Activation of Platelet αIIbβ3 by an Exogenous Peptide Corresponding to the Transmembrane Domain of αIIb*>

A transmembrane domain heterodimer, acting in concert with a membrane-proximal cytoplasmic domain clasp, is thought to maintain integrins in a low affinity state. To test whether helix-helix interactions between the αIIb and β3 transmembrane domains regulate the activity of integrin αIIbβ3, we synthesized a soluble peptide corresponding to the αIIb transmembrane domain, designated αIIb-TM, and we studied its ability to affect αIIbβ3 activity in human platelets. αIIb-TM was α-helical in detergent micelles and phospholipid vesicles, readily inserted into membrane bilayers, bound to intact purified αIIbβ3, and specifically associated with the transmembrane domain of αIIb, rather than the transmembrane domains of β3, α2, and β1, other integrin subunits present in platelets. When added to suspensions of gel-filtered platelets, αIIb-TM rapidly induced platelet aggregation that was not inhibited by preincubating platelets with the prostaglandin E1 or the ADP scavenger apyrase but was prevented by the divalent cation chelator EDTA. Furthermore, αIIb-TM induced fibrinogen binding to platelets but not the binding of osteopontin, a specific ligand for platelet αvβ3. The peptide also induced fibrinogen binding to recombinant αIIbβ3 expressed by Chinese hamster ovary cells, confirming that its effect was independent of platelet signal transduction. Finally, transmission electron microscopy of purified αIIbβ3 revealed that αIIb-TM shifted the integrin from a closed configuration with its stalks touching to an open configuration with separated stalks. These observations demonstrate that transmembrane domain interactions regulate integrin function in situ and that it is possible to target intra-membranous protein-protein interactions in a way that can have functional consequences.

The affinity of integrins for ligands appears to be regulated by interactions between the transmembrane (TM) 2 and/or cyto-plasmic domains of their ␣ and ␤ subunits (1). Thus, it is likely that a TM domain heterodimer, acting in concert with a membrane-proximal cytoplasmic domain clasp, maintains integrins in a low affinity state. The prototypic example of integrin regulation is platelet ␣IIb␤3. ␣IIb␤3, a receptor for the plasma proteins fibrinogen, von Willebrand factor, fibronectin, and vitronectin is maintained in an inactive state on circulating platelets, but following vascular trauma, it shifts allosterically to an active conformation, a prelude to the formation of hemostatic platelet aggregates (1).
Data supporting the heteromeric association of the ␣IIb and ␤3 TM domains in unstimulated platelets are largely indirect (2,3). Moreover, proteins containing these domains also associate homomerically in micelles and bacterial membranes (4,5). It is noteworthy that mutations that either enhance or disrupt homomeric ␣IIb and ␤3 TM domain interactions in vitro can activate the intact integrin expressed in Chinese hamster ovary (CHO) cells. These observations suggest a "push-pull" mechanism for ␣IIb␤3 regulation in which processes that destabilize heteromeric ␣IIb and ␤3 TM domain interactions push ␣IIb␤3 to its activated state, whereas processes that favor their homomeric association pull ␣IIb␤3 toward its active conformation (6).
In the absence of the extracellular portion of the protein, ␣IIb-TM domain peptides tend to form homodimers in micelles and Escherichia coli membranes, rather than associating with the ␤3 TM domain (5). Furthermore, the residues involved in ␣IIb homodimerization are also implicated in binding of the ␣IIb-TM domain to the ␤3 TM domain, implying that the two interaction sites overlap. Accordingly, we would expect that an ␣IIb-TM domain peptide might be able to bind to the TM region of ␣IIb in ␣IIb␤3-expressing cells, thereby disrupting heteromeric ␣IIb/␤3 TM domain helix-helix interactions. If helix-helix interactions between the ␣IIb and ␤3 TM domains are indeed critical for ␣IIb␤3 activation, disruption of these interactions should cause activation of the integrin. We show here that a synthetic peptide, designated ␣IIb-TM, corresponding to the wild type ␣IIb-TM domain, undergoes specific association with the ␣IIb-TM domain in micelles and bacterial membranes and induces platelet aggregation by interacting directly with ␣IIb␤3. These results provide strong support for the hypothesis that separation of the ␣IIb and ␤3 TM helices is required for ␣IIb␤3 activation. Furthermore, they demonstrate that it is possible to alter the function of an integrin in situ by targeting its TM domains, providing a new approach for the development of novel therapeutic agents. Finally, these results demonstrate a way to perturb TM-TM interactions in intact integrins, allowing determination of the role of these interactions in integrin activation and clustering.

