Identification of T w o Structurally and Functionally Distinct Sites on Human Platelet Membrane Glycoprotein IIb-IIIa Using Monoclonal Antibodies*

15,1982

derived from the study of platelets from patients with the inherited bleeding disorder, Glanzmann's thrombasthenia (6). These platelets do not bind fibrinogen (1,7,8) and do not aggregate, abnormalities which appear to be related to the deficiency of membrane glycoproteins known as IIb and IIIa. The experiments reported here provide evidence that IIb and IIIa are polypeptide subunits of a single membrane glycoprotein which together form the platelet membrane receptor for fibrinogen.
We previously reported using a monoclonal antibody, named Tab, to co-isolate both IIb and IIIa from Triton X-100solubilized platelets by affinity chromatography (9,10). We assumed that Tab bound either to IIb or IIIa and that these polypeptides were isolated together because they formed a complex in Triton X-100 and may also be associated in the intact platelet membrane. Experiments employing crossed immunoelectrophoresis of Triton X-100-solubilized platelet proteins against rabbit antiplatelet antibodies led to similar conclusions, since both IIb and IIIa could be eluted from a single immunoprecipitate (11). Kunicki et al. (12) subsequently employed CIE' to show that the association of IIb and IIIa was mediated by calcium ions (12). When platelets were solubilized in EDTA and then electrophoresed, the major IIb-IIIa immunoprecipitate dissociated into two new arcs from which IIb and IIIa, respectively, were eluted. Similar findings have been described by other investigators (13,14).
We earlier noted that Tab failed to block platelet aggregation induced by ADP or thrombin (9). Rabbit antibodies directed against purified, denatured IIb or IIIa also had no effect on platelet aggregation or on fibrinogen binding (15,16). In contrast, a human alloantibody directed against glycoprotein IIb-IIIa blocked both aggregation and fibrinogen binding (17,18,39). These data suggest that the localization of a specific antigenic determinant on the IIb-IIIa complex may help define the critical region for fibrinogen binding and platelet aggregation. We have applied the CIE technique to localize the site of Tab binding to the IIb polypeptide and also to localize the binding of a newly described monoclonal antibody, T10, to a determinant created only by the association of IIb and IIIa. Studies with these antibodies suggest that the association of IIb and IIIa in the platelet membrane is required, but not sufficient, for fibrinogen binding and platelet aggregation to occur.

5269
This is an Open Access article under the CC BY license.
X-100, Lubrol PX, ADP, and epinephrine were obtained from Sigma. Carrier-free NaTZ51 was from Amersham. Monoclonal anti-HLA antibody and goat IgG directed against mouse IgG were purchased from Bethesda Research Laboratories. DEAE-cellulose (DE52) was obtained from Whatman. Apiezon A oil was from J. G. Biddle Co. (Plymouth Meeting, PA). All other chemicals were reagent grade.
Cells-Human platelets were isolated by differential centrifugation from blood anticoagulated with acid/citrate dextrose and PGEl as described by George et al. (19). For platelet aggregation and binding studies, blood obtained from aspirin-free, normal donors was anticoagulated with 0.1 volume of 3.8% sodium citrate. The blood was centrifuged at 180 X g for 10 min and 8-10 ml of platelet-rich plasma were applied to a 50-ml Sepharose CL-2B column in a plastic syringe. The platelets were eluted with Tyrode's buffer (0.138 M NaC1, 0.029 M KCl, 0.012 M NaHC03, 0.036 M NaH2P04/H20, pH 7.4) containing 3.5 mg/ml of bovine serum albumin and 5.5 mM glucose. The platelet count of the pooled fractions was measured with a Thrombocounter (Coulter Electronics). Fibrinogen binding experiments and platelet aggregation studies were performed within 1 h after gel filtration.
