Complementary Roles for Receptor Clustering and Conformational Change in the Adhesive and Signaling Functions of Integrin α_IIbβ₃

A pCDM8 template containing full-length α_IIb (O'Toole et al., 1994) was subjected to PCR with Pfu polymerase (Stratagene, La Jolla, CA) to place XbaI and SpeI restriction sites at the 5′ and 3′ ends of α_IIb, respectively. The PCR product was cut with XbaI and SpeI and ligated into an XbaI-cut, CMV-based mammalian expression vector, pCF1E (ARIAD Pharmaceutical, Inc., Cambridge, MA). Plasmids with inserts in the correct orientation were amplified and purified for CHO cell transfections (Maxi-Prep; QIAGEN Inc., Chatsworth, CA). The resulting α_IIb(FKBP)/pCF1E plasmid encoded α_IIb fused in-frame to FKBP, which in turn was fused in-frame to a hemagglutinin epitope tag (see Fig. 1). To construct α_IIb fused to two tandem FKBP repeats (α_IIb(FKBP)₂), a single FKBP was removed from pCF1E with XbaI/SpeI and ligated into SpeI-cut α_IIb(FKBP)/pCF1E. The remaining α_IIb and β₃ cDNAs depicted in Fig. 1 were in pCDM8 (O'Toole et al., 1994). cDNA coding full-length human Syk was in EMCV (Gao et al., 1997). Plasmid inserts were analyzed by automated sequencing to confirm authenticity.

cDNAs were transfected into CHO-K1 cells with lipofectamine according to the manufacturer's instructions (GIBCO BRL, Gaithersburg, MD). Typically, 0.5–2 μg of each plasmid was used, supplemented when necessary with empty vector DNA (pCDNA3; Invitrogen, San Diego, CA) for a total of 4 μg per dish. Cells were maintained for 48 h for transient expression or subjected to antibiotic selection for stable expression. Stable cell lines were selected further for high integrin expression by single cell FACS^® sorting using an α_IIbβ₃-specific murine mAb, D57 (O'Toole et al., 1994).

Cell surface expression of recombinant α_IIbβ₃ was assessed by flow cytometry using biotin-D57 and FITC-streptavidin (Leong et al., 1995). α_IIb expression was also evaluated by Western blotting. 48 h after transfection, the cells were lysed in 66 mM Tris-HCl, pH 7.4, containing 2% SDS and 30 μg of protein were electrophoresed in SDS–7.5% polyacrylamide gels under nonreducing conditions, transferred to nitrocellulose, and then subjected to Western blotting with murine mAb B1B5 specific for α_IIb or antibody 12CA5 specific for the hemagglutinin epitope tag (Abrams et al., 1992). After addition of affinity-purified, HRP-conjugated goat anti– mouse IgG (Biosource International, Camarillo, CA), blots were developed for 0.1–1 min by enhanced chemiluminescence (ECL; Amersham, Arlington Heights, IL).

To establish whether AP1510 could induce clustering of α_IIb(FKBP)₂β₃ that was detectable morphologically, cells stably expressing α_IIb(FKBP)₂β₃ were incubated in the presence of 1,000 nM AP1510 (or 0.5% EtOH as a vehicle control) for 30 min at room temperature. Then, analogous to the method used by Yauch and co-workers to detect antibody-induced integrin clustering (Yauch et al., 1997), the cells were incubated with 10% goat serum for 30 min at room temperature, followed by 10 μM FITC-D57 or unlabeled D57 for 30 min on ice. After washing, the sample containing unlabeled D57 was incubated for 30 min with FITC-labeled goat anti–mouse heavy and light chains (1:500; Biosource International) to deliberately cluster the integrin as a positive control. All samples were fixed in 4% paraformaldehyde, resuspended in mounting medium (Fluorosave; Calbiochem-Novabiochem, San Diego, CA), and analyzed on glass slides with an MRC 1024 laser scanning confocal imaging system (Bio-Rad Laboratories, Hercules, CA).

