Platelet Membrane Glycoprotein IIb-IIIa Complex Incorporated into Phospholipid Vesicles

Platelet membrane glycoproteins (GP) IIb and IIIa have been identified as platelet aggregation These glycoproteins form a heterodimer complex (GP in the presence of To study the morphology of this glycoprotein complex in membranes, we incorporated GP IIb-IIIa into artificial vesicles of the GP IIb-IIIa Glycoprotein IIb-IIIa incorpora- tion into the vesicles was unaffected by ionic strength, a hydrophobic interaction glycoprotein the phospholipid. was detected on 5 to 20% SDS-polyacrylamide gradient gels under reducing conditions.

Platelet membrane glycoproteins (GP) IIb and IIIa have been identified as platelet aggregation sites. These glycoproteins form a heterodimer complex (GP IIb-IIIa) in the presence of Ca2+. To study the morphology of this glycoprotein complex in membranes, we incorporated GP IIb-IIIa into artificial phospholipid vesicles using a detergent (octyl glucoside) dialysis procedure. Phosphatidylserine-enriched vesicles (70% phosphatidylserine, 30% phosphatidylcholine) incorporated -90% of the GP IIb-IIIa as determined by sucrose flotation. Glycoprotein IIb-IIIa incorporation into the vesicles was unaffected by ionic strength, suggesting a hydrophobic interaction between the glycoprotein and the phospholipid. In both intact platelets or phospholipid vesicles, GP IIb was susceptible to neuraminidase hydrolysis, indicating that most of the glycoprotein complexes were oriented toward the outside of the platelets or vesicles. The morphology of GP IIb-IIIa in the phospholipid vesicles was observed by negative staining electron microscopy. Individual GP IIb-IIIa complexes appeared as spikes protruding as much as 20 nm from the vesicle surface. Each spike consisted of a GP IIb "head," which was distal to the vesicle and was supported by the GP IIIa "tails." The GP IIb-IIIa complex appeared to be attached to the vesicle membrane by the tips of the GP IIIa tails. Treatment of vesicles with EGTA dissociated the GP IIb-IIIa complex. The dissociated glycoproteins remained attached to the phospholipid vesicles, indicating that both GP IIb and GP IIIa contain membraneattachment sites. These data suggest a possible structural arrangement of the GP IIb-111s complex in whole platelets.
(to L. V. P.) and Grants HL28947 and HL32254 from the National * This work was supported by Postdoctoral Fellowship HL06886 Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore he hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence should be addressed Gladstone Foundation Laboratories, P. 0. Box 40608, San Francisco, CA 94140-0608. Several properties of purified GP' IIb and GP IIIa have been determined. When solubilized with Triton X-100 in the presence of Ca2+, these two glycoproteins form the heterodimer complex G P IIb-IIIa (15). The stability of the GP IIb-IIIa complex depends on divalent cations; chelating agents dissociate the complex into monomeric glycoproteins (1). The complex is asymmetrical, as indicated by hydrodynamic measurements (1). The asymmetry was confirmed in the companion paper by rotary shadowing and negative staining electron microscopy, which showed that the complex consists of two domains, an oblong portion or "head" with dimensions of -8 X 10 nm identified as GP IIb that was attached to two elongated "tails" of -15 nm identified as 1 molecule of GP IIIa (15). Upon removal of Triton X-100, the GP IIb-IIIa complexes aggregate by interacting at the tips of the GP IIIa tails (15). This finding suggests that the tips of these tails contain hydrophobic domains, which could be the sites on GP IIb-IIIa that interact with membranes.
The purpose of this study was to identify the sites on GP IIb-IIIa that interact with membranes and to characterize other structural properties of GP IIb-IIIa in membranes. Our approach was to study purified GP IIb-IIIa that had been incorporated into an artificial phospholipid vesicle membrane. This approach has provided structural information on several other membrane-associated proteins, the E l glycoprotein of the Semiliki forest virus (16), the human histocompatibility antigen (17), and the membrane attack complex of complement (18). The present study considers 1) the conditions necessary for the incorporation of GP IIb-IIIa into phospholipid vesicles, 2) the sidedness of GP IIb-IIIa in both the phospholipid vesicles and whole platelets, 3) the identification of sites on GP IIb-IIIa that interact with phospholipid vesicles, and 4) the ability of dissociated GP IIb and GP IIIa to incorporate into phospholipid vesicles. The implications that these findings have on the structural arrangement of GP IIb-IIIa in platelets are presented.
