 1% Triton X-100, 5 mM phenylmethylsulfonyl fluoride, 1 mM CaCl, and 10 Ϫ5 leupeptin. The lysate was centrifuged at 50,000 ϫ g and stored at Ϫ80°C until further use. The lysate was rapidly thawed at 37°C, centrifuged at 50,000 ϫ g, and passed over Con-A-Sepharose (Amersham Pharmacia Biotech). Proteins were eluted with Buffer A (20 mM Tris buffer, pH 7.0, 150 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 0.1% Triton X-100, and 0.05% NaN 3 ) containing 200 mM methyl ␣-D-mannopyranoside (Sigma) and 10 Ϫ5 M leupeptin (Sigma). Fractions containing ␣ IIb ␤ 3 were pooled and then depleted of the active form of the integrin (AS-2) by circulating over a column of KYGRGDS-Sepharose. To ensure that the flow through was depleted of AS-1, the lysate was passed over two RGD affinity columns in series. The AS-2 form of the integrin was eluted from these columns using soluble RGD peptide. AS-2 was concentrated, dialyzed against buffer A, and stored at Ϫ80°C. The flow through from the KYGR6DS-Sepharose column, containing the resting form of the integrin (AS-1) was circulated over heparin-agarose column to remove the thrombospondin. The effluent was concentrated and AS-1 further purified by gel filtration on Sephacryl S-300. The integrin was concentrated, dialyzed against Buffer A, and stored at Ϫ80°C. The purity and integrity of each form of the integrin was assessed by SDS-PAGE.
Peptides, Ligands, and Antibodies-Human fibrinogen was purchased from Enzyme Research Laboratories. Fab-9, a recombinant antibody containing the RGD motif (31,32), was purified by affinity chromatography on a column of goat anti-human IgG-Sepharose. Synthetic peptides were purchased from Coast Scientific and from Anaspec. Antibodies against the C-terminal domain of ␣ IIb , including PM1-1, anti-V41 (33) which are specific for C terminus of the heavy chain of ␣ IIb and the N terminus of the light chain of ␣ IIb, respectively, were generous gifts from Dr. Mark Ginsberg (Scripps Research Institute). Cells expressing recombinant variants of Fg lacking the RGD motifs were generously provided by Dr. David Farell (Milton Hershey Medical Center, Penn State University).
Construction of Ligand Affinity Resins-For affinity purification of ␣ IIb ␤ 3 , the peptide KYGRGDS was coupled to CNBr-Sepharose (Amersham Pharmacia Biotech) according to the manufacturer's specifications. To analyze the depth of the integrins ligand binding pocket, affinity resins were constructed in which the RGD peptide was coupled to the resin with an extended spacer. This was accomplished using a procedure reported by Yan et al. (34) with some modifications. Briefly, Sepharose was activated with epichlorohydrin at pH 12. Then 1,6hexanediamine was added and incubated with the resin at 45°C for 2 h with stirring. After washing with water, the resin was incubated with epichlorohydrin again for an additional hour. The activated resin was washed extensively with distilled water, and the peptide with sequence KYGRGDS was added at a concentration of 1 mM. The peptide was FIG. 1. Comparing the depth of the RGD binding pocket on purified AS-1 and AS-2. The resting and active conformers of ␣ IIb ␤ 3 were purified from outdated platelet lysates as described under "Materials and Methods." The purity of AS-1 (panel A, lane 1) and AS-2 (panel A, lane 2) was assessed by Coomassie staining of 7% acrylamide gels run under non-reducing conditions. The depth of the RGD binding site on AS-1 (panel B, lanes 1 and 2) and AS-2 (panel B, lanes 3 and 4) was measured by testing the ability of the integrin to bind to either KYGRGDS-Sepharose, (lanes 1 and 3) or to KYGRGDS-hexanediamine-Sepharose (lanes 2 and 4). Integrin (20 g) was mixed with 20 l of settled affinity resin and incubated for 18 h. The affinity resins were washed extensively to remove unbound integrin. Bound integrin was removed from the resin by incubation in SDS sample buffer. The eluted integrin was analyzed on SDS-PAGE. This experiment is representative of three repetitions, making use of three different batches of purified integrin in which identical results were obtained.

FIG. 2.
Binding of AS-1 and AS-2 to fibrinogen. The binding of AS-1 and AS-2 to Fg-linked plasmon resonance sensor chips was measured with the BIACore 3000. Binding was performed in buffer containing 1 mM Ca 2ϩ and 1 mM Mg 2ϩ . From a series of such sensorgrams in which the amount of integrin included in the analyte was varied, the association rate constant (k 1 ) between AS-2 and Fg was derived as described (35,36).