Ligand binding to α_IIbβ₃ in CHO cells was assessed by flow cytometry using a saturating amount of the fibrinogen-mimetic, murine monoclonal IgMκ antibody, PAC1 (Kashiwagi et al., 1997). CHO cells were resuspended to 10⁷ cells/ml in Tyrode's buffer supplemented with 2 mM CaCl₂ and MgCl₂ (O'Toole et al., 1994). For most experiments, 4 × 10⁵ cells were added to tubes containing a final concentration of 0.4% PAC1 ascites or 40 nM purified PAC1 in a final vol of 50 μl, and then incubated for 30 min at room temperature. In some experiments, monovalent recombinant PAC1 Fab produced in insect cells and purified by nickel-agarose chromatography was used instead of PAC1 IgM at a final concentration of 30 nM (Abrams et al., 1994). As indicated for each experiment, cell incubations were also carried out in the presence of one or more of the following reagents: 10–5,000 nM AP1510 (or vehicle buffer) to cluster α_IIb(FKBP)₂β₃ or α_IIb(FKBP)β₃ (Amara et al., 1997), 150 μg/ml anti-LIBS6 Fab to convert α_IIbβ₃ into a high affinity form through conformational changes (Du et al., 1993; Kashiwagi et al., 1997), and 10 μM integrilin, an α_IIbβ₃ antagonist to specifically block PAC1 binding (Scarborough et al., 1993). Preliminary experiments with AP1510 and anti–LIBS6 Fab indicated that a 10– 30-min incubation of cells with these reagents was sufficient to achieve their maximum effects. Cells were then washed and incubated on ice for 30 min with biotin-D57, followed by phycoerythrin-streptavidin and either FITC-labeled goat anti–mouse μ heavy chain antibody (to label PAC1 IgM) or FITC-labeled goat anti–mouse heavy and light chain antibody (to label PAC1 Fab) (both from Biosource International). Samples were diluted with 0.5 ml Tyrode's buffer containing 2 μg/ml propidium iodide (Sigma Chemical Co., St. Louis, MO) and analyzed on a FACSCalibur^® flow cytometer (Becton Dickinson Co., Mountain View, CA). After electronic compensation, PAC1 binding (FL1 channel) was analyzed on the gated subset of live cells (propidium iodide–negative, FL3) that was positive for α_IIbβ₃ expression (FL2). To control for variations in integrin expression from transfection to transfection, PAC1 binding, measured as mean fluorescence intensity in arbitrary units, was expressed relative to the levels of α_IIbβ₃, measured simultaneously with biotin-D57.

Immulon-2 microtiter wells (Dynex Laboratories, Chantilly, VA) were coated with purified fibrinogen (Enzyme Research Laboratories, South Bend, IN) or vWf (Ruggeri et al., 1983) overnight at 4°C at coating concentrations ranging from 0.01–2 μg/well, followed by blocking with 20 mg/ml BSA. CHO cells stably expressing α_IIb(FKBP)₂β₃ were labeled for 30 min at 37°C with 2 μM BCECF-AM (Molecular Probes, Eugene, OR). After washing, the cells were resuspended to 10⁶/ml, incubated for 30 min in the presence of 1,000 nM AP1510 and/or 150 μg/ml anti–LIBS6 Fab, and then 100-μl aliquots were added to the coated microtitre wells for 90 min at 37°C. After washing three times with 150 μl of PBS, cell adhesion was quantitated by cytofluorimetry (Leng et al., 1998).