Isolation of Glycoprotein Ilb-IIIa Complex-Platelet membranes were prepared as previously described (1) from clinically outdated human platelet concentrates. Briefly, washed platelets were sonicated in a buffer containing 0.02 M Tris, 0.5 mM CaC12, and 0.15 M NaCl (pH 7.4), with M leupeptin added to inhibit Ca2+-dependent proteases. Membranes were isolated from platelet lysates by centrifugation (3 h, 98,000 X g a t 4 "C in an SW 28 rotor) through an equal volume of 27% sucrose. In some cases, isolated membranes (-3 mg of membrane protein/ml) were labeled with sodium periodate and sodium [3H]borohydride by a previously described procedure (20). Glycoprotein IIb-IIIa was solubilized from labeled or unlabeled membrane preparations with 2% Triton X-100 in the presence of 0.5 mM CaC12 and purified by sucrose density centrifugation and Sephacryl S-300 gel filtration as previously described (1).

Incorporation of Glycoprotein Ilb-IIIa into Phospholipid Vesicles-
To incorporate G P IIb-IIIa into phospholipid vesicles, it was necessary to remove Triton X-100 from G P IIb-IIIa and resolubilize the protein in the dialyzable detergent, octyl glucoside. Triton X-100 was removed by adsorbing purified G P IIb-IIIa onto DEAE-Sephacel (column size, 7.7 X 0.7 cm) and washing the column with 30 ml of a 50 mM Tris buffer containing 0.5 mM CaClZ (pH 7.5). The adhering glycoproteins were then washed with 15 ml of the following buffer: 50 mM Tris, 0.5 mM CaC12, 2 mM NaN3, and 60 mM octyl glucoside (pH 7.5). Glycoprotein IIb-IIIa was eluted from the column with 15 ml of a buffer containing 50 mM Tris, 0.5 mM CaC12, 60 mM octyl glucoside, 2 mM NaN3, and 0.5 M NaCl (pH 7.5). Fractions containing G P IIb-IIIa were located by SDS-gel electrophoresis, pooled, and dialyzed overnight a t 4 "C to decrease the NaCl concentration to 0.1 M. The protein concentration was determined by the method of Lowry et al. (21). The efficiency of the above method for removing Triton X-100 was evaluated with I3H]Triton X-100. At least 99.95% of the Triton X-100 was removed, and -0.003 mg of Triton X-lOO/mg of protein remained (1.4 molecules of Triton X-lOO/molecule of G P IIb-IIIa).
Glycoprotein IIb-IIIa was incorporated into phospholipid vesicles by a modification of the procedure of Helenius et al. (16). Because several protein to phospholipid ratios were used, these ratios are given in the figure legends. Phospholipid, either bovine brain PS, egg yolk PC, or both, and octyl glucoside (four times the weight of lipid) were placed in the bottom of a glass test tube that had been washed with acid, boiled, and dried. In some experiments, trace amounts of ["C] PC were also added. The phospholipids were dried under a stream of filtered (filter from Alltech, Deerfield, IL) nitrogen, and twice dissolved in 0.2 ml of diethyl ether and redried.
Glycoprotein IIb-IIIa in a buffer composed of 50 mM Tris, 0.5 mM CaCI2, 60 mM octyl glucoside, 2 mM NaN3, and 0.1 M NaCl (pH 7.5) was added to the test tube and mixed with phospholipids a t ambient temperature, first by drawing the solution 20 times through a fine glass Pasteur pipette and then by sonicating it for 5 min in a bathtype sonicator (Laboratory Supplies, Co., Inc., Hicksville, NY). This mixture was then dialyzed in the dark for 17 h at room temperature against two changes of a 1000-fold excess of the above buffer without octyl glucoside. This buffer was deaerated under negative pressure and bubbled with nitrogen before use.
