A Molecular Basis for Affinity Modulation of Fab Ligand Binding to Integrin αIIbβ3

The Arg-Gly-Asp (RGD) sequence within the third complementarity-determining region (CDR3) of the heavy chain (H3) is responsible for the binding of the recombinant murine Fab molecules, AP7 and PAC1.1, to the platelet integrin αIIbβ3. AP7 binding is minimally influenced by the conformational state of this receptor, whereas PAC1.1 binds preferentially to the activated state of the receptor induced by platelet agonists. To study the molecular basis for this functional difference, we replaced the AP7 H3 loop (HPFYRGDGGN) with all or segments of the analogous sequence from PAC1.1 (RSPSYYRGDGAGP). AP7 Fd (VH domain + Cγ1 domain) segments containing these H3 loop sequences were expressed as active Fab molecules by coinfection of Spodoptera frugiperda cell lines with recombinant baculoviruses containing Fd and AP7 κ chain cDNA. Replacement of the entire AP7 H3 loop with that from PAC1.1 generated the mutant AP7.3 Fab molecule, which bound selectively to either activated, gel-filtered platelets or to purified αIIbβ3 in a manner identical to that of PAC1.1. Identical results were obtained when solely the sequences flanking the amino side of RGD within the respective H3 loops were exchanged. AP7.3 and PAC1.1 exhibited saturable but submaximal binding to activated gel-filtered platelets. Relative to AP7, the number of AP7.3 or PAC1.1 Fab molecules bound per platelet was 17% in the presence of 1 mM Ca2+ + 1 mM Mg2+ or 40% in the presence of 10 μM Mn2+. The ratio of Fab molecules bound after versus before activation (mean ± S.D.; n = 3) was: for AP7.3, 9.8 ± 0.6; for PAC1.1, 8.8 ± 0.3; and for AP7, 1.4 ± 0.2. In addition, AP7 bound to the stably expressed integrin mutant αIIbβ3(S123A), whereas AP7.3 and PAC1 did not. Because AP7.3 behaves in every respect like PAC1.1, we conclude that the ability of RGD-based ligands to distinguish activated from resting conformations of the integrin αIIbβ3 can be regulated by limited amino acid sequences immediately adjacent to the RGD tripeptide. Furthermore, those Fab molecules that exhibit increased selectivity for the activated conformation of αIIbβ3 bind to a subpopulation of this integrin on platelets that is modulated by divalent cations.

Several RGD-containing adhesive proteins, such as fibrinogen, fibronectin, or von Willebrand factor, when presented in soluble form, will bind to the platelet integrin ␣ IIb ␤ 3 only when the platelets are first stimulated by an appropriate agonist (1).
The molecular properties responsible for this selective binding of macromolecular ligands to ␣ IIb ␤ 3 on activated platelets are not fully understood. Ligand size is not a critical factor, because small RGD peptides and peptidomimetics, 5-10-kDa RGD-containing disintegrins, the 150-kDa monoclonal antibody OPG2, and the large 600-kDa tetramer of the murine monoclonal antibody 7E3 all bind to ␣ IIb ␤ 3 whether or not platelet activation has occurred, although platelet activation does result in a severalfold increase in apparent binding affinity for some of these ligands (2)(3)(4)(5). Ligand valency is also not a likely explanation for selective binding to activated conformations of ␣ IIb ␤ 3 , because the binding of the monoclonal antibody PAC1 to platelets remains dependent upon the activated state of ␣ IIb ␤ 3 whether the antibody is presented as a 50-kDa recombinant Fab molecule or a 900-kDa pentameric IgM (6).
To address the molecular basis of conformation-sensitive binding of an RGD ligand, we exploited the binding properties of two well characterized RGD ligands, AP7 and PAC1.1, which are the RGD-containing analogs of the previously described RYD-containing antibodies, OPG2 and PAC1, respectively (6,7). The characterization of PAC1.1 was accomplished in this study. AP7 Fab molecules bind with high affinity to either the nonactivated or the activated conformational states of ␣ IIb ␤ 3 . On the other hand, recombinant PAC1.1 Fab molecules bind more avidly to the activated conformation of ␣ IIb ␤ 3 , as reflected by an increase in affinity and number of sites. In this regard, the behavior of PAC1.1 is more similar to that of other macromolecular RGD ligands specific for ␣ IIb ␤ 3 , e.g. fibrinogen or von Willebrand factor (1,8,9).
