Affinity Modulation of Platelet Integrin α_IIbβ₃ by β₃-Endonexin, a Selective Binding Partner of the β₃ Integrin Cytoplasmic Tail

Mammalian expression vectors for green fluorescent protein (GFP)¹ (pS65T-C1 and pEGFP-C1) were obtained from Clontech (Palo Alto, CA). Monoclonal antibodies PAC1, A2A9, D57, anti-LIBS1, and anti-LIBS6 were obtained from ascites and purified as described (30). PAC1 was conjugated to phycoerythrin (PE) by first derivatizing it with N-succinimidyl S-acetylthioacetate (Pierce Chemical Co., Rockford, IL). SH groups were deprotected with hydroxylamine, and the antibody was then coupled to PE that had been derivatized with succinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate (Pierce Chemical Co.). PE:PAC1 conjugates (1:1 mol/mol) were isolated by sizing on a Superose 6 column (Pharmacia Fine Chemicals, Piscataway, NJ). In one experiment, PAC1 IgMκ was first reduced to the 185-kD monomer at pH 8.6 with 20 mM cysteine before conjugating to PE (46).

To express β₃-endonexin as a protein fused to the carboxy terminus of GFP, β₃-endonexin cDNA was excised from a yeast expression vector with XbaI and BamHI (63), and the recessed 3′ termini were filled in using Klenow (Boehringer Mannheim Biochemicals, Indianapolis, IN). The GFP vectors, pS65T-C1 and pEGFP-C1, were cut with XhoI, blunt-ended with Klenow, and ligated to β₃-endonexin with T4 DNA ligase (Boehringer Mannheim Biochemicals). After transformation of DH5α, clones in the correct orientation were selected by PCR using a sense primer in GFP and an antisense primer in β₃-endonexin.

Plasmid DNA encoding the cytoskeletal protein, VASP, was a gift from Ulrich Walter and Thomas Jarchau (25) (Medizinische Universitatsklinik, Wurzburg, Germany). The coding sequence of VASP was amplified with Pfu polymerase (Stratagene, La Jolla, CA) using a sense primer containing an XhoI site and an antisense primer with an HindIII site. The digested PCR fragment was subcloned into XhoI- and HindIII-cut pEGFP-C1 so that VASP would be expressed in-frame at the carboxy terminus of GFP. Plasmid DNA encoding FRNK, an autonomously expressed carboxy-terminal segment of pp125^FAK, was a gift from Michael Schaller (University of North Carolina, Chapel Hill, NC) (57). FRNK was amplified with Pfu polymerase with a sense primer containing a BglII site and an antisense primer containing an EcoRI site. The digested PCR fragment was subcloned into BglII- and EcoRI-cut pEGFP-C1 so that FRNK would be expressed in-frame at the carboxy terminus of GFP.

pCDM8 expression vectors encoding wild-type αIIb, β₃, a mutant form of β₃ containing a single amino acid substitution (S752P), and H-Ras (G12V) have been described (33, 49). Tac-β₃ and Tac-α₅ chimeras were in the vector, CMV-IL2R (7). All expression plasmids were amplified in Escherichia coli and purified (Plasmid Maxi Kit; Qiagen, Inc., Chatsworth, CA). Before use in transfection experiments, each plasmid was sequenced in the Scripps Research Institute DNA Core Facility to confirm the authenticity of the coding sequences.

cDNAs were transfected into CHO-K1 cells with Lipofectamine (GIBCO BRL, Gaithersburg, MD). A total of 5 μg of plasmid DNA and 20 μl of Lipofectamine solution was incubated for 10 min in 200 μl of DME and then diluted with 3.8 ml of DME. Unless otherwise indicated, the amount of DNA per transfection included 0.5 μg each of αIIb and β₃ and varying amounts of the GFP plasmids to obtain equivalent degrees of GFP expression (e.g., 4 μg of pS65T, 4 μg of pS65T/β₃-endonexin, 0.02 μg of pEGFP, 0.2 μg of pEGFP/β₃-endonexin, or 0.05 μg of pEGFP/VASP). When necessary, an empty vector (pcDNA3; Invitrogen, San Diego, CA) was included to equalize the amount of DNA transfected. In some experiments, 2 μg of the Tac-β₃, Tac-α5, or H-Ras (G12V) plasmid was cotransfected along with pEGFP/β₃-endonexin and the plasmids for αIIb and β₃. DNA/ Lipofectamine mixtures were added to CHO cells at 30–50% confluence in a 100-mm tissue-culture plate. 6 h later, the medium was changed to DME containing 10% FBS, 1% nonessential amino acids, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. 48 h after transfection, the cells were evaluated biochemically and by flow cytometry.