Stable cell lines expressing α_IIb(FKBP)₂β₃ were transiently transfected with EMCV-Syk and placed into complete DME with 10% FBS. 24 h after transfection, the amount of serum was reduced to 0.5%, and 48 h after transfection, the cells were resuspended to 3 × 10⁶/ml in DME and slowly rotated at 37°C for 45 min in the presence of 20 μM cycloheximide. Then cells were incubated for 10 min with one or more of the following: 1,000 nM AP1510 to stimulate receptor clustering, 150 μg/ml anti–LIBS6 Fab to increase integrin affinity, or 250 μg/ml fibrinogen to achieve ligand binding. As a positive tyrosine phosphorylation control, one aliquot of cells was allowed to attach for 60 min to a dish coated with α_IIbβ₃ antibody D57. Cells were lysed in RIPA buffer containing 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 158 mM NaCl, 10 mM Tris, pH 7.4, 1 mM Na₂EGTA, 0.5 mM leupeptin, 0.25 mg/ml Pefabloc, 5 μg/ml aprotinin, 20 mM NaF, 3 mM β-glycerophosphate, 10 mM sodium pyrophosphate, and 5 mM sodium vanadate. After clarification, 200 μg of protein were immunoprecipitated with rabbit antiserum specific for Syk or FAK (Gao et al., 1997). Immunoprecipitates were subjected to Western blotting with anti-phosphotyrosine mAbs, 4G10, and PY20 (Upstate Biotechnology Inc., Lake Placid, NY and Transduction Laboratories, Lexington, KY, respectively), followed by stripping and reprobing with mAb 4D10 to Syk or antiserum to FAK (Gao et al., 1997). Immunoreactive bands detected by ECL were quantitated by calibrated densitometry using a flatbed scanner, Power Center Pro 240 computer, and NIH Image software.

The purpose of these studies was to evaluate the possible contributions of integrin clustering and affinity modulation to the adhesive and signaling functions of α_IIbβ₃. A CHO cell model system, previously used to study factors that influence the ligand-binding affinity of human α_IIbβ₃ (O'Toole et al., 1994; Hughes et al., 1996; Zhang et al., 1996; Kashiwagi et al., 1997), has now been adapted to study integrin clustering. Full-length α_IIb was engineered to contain one or two FKBP repeats and a hemagglutinin epitope tag at the extreme COOH terminus of the cytoplasmic tail (Fig. 1). In theory, a protein containing a single FKBP may dimerize upon addition of a membrane-permeable, bivalent FKBP ligand, such as AP1510, and a protein with two or more FKBP repeats may form larger oligomers (Amara et al., 1997; Yap et al., 1997; Yang et al., 1998). Consequently, we reasoned that if α_IIb(FKBP) or α_IIb(FKBP)₂ were successfully coexpressed on the surface of CHO cells along with β₃, then α_IIbβ₃ heterodimers might be converted into a dimer of dimers ((α_IIbβ₃)₂) or into even larger oligomers in response to AP1510.

After transient or stable expression in CHO cells, both α_IIb(FKBP)β₃ and α_IIb(FKBP)₂β₃ were found to be expressed to the same extent as wild-type α_IIbβ₃, as determined by the binding of D57, an α_IIbβ₃-specific antibody (Fig. 2,A). In addition, Western blotting of the cell lysates with an antibody specific for the extracellular domain of α_IIb (B1B5) showed that α_IIb(FKBP) and α_IIb(FKBP)₂ exhibited the predicted slower electrophoretic mobilities compared with the smaller wild-type α_IIb subunit (Fig. 2,B). The α_IIb(FKBP) and α_IIb(FKBP)₂ fusion proteins also reacted on Western blots with an antibody to the COOH-terminal epitope tag, further suggesting that they represented full-length proteins (Fig. 2 B). Thus, the fusion of one or two FKBP repeats to the cytoplasmic tail of α_IIb does not interfere with the transient or stable expression of this subunit in CHO cells to form an α_IIbβ₃ complex. Therefore, in the following studies transient and stable cell lines were used as indicated, depending on the experimental protocol.