FIG. 3. Effect of Mn 2؉ on the binding of AS-1 and AS-2 to fibrinogen. To determine whether Mn 2ϩ could convert AS-1 to AS-2, the AS-1 form of the integrin was dialyzed into buffer containing 200 M Mn 2ϩ . A sample of this material was then dialyzed back into buffer containing 1 mM Ca 2ϩ . Then, the binding of AS-2 (q), AS-1 (Ⅺ), AS-1 dialyzed into Mn 2ϩ (f), and AS-1 dialyzed into Mn 2ϩ and then back into Ca 2ϩ (OE) to fibrinogen was measured with SPR. The resulting sensorgrams are shown. This experiment is one of three repetitions, making use of two separate batches of purified AS-1 and AS-2, each of which yielded nearly identical results.

TABLE I
Detection of tryptic peptides from AS-1 and AS-2 Integrin ␣ IIb ␤ 3 was separated on SDS-PAGE, and the heavy chain of ␣ IIb (section A), the light chain of ␣ IIb (section B) or the ␤ 3 subunit (section C) was excised and subjected to in-gel tryptic digestion as described under "Materials and Methods." Peptides were extracted from the gel slice and analyzed by MALDI-TOF mass spectrometry. The spectra between 800 daltons and 6000 daltons was examined for peptides. The monoisotopic mass of the predicted tryptic peptides are compared to the masses of peptides observed in the MALDI profile. The peptides observed in the spectra are in the column designated as "Native ␣ IIb ␤ 3 ." Some peptides could not be detected due to their glycosylation, so ␣ IIb ␤ 3 was enzymatically deglycosylated and analyzed in the same manner (column noted as "Deglycosylated"). Peptide fragments that were not observed due to size or detection limitations are noted as ND. The change in the mass of a fragment containing an N-linked sugar chain after deglycosylation (Asn 3 Asp) is an increase of 1 dalton.  Surface Plasmon Resonance-The kinetic parameters (association and dissociation rate constants, k 1 and k Ϫ1 , respectively) and the affinity constant (K d ) between Fg and ␣ IIb ␤ 3 , or Fab-9 and ␣ IIb ␤ 3 , were measured by SPR using methods we have described previously (35,36). To eliminate conformational changes, the ligands, either Fg or Fab-,9 were coupled to the sensor chip. Then, a solution containing integrin was applied to the surface and binding was measured as a function of time. Briefly, a CM5 sensor chip was activated by 15 l 1-ethyl-3-(3dimethylaminopropyl)carbodiimide/N-hydroxysuccinimide and 40 l of Fg at 50 g/ml in sodium acetate buffer, pH 5.87, was coupled to the chip. The surface was then treated with 10 l of ethanolamine to block uncoupled carboxylate groups on the chip. Fab-9 was linked to the sensor chip in the same manner except that the pH of the immobilization buffer was set at 4.5. Following binding, chips were regenerated with a vast excess of RGD peptide, or with 10 mM CaCl 2 . For all the reactions the working buffer was Tris-buffered saline, pH 7.4, 0.005% surfactant P20 and included divalent cations. Integrin was diluted from concentrated purified stocks into this working buffer just prior to use. Kinetic constants were fitted to a model that assumes a one-to-one binding relationship using the Langmuir equations in the BIAcore 3000 software package.
Peptide Mapping-Limited digestion of ␣ IIb ␤ 3 was accomplished with Asp-N protease (Roche Molecular Biochemicals, sequencing grade). Asp-N was added to purified ␣ IIb ␤ 3 (1 mg/ml) of 100 g in 100 l of 20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 0.1% Triton X-100. Protease was added in a ratio of 1:300 (weight to weight) and incubated with integrin at 37°C for 4 h. The reaction was stopped by chilling to 4°C and immediately analyzing the sample on 12% SDS-PAGE.
Protein and Peptide Mass Fingerprinting-Peptide mass fingerprinting (37,38) was used to define the composition of ␣ IIb and ␤ 3 and the composition of their proteolytic fragments. Proteins or peptides were separated on 12% SDS-PAGE under non-reducing conditions. Proteins were visualized by silver staining. Gel slices containing protein were excised and then reduced and alkylated by performing the following procedure three times. Gel slices were dehydrated with acetonitrile and subjected to reduction with 20 mM dithiothreitol (DTT) at 56°C for 1 h. Proteins embedded in the gel were then alkylated with 60 mM iodoacetamide at room temperature for 30 min. Gel slices were washed with 100 mM NH 4 HCO 3 , again dehydrated with acetonitrile. Gel slices containing reduced and alkylated integrin were dried in a vacuum centrifuge and then re-hydrated in 20 mM NH 4 HCO 3 containing 20 g/ml of trypsin (Roche Molecular Biochemicals, modified sequencing grade).