Expression of GFP/β₃-endonexin fusion proteins was confirmed by immunoblotting. 48 h after transfection, the cells were lysed for 30 min at 4°C in a lysis buffer containing 1% Triton X-100, 0.9% NaCl, 1 mM CaCl₂, 50 mM Tris, pH 7.2, and protease inhibitors (100 U/ml aprotinin, 0.5 mM leupeptin, 4 mM Pefabloc) (63). After clarification of the lysate in a microfuge, protein concentration was determined with a bicinchoninic acid reagent (BCA; Pierce Chemical Co.). 30 μg of each sample was then electrophoresed in 14% SDS-polyacrylamide gels under reducing conditions. After transfer to nitrocellulose, immunoblotting was performed with a rabbit polyclonal antibody reactive with GFP (Clontech) or rabbit antibodies reactive with β₃-endonexin (63). After the addition of affinity-purified, HRP-conjugated goat anti–rabbit IgG, the blots were developed for 0.1–1 min using the enhanced chemiluminescence reaction (ECL; Amersham Corp., Arlington Heights, IL).

To study the binding of GFP/β₃-endonexin to the β₃ integrin cytoplasmic tail, the 47–amino acid β₃ tail was expressed in bacteria with a (His)₆ tag at its amino terminus (pET His Tag System; Novagen, Inc., Madison, WI), and then immobilized on a nickel-agarose matrix. Transiently transfected CHO cells expressing equivalent amounts of GFP/β₃-endonexin or GFP were lysed in 0.4 ml of the Triton X-100 lysis buffer. After clarification, 0.35 ml was diluted with an equal volume of lysis buffer containing no Triton, and each sample was batch-incubated with 0.1 ml packed volume of the His-β₃ tail affinity matrix for 12 h at 4°C while shaking. After washing the matrices five times with 10 vol of lysis buffer, bound proteins were eluted into 0.7 ml of lysis buffer by addition of 1 M imidazole. Samples were electrophoresed on 12% SDS-polyacrylamide gels under reducing conditions and transferred to nitrocellulose. Immunoblotting was performed with the polyclonal anti-GFP antibody.

To evaluate whether GFP/β₃-endonexin could be specifically coimmunoprecipitated with the β₃ integrin subunit from CHO cells, immunoprecipitation studies were carried out in the Trition X-100 lysis buffer essentially as described (30). The β₃ integrin subunit was immunoprecipitated from 2-mg aliquots of cell lysate using 10 μg/ml of the murine mAb, SSA6, and protein A–Sepharose (1). Control immunoprecipitations were carried out with affinity-purified mouse IgG (Zymed Laboratories, Inc., South San Francisco, CA) and with an isotype-matched mAb against von Willebrand factor, RG 7 (a gift from Zaverio Ruggeri; Scripps Research Institute, La Jolla, CA). Samples were electrophoresed on 10% SDS-polyacrylamide gels under reducing conditions and subsequent Western blots were probed with the polyclonal anti-GFP antibody. To demonstrate equivalent recovery of the β₃ integrin subunit in the immunoprecipitates, blots were stripped and reprobed with Ab 8035, a rabbit polyclonal antibody specific for β₃.

48 h after transfection, CHO cells were resuspended at 1–2 × 10⁶ cells per ml in Tyrode's buffer containing 2 mM CaCl₂ and MgCl₂ (49). Cells were then incubated in the dark in 50-μl aliquots with 20 μg/ml PE-PAC1 for 30 min at room temperature. Some samples also contained an anti-β₃ antibody (2% anti-LIBS6 ascites) that stabilizes α_IIbβ₃ in a high affinity state (30). Others contained an α_IIbβ₃-selective inhibitor of ligand binding (either 2 μM Ro 43-5054 or 10 μM Integrilin) (2, 56). Samples were then diluted with 0.5 ml Tyrode's buffer containing 10 μg/ml propidium iodide (Sigma Chemical Co., St. Louis, MO) and analyzed on a FACS^®can or FACS^®Calibur flow cytometer (Becton Dickinson, San Jose, CA). In one set of experiments, the cells were preincubated for 30 min at room temperature with 4 mg/ml of 2-deoxy-d-glucose (Sigma Chemical Co.) and 0.2% sodium azide before incubation with PE-PAC1.