Large integrin oligomers might be detectable in CHO cells at the level of light microscopy (van Kooyk et al., 1994; Yauch et al., 1997). To determine if clusters of α_IIb(FKBP)₂β₃ could be detected morphologically, CHO cells stably expressing †this integrin were incubated for 30 min at room temperature with 1,000 nM AP1510 (or vehicle buffer as a control), stained with FITC-D57 on ice, fixed, and then examined by confocal microscopy. D57 staining was entirely surface associated, and cells that had been treated with buffer instead of AP1510 exhibited a finely patchy distribution of α_IIb(FKBP)₂β₃ (Fig. 3, A–C). In contrast, α_IIb(FKBP)₂β₃ in cells treated with AP1510 exhibited a coarse patchiness (Fig. 3, E–G), similar to that observed when the D57 antibody was deliberately cross-linked with a secondary antibody before cell fixation (Fig. 3 H). The same results with AP1510 were obtained with another independent α_IIb(FKBP)₂β₃ clone; in contrast, AP1510 caused no discernible clustering of wild-type α_IIbβ₃ in CHO cells (not shown). These results are consistent with the conclusion that oligomerization of α_IIb(FKBP)₂β₃ can be induced conditionally from within the cell using AP1510, enabling us to conduct a detailed study of the functional consequences of integrin clustering.

Activation of α_IIbβ₃ is required for the binding of soluble, macromolecular Arg-Gly-Asp–containing ligands, such as fibrinogen, vWf, and fibrinogen-mimetic antibodies, such as PAC1. To evaluate the contribution of clustering to α_IIbβ₃ activation, flow cytometry was used to quantitate the specific binding of PAC1 to transiently transfected CHO cells. Specific binding was defined as that inhibitable by 10 μM integrilin, an α_IIbβ₃-selective antagonist, and it was expressed relative to the amount of α_IIbβ₃ on the cell surface, determined simultaneously with antibody D57. In cells expressing α_IIb(FKBP)₂β₃, there was little binding of PAC1, indicating that, like α_IIbβ₃, this integrin is in a constitutive low affinity/avidity state. AP1510 caused a dose-dependent increase in PAC1 binding to α_IIb(FKBP)₂β₃ cells (Fig. 4, closed circles), without affecting the levels of surface expression of this integrin. However, PAC1 binding due to AP1510 appeared modest compared with binding in response to upregulation of integrin affinity by an activating antibody Fab fragment, anti–LIBS6 Fab (Fig. 4, open circles).

To evaluate possible mechanistic differences between integrin clustering and affinity modulation in the control of ligand binding, additional experiments were performed with cells expressing α_IIb(FKBP)₂β₃. First, we considered the possibility that AP1510 caused PAC1 binding by increasing integrin affinity rather than (or in addition to) avidity. However, AP1510 failed to stimulate the binding of a monovalent PAC1 Fab fragment to α_IIb(FKBP)₂β₃, although this form of PAC1 bound normally in response to anti–LIBS6 Fab (Fig. 5). Since a monovalent ligand might be expected to be sensitive to affinity modulation but less sensitive than a multivalent ligand to avidity modulation, this result suggests that AP1510 was indeed working by clustering the integrin. Second, PAC1 binding in response to AP1510 was completely prevented if the cells were preincubated for 30 min with 4 mg/ml of 2-deoxy-d-glucose and 0.2% sodium azide to deplete metabolic ATP (two separate experiments). Since oligomerization by AP1510 should be energy independent, this suggests that metabolic energy is needed to maintain the receptor in a proper conformation, even when ligand binding is triggered by receptor clustering. Third, the effect of a specific point mutation (β₃(S752P)) or a truncation (β₃(Δ724)) of the β₃ cytoplasmic tail were studied because both have been shown to disrupt affinity modulation of α_IIbβ₃ in platelets and CHO cells (Chen et al., 1992, 1994; O'Toole et al., 1994; Wang et al., 1997). Whereas β₃(S752P) abolished PAC1 binding in response to AP1510, β₃(Δ724) had no such effect (Fig. 6). Thus the β₃ cytoplasmic tail plays a role in ligand binding triggered by integrin clustering, but there must be differences in the structural features of β₃ required for affinity and avidity modulation.

