Swing-Out of the β3 Hybrid Domain Is Required for αIIbβ3 Priming and Normal Cytoskeletal Reorganization, but Not Adhesion to Immobilized Fibrinogen

Structural and functional analyses of integrin αIIbβ3 has implicated swing-out motion of the β3 hybrid domain in αIIbβ3 activation and ligand binding. Using data from targeted molecular dynamics (TMD) simulations, we engineered two disulfide-bonded mutant receptors designed to limit swing-out (XS-O). XS-O mutants cannot bind the high Mr ligand fibrinogen in the presence of an activating mAb or after introducing mutations into the αIIb subunit designed to simulate inside-out signaling. They also have reduced capacity to be “primed” to bind fibrinogen by pretreatment with eptifibatide. They can, however, bind the small RGD venom protein kistrin. Despite their inability to bind soluble fibrinogen, the XS-O mutants can support adhesion to immobilized fibrinogen, although such adhesion does not initiate outside-in signaling leading to normal cytoskeletal reorganization. Collectively, our data further define the biologic role of β3 hybrid domain swing-out in both soluble and immobilized high Mr ligand binding, as well as in priming and outside-in signaling. We also infer that swing-out is likely to be a downstream effect of receptor extension.


Introduction
Integrins belong to a cell adhesion molecular family that mediates cell-cell and cell-extracellular matrix interactions [1]. They signal bidirectionally through long-range allosteric changes, with proteins binding to the cytoplasmic domains initiating insideout signaling and ligands binding to the extracellular domain initiating outside-in signaling [2].
Integrin aIIbb3 is expressed on megakaryocytes and platelets and on cells early in hematopoietic stem cell development [3]. Platelet aIIbb3 contributes to hemostasis by supporting platelet aggregation at sites of vascular injury and pathological thrombosis by supporting platelet aggregation in atherosclerotic arteries, with the latter leading to myocardial infarction and stroke [4,5]. Physiological agonists such as ADP or thrombin initiate inside-out platelet signaling and induce aIIbb3 conformational changes that result in the binding of multimeric ligands, such as fibrinogen and von Willebrand factor. The simultaneous binding of either of these ligands to aIIbb3 receptors on two different platelets then results in platelet aggregation via crosslinking of platelets. Ligand binding also initiates outside-in signaling, leading to cytoskeletal reorganization and enhanced secretion [6]. The lifelong bleeding disorder Glanzmann thrombasthenia is an autosomal recessive disease in which patients either lack or have abnormal aIIbb3 receptors [3].
Similar to other integrins, activation of, and ligand binding to aIIbb3 is associated with large-scale global conformational rearrangements [2,[7][8][9][10][11][12][13]. Extensive structural and functional data have shown that aIIbb3 exists in at least three different conformations: a bent conformation with a closed headpiece (i.e., the b3 hybrid domain abuts the aIIb b-propeller), an extended conformation with a closed headpiece, and an extended conformation with an open headpiece (i.e., the b3 hybrid domain swings out from the aIIb b-propeller by 60-70u). Although all three conformations are capable of binding small ligands, the bent, closed conformation has low affinity for macromolecular physiologic ligands whereas both the extended, closed and extended, open conformations are associated with higher affinity for these ligands. The transition from the bent to the extended conformation, and from the closed to open conformation, can be achieved by adding peptides that contain the cell recognition Arg-Gly-Asp (RGD) sequence, which bind to the ligand binding site at the junction between the two head domains [8,13]. These peptides are thought to induce the open conformation by altering the structure around the b3 metal binding sites, leading to the downward movement of the a7 helix of the b I domain (b3 Inserted domain) (which connects the b I domain to the hybrid domain), which, in turn, initiates the swing-out motion of the hybrid domain away from aIIb [8]. Initial experimental support for the swing-out conformation having high ligand affinity came from data demonstrating that stabilizing the open headpiece conformation by introducing a disulfide bond in the b I domain [14] or engineering a new N-glycosylation site into the hybrid-b I domain interface to wedge the hybrid domain away from the b I domain [15] creates constitutively active receptors that do not require inside-out signaling to induce ligand binding.