After electronic compensation of the FL1, FL2, and FL3 fluorescence channels, PE-PAC1 binding (FL2) was analyzed on the gated subset of live cells (propidium iodide–negative, FL3 channel) that were positive for GFP expression (FL1 channel). PAC1 binding was expressed as an “activation index” calculated from median fluorescence intensity measurements (49). The activation index is defined as 100 × (Fx-Fi)/(Fm-Fi), where Fx is PAC1 fluorescence in the absence and Fi is PAC1 fluorescence in the presence of Ro 43-5054 or Integrilin. Fm is PAC1 fluorescence in the presence of anti-LIBS6.

Fibrinogen binding to GFP-positive cells was determined by flow cytometry using biotinylated anti-LIBS1, which recognizes a fibrinogen-sensitive epitope on the β₃ subunit (19). Cells were prepared as for the PAC1 binding studies and incubated for 30 min in the dark at room temperature with fibrinogen (250 μg/ml; Enzyme Research Laboratories, South Bend, IN), biotin-LIBS1 (20 μg/ml), and phycoerythrin-streptavidin (4% final dilution; Molecular Probes, Inc., Eugene, OR). To calculate the activation index, some aliquots were incubated with anti-LIBS6 to induce maximal fibrinogen binding, while others were incubated with the function-blocking anti-α_IIbβ₃ αntibody, A2A9, to determine nonspecific fibrinogen binding. Cells were then diluted with Tyrode's buffer containing 10 μg/ml propidium iodide, and analyzed by flow cytometry.

Fibrinogen-dependent aggregation of CHO cells was quantitated by flow cytometry as described (16), with minor modifications. First, CHO cells stably expressing α_IIbβ₃ (49) were labeled with a red fluorescent tracer, hydroxyethidine (Polysciences Inc., Junction City, OR). Then 250 μl of these cells (4 × 10⁶/ml) were added to siliconized glass cuvettes containing 250 μl of cells (2 × 10⁶/ml) that had been transfected with GFP/β₃-endonexin (or GFP) and α_IIbβ₃. After addition of 300 μg/ml fibrinogen, the cells were stirred with a magnetic stir bar at 1,000 rpm for 20 min at room temperature. In some cases, the incubations with fibrinogen were also carried out in the presence of 20 μg/ml A2A9 or 10 μM Integrilin to inhibit fibrinogen binding. Incubations were stopped by addition of 0.25% formaldehyde, and the samples were kept on ice for 30 min before flow cytometric detection of mixed red–green cellular aggregates.

48 h after transfection, CHO cells were cultured on fibrinogen-coated coverslips for 2 h at 37°C, and then processed and analyzed by fluorescence microscopy for expression of GFP, α_IIb, and β₃ as described (32). HMEC-1 human endothelial cells, which express α_Vβ₃, were similarly cultured on fibrinogen-coated coverslips, and then microinjected with plasmid DNA (0.5 μg/μl) encoding various GFP proteins (45). After 4 h at 37°C, the cells were processed and analyzed by fluorescence microscopy for expression of GFP, α_V, and β₃.

CHO cells provide a useful model system for characterizing the adhesive and signaling functions of ectopically expressed α_IIbβ₃ (43, 49, 50). Therefore, β₃-endonexin was transiently coexpressed with α_IIbβ₃ in these cells to study its effects on the ligand-binding function of this integrin. β₃-Endonexin cDNA was fused in-frame to the 3′ end of two different versions of GFP in a mammalian expression plasmid. One form (S65T) is red-shifted and the other (EGFP) is both red-shifted and codon-optimized for mammalian expression. 48 h after transfection, expression of recombinant proteins was assessed by Western blotting of cell lysates. GFP/β₃-endonexin was detectable using an anti-GFP antibody, and the codon-optimized plasmid provided higher levels of protein expression for a given amount of DNA transfected (Fig. 1). Subsequently, therefore, the amount of each plasmid used was adjusted to obtain roughly equivalent amounts of GFP/β₃-endonexin expression, and the plasmids were used interchangeably in the following experiments. GFP/β₃-endonexin was also detectable with polyclonal antibodies raised against either recombinant human β₃-endonexin or a synthetic peptide consisting of the carboxy-terminal 17 residues of the protein (Fig. 1). No hamster protein cross-reactive with these antibodies was detected in CHO cells. These results indicate that full-length GFP/β₃-endonexin can be expressed in CHO cells.