Thus far, the results support the validity of this model system to study integrin clustering, and they suggest that both clustering and affinity modulation can regulate ligand binding to α_IIbβ₃. Further studies were performed to determine the relative contributions of clustering and affinity modulation to ligand binding under conditions in which the effects of AP1510 and anti–LIB6 Fab were maximal (1,000 nM and 150 μg/ml, respectively). CHO cells were transiently transfected with either α_IIbβ₃, α_IIb(FKBP)β₃, α_IIb(FKBP)₂β₃, or α_IIb/α_6Aβ₃ (a constitutive, high affinity mutant [O'Toole et al., 1994]), and ligand binding was evaluated 48 h later. Whereas AP1510 had no effect on PAC1 binding to wild-type α_IIbβ₃, it increased binding to both α_IIb(FKBP)β₃ and α_IIb(FKBP)₂β₃ such that specific PAC1 binding was increased approximately twofold (P < 0.001) (Fig. 7). However, PAC1 binding induced by AP1510 amounted to only 50% of the binding observed with the high affinity α_IIb/α_6Aβ₃ chimera, and only 25% of the binding induced by anti–LIBS6 Fab (Fig. 7). Nevertheless, the PAC1 binding caused by clustering was statistically significant (P < 0.03) and approximately additive to the binding caused by affinity modulation (Fig. 7).

Fibrinogen and PAC1 binding to activated platelets is initially reversible by the addition of EDTA, but binding becomes progressively irreversible over 15–60 min (Peerschke, 1995a; Fox et al., 1996). In CHO cells that expressed α_IIb(FKBP)₂β₃ and were treated with both anti–LIBS6 Fab and AP1510 to achieve maximal PAC1 binding, the added component of ligand binding resulting from AP1510 was fully reversible at 10 min but irreversible at 30 min (Fig. 8). Similar results were obtained when FITC-fibrinogen was used instead of PAC1 to monitor ligand binding (not shown). This series of experiments demonstrates that affinity modulation is the predominant regulator of ligand binding to α_IIbβ₃. However, receptor clustering plays an additive role in promoting eventual irreversible binding of the ligand.

Activation of platelets by agonists leads to increased cell adhesion to the α_IIbβ₃ ligands, fibrinogen, and vWf (Savage et al., 1992). To determine the relative contributions of α_IIbβ₃ clustering and affinity modulation to cell adhesion, CHO cells that stably expressed α_IIb(FKBP)₂β₃ were loaded with BCECF as a fluorescent marker and incubated for 90 min in microtiter wells coated with fibrinogen or vWf. Cell adhesion was dependent on the coating concentration of fibrinogen (Fig. 9, left panel  ) and vWf (Fig. 9, right panel  ), as well as on the presence of α_IIb(FKBP)₂β₃, since it was blocked by 10 μM integrilin. AP1510 (1,000 nM) increased the extent of cell adhesion, but only very modestly and only at the higher coating concentrations of fibrinogen and vWf. On the other hand, increasing integrin affinity with anti–LIBS6 Fab (150 μg/ml) caused a more marked increase in cell adhesion, even at the lower ligand concentrations (Fig. 9, left and right panels). Thus under these assay conditions, receptor clustering is not as effective as affinity modulation in regulating cell adhesion via α_IIbβ₃.

In platelets and CHO cell transfectants, fibrinogen binding to α_IIbβ₃ leads to tyrosine phosphorylation and activation of Syk and FAK. The binding of soluble fibrinogen is sufficient to activate Syk, but additional post–ligand binding events, such as cytoskeletal reorganization, are necessary for activation of FAK (Huang et al., 1993; Clark et al., 1994; Gao et al., 1997). To study the role of integrin clustering in these events, CHO cells stably expressing α_IIb(FKBP)₂β₃ were transiently-transfected with human Syk, and tyrosine phosphorylation of Syk and endogenous hamster FAK was examined. Fig. 10,A shows the raw data for a single experiment and Fig. 10 B shows a summary of three separate experiments. Cells maintained in suspension for 10 min in the absence or presence of fibrinogen exhibited a low level of tyrosine phosphorylation of Syk and FAK. However, addition of 1,000 nM AP1510 caused an average 2.8-fold increase in tyrosine phosphorylation of Syk, even in the absence of fibrinogen (P < 0.05), and this response was greater still in the presence of fibrinogen (5.4-fold; P < 0.02). In contrast, in the absence of fibrinogen integrin clustering by AP1510 did not stimulate an increase in FAK tyrosine phosphorylation, but increased FAK phosphorylation was observed in the presence of fibrinogen (3.5-fold; P < 0.001). Thus, integrin clustering can trigger tyrosine phosphorylation of Syk even in the absence of fibrinogen binding, whereas phosphorylation of FAK requires both receptor clustering and fibrinogen binding. Affinity modulation by anti-LIBS6 is not necessary in either case.