Monoclonal Antibodies-Hybridomas were generated from a cell fusion performed previously and described in detail (9). The Tab antibody was derived from 1 of the 24 wells producing anti-platelet antibodies (9). The antibodies produced by the remaining hybridomas were tested for platelet specificity with a solid phase radioimmunoassay essentially as described by Stocker and Heusser (22). 10' platelets or 5 X lo5 CCRF-CEM cells or human mononuclear cells were fived to the wells of polyvinylchloride microtiter plates (Flow Laboratories) with 0.5% glutaraldehyde. The cells were incubated with hybridoma culture media, washed, and then incubated with 1251-labeled goat antimouse IgG. After washing, the wells were cut out and the radioactivity was measured. Positive and negative controls were, respectively, a monoclonal anti-HLA antibody and culture media from the parent myeloma cell line Sp2/O-Ag 14 (23). Hybridomas producing antibody only to platelets were cloned by limiting dilution, then rescreened. Monoclonal IgG from promising clones was purified from mouse ascites by (NH4),SO4 precipitation and DEAE-cellulose chromatography (9). Fab fragments from mouse IgG were prepared by papain digestion (24). The Fab fragments were separated from Fc fragments and undigested IgG by chromatography on Protein A-Sepharose (Pharmacia).
Protein Zodinution-IgG was labeled with Nal"I as described (9). The specific activity was 100-200 cpm/ng for platelet binding assays and 1000-1600 cpm/ng for CIE experiments. Human fibrinogen prepared by the method of Straughn and Wagner (25) was a gift from Dr. Philip Majerus (Washington University). The fibrinogen contained greater than 95% clottable protein. When analyzed by sodium dodecyl sulfate electrophoresis on a polyacrylamide gel, a single band of M, = 330,000 was seen with unreduced preparations. After disulfide bond reduction, three bands with the mobilities of the Aar, BP, and y chains were seen. The fibrinogen was radiolabeled with a commercial product, Enzymobeads (Bio-Rad, Richmond, CA), consisting of lacfibrinogen (10 mg of protein) was mixed with 100 pl of rehydrated toperoxidase and glucose oxidase coupled to inert beads. One ml of beads, 10 p1 (1 mCi) of NaIz5I, and 100 pl of 1% (w/v) P-D-glucose. After incubation at room temperature for 30 min, the mixture was centrifuged at 12,000 X g for 1.5 min and the supernatant was applied to a Sephadex G-25 column equilibrated in phosphate-buffered saline, pH 7.4. Radiolabeled fibrinogen had a specific activity of 100-200 cpm/ng and greater than 99% of the radioactivity was precipitable with 8% trichloroacetic acid. Aliquots were stored at -20 "C until use. as described by Hagen et al. (11) and Kunicki et al. (12).
Crossed Immunoelectrophoresis-CIE was performed essentially Rabbits were immunized with whole human platelets (26). Antisera collected during 2-3-month intervals were pooled and IgG was isolated by (NH4)*S04 precipitation and DEAE-cellulose chromatography (26). In some experiments, rabbit anti-platelet antibodies (12) generously provided by Drs. Thomas Kunicki (Blood Center of Southeastern Wisconsin) and Alan Nurden (H3pital Lariboisiere, Paris) were used. Whole human platelets were solubilized at 4 "C for 60 min in 0.038 M Tris, 0.1 M glycine, pH 8.7, containing 1% Triton X-100, and. in some cases. 5 mM Na2EDTA. Insoluble material was removed protein were electrophoresed at 10 V/cm at 12.5 "C for 60 min in a f i s t dimension gel consisting of 1% agarose and 1% Triton X-100 in the Tris/glycine buffer. Second dimension electrophoresis was performed at 0.5 V/cm for 18 h at 12.5 "C against an intermediate gel and then an upper gel, both containing 1% agarose and 1% Triton X-100 in Tris/glycine buffer. The intermediate gel contained lo6 cpm (0.1 pg/cm2) of 1251-monoclonal IgG. Polyclonal rabbit anti-platelet IgG (0.6 mg/cm2) was incorporated in the upper gel. In some experiments, 1% Lubrol PX was used instead of Triton X-100 with identical results. After electrophoresis, the gels were washed, dried, and stained with Coomassie blue R (27). Autoradiography was performed by exposing the dried gels to Kodak X-Omat AR film in a casette containing a DuPont Cronex Lightning Plus intensifying screen.