In a previous study we employed targeted molecular dynamics (TMD) simulations to study the pathway of the swing-out transition from the unliganded, closed to the liganded, open conformation of b3 integrins [19]. Stereochemically feasible pathways with candidate intra-domain and inter-domain interactions responsible for b3 integrin activation were explored and specific contacts were identified that are broken early during the swing-out process. Thus, creation of a new disulfide bond between these residues would be expected to prevent the normal swing-out mechanism. In this study we assessed the effect of creating two different mutant receptors, each of which contained two new cysteine residues designed to create a new disulfide bond that would prevent swing-out. Our data support and extend those of Kamata et al. [17] who studied a similar mutant receptor (aIIbb D319C/V359C). We also observed inhibition of the binding of large, but not small, activation-dependent soluble ligands. In addition, we were unable to rescue the abnormality in soluble ligand binding by introducing mutations into the aIIb subunit designed to simulate inside-out signaling. Moreover, we found that preventing swing-out prevented ligand binding induced by ''priming'' the receptor with low molecular weight ligands, but did not affect the receptor's ability to support adhesion to immobilized fibrinogen. Finally, we found that the mutations inhibited outside-in signaling after cell adhesion.

Targeted molecular dynamics (TMD) simulations
TMD simulations were performed to simulate the swing-out motion of the b3 hybrid (and PSI) domains as previously described [19]. The head and upper leg regions of aIIbb3 were simulated, including the b-propeller and thigh domains of aIIb and the b I, hybrid, and PSI domains of b3. Amino acid numbering is based on the mature protein without the leader sequence.

Assessment of disulfide bond formation from XS-O mutants by mass spectrometry
Cells expressing XS-O mutants and normal aIIbb3 were lysed using 1% Triton X-100, followed by immunoprecipitation with mAb 10E5. After resolving the proteins on a non-reduced SDS-PAGE gel, the purified protein bands from XS-O mutant 321/358 corresponding to the disulfide-bonded aIIbb3 heterodimer and the individual aIIb or b3 subunits were cut out and digested overnight with trypsin (2 mg/ml; Promega Madison, WI). Comparable purified protein bands from XS-O mutant 321/360 were digested with trypsin overnight, followed by digestion with endoproteinase Asp-N (2 mg/ml; Roche; Basel, Switzerland) for another two days. Samples were analyzed by LC-MS/MS using a nano-reversed phase column coupled online with an LTQ-Orbitrap mass spectrometer (ThermoFisher; Waltham, MA).

Spreading of adherent cells expressing normal aIIbb3 and XS-O mutants
Chamber slides were coated with 20 mg/ml fibrinogen for 1 hr and then washed and blocked with HBMT. Cells (1.5610 5 cells/ ml, 200 ml) in HBMT buffer with 2 mM Ca 2+ and 1 mM Mg 2+ were added to each chamber. In some experiments, 1 mM Mn 2+ was used to activate the cells. Cells were allowed to adhere to fibrinogen for 2 hr at 37uC. After washing with PBS, adherent cells were fixed with 4% formaldehyde in PBS and permeabilized with 0.5% Triton X-100 in PBS. After washing, adherent cells were stained with Alexa488-7H2 for 20 min, washed, and stained for F-actin with rhodamine phalloidin (Cytoskeleton Inc. Denver, USA) for 30 min. The cells were then washed and dried at room temperature (RT). A drop of anti-fade mounting medium (Dako, Carpinteria, CA) was added to the center of the chamber, a cover slip was placed on top of the cells, and the cover slip was sealed with nail polish. The specimens were imaged using a wide-field fluorescence/brightfield/DIC microscope (Zeiss) with a 40X objective. Image J (NIH) was used to analyze cell spreading.
For double staining of vinculin and actin, cells were allowed to adhere for 1 hr to wells pre-coated with 20 mg/ml fibrinogen at RT. After fixation and permeabilization, cells were reacted with an anti-vinculin antibody, followed by staining with a secondary FITC-anti-mouse antibody and TRITC-conjugated phalloidin simultaneously according to the manufacturer's protocol (Millipore). The specimens were imaged using a DeltaVision Image Restoration Microscope (Olympus) with a 60X objective. For time-lapse microscopy, live cells were added to cover-glass chamber slides pre-coated with 20 mg/ml fibrinogen at RT and differential interference contrast (DIC) images were obtained every 2 min for 1 hr using the DeltaVision microscope with a 20X objective.