EXPERIMENTAL PROCEDURES
Peptide Synthesis and Purification-Peptides were synthesized using a 430A peptide synthesizer (Applied Biosystems) at 0.25-mmol scales and on a Rink amide AM resin (200 -400 mesh) (Novabiochem) with a substitution level of 0.71 mmol/g. Activation of the free amino acids was achieved with N,N,NЈ,NЈ-tetramethyl-O-(7-azabenzotriazol-1-yl)uronium hexafluorophosphate as a 0.40 M solution in N,NЈ-dimethylformamide and using a reaction solvent containing 25% dimethyl sulfoxide (Me 2 SO) and 75% 1-methyl-2-pyrrolidinon. Side chain deprotection and simultaneous cleavage from the resin was performed with a mixture of trifluoroacetic acid/H 2 O/1,2-ethanedithiol/anisole (94:2.5:2.5:1 v/v) at room temperature under N 2 flow for 2 h. Crude peptides collected after precipitation with cold diethyl ether were dissolved in a mixture of 2-propanol/acetonitrile/ water (6:3:1) and then lyophilized. Peptides were then purified on a preparative reverse phase high pressure liquid chromatography system (Varian ProStar 210) with a C-8 semi-preparative column (Vydac) using a linear gradient of buffer A (0.1% trifluoroacetic acid in Millipore water) and buffer B, 2-propanol/ acetonitrile/water (6:3:1 v/v) containing 0.1% trifluoroacetic acid. The N terminus of nascent peptides was labeled with fluorescein isothiocyanate (FITC) using a modification of a standard protocol (7). To direct the modification to only the N-terminal amine, the reaction was conducted at the end of the automated peptide synthesis while the peptide was still attached to the resin and its lysine side chains protected by butoxycarbonyl groups. Briefly, N-⑀-fluorenylmethyloxycarbonyl-aminohexanoic acid (Fmoc-⑀-Ahx-OH; AnaSpec) was coupled to the N terminus of the peptide resin using manual peptide synthesis conditions. The fluorenylmethyloxycarbonyl protecting group was removed with 20% piperidine in N,N-dimethylformamide. Resin was then rinsed with N,N-dimethylformamide four times and swelled with dichloromethane. FITC (Aldrich) was dissolved in a mixture of pyridine/N,N-dimethylformamide/dichloromethane (12:7:5) to prepare a 0.1 M solution. The resulting solution was added to the resin, and the mixture was stirred at room temperature in the dark until a ninhydrin test indicated that the reaction was complete.
Analytical Ultracentrifugation-Equilibrium sedimentation was performed in a Beckman XL-I analytical ultracentrifuge (Beckman Coulter) at 25°C using six-channel carbon-epoxy composite centerpieces. Peptides, prepared at concentrations of 50 M, were dissolved in 20 mM HEPES buffer, pH 7.4, containing 5 mM C14-betaine (3-(N,N-dimethylmyristylammonio)propanesulfonate; Sigma) and 29% D 2 O previously determined to match the density of the detergent component (8). Partial specific volumes of peptides were calculated using a method described previously (9) and residue molecular weights corrected for the 29% D 2 O exchange expected for the densitymatched buffer. The solvent density (1.0245 g⅐ml Ϫ1 ) was measured using a Paar densitometer. Aqueous solution molar extinction coefficients were calculated using the program Sednterp and multiplied by the molar detergent concentration to provide mole fraction units. Data fitting and analysis were accomplished as described previously (10).
Tryptophan Fluorescence-Peptide insertion into lipid vesicles was assayed by monitoring changes in the emission spectrum of tryptophan as described previously (11). An aliquot of peptide from a concentrated stock in Me 2 SO was added to 10 mM HEPES buffer containing 150 mM NaCl, pH 7.0, to a final concentration of 2 M in 1.7 ml and was allowed to equilibrate for 5 min at which point the Trp emission spectra was recorded. The sample was then titrated with aliquots of vesicles (POPC/ POPG, 8:2). After incubation for 10 min with constant mixing, the Trp emission spectra were re-recorded.
Attenuated Total Reflection Infrared (ATR-IR) Spectroscopy-ATR-IR spectroscopy was performed as described previously (12). Briefly, ␣IIb-TM, solubilized in a 1:1 (v/v) mixture of 2-propanol/H 2 O, was mixed with lipid and dried under a stream of N 2 . The dried film was reconstituted in 5 mM HEPESbuffered D 2 O at pH 7.1 by vigorous vortexing. The peptide/ lipid suspension was then extruded 17 times using an Avestin liposofast mini extruder (Avestin Inc.) equipped with two stacked polycarbonate membranes with average pore diameter of 200 nm. The peptide-containing vesicles were deposited on the ATR crystal, gently spread with a Teflon bar to form a film, and dried under a gentle stream of N 2 . Infrared spectra were recorded on a Nicolet 4700 infrared spectrophotometer (Thermo-Electron Corp.) equipped with a DTGS detector and a ZnSe wire-grid polarizer. The internal reflection element was a zinc-selenide ATR crystal (80 ϫ 20 ϫ 3 mm) with an angle of 45°yielding 25 internal reflections. A total of 512 scans at polarizations of 0 and 90 o were collected for each sample. Spectra were recorded at 2 cm Ϫ1 resolution and analyzed using the OMNIC software package for peak deconvolution and area analysis. Helix orientation angle was calculated from the spectra as described previously (13) with the exception that the value used in this study for the crystal refractive index was 2.42 (ZnSe).