Binding Assays-Gel-filtered platelets were used for all binding assays except where other methods are specifically stated. Direct binding of 1251-monoclonal IgG was performed as described previously (9), except that reaction mixtures were in Tyrode's buffer, pH 7.4, containing 3.5 mg/ml of bovine serum albumin and 5.5 m M glucose.
In some experiments, binding of antibody was measured directly in platelet-rich plasma obtained as rapidly as possible after venipuncture in the presence of inhibitors added to prevent thrombin generation and platelet activation. Blood was collected directly into a syringe containing final concentrations of 0.38% sodium citrate, 2 units/ml of hirudin, 1 pg/ml of PGEI, 1 m~ dibutyryl CAMP, and 2 mM diisopropylphosphofluoridate. Platelet-rich plasma was immediately prepared by centrifuging blood 2.5 s at 12,000 X g in an Eppendorff microfuge. Aliquots of PRP were then added to Tyrode's buffer containing the above inhibitors and incubated with T251-monoclonal IgG for 7 min or 30 min at room temperature. After centrifugation over a 9 1 mixture of n-butyl phthalate and Apiezon A oil (9), the platelet pellets were counted. The experiments were completed within 15 min and 40 min, respectively, after venipuncture. The platelets remained unreactive as assessed by the inability of PRP to aggregate in response to 20 p~ ADP, 20 p~ epinephrine, and 1 unit/ml of thrombin.
Fibrinogen binding was performed as described by Bennett and Vilaire (l), with slight modifications. Suspensions of 5 X lo7 gelfiltered platelets were mixed with 1 mM CaC12 and labeled fibrinogen in 0.5-ml reaction mixtures and equilibrated at 37 "C. ADP, 10 pM, was then added. After incubating without stirring for 5 min at 37 "C, the platelets were centrifuged over 0.5 ml of the n-butyl phthalate-Apiezon A oil mixture at 12,000 X g. Nonspecific binding was determined with parallel tubes containing a 20-fold excess of unlabeled fibrinogen or with tubes in which ADP was omitted; identical results were obtained by either method. Nonspecific binding was subtracted from total binding to determine specific binding. In some experiments, unlabeled monoclonal IgG or Fab fragments were incubated with platelets and labeled fibrinogen before addition of ADP. Platelet Aggregation-Platelet aggregation was measured at 37 "C in siliconized glass cuvettes with constant stirring at 900 rpm. Platelet suspensions of 0.5 ml consisted either of citrated platelet-rich plasma or of 1.25 X lo8 gel-filtered platelets in Tyrode's buffer containing 3.5 mg/ml of bovine serum albumin, 5 m~ glucose, 1 mM CaC12, and 200 pg/ml of fibrinogen. Aggregation was initiated by adding 10 pM ADP or epinephrine. Thrombin, 1 unit/ml, was added to platelet suspensions without added fibrinogen. In some experiments, monoclonal IgG or Fab fragments were incubated with platelets for 15-30 min before addition of the agonist. Changes in light transmission were recorded with a Payton dual-channel aggregometer.
Other Methods-The protein concentration of solubilized platelets was determined according to Markwell et al. (28). The protein concentration of mouse IgG and Fab fragments were estimated from absorbance at 280 nm using an estimated EiZ of 14 and 15, respectively (29). Fibrinogen protein concentration was determined assuming a E:& of 15.1 (30). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis was performed as described by Laemmli (31). Reduced samples contained 5% 2-mercaptoethanol. Radioactive iodide was measured in a Tracor Model 1191 y counter.