Sucrose Flotation-Samples of vesicles in 50% sucrose were placed at the bottom of 1.27 X 5.08-cm cellulose nitrate centrifuge tubes (Beckman Instruments, Palo Alto, CA). Continuous 10 to 40% sucrose gradients were layered on the samples (5.3-ml total volume) and centrifuged at 55,000 rpm (290,000 X g) for 23 h at 4 "C in a Beckman SW 55 swinging-bucket rotor. Gradients were eluted from the bottom, and 20 to 25 fractions were collected. The bottoms of the drained centrifuge tubes were washed with 0.2 ml of 10% SDS to recover any adhering material.
The sucrose concentration of each fraction was determined by refractometry (American Optical, Buffalo, NY). The positions of I3H] G P IIb-IIIa and ["C]PC were determined by dual label counting in a Beckman LS 7500 scintillation counter.

Treatment of Glycoprotein Ilb-IIIa Vesicles with High Ionic
Strength-The NaCl concentration of the solution containing I3H] G P IIb-IIIa vesicles was increased to 0.5 M to disrupt potential electrostatic interactions of the protein with the vesicles (22,23). The vesicles in 0.5 M NaCl and control vesicles in 0.1 M NaCl were sonicated for 20 min in a bath-type sonicator to shear the vesicles so that any GP IIb-IIIa that may have been oriented to the inside of the vesicle could be released. After this treatment, the association of protein with phospholipid was analyzed by sucrose flotation.
Electron Microscopy-For visualization by electron microscopy, vesicles were negatively stained with 1% phosphotungstic acid on a carbon film and applied to a Formvar-coated grid. The grid was blotted and dried before being viewed in a Jeol 100 CX2 electron microscope at 80 kV.
Neuraminidase-catalyzed Hydrolysis-Freshly isolated washed platelets in Tyrode's buffer (24) or G P IIb-IIIa-containing vesicles were incubated with 5 units/ml neuraminidase for 15 min a t 37 "C (1.2 X lo7 platelets in 50 p1 or 44.8 fig of G P IIb-IIIa in phospholipid vesicles in a 50-pl final volume). Platelet samples were pelleted in an Eppendorf centrifuge for 5 min. The extent of neuraminidase-catalyzed hydrolysis of G P IIb-IIIa in platelet pellets and in vesicle samples was determined by examining the isoelectric points of G P IIb,' and GP IIIa by two-dimensional gel electrophoresis before and after neuraminidase treatment.
Two-dimensional gel electrophoresis was performed by the method of O'Farrell (25) as modified by Ames and Nikaido (26). For these experiments, platelet pellets were dissolved in 55 p1 of a buffer containing 2% SDS, 0.05 M Tris (pH 6.81, and 0.5 mM MgC12 and heated for 10 min at 100 "C. Glycoprotein IIb-IIIa vesicles were diluted 4:l (v/v) with the same buffer (concentrated 5-fold) and heated as above. As an internal standard for the control positions of G P IIb and GP IIIa, a trace amount of platelets labeled with lZ5I by the lactoperoxidase technique (27) was added to the vesicle sample. Vesicle or platelet samples were then diluted 1:2 (v/v) with a solution of 9.5 M urea, 5% p-mercaptoethanol, 8% Nonidet P-40, 0.75 mM EDTA, and 2% Ampholines (0.4% at pH 3.5 to 10 and 1.6% at pH 5 to 8). Reduced samples (70 pl of platelet or 100 pl of vesicle samples) were electrophoresed in the first dimension by isoelectric focusing on gels of 3.25% acrylamide containing 2% Ampholines (pH ranges as above) and, in the second dimension, on 5 to 20% SDS-polyacrylamide gradient gels (28). Proteins were located by Coomassie Blue staining or silver staining (29); 1Z51-labeled G P IIb and G P IIIa were detected by autoradiography on Cronex 4 film with a Cronex intensifying screen (DuPont, Wilmington, DE).