Clot retraction
Cells were harvested with trypsin and EDTA, washed sequentially with culture medium and HBMT buffer without cations, and resuspended to 6 X 10 6 cells/ml. A 0.5 ml aliquot of the cell suspension was placed in a 60 X 8 mm glass cuvette and then CaCl 2 (5 mM final), fibrinogen (200 mg/ml final) and thrombin (2 U/ml final) were added and the cells were mixed. The cuvettes were maintained at 37uC and photographed at timed intervals for up to 4 hrs.

TMD stimulation
In the unliganded, closed aIIbb3 structure, aIIbK321 in the bpropeller domain and b3E358 in the hybrid domain form a salt bridge at the interface between the two subunits (PDB: 3FCS) [7] ( Fig. 1A and C). Their Cb atoms are 7.7 Å apart. In the liganded, open structure, in contrast, their Cb atoms are 32.6 Å apart ( Fig. 1B and D) (PDB: 3FCU) [7]. This salt bridge breaks very early in the swing-out motion and thus we reasoned that preventing these residues from separating would stop the earliest conformational changes associated with swing-out. Similarly, in the unliganded, closed aIIbb3 structure, aIIbK321 is close to b3R360 in the hybrid domain, with the distance between Cb atoms being 11.7 Å (PDB: 3FCS) ( Fig. 1A and C). In the liganded structure, the Cb atom distance increases to 38.7 Å (PDB: 3FCU) ( Fig. 1B and E). As a result, we decided to make two different mutant receptors, aIIbK321C/b3E358C and aIIbK321C/ b3R360C.

Heterodimer disulfide bond formation in both XS-O mutants is supported by SDS-PAGE analysis and mass spectroscopy
As judged by the binding of anti-aIIbb3 mAb 10E5, aIIbb3 expression was similar on cells expressing the XS-O mutants and cells expressing normal aIIbb3 ( Fig. 2A). To assess whether the mutant proteins contained the expected disulfide bonds, we immunoprecipitated the biotin surface-labeled proteins and then analyzed them by SDS-PAGE and streptavidin staining. aIIb (with an Mr of 140 kD) and b3 (with an Mr of 95 kD) were immunoprecipitated with anti-aIIbb3 complex-specific mAb 10E5 from normal aIIbb3 transfected HEK293 cells (Fig. 2B). In contrast, an additional band of Mr ,250 kD, corresponding to the Mr expected if aIIb is covalently coupled to b3 was observed in samples of the two XS-O mutants. To assess whether the Mr 250 kD band contained both aIIb and b3, we performed immunoblotting with mAbs specific for each subunit. Both mAb PMI-1 (anti-aIIb) and 7H2 (anti-b3) reacted with the 250 kD band (Fig. S2), demonstrating the presence of both subunits. As expected, with reduction ( Fig. 2B), the normal aIIbb3 showed a decrease in Mr of aIIb and an increase in Mr of b3. With both mutants, reduction led to both complete loss of the Mr 250 kD band, indicating that it depended on disulfide cross-linking, and enhanced intensity of the aIIb and b3 subunits. The effects of different concentrations of DTT and different times of incubation on the XS-O mutants were assessed by immunoblotting with the aIIb-specific mAb PMI-1 (Fig. S3).