Because the heavy chain CDR3 (H3) 1 loops of either AP7 or PAC1.1 contain an RGD motif but differ otherwise in length and flanking amino acid composition, we asked whether differences in amino acid sequence of these flanking regions could be a major factor contributing to the ability of these Fab molecules to distinguish activated from resting conformations of the integrin. By comparing the binding properties of recombinant Fab molecules in which these flanking H3 sequences have been exchanged, we report here that this is indeed the case. An additional unexpected finding, albeit consistent with previous observations, is that the activation-dependent Fab molecules PAC1.1 and AP7.3 bind to a subpopulation of ␣ IIb ␤ 3 on activated platelets. This subpopulation is still distinguishable when the total integrin population is isolated from platelets. The proportion of the integrin that is bound by PAC1.1 or AP7.3 can be modulated by divalent cation composition.

Synthesis of Recombinant OPG2 Fd and Chain cDNA-The term
Fd denotes an Ig heavy chain segment that includes the V H domain, the C␥1 domain, and the hinge region up to and including the cysteine residue that forms a disulfide bond with the light chain (7). Fab molecules represent disulfide-linked heterodimers composed of Fdϩ.
AP7 and PAC1.1 Fd and chain cDNAs were prepared as described previously (6,7). An additional AP7 Fd construct was generated in which the oligonucleotide sequence (CATCAC) 3 was inserted just upstream of the TGA stop codon. This encodes a carboxyl-terminal (His) 6 sequence, which was employed to purify Fab or Fd molecules by nickel affinity matrix chromatography, as described (6).
The remaining constructs containing the H3 sequences depicted in Fig. 1 (AP7.1, AP7.2, and AP7.3) were generated by splice overlap extension polymerase chain reaction (7). AP7 Fd cDNA (7) was used as a template to generate AP7.1, AP7.2, and AP7.3 Fd cDNAs. As detailed below, each construct was generated by first producing two cDNA fragments, a 5Ј fragment, generally encompassing the Fd signal sequence up to a point within H3, and a 3Ј fragment, beginning in H3 and extending to the TGA stop codon. The 5Ј and 3Ј fragments were ligated, and a final amplification of each DNA with the primer pair HFOR and HREV was performed (Table I). A BglII/XbaI digest of the cDNA product was then ligated into pVL1392.
The 5Ј fragment of AP7.1 was obtained with the primers HFOR and CTRREV. The primer CTRREV adds an AflIII site while retaining the correct amino acid sequence. The primer AP7.1FOR adds the AflIII site and changes the first three amino acids (HPF) of AP7 H3 to the first five amino acids (RSPSY) of PAC1.1 H3. AP7.1FOR was used in combination with HREV to produce the 3Ј fragment of AP7.1. Both fragments were digested with AflIII and ligated. Ligated cDNA served as template for subsequent polymerase chain reaction reactions using HFOR and HREV to amplify AP7.1 cDNA.
The 3Ј fragment of AP7.2 was generated using the primer AP7.2FOR in combination with HREV. AP7.2FOR changes the last two amino acids (GN) of AP7 H3 to the last three amino acids (AGP) of PAC1.1 H3. The cDNA product and the 5Ј fragment of AP7 were then digested with SacII and ligated. Oigonucleotides HFOR and HREV were then used to amplify the AP7.2 cDNA.
AP7.3 was constructed by digesting AP7.1 cDNA and AP7.2 cDNA with SacII and ligating the 5Ј fragment of AP7.1 with the 3Ј fragment of AP7.2. Ligated cDNA was used as template for polymerase chain reaction employing the HFOR and HREV primer pair.
Cloning and Analysis of Recombinant Baculoviruses-Recombinant viruses were cloned by infection of Sf9 cells (Invitrogen) (2 ϫ 10 6 in 2 ml of complete Grace's medium) seeded in T25 culture flasks, as described (6,7). The sequence of each recombinant clone was confirmed prior to its use, using Sequenase version 2.0 (U.S. Biochemical Corp.). Recombinant viruses were used to coinfect either Spodoptera frugiperda Sf9 or High Five insect cells (Invitrogen, Inc.), and Fab molecules were harvested from the medium, normally after 72-h cultures, as described (6,7). Recombinant Fd and chains were detected by a quantitative Western blot assay using rabbit polyclonal anti-murine Fdϩ antibody, developed in our laboratory (7). Protein concentration was determined by the method of Markwell (10).