Platelet lysis, which may have occurred during the 15-min incubation of platelets with neuraminidase, was measured by the presence of extracellular lactate dehydrogenase (30).
Incorporation of Dissociated Glycoprotein IIb and Glycoprotein IIIa into Phospholipid Vesicles-To form vesicles with dissociated G P IIb and GP IIIa, 1 mM EGTA was added to I3H]GP IIb-IIIa (100 pg/ml) in 50 mM Tris, 0.035 mM CaCl2, 60 mM octyl glucoside, 2 mM NaN3, and 0.1 M NaCl (pH 7.5). The 3H-glycoprotein was then added to 70% PS:30% PC (w/w) (protein:phospholipid, 1:3.6, w/w) that had been previously dried onto the wall of a glass test tube. Vesicles were formed as described except that octyl glucoside was removed by G P IIb consists of two disulfide-linked subunits: G P IIb, ( M , = 116,000 under reducing conditions) and G P IIbs (M, = 23,000). It is the GP IIb, subunit that is actually identified on the reduced SDSpolyacrylamide gels in Figs. 3, 4, 6, and 7; G P IIb, is sensitive to hydrolysis by neuraminidase or thrombin. dialysis into a Tris buffer containing 1 mM EGTA and no CaC12. The dissociation of G P IIb from G P IIIa was verified by thrombin hydrolysis of [3H]GP IIbm2as described below. The extent of glycoprotein incorporation into vesicles was determined by sucrose flotation, with the sucrose dissolved in the EGTA dialysis buffer described above.
Treatment of Glycoprotein Ilb-llla Vesicles with EGTA-The effect of EGTA on the GP Ilb-IIIa complex incorporated into vesicles was determined by incubating [3H]GP IIb-IIIa vesicles (formed in the presence of 0.5 mM CaC12, 100 pg/ml [3H]GP IIb-IIIa, 70% PS:30% PC, and protein:phospholipid, 1:3.6, w/w) with EGTA (10 mM, pH 7.5) for 1 h at 37 "C. The effect of EGTA on the amount of [3H]GP IIb-IIIa complexes was determined by thrombin hydrolysis; the effect of EGTA on 3H-glycoprotein association with vesicles was determined by sucrose flotation, with the sucrose dissolved in a buffer of 50 mM Tris, 10 mM EGTA, 2 mM NaN3, and 0.1 M NaCl (pH 7.5).
Thrombin-catalyzed Hydrolysis-Thrombin-catalyzed hydrolysis of GP 114 was used as a probe to measure the extent of G P IIb dissociation from G P IIIa (10). Vesicles in the presence of 0.5 mM CaC12 (control) or vesicles in the presence of 1 or 10 mM EGTA were incubated with thrombin (10 p M ) or an equivalent volume of buffer (0.01 M Tris, 0.75 M NaCl, and 5% polyethylene glycol 6000, pH 7.4) for 1 h a t 37 "C. The hydrolysis of G P IIb,2 was detected on 5 to 20% SDS-polyacrylamide gradient gels under reducing conditions. the highest association of [3H]GP IIb-IIIa (-90%) occurred with vesicles formed from an initial ratio of 70% PS:30% PC (w/w) (Fig. 1D). In the peak fraction of ["C]PC and [3H]GP IIb-IIIa (Fig. lD), the ratio of PS:PC was actually 60%:40%, as determined by phosphorus analysis of PS and PC following separation of the phospholipids by TLC. The 13H]GP IIb-IIIa-containing vesicles in Fig. 1D floated at a density of 1.09 g/cm3, intermediate to that of phospholipid vesicles without GP IIb-IIIa (1.06 g/cm3, Fig. 1F) or GP IIb-IIIa without phospholipid (>1.20 g/cm3, Fig. 1E).
High ionic strength (0.5 M NaCl) did not decrease the amount of [3H]GP IIb-IIIa recovered in the fractions containing ["C]PC (Fig. 2B) below that recovered in the control (0.1 M NaC1) (Fig. 2 A ) . The inability of 0.5 M NaCl to disrupt the interaction of [3H]GP IIb-IIIa with the phospholipid vesicles indicated that this interaction was not electrostatic.