Mass spectroscopy of a trypsin digest of mutant 321/358 demonstrated a peak in the main spectrum at m/z = 488.2569, which corresponds to the m/z expected from the unique peptide resulting from a disulfide linkage between the peptides VELCVR and CLAEVGR (MH 4 3+ ) predicted from the creation of the engineered disulfide bond (Fig. 2C). Since the 321/360 mutant converted the trypsin cleavage site amino acid R360 to C, an alternative cleavage strategy was employed to assess this mutant, combining trypsin and Asp-N. A peptide was identified in the main spectrum at m/z = 653.2630 corresponding to the unique disulfide bonded peptide EVC-CLA (MH 4 + ) predicted from the creation of the engineered disulfide bond in mutant 321/360 (Fig. 2D). The identities of these two peptides were confirmed by tandem mass spectrometry. Neither peptide was present in normal aIIbb3. In contrast to normal aIIbb3, however, both mutants bound significantly less PAC-1 in the presence of PT25-2 (321/358: NNFI = 5.961.9, n = 5, p,0.001; 321/360: NNFI = 9.463.3, n = 5, p,0.001). DTT treatment partially or completely rescued the ligand binding ability of the XS-O mutants. Data with fibrinogen binding were similar to those with PAC-1 binding (Fig. 3A, right).

XS-O mutants
The aIIbF992A/F993Ab3 mutant (aIIbFFb3) has been reported to be constitutively active as a result of the mutations disrupting the association of the membrane-proximal portions of the a and b subunit cytoplasmic domains [31][32][33]. Similar disruption of the a and b subunit cytoplasmic domains is proposed to occur with inside-out activation of the receptor [34]. To assess whether the aIIbFF mutations could rescue the XS-O mutants' ability to bind high Mr ligands, we co-expressed the aIIbFF mutant with normal b3 and the aIIb mutant K321C, F992A, F993A with either the b3E358C or the b3R360C mutant. Immunoblot analysis of the surface membrane-labeled receptors demonstrated that crosslinked heterodimers containing the mutant aIIb and b3 chains were expressed on the cell surface and could be immunoprecipitated by mAb 10E5 (Fig. S4).
Consistent with our previous findings, there was little PAC-1 binding to cells expressing normal aIIbb3 in the absence of stimulation. In sharp contrast, PAC-1 binding to the aIIbFFb3 mutant in the absence of PT25-2 was comparable to that of cells expressing normal aIIbb3 in the presence of PT25-2 (NNFI: 23.463.9 vs 25.865.4) (Fig. 3B left). In the absence of stimulation, cells expressing both the aIIbFF mutations and the XS-O mutations bound somewhat more PAC-1 than did the normal aIIbb3, but much less than the aIIbFFb3 mutant. Adding PT25-2 greatly enhanced PAC-1 binding to normal aIIbb3, but had only modest impact on PAC-1 binding to the aIIbFFb3 mutant or the combined mutants. DTT treatment partially or completely rescued the ligand binding ability of the combined mutants. The fibrinogen binding results paralleled the PAC-1 binding results (Fig. 3B right).
Kistrin-induced AP5 binding to XS-O mutants 321/358 and 321/360 is less than to normal aIIbb3 or the FF321/ 358 and FF321/360 mutants Kistrin is a small RGD-containing snake venom protein. In the absence of DTT, kistrin bound similarly to normal aIIbb3, both XS-O mutants, and both FFXS-O mutants; it bound at higher levels, however, to the aIIbFFb3 mutant (P,0.001, n = 4) (Fig. 4A). DTT treatment enhanced kistrin binding to normal aIIbb3 and the four XS-O mutants, but the percentage increase was much less than with PAC-1 or fibrinogen binding (Fig. 3).
Since XS-O mutants can bind kistrin similarly to normal aIIbb3, we further evaluated anti-LIBS AP5 binding induced by kistrin. Relatively little AP5 bound to normal aIIbb3 or the XS-O mutant 321/358 in the absence of activation. AP5 binding increased to both receptors in the presence of kistrin, DTT, and kistrin+DTT (Fig. 4B). Kistrin produced a greater effect, however, on normal aIIbb3 than on the 321/358 mutant (p = 0.036, n = 3). A similar pattern was found with the 321/360 mutant (Fig. 4C), but this mutant bound significantly less AP5 than normal aIIbb3 (p = 0.003, n = 3) or the 321/358 mutant (p = 0.001, n = 3) in the presence of kistrin. DTT treatment partially or completely rescued the AP5 binding of the XS-O mutants.