Fluorescence Anisotropy-Full-length ␣IIb␤3 in 10 mM HEPES buffer containing 60 mM N-octyl-␤-D-glucopyranoside, 0.5 mM CaCl 2 , 0.02% NaN, pH 7.5, was prepared as reported previously (14). Fluorescence polarization experiments were conducted on an ATF105 spectrofluorometer (Aviv Instrument, Inc.) using a 0.3-cm path length cuvette. Spectra were measured at 25°C using 1.0 nm slit widths. Excitation at 485 nm was used for the FITC-labeled peptide, and the emission maximum at 515 nm was monitored. Anisotropy measurements were recorded by titrating a 64 nM FITC-␣IIb-TM peptide solution with increasing concentrations of purified ␣IIb␤3 protein.
A dissociation constant and base-line parameters were derived from the fluorescence anisotropy signal/concentration isotherm using methods described previously (15).
SDS-PAGE and Immunoblotting-The contents of the elution fractions from the size exclusion chromatography were identified using precast SDS-polyacrylamide gels (4 -12% NuPAGE BisTris gels; Invitrogen). Before electrophoresis, each sample was incubated at 90°C for 7 min. Electrophoresis was carried out at room temperature with NuPAGE MOPS SDS running buffer (Invitrogen). Immunoblotting was performed after electrophoretically transferring proteins to nitrocellulose paper (0.45 m, Schleicher & Schuell). The paper was then incubated with the anti-␤ 3 monoclonal antibody SSA6. Primary antibody binding was detected using ECL Western blotting detection reagents after 6000-fold dilution of the stock solution, followed by fluorography using Kodak Biomax MS film.
TOXCAT-TM peptide dimerization in membrane bilayers was measured by the TOXCAT assay using the expression vector pccKAN kindly provided by Donald M. Engelman (Yale University, New Haven, CT) (16). After changing the EcoRV restriction site between the TM region and the malB gene in pccKAN to a BamHI site, the vector was digested with NheI-BamHI, and cDNA encoding integrin TM helices were ligated into the vector in-frame. The resulting plasmids were transformed into E. coli MM39 cells. CAT synthesis was assayed by using a CAT-enzyme-linked immunosorbent assay kit (Roche Applied Sciences), as described (4). Chimeric protein expression was quantified from immunoblots using a Personal Densitometer SI (Amersham Biosciences) and was used to compare CAT expression by the various constructs.
Platelet Function Assays-Turbidometric measurements of platelet aggregation were performed in a Chrono-Log Lumi-Dual Aggregometer as described previously (17). Briefly, platelet-rich plasma, prepared from human blood anti-coagulated with 0.1 volume of 0.13 M sodium citrate, was gel-filtered on Sepharose 2B (Amersham Biosciences) using an elution buffer containing 137 mM NaCl, 2.7 mM KCl, 1 mM MgCl 2 , 5.6 mM glucose, 0.35 mg/ml bovine serum albumin, 3.3 mM NaH 2 PO 4 , and 4 mM HEPES, pH 7.4. 400-l aliquots of the gel-filtered platelet suspension were stirred at 900 rpm in siliconized aggre-gometer cuvettes and were supplemented with human fibrinogen (Enzyme Research Laboratories) and CaCl 2 to final concentrations of 200 g/ml and 1 mM, respectively, prior to adding platelet agonists.
Binding of FITC-conjugated fibrinogen to gel-filtered platelets was measured by fluorescence activated cell sorting (FACS) as described previously (18). Following the addition of ADP or the ␣IIb-TM peptide to platelet suspensions, the platelets were fixed with 0.37% formalin in PBS buffer for 10 min, washed, and examined by FACS analysis.
Peptide-induced Cell Lysis-Three methods were used to assess peptide-induced cell lysis. First, peptide-induced lysis of unilamellar vesicles containing the dye Tb(III)/dipicolinic acid was measured as described previously (19). Melittin and buffer alone were positive and negative controls, respectively. Second, peptide-induced lysis of erythrocytes was measured as described previously (20). Briefly, 1% (v/v) suspensions of washed human erythrocytes were incubated with peptide in 4 mM HEPES buffer containing 137 mM NaCl, 2.7 mM KCl, 1 mM MgCl 2 , 5.6 mM glucose, 0.35 mg/ml bovine serum albumin, and 3.3 mM NaH 2 PO 4 , pH 7.4, at 37°C for 3 min. After centrifuging the samples at 14,000 rpm, the optical density of the supernatant was measured at 405 nm. Third, peptide-induced release of [ 14 C]serotonin from nonaggregating platelets was measured as described previously (17). Platelets were loaded with [ 14 C]serotonin, gel-filtered, and incubated with peptide for 30 min in the absence of stirring. The [ 14 C]serotonin content of 100-l aliquots of platelet-free supernatant was then measured and compared with the [ 14 C]serotonin content of a 100-l aliquot of the original platelet suspension.