Each gel slice was incubated with enzyme in an ice-bath for 1 h, and then at 37°C for 18 h. Gel slices were washed with 20 mM NH 4 HCO 3 and then tryptic peptides were extracted with three changes of a solution of 5% formic acid, 50% acetonitrile, and 25% isopropanol. The solution containing peptides was evaporated to dryness. Dry peptide mixtures were dissolved in 0.1% trifluoroacetic acid. To detect some domains within ␣ IIb and ␤ 3 by mass spectrometry, it was necessary to deglycosylate the protein. Deglycosylation was carried out in 20 mM sodium phosphate buffer, pH 7.5 with O-glycanase, N-glycanase, and sialidase (Roche Molecular Biochemicals) at 37°C for 40 h.
In some cases proteins were digested in solution rather than in acrylamide gel slices. Because Triton X-100 suppresses peptide detection in MALDI mass spectrometry, it was necessary to remove this detergent. This was accomplished with the following procedure. Methanol (140 l) was added to 35 g of integrin in a volume of 35 l. The sample was vortexed and centrifuged at 9000 ϫ g for 10 s. Then, chloroform (35 l) was added to the solution and vortexed, and the sample was centrifuged again. Finally, water (90 l) was added and the sample vigorously vortexed. Phase separation was enacted by centrifugation for 1 min. The upper phase was discarded. The lower phase, containing chloroform, and the interface, comprised of precipitated ␣ IIb ␤ 3 , were retained. Methanol (90 l) was added to the remainder of the sample to precipitate the ␣ IIb ␤ 3 . Precipitated integrin was recovered by centrifugation for 2 min at 9,000 ϫ g. The supernatant was discarded and the protein pellet was air dried. Then the protein was resuspended in 20 mM NH 4 HCO 3 containing 0.2% octylglucoside. Lys-C protease (Roche Molecular Biochemicals, modified sequencing grade) was added to a final ratio of 1:10 (weight to weight), and the sample was incubated at 37°C for 18 h.
Mass Spectrometry-MALDI-MS spectra were obtained with a Voyager DE-RP MALDI-TOF mass spectrometer (PerSeptive Biosystems, Framingham, MA) equipped with a nitrogen laser (337 nm, 3-ns pulse). Spectra were collected in reflector mode. The accelerating voltage in the ion source was 20 kV. Data were acquired with a transient recorder with 2-ns resolution. The matrix used in this work was ␣-cyano-4hydroxycinnamic acid dissolved in water/acetonitrile (1:1, v/v) to give a saturated solution at room temperature. To prepare the sample for analysis, 1 l of the peptide solution (containing 1-10 pmol of protein in 0.1% trifluoroacetic acid) was added to 1 l of the matrix solution and applied to a stainless steel sample plate. The mixture was then allowed to air dry on the sample plate before being introduced into the mass spectrometer. Each spectrum was produced by accumulating data using 128 laser pulses. Mass assignments were assigned with an accuracy of approximately Ϯ 0.1% (Ϯ 1 Da/1000 Da). The computer program called Peptide Mass on the ExPASy Molecular Biology Server was used to calculate the masses of all possible peptides. The predicted mass of each peptide reflects the modifications to the mass made by reduction and alkylation. b Potential glycosylation sites were identified by the Swiss-Prot database; (I) reflects the presence of a potential glycosylation site whose usage has been substantiated by experimental evidence. (P) indicates that the peptide contains a potential glycosylation site that is predicted by consensus sequence.
c All peptides present in the spectra had masses greater that 200 daltons.

RESULTS
Purification of the Active and Resting Forms of ␣ IIb ␤ 3 from Platelet Lysates-The ability to purify two forms of ␣ IIb ␤ 3 from platelet lysates is well documented in the literature (28,29). Our procedure for obtaining the resting (AS-1) and active (AS-2) conformations of ␣ IIb ␤ 3 essentially paralleled the procedures established by others. With these methods, approximately 2.5 mg of AS-2 and 20 mg of AS-1 could be obtained from 50 units of outdated platelets. The purity of each form of the integrin was judged to be at least 90% by Coomassie staining of SDS-PAGE (Fig. 1A.) We found no evidence of the reproducible co-purification of any low M r proteins that could be associated with ␣ IIb ␤ 3 in a stoichiometric manner.
One of the hallmark differences between the resting and active forms of ␣ IIb ␤ 3 on the platelet surface is the depth of the RGD binding pocket. This depth has been gauged by measuring the ability of resting and activated platelets to adhere to beads coupled to RGD peptides via spacers of varying length (39). To verify that the purified forms of ␣ IIb ␤ 3 are similar to the conformers observed on the platelet surface, we assessed the depth of the RGD binding pocket on AS-1 and AS-2. We measured the binding of each conformer to affinity columns in which the RGD motif was linked to the resin in different ways. In one approach, the RGD peptide was linked directly to the resin, providing a short spacer. In another approach, the RGD peptide was bridged to the resin with a hexanediamine spacer of about 18 Å. Only AS-2 bound the RGD peptide linked directly to Sepharose (Fig. 1B), but both AS-1 and AS-2 bound the beads in which the RGD was displayed on an extended linker. These observations indicate that the differences in the exposure of the RGD binding pocket on purified AS-1 and AS-2 are similar to those observed on the platelet surface.