In the absence of activation, aIIbFFb3 bound significantly more AP5 than did normal aIIbb3 (p = 0.003, n = 5) or either XS-O mutant (p = 0.01 and 0.015 respectively) (Fig. 4D). Adding kistrin or DTT increased the binding of AP5 to all of the receptors, but the aIIbFFb3 mutant again bound more than any of the other receptors. Combining kistrin and DTT further increased binding to all of the receptors, but the aIIbFFb3 mutant still bound the most.
Mutant 321/358 and 321/360 can adhere to immobilized fibrinogen, adhesion of XS-O mutant 321/358 to immobilized fibrinogen is mediated by the covalentlyassociated heterodimer Cells expressing aIIbb3 receptors can adhere to fibrinogencoated surfaces in the absence of exogenous activation [35,36]. Cells expressing both XS-O mutants adhered to fibrinogen immobilized at low density (5 mg/ml) slightly better than cells expressing normal aIIbb3 (Fig. 5A). These small differences paralleled, and thus were probably caused by, the minor differences in receptor surface expression (normal aIIbb3:321/ 358:321/360 = 80:100:90). The mAb 7E3 inhibited adhesion of all three cell lines to similar extents. DTT enhanced the adhesion of all three cell types and 7E3 was able to inhibit the increased adhesion in the presence of DTT. Similar data were obtained when fibrinogen was immobilized at 50 mg/ml (data not shown).
To assess whether the adhesion to fibrinogen by the XS-O mutant 321/358 was mediated by the relatively small percentage of aIIbb3 receptors that may not have undergone disulfide bond formation, we repeated the adhesion experiments with dose-response inhibition of adhesion by mAb 10E5. Both the cells expressing normal aIIbb3 and the 321/358 mutant showed similar binding curves for mAb 10E5 when tested at multiple concentrations (Fig. 5B). In the absence of mAb 10E5, the adhesion of cells expressing normal aIIbb3 and the 321/358 mutant to immobilized fibrinogen were similar as judged by the absorbance values (Fig. 5C). We reasoned that if the cells expressing the 321/358 mutant adhered to fibrinogen via the ,10-15% of the receptors that did not undergo heterodimer formation (Fig. 2B), they would be more sensitive to the inhibitory effects of mAb 10E5. In fact, however, the mAb 10E5 doseresponse inhibition of adhesion of the cells expressing the 321/358 mutant was very similar to the dose-response inhibition for the cells expressing normal aIIbb3 (Fig. 5C). To assess whether aVb3 contributed to the adhesion, we also performed the experiments in the presence of mAb LM609 and found no effect on the adhesion of cells expressing normal aIIbb3 and less than a 4% decrease in the adhesion of the cells expressing the 321/358 mutant (data not shown). These data support the conclusion that covalently bonded mutant receptors mediated adhesion to fibrinogen.

Cells expressing the XS-O mutants 321/358 and 321/360 have defective cytoskeletal reorganization after adhering to immobilized fibrinogen
After adhering to fibrinogen, cells expressing normal aIIbb3 formed filopodia and lamellipodia, reorganized their actin into filaments detectable with phalloidin and became eccentric in shape (Fig. 6A upper). In contrast, the adherent cells expressing either the XS-O 321/358 or 321/360 mutant were nearly circular, with some irregular, short filopodia. When Mn 2+ was used to activate the cells, cells expressing normal aIIbb3 demonstrated increased cytoskeletal reorganization, forming more lamellipodia and adopting a more irregular shape. Cells expressing the XS-O mutants, however, did not demonstrate enhanced cytoskeletal reorganization in the presence of Mn 2+ (Fig. 6A lower). Cells expressing normal aIIbb3 formed multiple focal adhesions as shown by the colocalization of vinculin and actin filaments and the colocalization of b3 and actin filaments (Figs. S5 and S6). In both cases, the images show organized actin filaments connecting focal adhesions in the filopodia. In contrast, the mutant cells show a distinctive round shape with no concentration of vinculin in focal adhesions. Moreover, the actin filaments are arrayed radially and in loops at the periphery of the cell rather than spanning between focal adhesions.