Purified Integrin Enzyme-linked Immunosorbent Assay-Integrins ␣ IIb ␤ 3 and ␣ V ␤ 3 were purified, as described (7,11). Purified integrin heterodimers were adsorbed onto the wells of Immulon II microtiter plates (Dynatech, Inc., Chantilly, VA), and the ability of murine monoclonal Fab molecules or recombinant proteins to bind to each integrin was assessed by enzyme-linked immunosorbent assay (7).
Flow Cytometry Analysis of Platelets-Gel-filtered platelets were prepared as described (6,7). The recombinant Fab molecules in Tyrode's buffer were added to 5 ϫ 10 5 platelets in the presence of either 20 ng/ml PGE 1 or 0.2 M phorbol myristate (PMA). After a 15-min incubation at ambient temperature, fluorescein isothiocyanate-labeled goat anti-mouse IgG (F(abЈ) 2 -specific; Jackson Immunoresearch Laboratories, Inc., West Grove, PA) was added. After an additional 15-min incubation in the dark, samples were diluted 10-fold with Tyrode's buffer and analyzed on a Becton Dickinson FACScan apparatus, as described (6,7).
Flow Cytometry of CHO Cell Transfectants-Stably transfected CHO cell lines were established by electroporation with the wild type or mutant ␤ 3 constructs together with the wild type ␣ IIb construct CD2b (12). In this study, a mutant ␤ 3 was employed in which Ser 123 has been substituted with Alanine (␤ 3 (S123A)). Transfected cell lines were incubated (activated) in the presence of 10 M Mn 2ϩ for 10 min followed by excess recombinant Fab molecules. After 30 min, cells were washed and incubated with fluorescein isothiocyanate-goat anti-mouse Fab (HϩL) for 30 min. Bound Fab molecules were then analyzed, as described (12).

Immunochemical Characterization of Recombinant Anti-
Molecules-AP7 and PAC1 Fab molecules are completely distinct at the level of primary sequence other than the presence of an RGD sequence in the H3 of AP7 and an RYD in H3 of PAC1 (7) (Fig. 1). Because the number and composition of the flanking H3 amino acid residues are different in AP7 and PAC1,weaskedwhetherAP7couldbeendowedwiththeactivationdependent binding characteristics of PAC1 by replacing the H3 flanking residues in AP7 with those of PAC1. Fig. 1 depicts the recombinant Fab molecules that were constructed and expressed in insect cells to address this question. Note that all AP7 variants contained the AP7 scaffold, whereas PAC1.1 contained the PAC1 scaffold. In AP7.1, the amino-terminal H3 flank of AP7 was replaced with that of PAC1; in AP7.2, the carboxyl-terminal flank of AP7 was replaced with that of PAC1; and in AP7.3, both flanks were replaced with those of PAC1. 72 h after co-infection of High Five cells with the AP7 light chain baculovirus (pVL) and each of four pVLFd heavy chain viruses (AP7, AP7.1, AP7.2, and AP7.3), the average yields of Fab molecules were essentially equal for all of the constructs: AP7, 19.4 Ϯ 1.1 g/ml (mean Ϯ S.D., n ϭ 5); AP7.1, 20.5 Ϯ 4.2 g/ml (n ϭ 3); AP7.2, 22.8 Ϯ 1.8 g/ml (n ϭ 3); and AP7.3, 20.6 Ϯ 2.3 g/ml (n ϭ 3). Identical results were obtained when platelet and integrin binding studies detailed below were carried out with Fab molecules in protein-free culture supernatant or with hexahistidine-tagged versions of these Fab molecules purified by metal ion chelate chromatography.
Binding of Recombinant Fab Molecules to Platelets-Before exchanging AP7 flanking residues with those of PAC1, we tested whether substitution of a glycine for tyrosine in RYD of PAC1 exerted any effect on Fab function. The binding of recombinant PAC1.1 Fab molecules (containing RGD) to gelfiltered platelets was compared with the binding of PAC1 Fab (RYD) by flow cytometry. Both PAC1 and PAC1.1 bound selectively and equivalently to platelets activated by PMA, but neither antibody bound to unstimulated platelets. In both cases, Fab binding to activated platelets was inhibited by EDTA (Fig. 2). In contrast, as described previously (7), a PAC1 derivative in which RYD was replaced by RYE failed to bind to platelets under any conditions. Thus, the R(Y/G)D sequence in PAC1 is required for antibody binding to ␣ IIb ␤ 3 , but the tripeptide sequence by itself does not determine the activationdependent behavior of the antibody.