Comparison of the Fibrinogen Binding Properties of AS-1 and AS-2-In equilibrium binding studies, Kouns et al. (28) found that only AS-2 could bind to Fg. No binding between AS-1 and Fg could be detected. We sought to determine whether the failure to detect binding between AS-1 and Fg is a consequence of a slow association rate or a rapid dissociation rate between the two proteins. This distinction has important implications for understanding how cellular signaling events regulate integrin activation.
SPR was used to measure k 1 for the binding of AS-1 and AS-2 to Fg (Fig. 2). The association between AS-1 and fibrinogen could not be detected, even at high concentrations of integrin.
Under the conditions of this experiment, the lowest association rate that can be observed with the BIACore was 1 ϫ 10 3 M Ϫ1 s Ϫ1 . Therefore, the rate at which AS-1 associates with Fg is below this value. In contrast, AS-2 bound to Fg with an association rate constant of 1.3 ϫ 10 5 M Ϫ1 s Ϫ1 . In related experiments, we found that the association rate between AS-2 and the model RGD-ligand Fab-9 was also about 7-fold faster than the rate of association between AS-1 and Fab-9 (2.3 ϫ 10 5 M Ϫ1 s Ϫ1 versus 3.2 ϫ 10 4 M Ϫ1 s Ϫ1 ). Therefore, a primary difference between AS-1 and AS-2 is the rate at which ligands associate with their respective binding pockets.
Testing the Role of Mn 2ϩ as an Activator-The divalent ion Mn 2ϩ has been linked to the activation of integrins because it increases ligand binding affinity and promotes cell adhesion (20,40,41). It is widely thought that Mn 2ϩ mimics physiologic activation of integrin. To test this idea, we measured the ability of Mn 2ϩ to convert AS-1 to AS-2. AS-2 binds to Fg when Ca 2ϩ is the only available divalent ion, and this binding function is stable to prolonged dialysis. Therefore, these characteristics were set as criteria for the conversion of AS-1 to AS-2.
To perform the conversion test, AS-1, purified in Ca 2ϩ -containing buffer, was dialyzed into buffer containing 200 M Mn 2ϩ . Then, the ability of the integrin to bind to fibrinogen was measured by SPR. As shown in Fig. 3, Mn 2ϩ increased the association between AS-1 and Fg to the point where it could be detected by SPR. However, when the Mn 2ϩ -loaded AS-1 was returned to buffer containing only Ca 2ϩ , its ability to bind Fg was lost. This cation-exchange procedure could be repeated through several cycles, each resulting in an increase in k 1 for Fg when Mn 2ϩ was present, and a subsequent ablation of binding to Fg when only Ca 2ϩ was present. Thus, whereas Mn 2ϩ increases the association rate of AS-1 for ligand, it does not convert AS-1 to AS-2.
Probing the Chemical Structure of AS-1 and AS-2 by Peptide Mass Fingerprinting-There is no detailed information on how the structures of AS-1 and AS-2 differ. This leaves open the possibility that they are alternatively spliced forms of the integrin or differ in some key posttranslational modification. A comprehensive mass fingerprinting study was undertaken to probe for differences between AS-1 and AS-2. Because of the accuracy of mass spectrometry, this type of analysis is capable of revealing differences in amino acid sequence and in posttranslational modification. Mass fingerprinting was performed by separating the ␣ IIb and ␤ 3 subunits (derived from AS-1 and arrows) and Lys-C protease (bottom arrows). The monoisotopic mass of the predicted fragments are shown above each peptide. B, a portion of the mass spectra from ␤ 3 derived from AS-1 and digested with trypsin is shown. The peptide peaks corresponding to each of the peptides from within the cytoplasmic tail are labeled (arrows). A smaller section of the spectra from ␤ 3 derived from AS-1 but digested with Lys-C protease is shown in the inset, revealing a peptide fragment corresponding to the extreme C terminus of ␤ 3 . C, mass spectra derived from the AS-2 form of ␤ 3 treated in the same manner as described for panel B are shown. AS-2) by SDS-PAGE under reducing conditions. Bands corresponding to each subunit were excised and subjected to in-gel digestion with trypsin as described under "Materials and Methods." Then, the resulting mixtures of peptides were analyzed by MALDI-TOF mass spectrometry. From this analysis we were able to detect peptides comprising 86% of the entire sequence of ␣ IIb (Table I, sections A and B), and 85% of the sequence of ␤ 3 (Table I, section C). Perhaps surprisingly, no differences were observed in the peptides that were detected from AS-1 and AS-2. The majority of the undetected tryptic fragments were of very low mass (Ͻ300 daltons) and are not amenable to detection by MALDI-TOF. The only larger fragments that remained undetected by MALDI included tryptic fragments correspond-ing to residues 166 -276 from ␣ IIb and a fragment corresponding to residues 151-181 of ␤ 3 . These fragments were not observed in either AS-1 or AS-2. We also failed to detect a peptide encompassing the amino-terminal 12 residues of the ␣ IIb light chain, which is not unexpected because alternative processing of this terminus of the light chain of ␣ IIb has been reported before. Amino acid sequencing of this chain of ␣ IIb showed Val-13 to be the amino-terminal residue (42,43). We were also unable to Western blot ␣ IIb with the anti-V41 antibody which binds to this region of ␣ IIb (33).