We also conducted time-lapse photography of the adhesion and spreading of cells expressing normal or mutant aIIbb3 using differential interference microscopy (Videos S1, S2, S3, S4) and the results demonstrated that the cells expressing normal aIIbb3 adhere and spread with a discrete change in morphology at about 15 min as they begin to extend filopodia and lamellipodia and reorganize their cytoskeletons. In contrast, the cells expressing the mutant receptor maintain their round shape throughout and undergo repeated radial extension of the membrane, but without the development of mature, organized filopodia.
Quantitative analysis of cell area showed no differences between the cells expressing normal and the XS-O mutants (data not shown), but the cells expressing normal aIIbb3 spread in a more eccentric manner as judged by the elliptical form factor, reflecting differences in developing lamellipodia and focal adhesions (Fig. 6B, p,0.0001 for both mutants). As a control, cells expressing normal aIIbb3 or XS-O 321/358 were analyzed for their adhesion, spreading, and cytoskeletal reorganization on collagen (data not shown). Both cell types spread equally well and showed similar elliptical form factors (1.6660.26 and 1.6460.33; p = 0.78). Thus, the morphologic abnormalities found with adhesion to fibrinogen are not due to a generalized defect in cytoskeletal reorganization, but rather appear to be specific for the aIIbb3 mutations.

XS-O mutant 321/358 have reduced fibrinogen binding after priming
Cells expressing normal aIIbb3 bound little fibrinogen, but preincubating the cells with eptifibatide or the peptide RGDS primed the cells to bind fibrinogen (P = 0.02 and p = 0.001 respectively, n = 3) (Fig. 7). In contrast, the cells expressing the 321/358 mutant bound significantly less fibrinogen in the presence of each priming agent (n = 3; p = 0.03 for eptifibatide and p = 0.002 for RGDS). The small molecule aIIbb3 antagonists RUC-1 and RUC-2 served as controls since they both bind to aIIbb3, but do not induce fibrinogen binding [28][29][30]. To assess whether eptifibatide bound to the 321/358 mutant, we compared the ability of eptifibatide to inhibit the adhesion of cells expressing normal aIIbb3 or the 321/ 358 mutant to fibrinogen. The dose response was similar for both cell lines, with 3 mM eptifibatide inhibiting normal aIIbb3 adhesion by 62611% (n = 4) and the 321/358 mutant by 83610% (n = 4). At 10 mM eptifibatide the values were 9169% and 9963%, respectively. The RGDS compound could not be tested in the same manner because it did not inhibit adhesion of the normal aIIbb3 cells at 300 mM, presumably reflecting its lower affinity.

Cells expressing the XS-O mutant 321/358 can retract fibrin clots
Untransfected cells showed minimal clot retraction, but cells expressing both normal and XS-O 321/358 were able to retract fibrin clots to the same extent and at the same rate as judged by serial photographs over time (Fig. 8).

Discussion
Integrin b3 hybrid domain swing-out is closely associated with integrin activation and the adoption of the high affinity ligand binding state [8,9,15], but the precise contribution of swing-out to ligand binding and the sequence of events remain unclear. Using data from TMD simulations, we identified the aIIb b-propeller residue K321 and the b3 hybrid domain residues E358 and R360 as candidates for cysteine mutagenesis to create new disulfide bonds to restrict the swing-out motion [19]. The creation of the expected aIIb321-b3358 and aIIb321-b3360 heterodimers was confirmed by SDS-PAGE and mass spectroscopy. Binding of both activation-dependent high Mr ligands (PAC-1 and fibrinogen) to the XS-O 321/358 and 321/360 mutants in the presence of PT25-2 is significantly reduced when compared to high Mr ligand Binding was assessed via flow cytometry and expressed as net normalized fluorescence intensity (mean 6 SD; n = 4). B. AP5 binding to mutant 321/358. Alexa488-labeled anti-LIBS mAb AP5 was incubated with cells in the absence or presence of kistrin (200 nM), or DTT (5 mM), or both at 37uC for 1 h. AP5 binding was assessed by flow cytometry and expressed as NNFI, using excess unlabeled AP5 to assess nonspecific binding. C. AP5 binding to XS-O mutant 321/360. D. AP5 binding to aIIbFFb3 mutant and mutants FF321/358 and FF321/360 (mean 6 SD; n = 4). doi:10.1371/journal.pone.0081609.g004 binding to normal aIIbb3. The residual ligand binding to the XS-O mutants might reflect incomplete disulfide bond formation or relatively low affinity binding to the mutant receptor. We conclude that limiting b3 swing-out prevents the receptor from adopting the conformation(s) required to bind activation-dependent high Mr soluble ligands.