In this study, engraftment of one or two FKBP repeats onto the COOH terminus of the α_IIb subunit enabled us to cluster integrin α_IIbβ₃ in a conditional fashion by treating CHO cells with a synthetic, bivalent FKBP ligand, AP1510. This permitted us for the first time to conduct a detailed comparison of the functional effects of receptor clustering, initiated from within the cell, with the effects of increasing integrin affinity through conformational changes. The major conclusions of this work are: (a) Conformational changes play a predominant role in α_IIbβ₃ activation in CHO cells, as monitored by ligand binding and cell adhesion assays. (b) Clustering causes a modest increment in reversible and ultimately irreversible binding of multivalent ligands to α_IIbβ₃, and this binding is additive to that caused by affinity modulation. (c) Ligand binding resulting from receptor clustering is dependent on cellular metabolic energy and is sensitive to some, but not all, of the mutations or deletions in the β₃ cytoplasmic tail that block affinity modulation of the receptor. (d) Integrin clustering promotes ligand-independent tyrosine phosphorylation of Syk, and ligand- dependent phosphorylation of FAK, even when cells are maintained in suspension and even in the absence of deliberate affinity modulation. Thus, by being able to manipulate integrin clustering and affinity separately and in a controlled manner, we conclude that these two processes play complementary roles in the function of α_IIbβ₃.

Inside-out signaling is responsible for acute regulation of the ligand binding function of integrins. In the case of integrins that normally engage soluble adhesive ligands in vivo, such as α_IIbβ₃, inside-out signaling can be monitored directly using labeled ligands or ligand-mimetic antibodies, such as PAC1. Alternatively, it can be assessed by cell adhesion assays. Although physiologically relevant, cell adhesion is a more indirect measure of integrin activation because it can be influenced by factors, such as actin polymerization, cell spreading, and focal adhesion turnover, that may affect the overall strength of the adhesion process through mechanisms other than regulation of ligand binding (Burridge and Chrzanowska-Wodnicka, 1996; Yamada and Geiger, 1997; Hall, 1998). Thus, when possible, it is preferable to monitor inside-out signaling by ligand binding assays, as in the current study.

Ligand binding to integrins is thought to be regulated by a combination of affinity and avidity modulation (van Kooyk and Figdor, 1993; Diamond and Springer, 1994; Bazzoni and Hemler, 1998). In the case of α_IIbβ₃, platelet activation is believed to cause a modification of the conformation or orientation of the integrin cytoplasmic tails that is transmitted across the membrane, leading to increased access of the ligand to binding sites in the receptor (Loftus and Liddington, 1997; Shattil et al., 1998). However, cell activation appears to promote ligand binding to certain β₁ and β₂ integrins by also stimulating the lateral diffusion and clustering of these receptors (van Kooyk and Figdor, 1993; Kucik et al., 1996; Bazzoni and Hemler, 1998; Shattil et al., 1998), and the same might be true for α_IIbβ₃. Several experimental approaches have been used to cluster integrins, including treatment of cells with multivalent antibodies or chemical cross-linkers, incubation of cells with ligand-coated beads, and promotion of cell spreading (Kornberg et al., 1991; Dorahy et al., 1995; Hotchin and Hall, 1995; Miyamoto et al., 1995). Whereas each of these has provided important information about outside-in signaling, none is entirely suitable for studies of soluble ligand binding to α_IIbβ₃. The use here of AP1510 to cluster α_IIb(FKBP)β₃ or α_IIb(FKBP)₂β₃, while CHO cells were maintained in suspension demonstrates unambiguously that affinity and avidity modulation can complement one another with respect to the control of ligand binding. Given the wide variety of soluble, matrix- and cell-associated ligands that integrins must contend with, it is likely that the relative contributions of affinity and avidity modulation will vary with the integrin and the cell type.