Because so much attention has focused on the role of the integrin cytoplasmic tails in the process of activation, particularly the cytoplasmic domain of the ␤ subunit, the MALDI FIG. 6. Peptide maps of AS-1 and AS-2 digested with Asp-N protease. A, AS-2 (lane 1) and AS-1 (lane 2) were digested with Asp-N protease at an enzyme:protein ratio of 1:300 (w/w). Digestion was performed for 6 h at 37°C. Then, samples were separated on 12% acrylamide gels, and peptide fragments were visualized by Coomassie staining. B, peptides 1, 7, and 9 generated by Asp-N digestion of ␣ IIb ␤ 3 were excised from the gel shown in panel A and subjected to in-gel tryptic digestion as described under "Materials and Methods." Their composition was determined by peptide mass fingerprinting with MALDI-TOF mass spectrometry. This figure contains structural information required to interpret results of these peptide mass fingerprints from peptides 1, 7, and 9 (A). The amino-terminal domain of the ␤ 3 subunit (boxed) is linked to the cysteine-rich domain (boxed) via a long-range disulfide bond between Cys-5 and Cys-435 (arrows). Other known disulfide bonds within the domain are also noted with arrows. Peptides 1, 7, and 9 were generated by cleavage with Asp-N protease, which cleaves on the amino-terminal side of aspartic acid residues. The Asp-N fragments were then excised from the gel and subjected to peptide mass fingerprinting using trypsin, which cleaves at lysine and arginine residues. Therefore, the cleavage points for both Asp-N and trypsin are shown as vertical lines above and below the amino acid sequence. In addition, the monoisotopic mass of all fragments that were detected within the mass fingerprints by MALDI-TOF mass spectrometry is noted above the sequence. A key distinction between AS-1 and AS-2 is the presence of the tryptic peptide extending from residues 412-423 (bold text) in the mass fingerprints. This tryptic peptide is detected in mass fingerprints of Asp-N fragments 1 and 7 (AS-1) but is not detected in fragment 9 (AS-2). This tryptic peptide must be derived from the slightly larger Asp-N fragment, which was generated from the original digestion of ␣ IIb ␤ 3 and that extends Asp-393 to Asp-423 (underlined). spectra containing tryptic peptides derived from the tail of ␤ 3 are shown in Fig. 4. In fingerprints generated with trypsin, peptides encompassing ␤ 3 residues 717 to 762, all but the two most C-terminal amino acids, were observed in MALDI profiles (Fig. 4B). The presence of the carboxyl-terminal two residues on ␤ 3 was confirmed by digestion with Lys-C protease, yielding the predicted fragment with mass of 1561 daltons (Fig. 4, B and  C, insets). Phosphorylation of the cytoplasmic tails would be evident as an increase in the mass of tryptic peptides by 80 daltons per phosphate group. There was no reproducible evidence of phosphorylation of the peptides derived from either AS-1 or AS-2. This observation was supported by the inability to Western blot either form of ␣ IIb ␤ 3 with anti-phosphotyrosine antibodies (not shown). Similar observations were made with the cytoplasmic domain of ␣ IIb (Table I, section B). Based on these mass fingerprints, we conclude that the two conformers of ␣ IIb ␤ 3 have the same primary structure over the vast majority of their sequence.
A Difference in the Sensitivity of Disulfide Bonds in AS-1 and AS-2 to Reduction-Given that the chemical structure of AS-1 and AS-2 appeared to be identical, effort was focused on identifying the conformational distinctions between the two molecules. In one approach, we examined the sensitivity of disulfide bonds to reduction (Fig. 5). AS-1 was largely insensitive to mild reduction with DTT (Fig. 5A, lane 2). In contrast, the disulfide bond connecting the heavy and light chain of ␣ IIb was reduced using only mild reductant (Fig. 5B, lane 2). Therefore, a major conformational distinction between AS-1 and AS-2 is the exposure of the disulfide bond joining the heavy and light chains of ␣ IIb . With an increase in the amount of DTT, and with a prolonged incubation time, the disulfide bond connecting the heavy and light chain of ␣ IIb in AS-1 was partially reduced (Fig.  5A, lane 3). The addition of RGD peptide stabilized both forms of the receptor to reducing agent (Fig. 5, A and B, lanes 4), indicating a conformational connection between the labile disulfide bonds and the ligand binding pocket.