Rupture Force Spectroscopy (Laser Tweezers)-A custombuilt laser tweezers setup was used to measure the strength of fibrinogen and osteopontin binding to CHO cells or platelets (21,22). Purified human fibrinogen and recombinant osteopontin (OPN) were covalently bound to 1.87-m carboxylate-modified latex beads. An individual platelet was manually attached to a 5-m diameter silica pedestal coated with polylysine, whereas CHO cells were permitted to settle and attach spontaneously to the polylysine-coated bottom of the chamber. A ligand-coated bead, trapped by the laser light, was brought into proximity of an immobilized cell, oscillated at 10 Hz, and then brought into repeated intermittent contact with the cell by micromanipulation. Data collection was initiated at the first contact. Rupture forces following repeated contacts were collected into 5 pNor 10 pNwide bins. The percentage of events in a particular bin represented the probability of rupture events at that tension. Optical artifacts observed with or without trapped latex beads produced signals that appear as forces below 10 pN; rupture forces in this range were not considered when data were analyzed.
Transmission Electron Microscopy (TEM)-Electron microscopy of purified ␣IIb␤3 heterodimers was performed as described previously (23). Rotary-shadowed samples were prepared using a modification of standard procedures (24) by spraying a dilute solution of molecules in a volatile buffer (0.05 M ammonium formate) and glycerol (30 -50%) onto freshly cleaved mica and shadowing with tungsten in a vacuum evaporator (Denton Vacuum Co., Cherry Hill, NJ). All specimens were examined in an FEI/Philips 400 electron microscope (Philips Electronic Instruments Co., Mahwah, NJ), operating at 80 kV and at a magnification of 60,000.

RESULTS
Synthesis and Characterization of an ␣IIb-TM Peptide-We prepared a 22-residue peptide, designated ␣IIb-TM, encompassing residues Trp 968 -Lys 989 of the ␣IIb-TM domain, a length we determined previously to be optimal for ␣IIb-TM domain homodimerization (4). Two lysine residues were added at each terminus of the peptide to enhance its solubility, inhibit the formation of peptide aggregates, and facilitate its insertion into membranes (25) ( Table 1). We also prepared mutant peptides in which one or both of the Gly residues of the critical ␣IIb GX 3 G motif were changed to Leu, as well as the unrelated peptide MS-1, a model membrane peptide that forms dimers and trimers in membranes (26).
CD spectroscopy of ␣IIb-TM (Fig. 1, a and b) and LX 3 L (not shown) in micelles (2.5 mM DPC, CMC 1.0 mM) and phospholipid vesicles (500 M phospholipid, POPC/POPG 7:3) revealed minima at 208 and 222 nm, confirming that both peptides assume an ␣-helical secondary structure in micelles and phospholipid bilayers. Furthermore, analytical ultracentrifugation demonstrated that both ␣IIb-TM and FITC-labeled ␣IIb-TM assemble into homodimers in zwitterionic C14-betaine and DPC micelles, as we had shown previously for a peptide that encompassed the ␣IIb-TM and cytoplasmic domains (5). The pK d values for ␣IIb-TM and FITC-␣IIb-TM in betaine micelles (expressed in mole fraction units, the concentration of peptide in the detergent micellar phase) were 1.8 Ϯ 0.3 and 3.0 Ϯ 1.0, respectively (supplemental Fig. 1).