Identification of the Monoclonal Anti-platelet Antibody
TlO-On the basis of the screening radioimmunoassay as described under "Experimental Procedures," we selected hybridoma cells which were producing IgG antibodies apparently specific for platelets. These cells were cloned and rescreened and then monoclonal IgG was purified to homogeneity as determined by SDS-polyacrylamide gel electrophoresis. One by cmtrifugation at 35,000 x g for 60 min at 4 "C. One hundred pg of Platelet Membrane Glycoprotein IIb-IIIa 5271 of these clones, designated T10, was selected for detailed studies. T10 IgG was coupled to Protein A-Sepharose and Triton X-100-solubilized platelets were applied to the column, as described in our previous studies of the Tab antibody (10). Bound protein eluted with 3% SDS and analyzed on SDS gels revealed two bands with the characteristic mobilities (9, 10) of platelet membrane glycoproteins IIb and IIIa and a faint band co-migrating with platelet actin (not shown). Thus, the T10 antibody, like Tab, recognizes a determinant on the glycoprotein IIb-IIIa complex. Localization of Antigenic Determinants-We employed the technique of crossed immunoelectrophoresis as developed by Hagen et al. (11) and Kunicki et al. (12) to identify the areas on glycoprotein IIb-IIIa to which the Tab and T10 antibodies bind. to the residual IIb-IIIa complex, but also to the arc representing dissociated IIb (Fig. 1D). In contrast (Fig. 2D), '"I-T10 bound only to the undissociated IIb-IIIa complex. No binding of '251-T10 to either dissociated IIb or IIIa was seen even after prolonged exposure of the film. Identical results were obtained when CIE was performed using the previously characterized rabbit anti-platelet antibodies used by Kunicki et al. (12) and given to us for these experiments (data not shown). Thus, the Tab antibody recognizes a determinant on IIb, while T10 recognizes a determinant formed only after the calcium-mediated association of IIb and IIIa into the intact complex.
Binding of TI0 and Tab to Platelets-We next performed direct binding studies with '2sI-T10 IgG to determine the number of T10 binding sites on platelets. We used gel-filtered platelets for these studies, but have obtained similar results with platelet-rich plasma (see below) and with washed platelets. In preliminary experiments, we determined that binding of T10 was complete by 30 min. Therefore, all subsequent studies were performed using a 30-min incubation with 1251-T10. Fig. 3 illustrates a representative binding experiment. I2'I-TlO, as in previous studies with '251-Tab (9), bound with high affinity to a single class of binding sites on human platelets. Nonspecific binding was negligible since greater than 99.5% of bound 1251-T10 IgG was displaced when a 100-fold excess of unlabeled T10 IgG was added simultaneously. In this experiment, 1251-T10 bound to 53,000 sites/platelet with a dissociation constant of 8.5 nM.
In simultaneous experiments, gel-filtered platelets from six- the same as the 40,000 Tab binding sites reported previously with platelets isolated by differential centrifugation (9). The equivalent number of binding sites seen with each antibody suggests that there is one IIb polypeptide per IIb-IIIa complex and is consistent with IIb-IIIa being a heterodimer (32, 33).

Platelet Membrane Glycoprotein Ilb-IIIa
The binding sites for T10 and Tab do not appear to be adjacent to each other, since a 100-fold excess of either unlabeled antibody does not inhibit binding of the other radiolabeled protein (not shown). Binding ofboth TI0 and Tab was the same in unstimdated platelets and platelets treated with 10 /.LM ADP. However, only platelets stimulated with ADP were able to aggregate or bind '251-fibrinogen (see below). The binding of '251-Tab to platelets was unaffected by the addition of EDTA or CaC12. In contrast, addition of 5 mM EDTA to platelets reduced 1251-T10 binding by approximately 60% (Table I). T10 binding could be partially restored by addition of 10 mM CaC12 to the EDTA-treated platelets. We were unable to reduce binding of T10 below 40% of control values, even with platelets isolated in the continuous presence of 5 mM EDTA, as well as the metabolic inhibitors 0.1 M dibutyryl CAMP. and 1 pg/ml of PGEI. The same degree of inhibition was seen in Tyrode's buffer at pH 8.5 or at pH 7.4.
We considered the possibility that T10 might bind to platelets because of the association of IIb and IIIa in the membrane during the platelet isolation procedure. To examine this question, we rapidly prepared platelet-rich plasma in the presence of inhibitors designed to minimize platelet activation, then immediately measured T10 and Tab binding (see "Experimental Procedures"). Under these conditions, binding of T10 and Tab was not different (Table 11) and was comparable to that seen with washed platelets at the same antibody concentrations and incubation times. This suggests that the isolation procedure itself did not cause association of IIb and IIIa.