Mapping Conformational Differences in the ␤ 3 Subunit of AS-1 and AS-2-To examine the conformational differences between AS-1 and AS-2 in more detail, the two proteins were subjected to a peptide mapping study. This procedure differed from the mass fingerprinting study because the proteins were subjected to limited digestion with Asp-N protease in solution. Then the resulting fragments were subjected to mass fingerprinting to determine their composition.
Peptide maps that revealed fragments that distinguish AS-1 and AS-2 were generated by limited proteolysis. The Asp-N fragments unique to the digests of AS-1 and AS-2 were evident on SDS-PAGE (Fig. 6A). Each of the unique Asp-N fragments was excised from the SDS gel and subjected to in-gel digestion with trypsin and peptide mass fingerprinting as described under "Materials and Methods." From the masses of the peptides present in these MALDI spectra, the composition of these peptides and their location within the sequences of ␣ IIb and ␤ 3 was surmised. The Asp-N-generated fragments can be separated into two groups that are discussed below.
Fragments 1 and 7 from AS-1 and fragment 9 from AS-2 are derived from the ␤ 3 subunit (refer to Fig. 6B for a visual representation of this domain). These Asp-N fragments contain segments of a domain encompassing the amino terminus of ␤ 3 linked to the cysteine-rich domain through a long-range disulfide bond. The amino acid sequence of these domains, along with the Asp-N and trypsin cleavage points are shown in Fig.  6B. Peptides corresponding to the amino terminus of ␤ 3 (residues 1-62), along with peptides that correspond to residues 413-448 were evident in the mass fingerprint of fragment 1 (Table II, section A). Similarly, fragment 7 gave rise to tryptic peptides corresponding to the amino terminus of ␤ 3 (residues 1-62) along with peptides corresponding to ␤ 3 residues 413-578 (Table II, section B). Therefore, we conclude that fragments 1 and 7 contain segments of ␤ 3 that are connected by the disulfide bond that connects Cys-5 and Cys-435 (44). Fragment 1 migrates on SDS-PAGE at a mass consistent with the composition proposed in Table II, section A. However, fragment 7 migrates much higher than predicted based on the peptides detected in the mass fingerprint. This may indicate that fragment 7 contains additional peptide chain on its C terminus which is not detected by mass fingerprinting or indicate that it migrates anomolously on SDS-PAGE. The latter possibility is not entirely unexpected because the fragment is cysteine-rich and may remain partially folded during electrophoresis. Fragment 9, which is found only in the Asp-N digest of AS-2, is related to fragments 1 and 7 even though it migrates differently on SDS-PAGE (Fig. 6A). Tryptic mass fingerprinting of fragment 9 revealed the peptides shown in Table II, section C, and like peptides 1 and 7, it contains domains within the amino terminus linked to the cysteine-rich domain of ␤ 3 . However,

TABLE II
Mass fingerprints of fragments 1, 7, and 9 derived from ␤ 3 AS-1 and AS-2 were digested with Asp-N protease, and the resulting fragments were separated on SDS-PAGE. Peptide fragments 1 and 7 that are unique to AS-1 ( Figure 6A) and fragment 9 that is unique in AS-2 were excised and subjected to in-gel trypsin digestion to determine their composition. The mass fingerprint of each fragment was determined using MALDI-TOF mass spectrometry. The mass range extending from 800 to 6000 Da was examined. The predicted mass of each peptide and the corresponding fragment observed in MALDI are shown. Some tryptic peptides fall outside of the detectable mass range. Consequently, the peptides that are observed do not make up the contiguous sequence of each Asp-N fragment. fragment 9 differs from fragments 1 and 7 with regard to a tryptic peptide with a mass of 1123.8 daltons. This peptide, which is absent in the fingerprint of fragment 9, is consistently observed in the mass fingerprints of both fragments 1 and 7. This peptide corresponds to residues 413-422 (Fig. 6B, boldface). The presence of this tryptic peptide in fragments 1 and 7 indicates that they contain residues 393-423 which correspond to the Asp-N fragment that encompasses the tryptic peptide (Fig. 6B, underlined). The inability to detect this tryptic peptide in fragment 9 cannot be taken as proof of its absence. Nevertheless, the reproducible detection of the same tryptic peptide in several fingerprints