Sidedness of Glycoprotein ZZb-ZIZa in Intact Platelets and in Phospholipid Vesicles-Neuraminidase, which hydrolyzes sialic acid residues resulting in a shift of the isoelectric point of glycoproteins, was used as a probe to compare the sidedness of GP IIb-IIIa in platelets to that in phospholipid vesicles. In neuraminidase-treated platelets, the isoelectric point of GP IIb2 was much more basic than in control platelets (most of the GP IIb was between 5.60 and 5.90 for neuraminidasetreated platelets (Fig. 3B) versus 5.23 and 5.65 for control platelets (Fig. 3A)). This change occurred without platelet lysis; ~2 % of the platelet lactate dehydrogenase was present in the supernatant of platelets incubated with neuraminidase, indicating that neuraminidase was accessible only to cellsurface glycoproteins. Thus, most of the GP IIb in platelets was exposed to neuraminidase and, therefore, oriented to the outside of the membrane.
In contrast to its effect on GP IIb, neuraminidase caused only a slight shift in the isoelectric point of GP IIIa in whole platelets (5.35-5.55 in neuraminidase-treated platelets (Fig.  3B) uersus 5.20-5.50 in control platelets (Fig. 3A)), suggesting that either GP IIIa contained less sialic acid than GP IIb or that the sialic acid on GP IIIa was inaccessible to neuraminidase. However, incubation of purified G P IIb-IIIa (in 48 mM octyl glucoside) with neuraminidase resulted in a shift in the isoelectric point of G P IIIa (and G P IIb) similar to that observed in whole platelets (data not shown). This indicates that GP IIIa contains less sialic acid than GP IIb, a conclusion supported by observations showing that GP IIIa labels less intensely than G P IIb by the periodate/sodium ['Hlborohydride procedure (20). Thus, susceptibility of G P IIb, but not G P IIIa, to neuraminidase-catalyzed hydrolysis can be used as a measure of the sidedness of the GP IIb-IIIa complex in platelets and in phospholipid vesicles.
In neuraminidase-treated phospholipid vesicles, as in whole platelets, most of the detectable G P IIb2 had an isoelectric point that was more basic than that of the GP IIb in control vesicles (Fig. 4), indicating that the GP IIb-IIIa complex was oriented to the outside of the vesicles. Glycoprotein IIIa appeared to have an isoelectric point that was only slightly more basic than that of untreated G P IIIa (Fig. 4) as expected from the results obtained with whole platelets (Fig. 3) or purified G P IIb-IIIa.
Morphology of the Phospholipid Vesicles-Electron microscopic examination showed that vesicles prepared without G P IIb-IIIa were spherical with smooth perimeters and had diameters of 40 k 8 nm (S.D.) (Fig. 5A). Vesicles prepared with G P IIb-IIIa were irregularly shaped and had diameters of 43 A. Glycoprotein IIh-IIIa was incorporated into phospholipid vesicles (70% PS:30W PC (w/w), protein:phospholipid, 13.6 (w/w), and 400 pg/ml GP IIb-IIIa) and incubated with 5 units/ml neuraminidase. The reaction was stopped by heating, and a trace amount of InsIlabeled whole platelets was added as an internal standard for the control positions of G P IIb2 and G P IIIa. The sample was analyzed by t,wo-dimensional gel electrophoresis with silver stain ( A ) and autoradiography of the silver-stained gel (R). Dotted lines on the gel in A represent control positions of G P IIb2 and GP IIIa obtained from the autoradiogram in R. & 23 nm (S.D.) (Fig. 5B). Many of these vesicles were surrounded by spikes that extended up to 20 nm from the vesicle surface. Each spike consisted of a "head" connected to the vesicle membrane by two "tails" (Fig. 5B). The heads and tails resembled the components of the purified glycoprotein complex that were observed by electron microscopy in the companion paper and identified as GP IIb and GP IIIa, respectively (15). The tips of the tails appeared to be intercalated into the phospholipid vesicles. Occasionally, a structure resembling a GP IIb-IIIa aggregate or "rosette" was seen (Fig. 5B, large arrow) (15). The exclusion of negative stain from these vesicles suggested that they were sealed.