Previous studies have shown that β₃-endonexin binds in vitro to the β₃ integrin subunit from detergent-solubilized platelets and CHO cells (13, 63). To determine if β₃-endonexin retains its ability to bind to the β₃ integrin subunit after its fusion to GFP, lysates from CHO cells expressing GFP/β₃-endonexin were passed over an affinity matrix containing the bacterially expressed β₃ cytoplasmic tail. GFP/β₃-endonexin, but not GFP, was specifically retained by and eluted from this affinity matrix (Fig. 2,A). Moreover, GFP/β₃-endonexin and the β₃ integrin subunit could be specifically coprecipitated from CHO cell lysates (Fig. 2 B). Finally, CHO cells containing GFP/β₃-endonexin were strongly fluorescent in the FL1 channel of a flow cytometer (see below). Thus, fusion of β₃-endonexin to the carboxy terminus of GFP abrogates neither the integrin-binding function of β₃-endonexin nor the fluorescent properties of GFP.

48 h after cotransfection of CHO cells with expression plasmids encoding α_IIbβ₃ αnd GFP/β₃-endonexin, the affinity state of α_IIbβ₃ was determined by flow cytometry using a PE conjugate of the fibrinogen-mimetic mAb, PAC1. Since transfection efficiencies varied from 15–45%, data acquisition included only live cells positive for GFP fluorescence. About 75% of these cells were also positive for α_IIbβ₃, as assessed by staining with an antibody specific for the α_IIbβ₃ complex (D57). To standardize the results of PAC1 binding from experiment to experiment, binding was expressed as an activation index calculated from median fluorescence values (49). To obtain this index, nonspecific PAC1 binding was determined in the presence of a selective inhibitor of ligand binding to α_IIbβ₃ (either Ro 43-5054 or Integrilin). Maximal PAC1 binding was determined in the presence of an activating anti-β₃ antibody (anti-LIBS6) (49). After subtraction of nonspecific binding, this maximal binding was assigned an activation index of 100. Consequently, the activation index for PAC1 binding can range from 0 to 100.

Fig. 3 shows the results of a representative experiment. PAC1 binding to CHO cells transfected with GFP/β₃- endonexin and α_IIbβ₃ exhibited a relatively high activation index of 44 (Fig. 3,A). In contrast, PAC1 binding to cells transfected with GFP and α_IIbβ₃ exhibited a lower activation index of 18 (Fig. 3 D), a value similar to that observed previously for α_IIbβ₃ transfectants in the absence of GFP (31, 49). Thus, expression of β₃-endonexin appears to activate α_IIbβ₃ and increase its affinity for a cognate ligand.

This impression was confirmed by the series of experiments summarized in Fig. 4. Compared with cells expressing GFP, those expressing GFP/β₃-endonexin consistently showed an increase in PAC1 binding, and the difference was statistically significant (P < 0.03). In contrast, PAC1 binding to cells expressing an unrelated GFP fusion protein, GFP/VASP, was not increased despite similar levels of recombinant protein expression. VASP was chosen because it is present in platelets and localizes to integrin-rich focal adhesions (25). Although not shown, the PAC1 activation index for GFP/β₃-endonexin cells (44 ± 5) began to approach that for cells expressing a constitutively active form of α_IIbβ₃ (α_IIbα_6Aβ₃; 61 ± 6; n = 3) (49). Expression of GFP/β₃-endonexin or the other GFP proteins did not affect levels of surface expression of α_IIbβ₃, as determined by the binding of antibody D57. All together, these results indicate that expression of β₃-endonexin can increase the affinity state of α_IIbβ₃.

Platelets containing α_IIbβ₃ with a specific point mutation in the β₃ cytoplasmic tail at position 752 (S→ P) fail to bind fibrinogen or aggregate (6). Furthermore, the binding of β₃-endonexin to this mutant β₃ integrin subunit is markedly reduced (63). When α_IIbβ₃ (S752P) was coexpressed with GFP/β₃-endonexin in CHO cells, no increase in PAC1 binding was observed (Fig. 4). This suggests that β₃endonexin modulates the affinity state of α_IIbβ₃ in a structurally specific manner.