Next, the propensity of ␣IIb-TM to insert into phospholipid vesicles was examined using fluorescence spectroscopy. Measurement of the maximum fluorescence emission ( max ) of the Trp residues in ␣IIb-TM revealed a shift from 351 to 340 nm as the peptide inserted into the less polar environment of the lipid vesicle (Fig. 1c). A max value of 340 nm is consistent with emission from a Trp located in the polar head group region of a bilayer (12). Furthermore, at a peptide concentration of 2.0 M, the shape of the binding curve exhibited near linear titration behavior at low lipid concentrations, indicating that the peptide binds very tightly to the vesicles with an estimated K d value in the low micromolar range. Although the Trp fluorescence experiments indicated that ␣IIb-TM binds to POPC/POPG vesicles, polarized ATR-IR spectroscopy was used to confirm that ␣IIb-TM is capable of forming a TM ␣-helix in this environment. ATR-IR spectroscopy exploits the fact that in an ordered sample a given bond will absorb infrared radiation differentially, depending on the polarization of the light and the angle at which the bond is oriented relative to the polarized light. If the secondary structure of a peptide is known, the dichroic ratio (R ATR ) of the amide-I absorbance when the incident light is polarized at 0°to the amide absorbance when the light is polarized at 90°can be used to calculate the angle between the helical axis of the peptide and the bilayer normal. The amide-I vibration observed at 1656 cm Ϫ1 indicated that ␣IIb-TM adopted primarily an ␣-helical conformation in the POPC/POPG bilayers, consistent with the CD data (Fig. 1d). The R ATR  The ␣IIb-TM Peptide Binds to ␣IIb␤3 in Micelles and Membranes-Two approaches were taken to determine whether ␣IIb-TM can associate with full-length ␣IIb␤3. First, FITC-␣IIb-TM, either alone or mixed with purified ␣IIb␤3, was loaded onto a Superdex 200 HR 10/30 size exclusion column, and the amount of fluorescently labeled peptide in the eluted fractions was measured. As shown in Fig. 2a, the fluorescence intensity of eluted fractions decreased progressively when increasing amounts of ␣IIb␤3 were present, and a new peak of early eluting fluorescence appeared. SDS-PAGE and immunoblotting of the early peak revealed that it contained ␣IIb␤3, confirming that the elution of FITC-␣IIb-TM was affected by its association with ␣IIb␤3. By contrast, when the gel filtration was performed using LX 3 L instead of ␣IIb-TM, there was no decrease in the LX 3 L peak in the presence of increasing amounts of ␣IIb␤3 (data not shown), implying that the GX 3 G motif in ␣IIb-TM plays a critical role in its interaction with ␣IIb␤3.
Second, the physical association of FITC-␣IIb-TM with ␣IIb␤3 was detected by measuring the fluorescence anisotropy of peptide/␣IIb␤3 mixtures. As shown in Fig. 2b, measurement of FITC-␣IIb-TM anisotropy as a function of ␣IIb␤3 concentration resulted in a binding isotherm with an apparent K d of 2.6 Ϯ 1.1 ϫ10 Ϫ5 in mole fraction units, consistent with high affinity binding of ␣IIb-TM to full-length ␣IIb␤3.
The experiments described above demonstrated that ␣IIb-TM interacts with full-length ␣IIb␤3 in detergent micelles. The TOXCAT assay was then used to address whether the peptide can bind to the TM domain of ␣IIb in a membrane bilayer (16). In TOXCAT, a chimeric protein consisting of an N-terminal ToxRЈ DNA-binding domain, a C-terminal maltose-binding protein domain, and an interposed TM domain is expressed in the inner membrane of E. coli (16). TM domainmediated dimerization of the chimeric protein drives the transcriptional activation of a CAT reporter gene. The ability of exogenously added ␣IIb-TM to inhibit CAT synthesis was used as a measure of its association with the TM domain of the chimeric protein. ␣IIb-TM inhibited CAT synthesis driven by a chimeric protein containing the wild type ␣IIbTM domain by 44 Ϯ 3% (Fig. 3a). Replacing the first glycine in the GX 3 G motif of ␣IIb-TM with Leu (LX 3 G) significantly impaired the ability of the peptide to inhibit CAT synthesis (30 Ϯ 1%, p Ͼ 0.02), and replacing both glycines with Leu (LX 3 L) reduced inhibition to that of the solvent control (4.8 Ϯ 0.1 versus 4.7 Ϯ 0.1%, respectively). Thus, these data demonstrate that not only can ␣IIb-TM associate the ␣IIb-TM domain in a lipid bilayer but that the GX 3 G motif of the peptide plays an essential role in this interaction.
TOXCAT was also used to confirm that the interaction of ␣IIb-TM with the ␣IIb-TM domain was specific. TM domains from the integrin subunits ␤3, ␣2, and ␤1 were substituted for the ␣IIb domain in the TOXCAT chimera, and the ability of the exogenous ␣IIb-TM peptide to inhibit CAT synthesis was measured. As shown in Fig. 3b, ␣IIb-TM again significantly  inhibited CAT synthesis induced by the ␣IIb-TM domain-containing chimera, but there was no change in CAT synthesis when the chimera contained the ␤1 TM domain, and there were nonsignificant increases in CAT synthesis when chimeras contained either the ␤3 or the ␣2 TM domains (145 and 118%, p Ͻ 0.11 and p Ͻ 0.16, respectively).