Saturable binding to nonactivated platelets in the presence of 1 mM Ca 2ϩ plus 1 mM Mg 2ϩ was observed for AP7 Fab molecules at concentrations Ն2 g/ml (Fig. 3A). Following platelet activation with PMA, the number of AP7 Fab molecules bound at saturation increased by approximately 40%, almost certainly due to activation-induced surface expression of internal pools of ␣ IIb ␤ 3 (13,14). In sharp contrast to these results for AP7, the binding of AP7.1, AP7.3, or PAC1.1 Fab molecules to nonactivated platelets was negligible (Fig. 3A) and not significantly greater than the binding of nonspecific murine Fab molecules (not shown). Following platelet activation, how-ever, saturable binding of AP7.1, AP7.3, and PAC1.1 was observed at concentrations Ն10 g/ml, although the extent of binding was less than that of AP7. In each case, Fab binding was completely inhibited by RGDW (Ն10 M) or EDTA (5 mM) (not shown). AP7.2 Fab molecules failed to bind to platelets regardless of the state of platelet activation or divalent cation composition (Fig. 3, A and B). We conclude that the pattern of binding of AP7.1 and AP7.3 to platelets is identical to that of PAC1.1 rather than to that of the parent AP7 Fab molecule.
Divalent cations have a profound effect on the binding of fibrinogen or von Willebrand factor to ␣ IIb ␤ 3 , and the same was true for the Fab molecules studied in this report. In the presence of 10 M Mn 2ϩ , which is known to augment the ligand binding activity of ␣ IIb ␤ 3 (15), there was little effect on the binding of AP7, but PMA-induced increases in binding of AP7.1, AP7.3, and PAC1.1 to platelets were more pronounced than those observed in 1 mM Ca 2ϩ plus Mg 2ϩ . Saturable binding of these Fab molecules occurred at lower input concentrations (Ն2 g/ml), and the number of molecules bound at saturation increased (compare results in Fig. 3A with those in Fig.  3B). For example, relative to AP7, the number of AP7.3 or PAC1.1 Fab molecules bound per activated platelet was 17% in the presence of 1 mM Ca 2ϩ plus Mg 2ϩ and 40% in the presence of 10 M Mn 2ϩ . Fig. 4 provides a direct comparison of the activation-dependent behavior of these various Fab molecules when studied under identical conditions of divalent cation composition, i.e. in the presence of 1 mM Ca 2ϩ plus Mg 2ϩ .
Binding of Recombinant Fab Molecules to Expressed Stable Mutant ␣ IIb ␤ 3 Transfectants-For reasons that are not clear but not related to differences of affinity or activation state of the integrin, antibody OPG2, the parent molecule of AP7, binds to CHO cells stably expressing the mutant integrin ␣ IIb ␤ 3 (S123A) whereas PAC1 does not (12). This represents an independent peculiarity of these two antibodies. For this reason, we compared the binding of AP7, AP7.3, and PAC1 to CHO cells stably expressing wild type ␣ IIb ␤ 3 or mutant ␣ IIb ␤ 3 (S123A) (Fig. 5). The results show that AP7.3 behaved like PAC1 in that it failed to bind to the mutant ␣ IIb ␤ 3 (S123A). AP7, on the other hand, bound to the mutant receptor as did OPG2. This finding provides additional, independent evidence that AP7.3 behaves like PAC1 rather than AP7.
Binding of Recombinant Fab Molecules to Purified ␣ IIb ␤ 3 -The selectivity of each Fab molecule for ␣ IIb ␤ 3 as opposed to ␣ V ␤ 3 and the dependence of binding on the RGD sequence were confirmed using the purified integrins in an enzyme-linked immunosorbent assay (Fig. 6). The binding of AP7, AP7.3 or PAC1 Fab molecules to ␣ IIb ␤ 3 in the presence of 1 mM Ca 2ϩ plus Mg 2ϩ was completely inhibited by Ն10 M RGDW or 5 mM EDTA. As had been observed with intact platelets, AP7.2 failed to bind to purified ␣ IIb ␤ 3 .