To determine whether the changes in PAC1 binding induced by GFP/β₃-endonexin translate into increased binding of a physiological ligand, the binding of fibrinogen to CHO cells was studied by flow cytometry. Bound fibrinogen was detected with a biotinylated mAb (anti-LIBS1) specific for a fibrinogen-sensitive epitope on the β₃ subunit (19). Specific fibrinogen binding was defined as that inhibitable by a function-blocking anti-α_IIbβ₃ antibody, A2A9. CHO cells expressing wild-type α_IIbβ₃ bind little or no fibrinogen at a saturating concentration of ligand (250 μg/ml) (48). The same was true for cells expressing GFP and α_IIbβ₃. However, those expressing GFP/β₃-endonexin and α_IIbβ₃ bound increased amounts of fibrinogen (Fig. 5). Similar results were obtained when fibrinogen binding was measured directly with biotinylated fibrinogen (not shown). Thus, expression of GFP/β₃-endonexin can lead to an increase in fibrinogen binding to α_IIbβ₃.

When fibrinogen binds to activated α_IIbβ₃ on the surface of platelets or CHO cells under stirring conditions, the cells aggregate (4, 16). To determine whether GFP/β₃-endonexin can trigger this aggregation response, CHO cells expressing GFP/β₃-endonexin and α_IIbβ₃ were mixed with cells containing α_IIbβ₃ and a red fluorescent tracer, hydroxyethidine. After stirring for 20 min in the presence of 300 μg/ml fibrinogen, the formation of mixed, red–green cellular aggregates was monitored by flow cytometry. The rationale for this experimental design is that if fibrinogen first becomes bound to activated α_IIbβ₃ on the GFP/β₃- endonexin cells, this cell-bound fibrinogen should then be able to recruit the red fluorescent cells into mixed aggregates, even though the α_IIbβ₃ on the red fluorescent cells is initially in a low affinity state (Fig. 6 A) (16).

In the experiment shown in Fig. 6 B, it can be seen that GFP/β₃-endonexin promoted the formation of mixed aggregates (center), an effect that could be inhibited by the function-blocking antibody, A2A9 (right), or the cyclic peptide, Integrilin (not shown). In three such experiments, an average of 7.0 ± 1.6% of the cells expressing GFP/β₃endonexin were engaged in red–green aggregates, compared with 3.5 ± 1.9% of cells expressing GFP. While this effect may seem small, it was statistically significant (P < 0.01). Moreover, it should be emphasized that the extent of mixed aggregation was limited by the required use of red fluorescent cells expressing low affinity α_IIbβ₃. These results indicate that affinity modulation of α_IIbβ₃ by β₃- endonexin can cause fibrinogen-dependent cell aggregation.

Additional experiments were conducted to clarify the mechanism of action of GFP/β₃-endonexin. Although PAC1 is a multimeric IgM antibody, GFP/β₃-endonexin was also found to increase the binding of a monomeric form of PAC1 obtained by enzyme digestion. In addition, PAC1 binding because of GFP/β₃-endonexin was not affected by preincubation of the cells with 10 μM cytochalasin D, an inhibitor of actin polymerization (data not shown). Since actin polymerization promotes integrin clustering (12, 71), which would be expected to influence preferentially the binding of multivalent ligands, these results suggest that GFP/β₃endonexin is primarily a modulator of α_IIbβ₃ affinity rather than avidity.

Next, GFP/β₃-endonexin was studied in CHO cells expressing both α_IIbβ₃ and a β₃ cytoplasmic tail chimera containing the extracellular and transmembrane domains of the Tac subunit of the IL-2 receptor. We reasoned that the chimera, which does not dimerize with α_IIb (7, 40), would compete intracellularly with α_IIbβ₃ for β₃-endonexin. If so, it should prevent β₃-endonexin from binding to and modulating the function of α_IIbβ₃. Indeed, expression of the Tac/ β₃ chimera prevented GFP/β₃-endonexin from activating α_IIbβ₃ (Fig. 7). In contrast, a Tac chimera containing the structurally unrelated α₅ cytoplasmic tail exhibited no such effect. This is consistent with the idea that β₃-endonexin modulates integrin affinity through an interaction with the β₃ cytoplasmic tail.

Affinity modulation of α_IIbβ₃ by platelet agonists requires metabolic energy (68). In CHO cells, PAC1 binding induced by GFP/β₃-endonexin was not observed if the cells were pretreated with sodium azide and 2-deoxy-dglucose to inhibit oxidative metabolism (Fig. 7). In this respect, the effect of β₃-endonexin in the CHO cell system is similar to that of excitatory agonists in the platelet system.