Inhibition of Fibrinogen Binding by TlO-To see if Tab and T10 might identify domains of functional significance in glycoprotein IIb-IIIa, we investigated the effects of both antibodies on fibrinogen binding to platelets. We found that 1251fibrinogen bound to a single class of binding sites with approximately 45,000 rnqlecules bound per platelet at saturation, confiiing the data $f other investigators (1-3). Fibrinogen binding was calcium-dependent and required platelet stimulation with an agonist such as ADP. When platelets were preincubated with T10 IgG, 1251-fibrinogen binding was reduced to approximately 35% of control values (Fig. 4). T10 Fab fragments also inhibited fibrinogen binding but, in some experiments, to a lesser extent than the intact IgG molecule. The less effective inhibition of fibrinogen binding by the Fab fragments was not due to alteration of Fab specificity by

TABLE I1
Binding of Tab and TI0 in platelet-rich plasma PRP was rapidly prepared by centrifuging citrated blood containing 2 units/& of hirudin, 2 mM diisopropylphosphofluoridate, 1 mM dibutyryl CAMP, and 1 pg/ml of PGEI at 12,000 X g for 2.5 s. Aliquots of PRP were then incubated with 2 pg/ml of "'I-Tab or 1251-T10 IgG for 7 min or 30 min and then centrifuged and bound radioactivity was measured. Results are means of duplicate determinations and are representative of two similar experiments. The range of the duplicates was within 10% of the mean. papain digestion, since the inhibition of '251-T10 IgG binding to platelets was the same with unlabeled intact T10 or T10 Fab fragments.
In contrast to T10, Tab or nonimmune mouse IgG or Fab fragments either had no effect on fibrinogen binding or slightly augmented binding (Table 111). The mechanisms for augmentation of binding were not investigated, but similar findings have been reported for nonimmune human IgG (18) and rabbit F(ab') 2 fragments directed against IIb (16). TlO- Fig. 5 illustrates the effects of Tab and T10 IgG on ADP-induced aggregation in citrated platelet-rich plasma. The same results were

TABLE I11 Effects of antibodies on fibrinogen binding to platelets
Platelets, lO*/ml in Tyrode's buffer, pH 7.4, containing 3.5 m g / d of bovine serum albumin, 1 m g / d of glucose, and 1 m~ CaCL were preincubated at 37 "C for 5 min with 25 pg/ml of the indicated IgG or Fab fragment. '251-fibrinogen, 200 pg/ml, followed by 10 p~ ADP, was then added. After incubation for 5 min at 37 "C, specific fibrinogen binding was measured. Mean values of duplicate assays are shown and are representative of two similar experiments. The range of the duplicates was within 15% of the mean. Effect of antibodies on ADP-induced platelet aggregation. Citrated platelet-rich plasma was preincubated for 15 min at 37 "C with 100 p g / d of Tab IgG or 20 p g / d of T10 IgG. ADP, 10 p~ final concentration, was then added and changes in light transmission followed. The control tracing represents superimposable curves obtained when PRP was preincubated with no antibody or with 20 p g / d of nonimmune mouse IgG or 100 p g / d of nonimmune Fab fragments. seen with washed platelets. Tab IgG either had no effect, or, as seen in Fig. 5, slight inhibition of the rate and total extent of platelet aggregation. Tab Fab fragments and nonimmune mouse IgG or Fab fragments had no effect. In contrast, T10 IgG completely blocked aggregation except for a small initial wave of reversible aggregation. T10 Fab fragments caused a modest, but less consistent inhibition of aggregation. In general, thrombin-induced aggregation of platelets was inhibited in similar fashion, although in one experiment aggregation reached nearly normal levels after a delay of 8-10 min (not shown). The oscillating baseline tracing seen before the addition of ADP is characteristic of unstimulated discoid platelets which randomly interfere with light transmission as they are stirred. After activation, platelets undergo a shape change becoming spherical cells with small pseudopods. The oscillating pattern on the tracing is no longer seen, because the spherical platelets now uniformly affect light transmission during stirring. T10 IgG did not affect shape change, as seen by the normal flattening of the aggregometer tracing after the addition of ADP.