The ␣IIb-TM Peptide Induces Platelet Aggregation by Directly Activating ␣IIb␤3-The ability of ␣IIb-TM to interact with the ␣IIb-TM domain in situ in platelets was assessed by its ability to induce platelet aggregation. Platelets undergo fibrinogen-dependent, ␣IIb␤3-mediated aggregation when stimulated by agonists such as ADP, a process preceded by a change in platelet shape from a disk to a spiny sphere (17). Adding micromolar concentrations of ␣IIb-TM to suspensions of gel-filtered human platelets rapidly induced platelet aggregation (Fig. 4a). Like ADP-stimulated aggregation, ␣IIb-TM-induced aggregation was inhibited by EDTA. But unlike ADP, peptide-induced aggregation was unaffected by the adenylyl cyclase agonist PGE 1 or the ADP scavenger apyrase and was not accompanied by platelet shape change (Fig. 4, a-c). This suggests that ␣IIb-TM induced platelet aggregation independent of platelet signal transduction or secreted ADP by interacting directly with ␣IIb.
To verify that ␣IIb-TM interacts specifically with ␣IIb, the ability of ␣IIb-TM LX 3 L and the unrelated peptide MS1 to induce platelet aggregation was tested. As shown in Fig. 4d, neither 3 M LX 3 L nor MS1 induced platelet aggregation, although increasing the LX 3 L concentration nearly 7-fold to 20 M had a partial effect. Thus, these results indicate that the interaction of ␣IIb-TM with ␣IIb in platelets is specific and requires the presence of a GX 3 G motif.
Because fibrinogen binding to ␣IIb␤3 is a prerequisite for platelet aggregation (1), ␣IIb-TM-induced binding of FITC-labeled fibrinogen to platelets was assessed using flow cytometry (18). As shown by the histograms in Fig. 5, both 20 M ADP and 2.0 M ␣IIb-TM induced FITC-fibrinogen binding to gel-filtered platelets. In both instances, fibrinogen binding was prevented by EDTA, consistent with fibrinogen binding to ␣IIb␤3.
Platelets express a second ␤ 3containing integrin, ␣v␤3, that mediates agonist-stimulated platelet adhesion to surfaces coated with the matrix protein OPN (27). Because agonist-induced binding of OPN to ␣v␤3 on platelets can be specifically measured using laser tweezers (22), this methodology was used to address whether ␣IIb-TM can also activate ␣v␤3. However, in contrast to ADP, ␣IIb-TM did not induce a peak of specific rupture forces between platelets and beads coated with OPN (supplemental Fig. 2). Thus, these results demonstrate that ␣IIb-TM specifically recognizes the TM domain of ␣IIb despite the presence of the closely related integrin ␣ subunit ␣v.
Although ␣IIb-TM-induced platelet aggregation was not inhibited by preincubating platelets with PGE 1 or apyrase, it is still conceivable that  the aggregation tracings resulted from peptide-induced platelet lysis, rather than platelet aggregation. This possibility was addressed in three ways (supplemental Fig. 3). First, ␣IIb-TM did not induce leakage of the small fluorescent molecule Tb(III)/dipicolinic acid from phospholipid vesicle at peptide/ lipid ratios up to 1:10. Second, the peptide did not cause lysis of erythrocytes at concentrations as high as 10 M. Third, the peptide did not cause the leakage of [ 14 C]serotonin from platelet-dense granules in the absence of platelet aggregation.
␣IIb-TM Activates Recombinant ␣IIb␤3 Expressed by CHO Cells-The ability of ␣IIb-TM to activate ␣IIb␤3 independent of signal transduction was corroborated by measuring fibrinogen binding to recombinant ␣IIb␤3 expressed by CHO cells using laser tweezers-based force spectroscopy (28). The results of these experiments are shown in Fig. 6. CHO cells readily express recombinant ␣IIb␤3 that cannot be activated by cellular agonists. Thus, in the absence of ␣IIb-TM, rupture forces between fibrinogencoated beads and ␣IIb␤3-expressing CHO cells were Ͻ20 pN and resulted from nonspecific protein-protein interactions (Fig. 6a) (28). By contrast, when the cells were incubated in Mn 2ϩ -containing media, up to 10% of rupture forces were Ͼ20 pN, with a characteristic peak at 70 pN (Fig. 6b). The effect of adding ␣IIb-TM was similar to Mn 2ϩ ; in this case Ϸ7% of rupture forces were Ͼ20 pN with a rupture force maximum at 70 pN (Fig. 6c). Furthermore, there was a substantial reduction in the rupture force peak when ␣IIb-TM was added along with either the ␣IIb␤3 antagonist abciximab (Fig. 6d) or the divalent cation chelator EDTA (Fig. 6e), con-firming that the peak resulted from the rupture force of fibrinogen bound to activated ␣IIb␤3.