Important differences in the binding of these Fab molecules to purified ␣ IIb ␤ 3 were observed when the effect of divalent cations was analyzed. In the presence of 1 mM Ca 2ϩ plus Mg 2ϩ , comparable binding of AP7.1, AP7.3, and PAC1.1 molecules to ␣ IIb ␤ 3 was observed (optical density, 0.3-0.5) (Fig. 7A). In contrast, the absolute number of AP7 Fab molecules bound at saturation (optical density, 1.3-1.4) was 2-3-fold higher than that observed for any of the other three Fab molecules. However, in the presence of 10 M Mn 2ϩ , the maximal extent of binding of all four Fab molecules to ␣ IIb ␤ 3 was now equivalent (Fig. 7B). Nonetheless, significant differences in apparent affinity persisted, such that the concentrations of each Fab molecule at which half-saturation was achieved was markedly disparate. From the representative experiment shown in Fig.  7B, these approximate values were 0.2 g/ml for AP7, 2.5 g/ml for AP7.3, 10 g/ml for AP7.1, and 20 g/ml for PAC1.1. The differences in the numbers of ␣ IIb ␤ 3 molecules occupied by AP7.3 (or PAC1.1) Fab in the absence versus the presence of Mn 2ϩ must be due to the presence in the former case of stable conformers of ␣ IIb ␤ 3 in which the epitope remains inaccessible to the antibody.

DISCUSSION
This study is the first to show that the amino acid environment immediately adjacent to the RGD tripeptide in macromolecular ligands can determine whether ligand binding is sensitive to the activation state of ␣ IIb ␤ 3 . This finding excludes ligand size or valency as requisite factors responsible for affinity modulation of the receptor. The only difference between AP7, which does not bind selectively to activated ␣ IIb ␤ 3 , and AP7.3, which does, is the length and composition of flanking sequences in H3: AP7 contains HPFYrgdGGN, whereas AP7.3 contains RSPSYYrgdGAGP. The fact that AP7.3 behaves identicallyinparallelbindingexperimentstotheprototypeactivationdependent ligand PAC1 argues that a very subtle change in the position and/or orientation of the RGD sequence in this CDR loop is sufficient to modify the dependence of ligand binding on the conformational state of the receptor. These sequence differences would be expected to modify a number of conformational properties of the H3 loop, including the distance of the RGD sequence from the apex of the loop to the plane of the antigen-binding face and the orientations of the Arg and Asp side chains themselves.
X-ray crystallographic studies of OPG2, the parent molecule of AP7, have established that the distance from the apex of the H3 loop to the plane of the antigen-binding face is about 14 Å (16). The insertion of an additional two amino acids to the amino side of the H3 loop, as in the case in AP7.1 and AP7.3, may shift the position of the RGD sequence from the apex of the loop to a position closer to the carboxyl side of the loop. This would decrease the distance between the RGD sequence and the antigen binding face. This interpretation would support the contention of Beer et al. (17), who proposed that the ability of the RGD sequence to extend into a receptor pocket influences its ability to distinguish conformations of ␣ IIb ␤ 3 . Using a peptide approach, they demonstrated a direct relationship between the platelet activation dependence of peptide binding and the length of the spacer (Gly) n sequence between RGDF and the polyacrylonitrile beads to which the peptide was conjugated. Longer peptides (ՆG 9 RGDF) bound equally well to both activated and nonactivated platelets, whereas intermediate length peptides (e.g. G 3 RGDF) exhibited the greatest sensitivity to the conformation of ␣ IIb ␤ 3 (17). This raises the possibility that RGD recognition sites may be recessed from the ligand-binding surface of the ␣ IIb ␤ 3 molecule, perhaps situated within a shallow groove. In that case, platelet activation may alter the depth or size of this groove and thereby modulate accessibility of the relevant integrin contact sequences to RGD ligands.