In contrast to the data with the activation-dependent high Mr ligands, the XS-O mutants are able to bind the lower Mr snake venom protein kistrin. Thus, swing-out is not required for the binding of this ligand, which may reflect its smaller size and/or higher affinity. Kistrin binding to the XS-O mutants fails, however, to fully expose the AP5 binding site, indicating that the exposure of the AP5 binding site requires some contribution from swing-out. Similarly, the aIIb activating mutations increased exposure of the AP5 binding site on normal aIIbb3, whereas it produced less increase in AP5 binding to the XS-O mutants, supporting a role for swing-out in the exposure of the AP5 epitope. XO-mutant 321/360 bound much less AP5 than mutant 321/358 in the presence of kistrin, possibly because 321/360 adopts a more compact conformation than 321/358, which is perhaps reflected in the fact that XO-mutant 321/360 is more resistant to DTT treatment (Fig. S3). Collectively these data are similar to those we obtained with an aIIb mutant designed to prevent extension of the aIIb subunit around the genu [16]. The defect in soluble ligand binding to that mutant could, however, be overcome by introducing an activating mutation in the b3 b I domain thought to induce swing-out. This raises the possibility that the major impact of receptor extension is to facilitate swing-out, perhaps by diminishing restrictive headpiece-tailpiece interactions. In this model, swing-out is downstream from extension and proximate to ligand binding.
aIIbb3 activation and ligand binding in platelets is thought to be initiated by inside-out signaling, leading to separation of the aIIb and b3 cytoplasmic and transmembrane domains, and terminating in conformational changes in the head region that lead to extension and swing-out. To simulate this process in the HEK293 cell line, we introduced the F992A/F993A mutations into aIIb to produce a constitutively active receptor [33,37,38]. When combined with normal b3, the aIIbFFb3 mutant receptor binds PAC-1 and fibrinogen constitutively. Combining the F992A/F993A mutations with the XS-O 321/358 or 321/360 mutations does not, however, rescue the XS-O mutants' ability to bind PAC-1 or fibrinogen, either constitutively or in the presence of PT25-2. If the aIIb F992A/F993A mutations do induce receptor extension, these data are consistent with the above model in which the swing-out motion is required for the binding of select activation-dependent high Mr ligands and is downstream from both aIIbb3 leg separation induced by inside-out signaling and the conformational change(s) induced by the binding of mAb PT25-2 to the aIIb b-propeller domain. Alternatively or additionally, b3 subunit extension may provide large soluble ligands greater access to the ligand binding pocket.
Low molecular weight ligands patterned after the RGD sequence and RGD-containing peptides can ''prime'' the aIIbb3 receptor such that after washing away the free low molecular weight ligand the receptor can bind fibrinogen [28,30,[39][40][41]. This activation of the receptor has been hypothesized to contribute to the paradoxical increase in deaths ascribed to several oral aIIbb3 antagonists patterned after the RGD sequence [42,43]. To assess whether ''priming'' requires b3 integrin subunit swing-out, we tested the priming effect of eptifibatide and the peptide RGDS on cells expressing normal aIIbb3 and the XS-O 321/358 mutant. We found that the mutant receptor had a markedly reduced ability to bind soluble fibrinogen after priming. Thus, it appears that receptor priming requires b3 swing-out, raising the possibility that therapeutic agents do not induce swing-out may have a reduced risk of paradoxical receptor activation [28,30,42,43].