In platelets, heterotrimeric GTP-binding proteins have been implicated in affinity modulation. On the other hand, the role of the small GTPase, H-Ras, which is also present in these cells, has not been examined. Recently, Hughes and co-workers found that a constitutively active form of H-Ras (G12V) acts as a general suppressor of integrin adhesive function in CHO cells (33). Similarly, we found that the expression of H-Ras (G12V) inhibited the effects of GFP/β₃-endonexin on PAC1 binding (Fig. 7). Taken together with the energy depletion experiments, this indicates that the function of GFP/β₃-endonexin is subject to metabolic regulation.

In order for β₃-endonexin to directly influence the function of the β₃ integrin cytoplasmic tail, these proteins must be located together in the cell. To address this question, HMEC-1 human endothelial cells, which attach and spread on immobilized fibrinogen through α_Vβ₃, were microinjected with DNA encoding GFP/β₃-endonexin or GFP. 4 h later, specific green fluorescence could be observed diffusely in the cytoplasm and the nucleus. The degree of nuclear fluorescence was much greater in the case of GFP/ β₃-endonexin (Fig. 8). An identical pattern of GFP/β₃- endonexin localization was observed in CHO cells that had been allowed to spread on fibrinogen through α_IIbβ₃ (not shown). These results are consistent with a generalized cytoplasmic distribution of GFP/β₃-endonexin and with a nuclear localization that may be promoted by a consensus nuclear localization signal in β₃-endonexin (see Discussion).

When CHO cells containing α_Vβ₃ or α_IIbβ₃ are allowed to spread on fibrinogen, the β₃ cytoplasmic tail is necessary and sufficient for localization of the β₃ integrins to focal adhesions (40, 72). Immunostaining of HMEC-1 cells revealed that α_V and β₃ were localized both in a diffuse pattern consistent with a generalized plasma membrane distribution and in discrete foci characteristic of focal adhesions (Fig. 9). There was no strong or consistent localization of GFP/β₃-endonexin to these focal adhesions, excluding the possibility that β₃-endonexin might associate tightly with the β₃ cytoplasmic tail during cytoskeletal assembly. However, some weak staining of β₃-endonexin in focal adhesions was observed, suggesting that a weaker or more transient association may occur (Fig. 9, arrowheads). No localization of GFP to focal adhesions was detected. As a positive control, GFP was fused to FRNK, an autonomously expressed segment of pp125^FAK that contains a focal adhesion targeting sequence (57). After microinjection, GFP/FRNK significantly localized to focal adhesions, demonstrating that a GFP fusion protein can target to these structures under the experimental conditions used here (Fig. 8). Thus, GFP/β₃-endonexin is not strongly or consistently concentrated in focal adhesions.

These studies demonstrate that: (a) in CHO cells, expression of β₃-endonexin as a fusion protein with GFP is associated with an increase in the affinity state of integrin α_IIbβ₃. This affinity change enables the cells to undergo fibrinogen-dependent aggregation. (b) Affinity modulation of α_IIbβ₃ by GFP/β₃-endonexin is structurally specific in that other GFP proteins (GFP; GFP/VASP) do not promote this response. Furthermore, GFP/β₃-endonexin does not affect the function of α_IIbβ₃ (S752P), a mutant integrin that is defective in binding to β₃-endonexin and in integrin signaling. (c) Affinity modulation by β₃-endonexin may be the consequence of its direct interaction with α_IIbβ₃ since it is prevented by coexpression of a Tac-β₃ cytoplasmic tail chimera. (d) The effect of β₃-endonexin on α_IIbβ₃ may be subject to metabolic regulation since it is not observed if cellular energy is depleted or if the cells are cotransfected with an activated form of H-Ras. (e) GFP/β₃-endonexin is found in both the nuclear and cytoplasmic compartments after CHO cells or HMEC-1 cells have spread on a fibrinogen matrix via a β₃ integrin. Taken together, these results indicate that β₃-endonexin may play a significant role in cell adhesion and signaling through integrin α_IIbβ₃.