of fragments 1 and 7, and from several distinct batches of purified AS-1, its absence in the fingerprint of fragment 9 is striking. We suggest that Asp-393 and Asp-423, the Asp-N cleavage points that flank this peptide are exposed to solvent in AS-2 and are liberated by Asp-N protease. In contrast, this fragment in AS-1 remains connected to the amino terminus during Asp-N digestion, indicating that it is buried. This distinction, combined with the simple fact that peptides 1,7, and 9 all migrate at different positions on SDS-PAGE, show that the conformation of AS-1 and AS-2 differ at the disulfide knot that connects the N terminus of ␤ 3 to the cysteine-rich domain. Mapping Conformational Differences in the ␣ IIb Subunit of AS-1 and AS-2-The other group of unique fragments generated by digestion with Asp-N protease arise from cleavage of the divalent cation binding domains within ␣ IIb . These include fragments 2-6 and 8 which are observed only in AS-1 (Fig. 6A). Mass fingerprinting of these fragments showed that they each contained peptides derived from the cation binding domains of ␣ IIb (Table III). No Asp-N generated fragments encompassing these domains of ␣ IIb were detected in AS-2. Consequently, we conclude that the divalent cation binding domains within AS-2 are highly sensitive to proteolytic attack by Asp-N protease. This is consistent with the fact that the divalent ion binding domain is rich in aspartate residues, a property that probably leads to the complete digestion of this domain to small peptide fragments. DISCUSSION Despite advances in our understanding of the signaling pathways that enact activation of integrins, little attention has been given to the changes in the kinetics of ligand binding that lead to activation. Similarly, the identity of chemical modifications, or positions of conformational change, that accompany activa-TABLE III Mass fingerprints of fragments 2-6 and 8 derived from ␣ IIb AS-1 and AS-2 were digested with Asp-N protease, and the resulting fragments were separated on SDS-PAGE. Peptide fragments 2-6 and 8 that are unique to AS-1 ( Figure 6A) were excised and subjected to in-gel trypsin digestion to determine their composition. The mass fingerprint of each fragment was determined using MALDI-TOF mass spectrometry. The mass range extending from 800 to 6000 Da was examined. The predicted mass of each peptide and the corresponding fragment observed in MALDI are shown. Some tryptic peptides fall outside of the detectable mass range. Consequently, the peptides that are observed do not make up the contiguous sequence of each Asp-N fragment. Fragments 2-4 migrate as a triplet on SDS-PAGE but show the same mass fingerprint.  tion have not been mapped. Here, we address these issues by using the platelet integrin ␣ IIb ␤ 3 as a model. As an initial step toward understanding the differences between AS-1 and AS-2, we compared their ligand binding properties. Prior study shows that ␣ IIb ␤ 3 contains two ligand binding sites, one for Fg and another site that binds to RGD (35,45). These ligand binding sites are physically separate but are linked kinetically. For example, RGD ligand blocks the association of Fg with its binding site and has the ability to dissociate pre-bound Fg (35). Our finding that the RGD binding site is present on AS-1 and AS-2 confirms the prior observation by Kouns (28). The fact that AS-1 and AS-2 have RGD binding sites with different degrees of exposure, or of depth, is also consistent with the observation that the depth of this site is regulated by activation on the platelet surface (39). The fact that the RGD site is more accessible to ligand on AS-2 also provides a structural basis for the more rapid rate of association with the model RGD ligand Fab-9.
No study has measured the association and dissociation rates between AS-1 and Fg. Therefore, the lack of binding between AS-1 and Fg under equilibrium conditions could have resulted from a high dissociation rate rather than a lack of association. In the present study, no association between As-1 and Fg could be detected, even at very high concentrations of Fg. This observation suggests that the Fg binding site may be entirely absent on resting integrin and may come to exist only when conformational rearrangements create such a site. This is in contrast to the RGD binding site, which can be detected on both AS-1 and AS-2 and which appears to be displayed at different depths on the two conformers.