DISCUSSION
In this study, we have employed crossed immunoelectrophoresis to localize the antigenic determinants of two monoclonal antibodies to glycoprotein IIb-IIIa. As expected, both Tab and T10 bound to the immunoprecipitate representing the intact IIb-IIIa complex. After dissociating IIb and IIIa with EDTA, we identified specific binding of ?-Tab only to IIb. In contrast, '251-T10 failed to bind to either IIb or IIIa separately. Thus, T10 recognizes a determinant formed only after the association of IIb and IIIa, most likely a region where the two polypeptides interact. Alternatively, the association of the two polypeptides creates a new antigenic determinant on IIb or IIIa remote from the actual regions of contact. One of the advantages of the CIE system is that platelet proteins are analyzed under nondenaturing conditions. This may be of particular importance in examining the binding of monoclonal antibodies with restricted specificity to secondary or tertiary structures. We were unsuccessful in identifying the site of binding of Tab by "Western blot" techniques (34,35) in which I-Tab was overlaid over platelet proteins separated on SDSgels and transferred to activated paper (not shown), presumably because of denaturation of the antigenic determinant on IIb. Therefore, a negative result with the blotting procedure would not have distinguished the antigenic specificities of Tab and T10.
Recent analyses of Triton X-100-solubilized glycoprotein IIb-IIIa by gel fitration and sucrose gradient ultracentrifugation suggest that the IIb-IIIa complex is a heterodimer (32, 33). The equivalent number of platelet binding sites that we find for both lZ5I-Tab and 1251-T10 is consistent with the data of these investigators. It is notable that neither antibody inhibits binding of the other, which indicates that the externally oriented portion of IIb-IIIa encompasses enough area to allow simultaneous binding of at least two intact IgG molecules.
The reversible inhibition of '"I-TlO binding in intact platelets by EDTA indicates that the calcium-mediated association of IIb and IIIa occurs in the membrane itself and is not a phenomenon occurring only after solubilization in nonionic detergents. We were unable to inhibit T10 binding completely with EDTA. This is probably due to inaccessibility to the chelator of a portion of the calcium ions bound to IIb-IIIa within the membrane. Investigators using CIE have also pointed out that complete dissociation of IIb and IIIa may require prolonged incubation with EDTA and may be more effective at alkaline pH (12,36). We noted no difference in '251-T10 binding in Tyrode's buffer containing 5 mM EDTA at pH 8.5 or pH 7.4, although it is conceivable that local pH within the membrane was not affected by the change in buffers. We did not incubate platelets with EDTA longer than 2 h. The reasons for incomplete restoration of T10 binding with added CaC12 are also not clear. Jennings and Phillips (32), using purified IIb and IIIa in solution in Triton X-100, also noted that reformation of the complex was incomplete after addition of CaC12.
We originally hypothesized that IIb and IIIa are polypeptide subunits of a single membrane glycoprotein (9, 10). After platelet stimulation, some additional change, perhaps a slight conformational shift in one or both of the subunits, would allow fibrinogen to bind to the glycoprotein. An alternative hypothesis is that IIb and IIIa exist as discrete glycoproteins in the resting platelet, but associate together after platelet activation, perhaps as a result of shifts in membrane calcium. In this view, solubilization of platelets in nonionic detergent would promote complex formation of IIb and IIIa because of small amounts of calcium made available during solubilization. Polley et al. (37) used immunoelectron microscopy with anti-IIb and anti-IIIa antibodies to demonstrate the clustering of IIb and IIIa in platelet membranes after thrombin activation. However, the resolution obtained with their electron micrographs would not distinguish association of IIb and IIIa 125 Platelet Membrane Glycoprotein IIb-IIIa polypeptides into heterodimers or the clustering of pre-existing heterodimers into localized areas of the membrane. Since identical numbers of T10 binding sites are seen in unstimulated platelets and platelets stimulated with ADP, we conclude that platelet activation is not required for association of IIb and IIIa in the membrane. Our "unstimulated" platelets were not activated prematurely by gel fiitration, since they did not bind fibrinogen or aggregate until ADP