We interpret the effect of GFP/β₃-endonexin on PAC1 and fibrinogen binding to CHO cells to represent an example of inside-out signaling in which β₃-endonexin increases the affinity of individual α_IIbβ₃ heterodimers for specific ligands. An alternative interpretation that β₃-endonexin triggers oligomerization of α_IIbβ₃ complexes and therefore increases receptor avidity cannot be excluded, but it seems less likely for several reasons. First, changes within α_IIbβ₃ that enable the binding of RGD-containing macromolecular ligands have been detected with both a monovalent Fab fragment of PAC1 as well as with the native, multivalent antibody (1). This indicates that regulated ligand binding to α_IIbβ₃ is not absolutely dependent on the valency of the ligand or, presumably, the receptor. Second, the effect of GFP/β₃-endonexin on α_IIbβ₃ was detected using either native PAC1 or a monomeric fragment of the antibody. Third, agonist-induced clustering of β₂ integrins in leukocytes and possibly α_IIbβ₃ in platelets is facilitated by polymerization of F-actin (12, 14, 71). However, cytochalasin D, an inhibitor of actin polymerization, had no effect on PAC1 binding induced by GFP/β₃-endonexin. While it is not possible to quantitate precisely the relative contributions of affinity and avidity regulation, based on the above considerations, we speculate that β₃-endonexin can regulate reversible fibrinogen binding through affinity modulation. Other factors, including actin polymerization and cytoskeletal reorganization, may enhance cell adhesion by promoting receptor clustering and irreversible ligand binding. Consistent with this idea, cytochalasin D has been reported to inhibit primarily the later, irreversible phase of fibrinogen and PAC1 binding to platelets and CHO cells (14, 53, 54).

The present results were obtained by expressing β₃- endonexin ectopically in CHO cells. Therefore, it is possible that the function of the endogenous protein in platelets or other cells differs quantitatively or qualitatively from that described here. Despite this caveat, a number of observations indicate that affinity modulation may result directly from the interaction of β₃-endonexin with the cytoplasmic tail of the β₃ integrin subunit. A mutational analysis of the β₃ tail has shown that membrane-distal residues near the carboxy terminus of the tail (N⁷⁵⁶ITY) are required for the interaction with β₃-endonexin (13). Mutation or deletion of these same residues also disrupts insideout integrin signaling in platelets and CHO cells (50; Wang, R., D.R. Ambruso, and P.J. Newman. 1994. Blood. 84:244a). Moreover, coexpression of a β₃ tail chimera, but not an α₅ tail chimera, prevented affinity modulation by β₃-endonexin, possibly because the former chimera but not the latter could compete with α_IIbβ₃ for binding to β₃endonexin. Finally, when other recombinant GFP proteins such as GFP and GFP/VASP were expressed in CHO cells, they failed to increase α_IIbβ₃ affinity.

Additional studies will be required to determine how cellular energy depletion or coexpression of activated H-Ras inhibits affinity modulation by β₃-endonexin. Nonetheless, these results imply that this function of β₃endonexin is subject to metabolic regulation. In this context, studies with platelets have suggested that serine-threonine kinases (61), tyrosine kinases (23), and PI 3-kinase (38, 74; Kovacsovics, T.J., J.H. Hartwig, L.C. Cantley, and A. Toker. 1995. Blood. 86:454a) are involved in promoting fibrinogen binding to α_IIbβ₃. In contrast, compounds that activate protein kinase A or G inhibit fibrinogen binding (22, 28). Perhaps β₃-endonexin is a direct substrate of specific kinases or phosphatases or is a target of downstream effectors of these enzymes. For example, β₃-endonexin contains several serine and threonine residues in favorable contexts for phosphorylation by protein kinase C and A (63).

Suppression of integrin activation in CHO cells by activated H-Ras involves a Raf-1–initiated MAP kinase pathway and is transcription independent (33). In contrast, an activated variant of R-Ras was recently implicated in the promotion of integrin-mediated cell adhesion (75). Since the reported opposite actions of activated R-Ras and H-Ras affect both β₁ and β₃ integrins, it seems unlikely that the pathways triggered by these GTPases converge directly on β₃-endonexin. Although platelets contain both H-Ras and R-Ras, and platelet stimulation by thrombin activates H-Ras, the functions of these GTPases in this terminally differentiated cell are unknown (64).