Fortunately, the fusion of single or tandem FKBP repeats to the α_IIb cytoplasmic tail did not interfere with α_IIbβ₃ expression or function in CHO cells. Perhaps this means that the very COOH terminus of the α subunit is dispensable for the integrin functions that were assessed. On the other hand, direct attachment of FKBP to the β₃ tail interferes with energy-dependent affinity modulation of α_IIbβ₃, possibly by disrupting necessary interactions of the β₃ tail with regulatory proteins (Hato, T., and Shattil, S.J., unpublished observations). We ascribe any functional effects of AP1510 on α_IIb(FKBP)β₃ and α_IIb(FKBP)₂β₃ to receptor clustering. Although the evidence for this is strong, it is largely indirect. First, AP1510 only affected those forms of α_IIbβ₃ that contained FKBP repeats (Figs. 6 and 7). Second, confocal microscopy showed that AP1510 treatment was associated with the appearance of coarse patches of integrin staining in the surface membrane (Fig. 3). Finally, AP1510 caused binding of a multivalent but not monovalent form of PAC1, precisely what might be expected in the case of integrin clustering (Fig. 5). Interestingly, the effect of AP1510 on PAC1 binding to α_IIb(FKBP)β₃ and α_IIb(FKBP)₂β₃ was nearly equivalent (Fig. 7). Assuming that AP1510 can only dimerize α_IIb(FKBP)β₃, this suggests that formation of a dimer of dimers, e.g., (α_IIb(FKBP)β₃)₂, may be sufficient to initiate some ligand binding to α_IIbβ₃.

What are the biological implications of α_IIbβ₃ clustering during inside-out signaling? Ligand binding resulting from clustering of α_IIb(FKBP)₂β₃ required a normal β₃ cytoplasmic tail since a tail point mutation (S752P) disrupted this process (Fig. 6). This same mutation disrupts energy- dependent affinity modulation of α_IIbβ₃ in CHO cells and in human platelets, where it is responsible for a bleeding diathesis (Chen et al., 1992, 1994). In contrast, another β₃ cytoplasmic modification, a truncation starting at residue Arg 724, did not disrupt ligand binding induced by clustering of α_IIb(FKBP)₂β₃, although it does abolish affinity modulation of α_IIbβ₃ in CHO cells and platelets (O'Toole et al., 1994; Wang et al., 1997). This suggests that there are differences in the structural elements of the β₃ tail that are needed for affinity and avidity modulation. A growing number of proteins have been shown to interact directly with the cytoplasmic tails of α or β integrin subunits, at least in vitro, and overexpression of some of them, including calreticulin (α tails) (Coppolino et al., 1997), cytohesin-1 (β₂ tail) (Kolanus et al., 1996), and β₃-endonexin (β₃ tail) (Kashiwagi et al., 1997) can affect ligand binding and cell adhesion. When overexpressed in CHO cells, several other signaling proteins, including H-ras (Hughes et al., 1996), R-ras (Zhang et al., 1996), and CD98 (Fenczik et al., 1997) have been shown to modulate the ligand binding properties of α_IIbβ₃; however, it is not known if any of these proteins interact directly with the integrin. The relative effects of potential regulatory molecules such as these on receptor clustering and receptor affinity remain to be determined.