TEM of ␣IIb␤3 in the Presence or Absence of ␣IIb-TM-Integrins are inactive when their TM-containing stalks are in proximity and active when the stalks separate (23,29). Thus, as we showed previously, TEM of inactive purified ␣IIb␤3 in buffer containing octyl glucoside and 1 mM CaCl 2 revealed that the majority of the ␣IIb␤3 molecules had a closed configuration with their stalks touching at the their tips (Fig. 7, CaCl 2 ; Table 2) (23). By contrast, when ␣IIb-TM was present, most of the ␣IIb␤3 molecules had an open configuration with separated stalks (Fig. 7, CaCl 2 ϩ␣IIb-TM). These observations are consistent with the notion that ␣IIb-TM activates ␣IIb␤3 by disrupting the TM heterodimer maintaining the integrin in its inactive state. It is noteworthy that when ␣IIb␤3 is activated by Mn 2ϩ (23), nearly 40% of the open molecules are present as dimers and higher order oligomers. However, when ␣IIb␤3 was shifted to an open configuration by ␣IIb-TM, only 0.8% of the molecules were present as dimers, and no higher order oligomers were observed. This result suggests that although ␣IIb-TM is able to induce an active ␣IIb␤3 conformation, oligomerization of the active molecules cannot occur because the required oligomerization site on ␣IIb is occupied by the peptide.

DISCUSSION
The ability of integrins to bind to ligands with high affinity is regulated by cell metabolism (1). This is most apparent for inte- grins on circulating blood cells where integrin function is desirable under only specific circumstances; it is less apparent for adherent cells but is equally important, because cycles of integrin activity and inactivity are necessary for cell movement (30). Integrins reside on cell membranes in an equilibrium between low affinity (inactive) and high affinity (active) conformations (31). The agonist-mediated shift between these conformations appears to involve the disruption of intra-molecular interactions involving subunit TM and/or cytoplasmic domains (1). The most extensively studied intra-molecular interaction involves conserved membrane-proximal cytoplasmic domain sequences (32). These sequences are thought to form a low affinity activation-constraining "clasp," a notable feature of which is a salt bridge between a conserved ␣ subunit Arg and a conserved ␤ subunit Asp (33). There is also mounting evidence that TM domain sequences participate in integrin regulation. These domains, including those of ␣IIb and ␤3, can undergo both heteromeric (2,34,35) and homomeric (5) interactions, but it is likely that heteromeric interactions are paramount in constraining integrin activity. In the case of ␣IIb␤3, for example, single disruptive amino acid replacements in the TM domain of either ␣IIb or ␤3 are sufficient to induce constitutive ␣IIb␤3 activity when the mutants are expressed in tissue culture cells (36).
Despite the mutational data cited above, proteins corresponding to the wild type ␣IIb and ␤3 TM and cytoplasmic domains undergo homomeric, not heteromeric, association in micelles (5). This raises the possibility that an exogenous ␣IIb-TM peptide would bind in a homomeric manner to the ␣IIb-TM domain and cause platelet aggregation by disrupting the constraining ␣IIb/␤3 TM domain heterodimer. However, TM segments are inherently hydrophobic and would not be expected to be soluble in aqueous buffers. Nonetheless, peptides corresponding to such segments can be solubilized in aqueous buffers by appending polar residues (e.g. lysine) to their N and C termini (25,37,38). Such peptides insert spontaneously into lipid membranes where they fold into ␣-helices and assume their native monomeric or oligomeric states (25,38). Based on these observations, we synthesized a 22-residue ␣IIb-TM peptide and added two lysine residues to each end to enhance its solubility. As expected, the peptide was soluble in aqueous buffer and rapidly inserted into small unilamellar phospholipid vesicles where it assumed an ␣-helical conformation. Furthermore, the peptide formed homodimers in detergent micelles, bound with high affinity to purified ␣IIb␤3, and as shown by the TOXCAT assay, interacted specifically with the ␣IIb-TM domain rather than the TM domains of ␤3, ␤1, or ␣2.
The plasma membrane of platelets is a mosaic of at least 83 different proteins (39), including five different integrins (1). Nonetheless, ␣IIb␤3 is the only platelet protein known to support platelet aggregation. Thus, the ability of ␣IIb-TM to cause platelet aggregation implies that the peptide can recognize ␣IIb in this complex environment. Normally, ␣IIb␤3 activity is regulated by "inside-out" signals initiated by specific platelet agonists, several of which, notably ADP and thromboxane A 2 , are either released or generated by the platelets themselves (40). Thus, it is essential to be certain that ␣IIb-TM did not activate ␣IIb␤3 by causing inside-out signaling. Agonist-stimulated platelet function can be prevented when signal transduction is interrupted by increases in the platelet content of cAMP. This is usually accomplished by exposing platelets to the prostaglandins PGI 2 or PGE 1 . However, preincubating platelets with PGE 1 had only a minimal effect on ␣IIb-TM-induced aggregation, strong evidence that the effect of the peptide was independent of signal transduction. Furthermore, as would be predicted from the lack of a PGE 1 effect, the ADPase apyrase did not inhibit ␣IIb-TM-induced aggregation nor was ␣IIb-TM-induced aggregation accompanied by platelet shape change, an agonist-induced phenomenon.