Incorporation of the Dissociated Glycoproteins into Phospholipid Vesicles-To determine whether the dissociated GP IIb and GP IIIa could be incorporated into phospholipid vesicles, "H-glycoprotein-containing vesicles were formed in the presence of EGTA (1 mM). The dissociated glycoproteins did incorporate into the phospholipid vesicles; both ['HH]GP IIb and ["HIGP IIIa floated through a sucrose gradient (Fig. 6A). In a control experiment, vesicles containing the ["HIGP IIb-IIIa complex were formed in the presence of CaC12. The ['HI G P IIb-IIIa complex floated to a similar density (Fig. 6B) as that of the dissociated glycoproteins in Fig. 6A. Analysis of the individual gradient fractions on SDS-polyacrylamide gels showed that ['HIGP IIb floated with ['HIGP IIIa in both gradients (data not shown).
To confirm that ['HIGP IIb and ["HIGP IIIa were either dissociated (Fig. 6A) or complexed (Fig. 6B), samples of the phospholipid vesicles were treated with thrombin. Thrombin can discern complexed from dissociated glycoproteins since G P IIb2 is a thrombin substrate only when dissociated from G P IIIa (10). The ['HIGP IIb was completely hydrolyzed by thrombin in vesicles formed in the presence of EDTA (Fig.  6A, inset) but was only partially hydrolyzed in vesicles formed in the presence of CaC12 (Fig. 6B, inset). These results demonstrate that the vesicles prepared with EDTA contained dissociated glycoproteins, and vesicles prepared in CaC12 contained primarily complexed glycoproteins.

Effect of EGTA on the Glycoprotein Ilb-IIIa Complex in
Phospholipid Vesicles-Previous studies have shown that treatment of whole platelets with Ca2+ chelators dissociates the Ca"-dependent GP IIb-IIIa complex2 (33) but does not cause a loss of either glycoprotein from the platelet surface: T o determine whether these conditions also hold true for G P IIb-IIIa in phospholipid vesicles, ["HJGP IIb-IIIa-containing vesicles were incubated with EGTA (10 mM) under conditions known to dissociate the Ca"-dependent GP IIb-IIIa complex. Dissociation of the complex was confirmed by thrombincatalyzed hydrolysis of [:3H]GP IIb' in the EGTA-treated vesicles (Fig. 7A, inset) (10). The flotation of the dissociated ["HIGP IIb and ["HIGP IIIa through the sucrose gradient (Fig. 7A) and the analysis of these sucrose fractions by SDSpolyacrylamide gel electrophoresis (data not shown) demonstrated that ["HIGP IIb and ["HIGP IIIa remained attached to the phospholipid vesicles. In a control experiment, ["HIGP IIb-IIIa vesicles were incubated with CaCl, instead of EGTA.
The "H-glycoprotein in these vesicles floated to a density (Fig.  7 R ) similar to that of the dissociated "H-glycoproteins in Fig.  SA. Most of the ["HIGP IIb' in the control vesicles was resistant to thrombin hydrolysis (Fig. 7R, inset), demonstrating the presence of the ["HIGP IIb-IIIa complex.
Glycoprotein-containing vesicles treated with EGTA as described above were examined by electron microscopy as in Fig. 5 to determine their morphology. However, no structures could be clearly resolved on these vesicles (data not shown).

DISCUSSION
The present study demonstrated that 1) G P IIb-IIIa is incorporated into phosphatidylserine-enriched vesicles by the tips of the GP IIIa tails, an interaction that is most likely hydrophobic; 2) G P IIb-IIIa is oriented to the outside of these vesicles, an orientation that also exists in platelets; 3) dissociated G P IIb and IIIa can be incorporated into phospholipid vesicles; and 4) the GP IIb-IIIa complex in phospholipid vesicles is dissociated by EGTA.