was added. Furthermore, T10 and Tab binding were identical in plateletrich plasma obtained rapidly after venipuncture in the presence of several inhibitors of thrombin generation and platelet activation. Finally, we have obtained similar binding data using I2'I-Fab fragments (not shown), indicating that membrane association of IIb and IIIa was not produced by crosslinking effects from divalent IgG. Our data therefore suggest that IIb and IIIa are associated in circulating unstimulated platelets and thus can be accurately defied as subunits of a single membrane glycoprotein. Glycoprotein IIb-IIIa is analogous to proteins such as hemoglobin and the acetylcholine receptor which consist of distinct polypeptide subunits associated by noncovalent interactions. It is not yet known if IIb and IIIa are derived from common or separate genes, although structural studies of the peptides suggest that the latter possibility is more likely (9, 15). There is considerable evidence supporting the role of glycoprotein IIb-IIIa in the binding of fibrinogen to platelets, but it has been unclear as to whether one or both polypeptide subunits are required. Using an enzyme-linked immunosorbent assay, Nachman and Leung (16) demonstrated calciumdependent binding of fibrinogen to a partially purified preparation of IIb-IIIa. However, they were unable to prepare separated IIb and IIIa without protein denaturation and could not determine whether one or both polypeptides were required for fibrinogen interaction. Platelet binding studies with fibrinogen modified with a photoreactive cross-linker provided evidence for specific binding of fibrinogen to IIIa, but did not rule out an interaction with IIb undetected by the cross-linker (38). It is significant that non-cross-reacting rabbit antibodies to IIb and to IIIa, prepared by immunization with the individual polypeptides, failed to inhibit fibrinogen binding or platelet aggregation (15,16). This is consistent with the lack of inhibition of these functions by the anti-IIb monoclonal antibody, Tab. Our fiiding that only TI0 inhibits platelet fibrinogen binding suggests that IIb and IIIa must interact in order for fibrinogen binding to occur. Our results are supported by experiments with a human alloantibody to glycoprotein IIb-IIIa which inhibits fibrinogen binding and platelet aggregation (17,18). When analyzed by CIE, this antibody also appears to bind to a determinant created only by the association of IIb and IIIa (39). Preliminary reports have appeared of other monoclonal antibodies to glycoprotein IIb-IIIa that inhibit fibrinogen binding, but the specific antigenic determinants were not identified (40,41). Recently, Gogstad et al. (42) demonstrated calcium-dependent binding of '251-fibrinogen to the IIb-IIIa immunoprecipitate in CIE gels. Immunoprecipitates of dissociated IIb and IIIa did not react with fibrinogen. These results also suggest a requirement for association of IIb and IIIa in order for fibrinogen binding to occur. It is likely that fibrinogzn binds to a domain where IIb and IIIa are in close proxirmty, although the association of the subunits might cause a conformational change in another region of either IIb or IIIa where fibrinogen would then interact.
It appears that T10 binds to a region near to, but not precisely corresponding to, the fibrinogen binding site of IIb-IIIa, since fibrinogen binding is not totally inhibited by T10 IgG and even less impaired by Fab fragments. Incomplete inhibition may also be explained by the observation of Mar-guerie et al. (43) that fibrinogen becomes irreversibly bound to platelets with increasing time of incubation. If a similar phenomenon occurred in our system, some bound fibrinogen molecules might become resistant to competitive displacement by T10.
In summary, our results suggest that IIb and IIIa exist as polypeptide subunits of a single membrane glycoprotein in unstimulated platelets. Fibrinogen appears to bind to the glycoprotein at or near a region created by the interaction of the two subunits. Platelet activation with an agonist such as ADP is not required for association of IIb and IIIa but is necessary to promote fibrinogen binding and platelet aggregation. Therefore, some additional event in addition to membrane association of IIb and IIIa must be required for fibrinogen binding to proceed. Possible events include a conformational shift in glycoprotein IIb-IIIa, clustering of IIb-IIIa heterodimers thefeby increasing their binding affinity for fibrinogen (37), or interaction of other unknown membrane components with the glycoprotein.