Based on the present results, we propose that the interaction of β₃-endonexin with the β₃ cytoplasmic tail triggers a structural change in the integrin to reconfigure the extracellular face of the receptor so that it can engage fibrinogen. While the nature of this propagated change is unknown, one possibility is that there is a reorientation of the β₃ subunit relative to the α_IIb subunit. This is plausible given the biophysical evidence for interactions between the cytoplasmic tails of α_IIb and β₃ (24) and for agonist- induced structural changes in the extracellular domains of α_IIbβ₃ in platelets (65). In a similar manner, relatively subtle changes within preexisting dimers may play a role in ligand-triggered, “outside-in” signaling across other plasma membrane receptors, including the bacterial aspartate receptor (66) and the mammalian EGF receptor (15).

A complete understanding of the proximate events in inside-out signaling will require identification of all relevant integrin-binding proteins and a more refined knowledge of integrin structure. Progress is beginning to be made in both of these areas (11, 42, 55). Several proteins have been described that bind directly to integrin cytoplasmic tails, at least in vitro. These include structural proteins of the cytoskeleton, such as F-actin (specific for the α₂ tail) (36), α-actinin (β tails) (51), talin (β) (29), and filamin (β₂) (60), and potential signaling molecules, such as calreticulin (α tails) (10), pp125^FAK (β) (58), integrin-linked kinase (β) (26), and cytohesin-1 (β₂) (37). Of note, the expression of calreticulin and cytohesin-1 appears to stimulate or stabilize a high affinity state of integrins α₂β₁ and α_Lβ₂, respectively (10, 37). There is no sequence similarity between either of these proteins and β₃-endonexin. Thus, a structurally diverse group of cytoplasmic tail-binding proteins may function to regulate integrins. Some like cytohesin-1 and β₃-endonexin may be restricted in their action because of their binding specificities, while others like calreticulin, which recognizes a conserved motif in all integrin α tails, may be less specific.

While not relevant to platelets, the localization of β₃- endonexin to the cytoplasm and nucleus of HMEC-1 and CHO cells suggests that this protein may have more than one function. The nuclear localization may be explained, in part, by the presence of a consensus nuclear localization signal in β₃-endonexin (K⁶²RKK) (35, 63). Interestingly, several proteins implicated in cell adhesion and adhesive signaling, including ZO-1, β-catenin, zyxin, c-Abl, and HEF1, either exhibit cytoplasmic and nuclear localization or shuttle between the cytoplasm and the nucleus depending on the adhesive state of the cell (Nix, D.A., and M.C. Beckerle. 1995. Mol. Biol. Cell. 6:366a; 3, 21, 41, 44). The identification of other proteins that can bind to β₃-endonexin should help to explain its pattern of subcellular localization.

Focal adhesions are dynamic structures containing integrins, cytoskeletal elements, and signaling molecules that form on the basal surfaces of many types of cells in culture and in platelets during spreading on fibrinogen (47). These macromolecular assemblies may function to optimize traction during cell motility and to promote information flow from the extracellular matrix to the nucleus (5, 17, 18). The lack of consistent and strong localization of β₃-endonexin to β₃-rich focal adhesions suggests that it may interact most strongly with the β₃ cytoplasmic tail while cells are in suspension or are in the early phases of adhesion. Thus, it is attractive to speculate that β₃-endonexin may participate in integrin activation but may dissociate at later times to permit cytoskeletal interactions with the integrin tails.

These studies provide the first clues about the functions of β₃-endonexin, but they leave several questions unanswered. Does β₃-endonexin influence outside-in signaling events, such as protein tyrosine phosphorylation (8)? Is β₃endonexin subject to posttranslational modifications in vivo, and does this affect its subcellular localization or function? Does β₃-endonexin modulate the adhesive function of α_Vβ₃, which like α_IIbβ₃ appears to be subject to rapid regulation in some cell types (67, 73)? Does β₃-endonexin regulate α_IIbβ₃ in platelets?

We thank David Phillips for providing Integrilin, Michael Schaller for FRNK cDNA, Beat Steiner for Ro 43-5054, Zaverio Ruggeri for antibody RG 7, and Ulrich Walter and Thomas Jarchau for VASP cDNA.

These studies were supported by research grants from the National Institutes of Health (HL 56595, HL 48728, AR 27214) and from Cor Therapeutics, Inc. H. Kashiwagi is the recipient of a Banyu Fellowship in Lipid Metabolism and Atherosclerosis sponsored by Banyu Pharmaceutical Co., Ltd. and the Merck Company Foundation. M. Eigenthaler is the recipient of a fellowship from the American Heart Association (CA Chapter). This is manuscript 10536-VB from the Scripps Research Institute.