Glycoprotein IIb-IIIa was readily incorporated into vesicles enriched in phosphatidylserine but was poorly incorporated into vesicles enriched in phosphatidylcholine. Maximal GP IIb-IIIa incorporation was achieved with vesicles formed from 70% PS and 30% PC. The effectiveness of PS in promoting protein incorporation into phospholipid vesicles is not limited to GP IIb-IIIa; other proteins, such as coagulation factors V and Va (34) and the acetylcholine receptor (35), are also incorporated readily into PS-containing vesicles. Although it is not known why PS promotes GP IIb-IIIa incorporation into phospholipid vesicles, we have considered several possible explanations. One is that electrostatic interactions may occur between the acidic phospholipid (PS) and the positive charges on the glycoprotein. This is not likely to be a major factor in incorporation, however, because 0.5 M NaCl, which would neutralize such interactions (22, 23), did not dissociate GP IIb-IIIa from the phospholipid vesicles. A second possibility is that acidic phospholipids such as PS tend to form defects in the bilayer structure, which may be required for glycoprotein incorporation (23). A third possibility is that in platelet membranes, PS or similar phospholipids may be specifically bound to GP IIb-IIIa. In artificial membranes, high concentrations of PS may be required to achieve concentrations that may be equivalent to local concentrations in the platelet membrane. At present, it is not possible to distinguish between these latter two possibilities.
The hydrolysis of glycoproteins by neuraminidase has been used extensively to determine the sidedness of glycoproteins in the membranes of cells and vesicles (36-38), because neuraminidase specifically hydrolyzes sialic acid residues but does not permeate membranes. In the present study, neuraminidase-catalyzed hydrolysis was used to compare the orientation of GP IIb-IIIa in whole platelets to that in phospholipid vesicles. The results of experiments with whole platelets and phospholipid vesicles were similar; neuraminidase treatment completely shifted the isoelectric point of GP IIb' to a more basic position, as detected by two-dimensional gel electrophoresis. Because this enzyme did not cause cell lysis (as determined by the lack of extracellular lactate dehydrogenase), it can be assumed that neuraminidase-catalyzed hydrolysis is restricted to the outer membrane surface (and possibly the open canalicular system) of platelets. These results demonstrate that most of the GP IIb is oriented to the outside of both platelets and phospholipid vesicles. Since considerable overlap exists between the isoelectric points of control and neuraminidase-treated G P IIIa, no conclusion can be made concerning the sidedness of GP IIIa. However, because GP IIb and GP IIIa exist as a complex in whole platelets" (33) and in phospholipid vesicles, it is likely that not only GP IIb but also the GP IIb-IIIa complex is oriented to the outside of platelets and phospholipid vesicles. The exclusive orientation of GP IIb-IIIa to the outside of vesicles is not surprising given the small diameter of the vesicles (-40 nm) and the limited internal space to accommodate 20-nm protrusions of G P IIb-IIIa complexes.
In the companion paper (15), it was reported that the GP IIIa tails of purified G P IIb-IIIa interact with each other in the center of rosettes that form in the absence of detergent. This observation indicates that the tails are hydrophobic and are potential sites a t which G P IIb-IIIa could interact with membranes. Indeed, this seems to be the case. Vesicles formed with GP IIb-IIIa were found by electron microscopy to have spikes protruding as much as 20 nm from the membrane surface. The morphological similarity of these spikes to purified G P IIb-IIIa (15) suggests that each spike is a G P IIb-IIIa complex attached to the membrane by the ends of the GP IIIa tails; the G P IIb head is distal to the membrane and supported by the tails.