Turbidometric platelet aggregation measures the increase in light transmission that occurs when platelets in suspension aggregate. However, platelet lysis, by increasing light transmission, could be mistaken for aggregation and could be an alternative explanation for the aggregometry results discussed  above. But we found that ␣IIb-TM did not cause the release of serotonin from nonaggregating platelets, as would occur if the peptide had disrupted the integrity of the platelet plasma membrane. Nor did it disrupt the integrity of unilamellar liposomes or cause red cell lysis. Thus, platelet lysis is not a tenable explanation for aggregation tracings shown in Fig. 4. Mutating the GX 3 G motif in ␣IIb-TM impaired its ability to bind to the ␣IIb-TM domain and to activate ␣IIb␤3. First recognized as a framework for the homomeric association of the glycophorin A TM domain, GX 3 G has been consistently identified as the most over-represented sequence motif in TM domain data bases (41). In glycophorin A, GX 3 G permits extensive backbone-backbone contacts at the homodimer interface because a groove created by the glycines in one monomer packs tightly against a ridge created by the side chains of the residues that succeed each glycine in the other. Previously, we expressed an extensive series of single site mutations of the ␣IIb-TM domain in TOXCAT and found that its GX 3 G motif was essential for dimerization (4). The mutagenesis data were then used to construct an atomic model for an ␣IIb-TM domain dimer that placed GX 3 G in the dimerization interface, although in this model, residues preceding the glycines formed the ridge that packed against the glycine-created groove (42). GX 3 G mutations in ␣IIb also resulted in constitutive ␣IIb␤3 activation when ␣IIb␤3 was expressed in CHO cells, implying that GX 3 G is not only involved in homomeric ␣IIb-TM domain interactions, but in heteromeric ␣IIb-␤3 TM domain interactions as well (6). That being the case, the ability of ␣IIb-TM to activate ␣IIb␤3 indicates that the tendency of the ␣IIb-TM domain to interact homomerically is substantially greater than its tendency to interact heteromerically with ␤3, a suggestion consistent with the behavior of small proteins corresponding to the ␣IIb and ␤3 TM and cytoplasmic domains. It also suggests that sequestration of the ␣IIb GX 3 G motif in an ␣II␤/␤3 heterodimer is an important factor for maintaining ␣IIb␤3 in an inactive state.
There is a substantial body of experimental evidence that integrins are inactive when their TM and/or cytoplasmic domains are in proximity. Thus, replacing the cytoplasmic domains of ␣L␤2 and ␣5␤1 with complementary acidic and basic peptides maintains the integrins in a low affinity state (43,44). Likewise, using TEM, we found that the tips of the stalks of most ␣IIb␤3 molecules purified from unstimulated platelets appeared to be touching in the absence of ␣IIb-TM. But when the peptide was present at a concentration sufficient to activate ␣IIb␤3 in platelets and CHO cells, the majority of molecules had separated stalks. These observations are entirely consistent with the push-pull hypothesis for integrin regulation and imply that TM domain interactions alone are at least sufficient to regulate integrin activity.
Clusters of ␣IIb␤3 molecules have been detected on thrombin-stimulated platelets (45), and we found ␣IIb␤3 present in patches on the surface of CHO cells expressing constitutively active ␣IIb␤3 because of mutations in the ␣IIb or ␤3 TM domains (6,36). TEM images of Mn 2ϩ -activated ␣IIb␤3 revealed that nearly 40% of the molecules formed dimers, trimers, and higher order oligomers via interactions involving the distal ends of ␣IIb and ␤3 stalks (23). It is noteworthy then that ␣IIb␤3 oligomers were rarely seen when purified ␣IIb␤3 was incubated with ␣IIb-TM, even though most of the molecules were in an open active conformation. Because small proteins corresponding to the ␣IIb and ␤3 TM and cytoplasmic domains tend to undergo homomeric, rather than heteromeric, association (5), it is likely that the clustering of Mn 2ϩactivated ␣IIb␤3 resulted from the homomeric interaction of these domains in the intact protein. The paucity of ␣IIb␤3 oligomers when ␣IIb␤3 is incubated with ␣IIb-TM is consistent with this conclusion because, in the presence of the peptide, there would be no free ␣IIb-TM domains to support polymerization. Whether homomeric ␣IIb and ␤3 TM domain association participates in ␣IIb␤3 clustering on agonist-stimulated platelets is not known. Nonetheless, these observations suggest that such interactions can occur and could be responsible for the formation of the signaling complexes that mediate signal transduction following ligand binding to integrins.
In summary, we report that a lysine-flanked peptide corresponding to the ␣IIb-TM domain not only binds to ␣IIb␤3 when the purified integrin is present in detergent micelles and in bacterial membranes, but also induces platelet aggregation by interacting directly with ␣IIb␤3 in platelet membranes. Thus, these observations demonstrate that TM domain interactions regulate integrin function in situ and that it is possible to target intra-membranous protein-protein interactions in a way that can have functional consequences.