Our data on soluble fibrinogen, PAC-1, and LIBS mAb binding to the XS-O mutants are similar to those reported by Luo et al. with aIIbb3 containing a b3T329C/A347C double mutation designed to prevent the motion of the b I domain a7 helix associated with b3 swing-out [14]. Their mutant differs from ours, however, in being unable to support cell adhesion to immobilized fibrinogen. Since their mutation introduces constraints within the b I domain whereas ours primarily constrains the movement of the hybrid domain away from the b I domain, intra-b I conformational changes may be required for binding immobilized fibrinogen.
While our studies were in progress, Kamata et al. reported the effect of creating cysteine mutations in aIIbD319 and b3V359 to create a disulfide bond similar to ours to prevent swing-out of the hybrid domain [17]. Their mutant, like ours, did not bind soluble fibrinogen in the presence of the activating mAb PT25-2. This defect in ligand binding could not be overcome by introducing mutations into the b3 subunit (Q595N/R597T) designed to induce receptor extension. Thus, their data also support the hypothesis that swing-out is downstream from receptor extension. They concluded that swing-out is required for ''high affinity ligand binding,'' but we would temper this conclusion by restricting the requirement for swing-out to the binding of select activationdependent high Mr soluble ligands. For example, as in our studies, they found that the inability to bind soluble ligands was selective, since the mAb OP-G2 was able to bind to the mutant. Moreover, the small GRGDS peptide was presumed to bind to their 319/359 mutant receptor.
Our data on cell adhesion and cytoskeletal reorganization extend those of Kamata et al. by examining the effect of limiting the swing-out motion on the interaction of aIIbb3 with immobilized ligand. Of particular note, we found that the XS-O mutant receptors are capable of supporting cell adhesion to immobilized fibrinogen. These data also echo those we obtained with the aIIb mutant designed to prevent extension about the aIIb genu [16], reinforcing that inhibiting extension and swing-out produce similar functional defects. Potential explanations for the XS-O mutant's ability to support cell adhesion to immobilized fibrinogen include: 1. immobilizing fibrinogen at high density increases receptor avidity enough to compensate for decreased affinity [35], 2. immobilizing fibrinogen alters its conformation so that it can: a) bind with high affinity even without b3 swing-out, b) bind to another site on aIIbb3, or c) bind to another receptor [44].  To address the first possibility, we used two different concentrations of fibrinogen, including one designed to be nearly limiting in density [45], and did not observe a difference in the relative cell adhesion by the mutants. It appears unlikely that the immobilized fibrinogen is binding to another site on aIIbb3 or to another receptor since both eptifibatide and the mAb 7E3 inhibited the adhesion of cells expressing normal aIIbb3 or the XS-O mutants.
We also found that the mutant receptors are unable to support normal outside-in signaling required for normal cytoskeletal reorganization on immobilized fibrinogen via lamellipodia and the formation of focal adhesions. Thus, b3 hybrid domain swingout is required for the outside-in signaling that results in cytoskeletal rearrangements [46].
In conclusion, our study demonstrates that b3 hybrid swing-out is necessary for the activation-dependent binding of select high Mr ligands to aIIbb3, but not for the binding of aIIbb3 to immobilized fibrinogen. It is likely that swing-out is downstream from receptor extension since interventions designed to initiate extension cannot rescue ligand binding to receptors that have limited ability to undergo swing-out. Finally, swing-out is necessary for receptor priming by low molecular weight ligands and integrinmediated outside-in signaling, the latter suggesting that the swingout motion transmits information to the cytoplasmic domain(s) of one or both subunits. The precise structural changes that connect swing-out to changes in ligand affinity and outside-in signaling and the precise sequence of events remain to be defined. Figure S1 HEK293 cells expressing normal aIIbb3 express little or no aVb3. HEK293 cells expressing either normal aIIbb3 or aVb3 were lysed with Triton X-100 and the proteins in the lysates were resolved on SDS-PAGE. Integrin subunits were immunoblotted with anti-aIIb mAb PMI-1, an anti-aV antibody, or anti-b3 mAb 7H2. the cell and measuring both the major and minor axes. Eccentricity was defined as the ratio of the major axis (b) to the minor axis (a) and expressed as the elliptical form factor. * p,0.0001 compared to normal aIIbb3 (n = 20). doi:10.1371/journal.pone.0081609.g006