The extent of PAC1 binding observed in response to integrin clustering was only a fraction of that observed with affinity modulation. Yet ligand binding resulting from clustering and affinity modulation was additive (Fig. 7). Clustering also facilitated CHO cell adhesion to immobilized fibrinogen and vWf, but once again, this effect was relatively minor compared with the effect of affinity modulation and it was apparent only at the higher coating concentrations of the ligands (Fig. 9). On the other hand, if clustering of α_IIbβ₃ were to cause increased ligand binding in platelets as it does in CHO cells, it could affect the ultimate size of a platelet aggregate and hence the delicate balance between adequate and inadequate hemostasis or the difference between partial and total arterial occlusion by a platelet-rich thrombus. Conceivably, the effects of integrin clustering on ligand binding may be even more pronounced in platelets than in the CHO cell model system because receptor density may be higher in platelets and clustering may be stimulated by agonists that trigger integrin interactions with multivalent, polymerizing ligands on both sides of the plasma membrane (Hartwig, 1992; Fox et al., 1996; Simmons and Albrecht, 1996). Furthermore, any increase in adhesive strength promoted by α_IIbβ₃ clustering in vivo might help platelets resist detachment from sites of vascular injury in response to hemodynamic forces (Savage et al., 1996).

A potential limitation of the chemical dimerization approach used here is that it may not reflect or trigger the types of interactions between α_IIbβ₃, cytoskeletal proteins, and signaling molecules that take place normally during outside-in signaling. For example, in platelets, the binding of fibrinogen to α_IIbβ₃ is sufficient to trigger tyrosine phosphorylation and activation of Syk, whereas tyrosine phosphorylation of FAK requires additional post–ligand binding events that occur during platelet aggregation or spreading (Haimovich et al., 1993; Huang et al., 1993). In this regard, clustering of α_IIb(FKBP)₂β₃ in CHO cells by AP1510 caused significant tyrosine phosphorylation of Syk, even when the cells were maintained in suspension without fibrinogen (Fig. 10). Since integrin-dependent tyrosine phosphorylation of Syk correlates with induction of Syk kinase activity in both platelets and CHO cells (Clark et al., 1994; Gao et al., 1997), these results suggest that the binding of multivalent fibrinogen to α_IIbβ₃ triggers Syk activation, at least in part, by inducing integrin clustering.

In contrast to the results for Syk, clustering of α_IIb(FKBP)₂β₃ by AP1510 was not sufficient to cause tyrosine phosphorylation of FAK in cells maintained in suspension. However, fibrinogen binding together with receptor clustering were sufficient to induce the response (Fig. 10). These results highlight the apparent differences in coupling mechanisms between α_IIbβ₃ and Syk and α_IIbβ₃ and FAK (Gao et al., 1997). At the same time, they demonstrate unambiguously that conditional clustering of α_IIbβ₃ in CHO cells can recapitulate a pattern of outside-in signaling that is characteristic of platelets. In nucleated cells, integrin and growth factor signaling pathways collaborate to regulate gene expression, cell adhesion, and motility (Schwartz et al., 1995; Juliano, 1996; Sastry and Horwitz, 1996; Yamada and Geiger, 1997). One hallmark of integrated signaling networks is tight control of enzyme activity and protein subcellular localization though regulated protein–protein interactions (Pawson and Scott, 1997). Chemical inducers of dimerization can be used to promote controlled homodimerization and heterodimerization of proteins in vivo as well as ex vivo (Rivera et al., 1996; Spencer et al., 1996; Clackson, 1997; Yang et al., 1998). Consequently, they should prove useful in evaluating diverse aspects of integrin signaling.

The authors are grateful to J. Amara and V. Rivera (ARIAD Pharmaceuticals, Inc., Cambridge, MA) for supplying pCF1E and AP1510; and to several colleagues for their gifts of other reagents: J. Brugge (Harvard Medical School, Boston, MA [EMCV-Syk]); M. Ginsberg (Scripps Research Institute [antibodies D57 and No. 2308, α_IIb and β₃ pCDM8 plasmids]); D. Phillips (Cor Therapeutics, Inc., South San Francisco, CA [Integrilin]); and Z. Ruggeri (Scripps Research Institute [von Willebrand factor]).

ECL

enhanced chemiluminescence

FKBP

FK506 binding protein, or FKBP12

vWf

von Willebrand factor