The conclusion that the tips of the GP IIIa tails on the GP IIb-IIIa heterodimer complex are hydrophobic and interact with the phospholipid membrane is consistent with three observations. First, G P IIb-IIIa binds 0.30 mg of Triton X-100/mg of protein (15), as determined by the method of Clarke (39). This amount is lower than that of proteins in the erythrocyte membrane, i.e. band 3 (0.77 mg of Triton X-loo/ mg of protein) and PAS-1 (1.12 mg of Triton X-lOO/mg of protein) (39). Based on these binding data, it was calculated that <18% of the protein surface is hydrophobic (15), suggesting that the hydrophobic domains exist on limited regions of the GP IIb-IIIa complex. Second, aggregates of GP IIb-IIIa prepared in the absence of detergent were shown to consist of rosettes of GP IIb-IIIa complexes interacting at the tips of the GP IIIa tails (15). This condition suggests that the hydrophobic domains are restricted to the tips of the tails of the G P IIb-IIIa complex. Third, the results of the present study show that GP IIb-IIIa is attached to the phospholipid vesicle membrane by the tips of the GP IIIa tails, as observed by electron microscopy. Thus, the hydrophobic domains of GP IIb-IIIa, which exist primarily at the tips of the GP IIIa tails, provide a membrane-attachment site for the heterodimer complex.
The model of GP IIb-IIIa interaction with membranes presented above suggests that only G P IIIa interacts directly with membranes. However, when glycoprotein-containing vesicles were formed in the presence of EGTA, a condition that dissociates G P IIb from G P IIIa, both glycoproteins were still incorporated into the phospholipid vesicles. This result suggests that GP IIb also contains sites that interact with membranes. We have considered two possible explanations for this finding: 1) GP IIb, in both the complexed or dissociated state, may attach to the vesicle membrane by a hydrophobic structure that is not resolved by the electron micro-scopic technique used; or 2) the dissociation of GP IIb from GP IIIa may cause hydrophobic sites on GP IIb to be expressed, which allows G P IIb to interact with membranes. At this time, it is not possible to distinguish between these two possibilities.
It has recently been shown that treatment of whole platelets with Ca2+ chelators results in the loss of the GP IIb-IIIa CaZ+dependent complex3 (33) and that both dissociated glycoproteins remain attached to the late let.^ To determine if the same holds true for GP IIb-IIIa complex in phospholipid vesicles, vesicles were treated with EGTA under conditions known to dissociate the Ca2+-dependent complex. Susceptibility of G P IIb2 to hydrolysis by thrombin demonstrated that the GP IIb-IIIa in these vesicles was dissociated by EGTA. Thus, the Ca2+-sensitive site on G P IIb-IIIa that is responsible for the heterodimer complex formation is accessible to EGTA in phospholipid vesicles, as it is in platelets. In addition, treatment of G P IIb-IIIa vesicles with EGTA did not result in a loss of either glycoprotein from the vesicle, a result similar to that obtained with platelets.
At present, it is not possible to determine whether the GP IIb-IIIa structure observed in phospholipid vesicles also exists in the platelet membrane. This is because the platelet membrane is surrounded by the glycocalyx, a layer composed primarily of glycoproteins, which extends 15 to 20 nm from the membrane surface and appears as a complex structure in electron micrographs. However, several lines of evidence suggest that the structural arrangement of GP IIb-IIIa observed in vesicles may represent that in the platelet membrane: 1) GP IIb-IIIa in phospholipid vesicles and in whole platelets is accessible to neuraminidase; 2 ) the Ca2+-sensitive site responsible for G P IIb-IIIa complex formation is accessible to Ca2+ chelators in both phospholipid vesicles and whole platelets; 3) Ca2+ chelators do not affect the amount of G P IIb or GP IIIa that remains attached to the membranes of vesicles or whole platelets; and 4) in simpler systems, the morphology of proteins in reconstituted vesicles has been shown to reflect that of proteins in the native membrane. For example, the 7.7-nm spikes in phospholipid vesicles containing the E l glycoprotein resemble projections on the native Semiliki forest virus membrane (16), and the "fuzzy" coat on reconstituted G protein-lipid vesicles resembles the native vesicular stomatitis virus (40). Thus, the structural arrangement of GP IIb-IIIa observed in artificial phospholipid vesicles probably exists in platelet membranes, and the glycocalyx observed on intact platelets probably includes protrusions of GP Ilb-IIIa heterodimer complex. These results show that most of the GP IIb-IIIa molecule could be accessible to external ligands. This accessibility may explain the functional diversity of this glycoprotein. We suggest that these reconstituted vesicles, which contain surface-oriented glycoproteins, may be useful for studying the functional properties of G P IIb-IIIa in a defined environment.