The decrease in apparent affinity of AP7.3 relative to that of AP7 recapitulates the observed difference in apparent affinity between PAC1 and AP7. In a previous study, we showed that PAC1 Fab molecules exhibit a 60-fold decrease in apparent K d relative to the parent IgM (6). This difference was reasonably ascribed to loss of multivalency. However, in the present study, we observed that the conversion of AP7 to AP7.3 results in a comparable decrease in affinity. Half-maximal binding of AP7 Fab molecules to activated platelets in the presence of manganese occurred at roughly 1 g/ml (20 nM), whereas half-maximal binding of AP7.3 Fab molecules occurred at roughly 30 g/ml (600 nM) (Fig. 7B). Thus, a 30-fold decrease in affinity resulted simply by exchanging the H3 of PAC1.1 for the H3 of AP7. Not surprisingly, the affinity of AP7.3 Fab molecules was similar to that of recombinant PAC1.1 Fab molecules, in side by side comparisons. From these results, we conclude that there is a relationship between the affinity of an RGD ligand and its sensitivity to the conformational state of the integrin to which it binds. This is consistent with our hypothesis that increased accessibility of the ␣ IIb ␤ 3 recognition sites facilitates binding of even lower affinity ligands. The modified behavior of AP7.3, relative to AP7, should not be misconstrued as a coincidence resulting merely from a decrease in affinity. AP7.3 also acquires an altered specificity for the mutant integrin ␣ IIb ␤ 3 (S123A) that is identical to that of PAC1 but unlike that of AP7. Mutagenesis experiments (12) have clearly shown that alanine substitutions of clustered oxygenated ␤ 3 residues Asp 119 , Ser 121 , or Ser 123 do not influence heterodimer formation or surface expression but do alter fibrinogen, PAC1, and OPG2 binding. Although alanine substitution of Asp 119 or Ser 121 abolished binding of all three ligands, alanine substitution of Ser 123 abolished binding of fibrinogen and PAC1 but not OPG2. In this study, we show that AP7 behaves like OPG2 in this regard, because AP7 binding is abolished by alanine substitution of Asp 119 or Ser 121 (not shown) but not affected by alanine substitution of Ser 123 . The most likely explanation for these findings is that OPG2 and AP7 recognize a region of the integrin binding pocket that is slightly different from the RGD or PAC1 binding site. Because translocation of the PAC1 H3 loop sequence into AP7 confers PAC1 specificity to the resultant Fab molecule AP7.3, we conclude that the observed difference in specificity for the mutant S123A is due to amino acid sequence differences and/or the resultant structure differences of the H3 loop. These results provide independent confirmation that this limited sequence flanking the RGD site dictates the binding properties of AP7 versus AP7.3.
An intriguing result of our study was that activation-dependent ligands, such as AP7.3 and PAC1, bound to less than the total population of ␣ IIb ␤ 3 molecules on the platelet surface or within a preparation of isolated, purified ␣ IIb ␤ 3 . The maximum number of Fab molecules bound was dependent not only on the state of platelet activation but also on the divalent cation composition of the medium. With intact platelets, in the presence of millimolar calcium and magnesium, the number of AP7.3 molecules bound per platelet at saturation was only about 17% of the number of AP7 molecules that bound per platelet. In the presence of 10 M manganese, the total number of sites occupied by AP7.3 increased to 40% of that occupied by AP7. This difference between AP7 and AP7.3 was also seen when purified ␣ IIb ␤ 3 adsorbed to plastic microtiter wells was the antigen target. However, in the case of the purified integrin, the presence of manganese resulted in a comparable saturation of binding by both AP7 and AP7.3. These results provide direct evidence for the existence of subpopulations of platelet ␣ IIb ␤ 3 that assume distinctive phenotypes distinguishable by ligand and/or antibody binding, predicted by a number of prior studies (17,18). Presumably, ␣ IIb ␤ 3 can exist in distinct stable conformations even after it is isolated from the platelet membrane. These conformational states of ␣ IIb ␤ 3 remind one of similar findings by Chan and Hemler (19) with another integrin, ␣ 2 ␤ 1 .
In its basal state, ␣ IIb ␤ 3 does not engage soluble ligands and mediates platelet attachment selectively to surfaces coated with fibrinogen (20). Once the platelet is activated, however, the conformation and activity of ␣ IIb ␤ 3 changes, such that it can engage fibrinogen or other soluble ligands and mediate attachment to surfaces coated with von Willebrand factor, fibronectin, or vitronectin (21,22). Comparative binding studies of soluble fibrinogen and fibronectin suggest that conformational changes in ␣ IIb ␤ 3 are not "all or none" but that intermediate states in the conformational range of this receptor may further modulate the selectivity of the receptor for the soluble versus surface-bound conformation of any one ligand. The ultimate strength of the adhesive interaction between ligand and receptor in vivo may be influenced by structural determinants in the ligand other than the RGD sequence (23)(24)(25)(26), by post ligand binding events, including outside-in signaling (27)(28)(29), by multiple contact points on the receptor (12, 30 -33), and by hemodynamic forces (34). Despite this complexity, the similarities in binding between AP7.3, PAC1, and other natural ligands that contain RGD sequences and the fact that the structure of OPG2 has been solved by x-ray crystallography make the AP7 derivatives an excellent paradigm for future studies of the molecular basis of RGD ligand specificity and activation-induced changes in integrin conformation.