Many artificial stimuli are used to mimic cellular activation in the study of integrin activation. Among these are antibodies that alter the integrins affinity state, reducing agents like dithiothrietol and the divalent ion Mn 2ϩ . Mn 2ϩ has been used extensively as a stimulus because it binds to integrins and increases the ligand association rate (20,40,41). Because of this property, Mn 2ϩ has come to be considered a "universal" integrin activator. Given this prevailing notion, we suspected that Mn 2ϩ might be able to convert AS-1 to AS-2. The conformations of AS-1 and AS-2 are stable to prolonged dialysis, so we were able to perform a series of experiments to test the ability of Mn 2ϩ to convert AS-1 to AS-2. The two parameters that were considered as benchmarks of conversion were 1) the ability to bind to Fg in the presence of Ca 2ϩ , and 2) the stability of this property to dialysis. Dialysis of AS-1 into buffer containing Mn 2ϩ increased the rate of association for Fg, but this effect was lost when the integrin was dialyzed back into buffer containing Ca 2ϩ . Therefore, we conclude that Mn 2ϩ enhances the rate of association with ligand but fails to induce the conformational changes that are equivalent to activation. We have also observed that treatment of AS-1 with mild reductant, like DTT, will increase the rate of association with ligand. Yet, our observations indicate that DTT also fails to convert AS-1 to AS-2. 2 Prior study established purification protocols for AS-1 and AS-2, but no detailed comparison of their chemical structures had been performed. This left open the possibility that the two variants of ␣ IIb ␤ 3 were actually alternatively spliced forms of the integrin, especially because alternative splicing has been observed for both ␣ IIb and for ␤ 3 (46 -48). By using peptide mass fingerprinting, we were able to detect tryptic peptide fragments that constitute more than 85% of the sequence of ␣ IIb and ␤ 3 in both the AS-1 and AS-2 forms. Most of the peptides that were not detected by MALDI were of very low mass. These studies showed the amino acid sequence of the two proteins to be virtually identical, and indicate that AS-2 is not an alternatively spliced form of the integrin present in platelets at low abundance.
We anticipated that the mass fingerprinting study would reveal some type of post-transitional modification that would account for the difference in activity of AS-1 and AS-2. Of particular interest were potential phosphorylations within the cytoplasmic domain of ␤ 3 . This subunit contains two consensus tyrosine phosphorylation sites and several serine and threonine residues that are candidates for modification by phosphate. We interpret the lack of phosphorylation on AS-1 or AS-2 to show that phosphorylation is not necessary to maintain the integrin in an active and stable conformation. We cannot exclude the possibility that phosphorylation within the cytoplasmic domains provides the activation energy to guide the integrin through a transition state that ultimately "decays" to a stable conformation like AS-2. There is certainly precedent for stepwise activation because some platelet agonists induce reversible activation of ␣ IIb ␤ 3 but others cause an irreversible activation (49,50). Similarly, the activation of the ␣ 5 ␤ 1 integrin is a two-stage event (51).
Even though the chemical structure of AS-1 and AS-2 are virtually identical, they have significant differences in conformation. One conformational difference centered around the disulfide bond that connects the heavy and light chain of ␣ IIb . Within AS-2, this disulfide bond is sensitive to mild reducing conditions, whereas the same bond in AS-1 is stable. Interestingly, the binding of RGD ligand stabilizes this bond to reduction, providing a conformational link to the RGD binding site. The association of this segment of the integrin with modulation of ligand binding affinity is consistent with prior work showing that monoclonal antibodies that bind in this vicinity can activate the ligand binding function of ␣ IIb ␤ 3 (52). It has also been reported that the cleavage of ␣ IIb in this region by neutrophil elastase can induce platelet aggregation (19). The association of the junction between the heavy and light chain of ␣ IIb with activation is also of interest because the junction is generated by intracellular proteolysis during maturation of the integrin. Most integrin ␣ subunits undergo similar processing at dibasic residues within the ␣ subunits. Although this cleavage is not required for ligand binding, it may be important for activation. Mutations in the integrin ␣ 6 gene that eliminate the proteolytic processing site render the ␣ 6 ␤ 1 integrin resistant to activation and kept the integrin in a low affinity state on the cell surface (53). Along with the differences that AS-1 and AS-2 display in this region of the integrin, all of these observations strongly implicate the domain encompassing the junction of the heavy and light chains of integrin ␣ subunits as a key site where conformational changes accompany activation.
Peptide mapping also showed that AS-1 and AS-2 exhibit a difference in conformation in a region of the ␤ 3 subunit that can be considered a disulfide "knot" connecting three regions of the protein that are well separated in the linear sequence. These are the extreme amino terminus (residues 1 to 62), the cysteine-rich domain (roughly residues 420 to 620), and a region proximal to the membrane spanning segment (residues 635 to 663). Within this knotted domain, a primary difference between AS-1 and AS-2 appears to center on a peptide extending from residues 393 to 423 that appears to be exposed in AS-2 and buried in AS-1. This segment of ␤ 3 is directly adjacent to Cys-435 which forms the disulfide bond to Cys-5, and it encompasses Cys-405 which links to Cys-655. Interestingly, the epitopes for many LIBS antibodies, which can activate the integrin and induce platelet aggregation, bind to these domains (54,55). Consequently, the evidence linking conformational 2 B. Yan and J. Smith, forthcoming report. changes within the disulfide knotted core of ␤ 3 to the process of activation is compelling.
Within the limits of detection, the resting and active forms of ␣ IIb ␤ 3 appear to have the same chemical structure. However, the two conformers have significant differences in conformation. These conformational differences primarily effect the rate of ligand association and, in fact, may create of the Fg binding site. These findings suggest that future effort should focus on understanding the mechanism by which structural transitions within the ectodomain of the integrin are brought about by inside-out signaling.